Energy Harvesting and Energy Efficiency: Nicu Bizon Naser Mahdavi Tabatabaei Frede Blaabjerg Erol Kurt Editors

Lecture Notes in Energy 37
Nicu Bizon
Naser Mahdavi Tabatabaei
Frede Blaabjerg
Erol Kurt Editors
Energy
Harvesting
and Energy
Efficiency
Technology, Methods, and Applications
Lecture Notes in Energy
Volume 37
Lecture Notes in Energy (LNE) is a series that reports on new developments in the
study of energy: from science and engineering to the analysis of energy policy. The
series’ scope includes but is not limited to, renewable and green energy, nuclear,
fossil fuels and carbon capture, energy systems, energy storage and harvesting,
batteries and fuel cells, power systems, energy efficiency, energy in buildings,
energy policy, as well as energy-related topics in economics, management and
transportation. Books published in LNE are original and timely and bridge between
advanced textbooks and the forefront of research. Readers of LNE include
postgraduate students and non-specialist researchers wishing to gain an accessible
introduction to a field of research as well as professionals and researchers with a
need for an up-to-date reference book on a well-defined topic. The series publishes
single and multi-authored volumes as well as advanced textbooks.
More information about this series at http://www.springer.com/series/8874

Nicu Bizon Naser Mahdavi Tabatabaei
•
Frede Blaabjerg Erol Kurt

•
Editors
Energy Harvesting
and Energy Efficiency
Technology, Methods, and Applications
123
Editors
Nicu Bizon Frede Blaabjerg
Faculty of Electronics, Communication, Department of Energy Technology
and Computers Aalborg University
University of Piteşti Aalborg East
Piteşti Denmark
Romania
Erol Kurt
Naser Mahdavi Tabatabaei Department of Electrical and Electronics
Electrical Engineering Department, Engineering, Faculty of Technology
Faculty of Engineering Gazi University
Seraj Higher Education Institute Ankara
Tabriz Turkey
Iran
ISSN 2195-1284 ISSN 2195-1292 (electronic)

Lecture Notes in Energy
ISBN 978-3-319-49874-4 ISBN 978-3-319-49875-1 (eBook)
DOI 10.1007/978-3-319-49875-1
Library of Congress Control Number: 2016959749
© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Foreword
Energy efficiency has always been a major challenge for the scientist and the
engineers. However, in recent years, the increased public concern for the preser-
vation of natural resources and the protection of the environment has strongly
stimulated the research and development activities in this area. More than 20,000
technical papers written in 2016 and stored in “Scopus” database have “energy
efficiency” in their title, in their abstract or in their list of keywords. There were less
than 4000 in 1996.
Energy harvesting is a much more recent topic. Less than 100 papers addressed
it twenty years ago. More than 2500 articles were published in top-level journals
from January to October 2016. The reason behind this spectacular growth is simple:
Both academia and industry are interested in the design and engineering of
energy-autonomous small electronic devices that can harvest the various forms of
energy available in the environment (solar, eolian, and hydraulic) and convert them
to electric power. The development of energy-harvesting applications is driven by
the increased need of autonomous wireless electronic systems in various fields of
human activities, ranging from medicine and aeronautics to civil engineering and
animal tracking.
Textbooks and monographs are already available for anyone who wants to learn
more on either “energy efficiency” or “energy harvesting.” The merit of this book is
that it brings together the two topics, which are more and more interrelated. The
editors carefully selected the topics to be treated, and each chapter of this book is
written by well-recognized experts in the field. This book introduces the reader to
up-to-date research on nonlinearity of energy-harvesting systems, energy efficiency
of hybrid power systems, and optimal design of autonomous electronic systems. It
also contains instructive case studies and examples of experimental validation of the
novel energy-saving or energy-harvesting techniques.
v
vi Foreword
This book can be used in the classroom, to teach energy management courses to
graduate students, and be suggested as further reading to undergraduate students in
engineering sciences. It will also be a valuable information resource for the
researchers and engineers concerned by energy efficiency issues or involved in the
development and application of energy-harvesting techniques.
October 2016 Lucian Dascalescu

IEEE Fellow, Distinguished Professor
of Electrical Engineering
University of Poitiers, University Institute
of Technology, Angoulême, France
Preface
Energy harvesting and energy efficiency are two key topics for today’s power
community. In the development of modern society, one of the key factors is to save
energy in order to become more independent of other resources. Two important
approaches can be taken—one is to change behavior and thereby save energy and
the second is to develop new technology which is able to save energy in different
applications. Chapter 1 gives an overview of challenges and possibilities in terms of
energy saving and also energy efficient use.
Initially, the first key topic—energy harvesting—becomes one of the most
motivated fields of the multidisciplinary science due to the complicated features
of the harvester materials, dependences on various mechanical, electrical, and
magnetic parameters, rich responses on different external excitation frequencies and
strength. Strictly speaking, vibrations stem from either man-made systems or nat-
ural processes can be used as an important electric resource for
low-power-consuming electronic devices such as transducers and wireless sensors.
That can contribute at the batteryless applications for much sustainable and
renewable power generation, whereas some technical problems should be solved to
achieve the expectations of the electronics society. Although conventional har-
vesters work on the basis of linear resonance, there exist certain parametrical
limitations on their power generation. Indeed, excitation frequency, electrical load,
manufacturing tolerance, and ambient temperature play important roles in order to
determine the optimized energy generation. Besides, the nonlinear nature of the
vibration phenomena contributes at the power, and these nonlinear effects cannot be
neglected for an optimized harvester system. Thus, Part I of this book initially gives
an outline to the reader on the electromagnetic and piezoelectric energy-harvesting
systems and then focuses on the theoretical and experimental techniques by
introducing different harvester systems.
In that context, Chap. 2 describes the harvesting sources with classical and novel
types for the use of electromagnetic and piezoelectric hybrid structures. Various
vii
viii Preface
experimental systems are described in detail in order to compare their output

powers and their relation to the system parameters.
The batteryless applications of microscale harvesters have been explained in
Chap. 3 for information technologies (IT). The importance of low-power harvesting
systems for IT applications is particularly emphasized, and model systems have
been discussed. One of the important practice areas of the harvesters is wireless
sensor application. Therefore, a specific chapter (i.e., Chap. 4) is dedicated to the
problems of electromagnetic and piezoelectric harvesters in wireless devices. This
chapter gives both experimental and theoretical details on the matter. As the har-
vester systems have complicated equilibrium features for their time- and
space-dependent nature, nonlinearity plays an important role to identify their
dynamic behavior and power-generation strategy. Therefore, Chap. 5 is devoted to
the nonlinear problems of the harvesters. Although the energy-harvesting issues
mostly cover the systems related to the piezoelectric and electromagnetic ones in
low power range, the most frequent energy-generation system—photovoltaics
(PV) has been an important topic. Therefore, Chap. 6 focuses on the control phe-
nomena of PV hybrid systems.
It can be emphasized that the chapters mentioned above provide a good back-
ground to the reader on the harvester systems and their applications. Both experi-
mental and theoretical approaches to different harvesting problems help to
understand the advanced problems and cutting-edge information, world widely,
thereby the readers at different educational levels from undergraduate to the pro-
fessionals can find interesting research topics in order to apply in their own studies.
Other main topic of this book is the energy efficiency. Due to the increasing
population and industrial growth, energy efficiency has become a popular topic for
every level of communities from ordinary to technical. There exist many attempts
today that the energy efficiency itself can be counted as a new energy resource.
Thus, interdisciplinary studies, which have been carried out in the fields of
renewable energy, focus on different mechanisms that decrease the losses of the
energy in methodological ways. In light of the present technology, the efficiency
cannot be considered detached from the cost. Strictly speaking, the balance between
the efficiency and system cost should be ascertained. With that respect, many
energy systems such as solar, wind, and tidal can make use of good-quality
materials or efficiency techniques if they are financially appropriate. Therefore,
Chaps. 7–9 are devoted to the sun-tracking applications and maximal power point
tracking (MPPT) techniques in PVs. In these chapters, both practices and theoretical
backgrounds on the tracking mechanisms are presented including the case studies.
Chapter 10 mentions the partial shading effect on the PV systems and clarifies the
methodology on the solution of MPPT for those systems. The applications on solar
cars are presented in Chap. 11. This chapter also sheds a light on the polymer
composite materials in order to enhance the efficiency and gives some information
on the charging stations.
The increasing demand for electricity supply along with higher requirements for
power quality and system reliability, restrictions to use the available fossil fuels,
and minimization of the environmental pollutants leads to the aggregation of clean
Preface ix
energy sources (renewable energy sources, fuel cell, etc.) in distributed generation
systems and developing microgrids. Consequently, the energy efficiency of hybrid
power system that integrates such clean energy sources must be improved through
appropriate energy management strategies. Thus, the remaining parts of the book,
namely II, III, and IV, analyze the energy efficiency based on fuel cell, PV, wind,
and hybrid power systems.
The term “hybrid” means the use of other energy storage devices, or multiple
input energy sources in hybrid power sources to sustain the load demand. Thus, the
use of fuel cell system as energy source or energy storage devices in conjunction
with an electrolyzer is analyzed in Chaps. 12 and 13. While Chap. 12 analyzes the
possibility to use the extremum seeking control schemes for the reduction of
hydrogen consumption in fuel cell hybrid power sources, Chap. 13 analyzes the
efficiency of a fuel cell hybrid power source required for an automotive application.
Chapter 14 proposes a stochastic model to analyze the microgrids with the goal of
profit maximization and imbalance cost minimization. In this framework, a new
method based on neural network theory is proposed for predicting wind speed and
solar radiation. Other chapter (i.e., Chap. 15) analyzes the energy efficiency of a
micro-combined cooling, heating, and power system driven by a solar dish stirling
heat engine that is used for residential buildings. A novel methodology was
introduced for short-term scheduling of small-scale trigeneration system, which can
be used optimally and efficiently to provide cooling, heating, and power for resi-
dential applications, being environmentally friendliness, cost-cutting, and on-site
applied.
The last part of this book is dedicated to some technical strategies, efficient
methods, and applications in field of energy efficiency, so it will be of interest for all
current researchers and specialists in that field as well as for technicians.
Chapter 16 presents wired and wireless communication systems in smart homes
and buildings based on the recent developments proposed in applications. The basic
principles of the smart homes and energy efficient buildings are introduced firstly in
order to provide basic knowledge for readers and the chapter also gives an idea on
the communication systems used for outdoor and indoor scenarios. Chapters 17 and
18 propose new flexible hybrid architecture for the power-conditioning unit for
small satellites. Since the space agencies all over the world are interested today in
very small satellites due to their advantages compared to heavier satellites, the
advanced techniques are discussed including their converter and storage systems.
The batteries are unavoidable for any electricity system. Therefore, while
making a discussion on the efficiency issues, one should also consider the storage
techniques. Chapter 19 introduces a new method for determining the optimal model
of batteries, puts a starting point in analyzing their discharge profiles, and employs a
multicriteria analysis for processing the experimental data.
While considering the efficiency in solar, fuel cell and related hybrid systems,
the energy efficiency in wind and water distribution systems should also be men-
tioned. In this manner, the optimal planning and operation of water distribution is
presented in Chap. 20. This problem mainly involves the establishment of the
operation schedule for all water hydrophore stations and uses a database of 85 urban
x Preface
water hydrophore stations as a case study. Finally, the last chapter (i.e., Chap. 21)
provides an overview about available knowledge, references, and investigations on
the active and passive flow control devices, initially developed for aeronautical
industry that are currently being investigated and introduced on wind turbines in
order to improve their efficiency.
As a conclusion, a sustained research in the field of energy efficiency does not
only give more chances to significant reduction of carbon dioxide, greenhouse gas
emissions, and environmental pollution, but also increases the economic saving in
fuel consumption and use of energy sources. Therefore, this book tries to highlight
the difficulties of the basic methods on energy harvesting and energy efficiency and
proposes advanced methods to solve these issues. All proposed methods were
validated through simulation and experimental results. These “hot subjects” will be
of interest for many decades and, at the same time, will be a challenge and hard task
for the researchers all over the world, considering the new energy policies due to
energy crisis.
We hope that this book will be very efficient for students and engineers who
learn and wish to work in this field, because the chapters of this book cover all
important and challenging subjects related to energy harvesting and energy effi-
ciency. The book comprises the knowledgeable and up-to-date contents that present
the state-of-the-art equipment and methods used for the energy harvesting and
energy efficiency. Finally, the main arguments that may recommend this book to be
read are the following: (1) It is the first comprehensive book on energy harvesting
and energy efficiency of the power hybrid systems; (2) covers the operating prin-
ciples, design methods, and real applications; (3) enables the low power for
autonomous electronic system design; (4) introduces the high-power density
technology and adiabatic concept to efficiently design the mission critical systems;
(5) provides a much-needed system approach to hydrogen energy applications;
(6) provides a comprehensive overview of the fundamentals of renewable power
generation, conversion, and storage; and the last, but not the least, (7) can be used
as a course text.
The editors and authors made all efforts to have a good book, and we hope
interested readers to enjoy by reading this book and to be satisfied by its content.
Piteşti, Romania Nicu Bizon

Tabriz, Iran Naser Mahdavi Tabatabaei
Aalborg, Denmark Frede Blaabjerg
Ankara, Turkey Erol Kurt
Contents
1 Energy Saving and Efficient Energy Use By Power Electronic

Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Frede Blaabjerg, Huai Wang, Pooya Davari, Xiaohui Qu
and Firuz Zare
Part I Energy Harvesting

2 Hybrid Energy Harvesters (HEHs)—A Review. . . . . . . . . . . . . . . . . 17
Nazenin Gure, Abdulkerim Kar, Erturul Tacgin, Alper Sisman
and Naser Mahdavi Tabatabaei
3 Micro-scale Energy Harvesting for Batteryless Information
Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Ali Muhtaroğlu
4 Efficient Energy Harvesting Systems for Vibration
and Wireless Sensor Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Mustafa Doğan, Sıtkı Çağdaş İnam and Ö. Orkun Sürel
5 Nonlinear Problems in Piezoelectric Harvesters Under Magnetic
Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Erol Kurt and Yunus Uzun
6 Energy Harvesting from the Photovoltaic Hybrid Power Source
Based on Extremum Seeking Control Schemes . . . . . . . . . . . . . . . . . 143
Nicu Bizon, Marian Raducu, Luminita-Mirela Constantinescu
and Mihai Oproescu
Part II Energy Efficiency of the Photovoltaic Systems

7 Improving Tracking Efficiency of Two-Axis Sun Tracking
Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Fevzi Kentli and Musa Yilmaz
xi
xii Contents
8 Maximum Power Point Tracking (MPPT) Algorithms

for Photovoltaic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Ersan Kabalci
9 Photovoltaic System: Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Ali Durusu, Ismail Nakir and Mugdesem Tanrioven
10 Maximum Power Point Tracking Algorithms for Partial
Shaded PV Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Ibrahim Sefa, Necmi Altin and Saban Ozdemir
11 Solar Energy Harvesting in Electro Mobility . . . . . . . . . . . . . . . . . . 293
Aytaç Gören
Part III Energy Efficiency of the Hybrid Power System

12 Energy Harvesting from the Fuel Cell Hybrid Power Source
Based on Extremum Seeking Control Schemes . . . . . . . . . . . . . . . . . 329
Nicu Bizon
13 Energy Efficiency of PEM Fuel Cell Hybrid Power Source . . . . . . . 371
Nicu Bizon and Mircea Raceanu
14 Integration and Management Technique of Renewable
Energy Resources in Microgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Hossein Shayeghi and Elnaz Shahryari
15 Optimal Planning of a Micro-combined Cooling, Heating
and Power System Using Air-Source Heat Pumps
for Residential Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Farkhondeh Jabari, Behnam Mohammadi-Ivatloo
and Mohammad Rasouli
Part IV Technical Strategies, Efficient Methods and Applications

16 Communication Methods for Smart Buildings
and Nearly Zero-Energy Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . 459
Yasin Kabalci
17 Power Architectures and Power Conditioning Unit
for Very Small Satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
Sergiu Oprea, Constantin Radoi, Adriana Florescu,
Andrei-Stefan Savu and Adrian-Ioan Lita
18 Power Conversion and Energy Management
for Mission-Critical Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541
Andrei-Stefan Savu, Adrian-Ioan Lita, Constantin Radoi,
Adriana Florescu, Sergiu Oprea and Ioan Lita
Contents xiii
19 Determining the Optimal Battery Model for a Specific

Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573
Bogdan-Adrian Enache
20 Electrical Energy Consumption Forecasting to Improve
Energy Efficiency of Water Distribution Systems . . . . . . . . . . . . . . . 599
Gheorghe Grigoras
21 Flow Control Devices for Wind Turbines . . . . . . . . . . . . . . . . . . . . . 629
Iñigo Aramendia, Unai Fernandez-Gamiz,
Jose Antonio Ramos-Hernanz, Javier Sancho,
Jose Manuel Lopez-Guede and Ekaitz Zulueta
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
Contributors
Necmi Altin Department of Electrical and Electronics Engineering, Faculty of

Technology, Gazi University, Ankara, Turkey
Iñigo Aramendia Department of Nuclear Engineering and Fluid Mechanics,
University of the Basque Country, Vitoria-Gasteiz, Araba, Spain
Nicu Bizon Department of Electronics, Computers and Electrical Engineering,
University of Piteşti, Piteşti, Romania; University Politehnica of Bucharest,
Bucharest, Romania
Frede Blaabjerg Department of Energy Technology, Center of Reliable Power
Electronics (CORPE), Aalborg University, Aalborg, Denmark
Luminita-Mirela Constantinescu University of Piteşti, Piteşti, Romania
Pooya Davari Department of Energy Technology, Center of Reliable Power
Mustafa Doğan Department of Electrical and Electronics Engineering, Baskent
University, Ankara, Turkey
Ali Durusu Yildiz Technical University, Istanbul, Turkey
Bogdan-Adrian Enache Faculty of Electronics, Communications and Computers,
University of Piteşti, Piteşti, Romania
Unai Fernandez-Gamiz Department of Nuclear Engineering and Fluid
Mechanics, University of the Basque Country, Vitoria-Gasteiz, Araba, Spain
Adriana Florescu University POLITEHNICA of Bucharest, Bucharest, Romania
Gheorghe Grigoras Department of Power System, Faculty of Electrical
Engineering, “Gheorghe Asachi” Technical University of Iasi, Iasi, Romania
Nazenin Gure Department of Mechanical Engineering, Faculty of Engineering,
Marmara University, Istanbul, Turkey; Enhas R&D Energy Systems Ind. Co. Ltd,
Istanbul, Turkey
xv
xvi Contributors
Aytaç Gören Automatic Control and Robotics Laboratories, Department of

Mechanical Engineering, Dokuz Eylul University, Izmir, Turkey
Sıtkı Çağdaş İnam Department of Electrical and Electronics Engineering,
Baskent University, Ankara, Turkey
Farkhondeh Jabari Faculty of Electrical and Computer Engineering, University
of Tabriz, Tabriz, Iran
Ersan Kabalci Department of Electrical and Electronics Engineering, Faculty of
Engineering and Architecture, Nevsehir HBV University, Nevsehir, Turkey
Yasin Kabalci Department of Electrical and Electronics Engineering, Faculty of
Engineering, Omer Halisdemir University, Nigde, Turkey
Abdulkerim Kar Department of Mechanical Engineering, Faculty of Engineering,
Marmara University, Istanbul, Turkey
Fevzi Kentli Department of Mechatronics Engineering, Technology Faculty,
Erol Kurt Department of Electrical and Electronics Engineering, Technology
Faculty, Gazi University, Teknikokullar, Ankara, Turkey
Adrian-Ioan Lita University Politehnica of Bucharest, Bucharest, Romania
Ioan Lita University of Piteşti, Piteşti, Romania
Jose Manuel Lopez-Guede Department of Systems Engineering and Automatics,
Behnam Mohammadi-Ivatloo Faculty of Electrical and Computer Engineering,
University of Tabriz, Tabriz, Iran
Ali Muhtaroğlu Department of Electrical and Electronics Engineering, Middle
East Technical University, Güzelyurt, Mersin, Turkey
Ismail Nakir Yildiz Technical University, Istanbul, Turkey
Sergiu Oprea University POLITEHNICA of Bucharest, Bucharest, Romania
Mihai Oproescu University of Piteşti, Piteşti, Romania
Ö. Orkun Sürel Department of Electrical and Electronics Engineering, Baskent
Saban Ozdemir Vocational School of Technical Sciences, Gazi University,
Ankara, Turkey
Xiaohui Qu School of Electrical Engineering, Southeast University, Nanjing,
China
Contributors xvii
Mircea Raceanu National Research and Development Institute for Cryogenics

and Isotopic Technologies, Râmnicu Vâlcea, Romania; University Politehnica of
Bucharest, Bucharest, Romania
Constantin Radoi University POLITEHNICA of Bucharest, Bucharest, Romania
Marian Raducu University of Piteşti, Piteşti, Romania
Jose Antonio Ramos-Hernanz Department of Electrical Engineering, University
of the Basque Country, Vitoria-Gasteiz, Araba, Spain
Mohammad Rasouli Department of Electrical and Computer Engineering, School
of Engineering, Penn State University, Erie, PA, USA
Javier Sancho Department of Nuclear Engineering and Fluid Mechanics,
Andrei-Stefan Savu University POLITEHNICA of Bucharest, Bucharest,
Romania
Ibrahim Sefa Department of Electrical and Electronics Engineering, Faculty of
Technology, Gazi University, Ankara, Turkey
Elnaz Shahryari Department of Technical Engineering, University of Mohaghegh
Ardabili, Ardabil, Iran
Hossein Shayeghi Department of Technical Engineering, University of
Mohaghegh Ardabili, Ardabil, Iran
Alper Sisman Department of Electrical and Electronics Engineering, Faculty of
Engineering, Marmara University, Istanbul, Turkey
Naser Mahdavi Tabatabaei Department of Electrical Engineering, Faculty of
Engineering, Seraj Higher Education Institute, Tabriz, Iran
Erturul Tacgin Department of Mechanical Engineering, Faculty of Engineering,
Mugdesem Tanrioven Yildiz Technical University, Istanbul, Turkey
Yunus Uzun Department of Electrical and Electronics Engineering, Faculty of
Engineering, Aksaray University, Aksaray, Turkey
Huai Wang Department of Energy Technology, Center of Reliable Power
Musa Yilmaz Department of Electronics and Communications Engineering,
Technology Faculty, Batman University, Batman, Turkey
Firuz Zare Power and Energy Group, The University of Queensland, Brisbane,
QLD, Australia
Ekaitz Zulueta Department of Systems Engineering and Automatics, University
of the Basque Country, Vitoria-Gasteiz, Araba, Spain
Abbreviations and Acronyms
AAHP Air-to-Air Heat Pump

ABC Artificial Bee Colony
AC Alternative Current
ACA Ant Colony Algorithm
AcVG Actuator Vortex Generator
ADC Analog-to-Digital Conversion
AES Advanced Encryption Standard
aESC Advanced Extremum Seeking Control
AF Active Filtering
AFC Active Flow Control
AFE Active Frontend
AGS Automatic Generating System
AI Artificial Intelligence
AJVG Air Jet Vortex Generator
AM Amplitude Modulation
AMI Advanced Metering Infrastructure
AMM Automatic Meter Management
AMR Automatic Meter Reading
ANN Artificial Neural Network
AP Access Point
ARIB Association of Radio Industries and Businesses
ASD Adjustable Speed Drive
ASHP Air Source Heat Pump
a-Si Amorphous Silicon
ASIC Application-Specific Integrated Circuit
ASM Assembly
BAN Body Area Network
BB Broadband
BCDMU Battery Charge/Discharge Monitor Unit
BCU Battery Charger Unit
xix
xx Abbreviations and Acronyms
BDR Battery Discharge Regulator

BFV Best Fixed Voltage
BGA Ball Grid Array
BLDCM Brushless Direct Current Motor
BMS Battery Management System
BOL (Battery’s) Beginning of Life
BoP Balance of Plant
BP Belief Propagation
BPF Band-Pass Filter
bpfESC Band-Pass Filter ESC
BPSK Binary Phase Shift Keying
BSN Body Sensor Nod
BU Battery Unit
CAD Computer-Aided Design
CART Classification and Regression Trees
CC Constant Current
CCHP Combined Cooling Heating and Power
CCK Complementary Code Keying
CCM Continuous Current Mode
CCP Combined Cool and Power
CD Charge Depletion
CDRG Coulomb-Damped Resonant Generator
CFD Computational Fluid Dynamics
CFL Compact Fluorescent Lamp
CHP Combined Heat and Power
CI Charge Increasing
CMMR Common Mode Rejection Ratio
COA Centroid of Area
COE Cost of Energy
COFDM Coded OFDM
COTS Commercial Off-the-Shelf
CP Cyclic Prefix
CS Charge Sustaining
CSMA Carrier Sense Multiple Access
CSMA/CA Carrier Sense Multiple Access/Collision Avoidance
CSMA/CD Carrier Sense Multiple Access/Collision Detection
CST Current Sweep Technique
CT Copper Thickness
CU Chargers’ Unit
CV Constant Voltage
CVaR Conditional Value at Risk
D8PSK Differential 8-Phase Shift Keying
DA Day Ahead
DAQ Data Acquisition
DBPSK Differential Binary Phase Shift Keying
Abbreviations and Acronyms xxi
DC Direct Current
DCM Discontinuous Conduction Mode
DCSK Differential Chaos Shift Keying
DET Direct Energy Transfer
DG Distributed Generation
DM Damper Motor
DMS Data Management System
DNLP Discontinuous Nonlinear Program
DO Disjoint Operation
DOD Depth of Discharge
DOF Degree of Freedom
DPE Direct Piezoelectric Effect
DQPSK Differential Quadrature Phase Shift Keying
DRIBA Double-Regulated Intermediate Bus Architecture
DS Delay Stall
DSL Digital Subscriber Line
DTU Denmark Technical University
DVS Dynamic Voltage Scaling
e.i.r.p. Equivalent Isotropic Radiated Power
EA Evolutionary Algorithms
EH Energy Harvester
EHing Energy Harvesting
EM Electromagnetic
EMF Electromotive Force
EMHs Electromagnetic Energy Harvesters
EMI Electromagnetic Interference
EMP Electromagnetic Pulse
EMS Energy Management Strategy
EOL (Battery’s) End of Life
EP Evolutionary Programming
EQ Equivalence
ES Energy Sources
ESA European Space Agency
ESC Extremum Seeking Control
ESR Equivalent Series Resistance
ESS Energy Storage System
EU European Union
EV Electric Vehicle
EWEA Energy Wind Energy Association
FC Fuel Cell
FCC Federal Communications Commission
FCHPS Fuel Cell Hybrid Power Source
FEA (FEM) Finite Element Analysis (Finite Element Method)
FEC Forward Error Correction
FEM Finite Element Method
xxii Abbreviations and Acronyms
FFT Fast Fourier Transform

FLC Fuzzy Logic Controller
FSK Frequency Shift Keying
FT Fuzzy Techniques
FW Freewheeling Diode
GA Genetic Algorithm
GAMS General Algebraic Modeling System
GaN Gallium Nitride
GEO Geosynchronous Earth Orbit
GFSK Gaussian Frequency Shift Keying
GMPP Global Maximum Power Point
GMPPT Global Maximum Power Point Tracking
GP Global Peak
GSTP General Support Technology Program
GUI Graphical User Interface
H1 First Harmonic
HAN Home Area Network
HAWT Horizontal Axis Wind Turbine
HC Hill Climbing
HEHs Hybrid Energy Harvesters
HEO Highly Elliptical Orbit
HF High Frequency
HID High Intensity Discharge
hoESC High-Order Extremum Seeking Control
HPF High-Pass Filter
HPS Hybrid Power Source
HRTHs Hybrid Rotary-Translational Harvesters
HSAT Horizontal Single Axis Tracking
HT Hydro-Turbines
IC Integrated Circuit
ICE Internal Combustion Engine
ICT Information and Communication Technologies
IEA International Energy Agency
IGBT Insulated-Gate Bipolar Transistor
IncCond Incremental Conductance
IOD In-Orbit Demonstration
IRRWBF Implementation-Efficient Reliability Ratio-Based Weighted Bit
Flipping
ITU International Telecommunication Union
IU Isolation Unit
I-V Current–Voltage
IWBF Improved Weighted Bit Flipping
LAN Local Area Network
LDPC Low-Density Parity Check
LDR Light-Dependent Resistor
Abbreviations and Acronyms xxiii
LE Leading Edge
LED Light-Emitting Diode
LEO Low-Earth Orbit
LET Linear Energy Transfer
LF Low Frequency
LFP LiFePO4 Battery
LLR Log-Likelihood Ratio
LP Local Peak
LPF Low-Pass Filter
LRCM Linear Reoriented Coordinates Method
M2M Machine-to-Machine
MAC Media Access Control
MAPE Mean Absolute Percentage Error
MC Mid-Chord
MCA Multicriteria Analysis
MCP Marginal Clearing Price
MCU Microcontroller Unit
MDO Multidisciplinary Design Optimization
MEMS Micro-Electro-Mechanical Systems
MEO Mid-Earth Orbit
MEP Maximum Efficiency Point
mESC Modified Extremum Seeking Control
MG Microgrid
MILP Mixed-Integer Linear Programming
MLI Multilevel Inverter
MLP Multilayer Perceptron
MPP Maximum Power Point
MPPT Maximum Power Point Tracker
MSE Mean Square Error
MT Micro-Turbine
MTBF Mean Time between Failures
NB Narrowband
NGSO Non-Geostationary Orbit
NN Neural Network
NOCT Normal Operating Cell Temperature
NREL National Renewable Energy Laboratory
NTC Negative Temperature Coefficient
O&M Operation and Maintenance
OC Only Current Photovoltaic
OCC One-Cycle Control MPPT
OCV Open-Circuit Voltage
ODE Ordinary Differential Equations
OFDM Orthogonal Frequency Division Multiplexing
OOK On–Off Keying
OPGW Optical Power Ground Wire
xxiv Abbreviations and Acronyms
O-QPSK Offset Quadrature Phase Shift Keying

ORC Organic Rankine Cycle
ORING OR-ing Output Logic Function
OV Open Voltage
P&Q Perturb and Observe
PAGV Power Augmentation Guide Vane
PAM Pulse Amplitude Modulation
PAN Personal Area Network
PASAT Polar-Aligned Single Axis Tracking
PC Personal Computer
PCB Printed Circuit Board
PDE Partial Differential Equations
PE Piezoelectric
PEC Packet Error Check
PEHs Piezoelectric Energy Harvesters
PEM Proton Exchange Membrane
PEMFCS Proton Exchange Membrane Fuel Cell Stack
PET Polyethylene Terephthalate
PF Passive Filtering
PFC Power Factor Correction
PG Pressure Growth
PI Proportional Integral
PID Proportional Integral Derivative
PLC Power Line Communication
POL Point of Load
POL/D Point-of-Load/Distribution
POU Point of Use
PPM Pulse Position Modulation
PSC Partially Shading Condition
PSO Particle Swarm Optimization
PV Photovoltaic
P-V Power-Voltage
PVGJ Pulsed Vortex Generator Jet
PVHPS Photovoltaic Hybrid Power Source
PWM Pulse Width Modulation
PZT Lead Zirconate Titanate
QAM Quadrature Amplitude Modulation
QPSK Quadrature Phase Shift Keying
RANS Reynolds-Averaged Navier Stokes
RC Resistance–Capacitor
RCC Ripple Correlation Control
RES Renewable Energy Sources
RF Radio Frequency
RLP Representative Loading Profiles
RMS Root Mean Square
Abbreviations and Acronyms xxv
RS Reed Solomon
RWT Reference Wind Turbine
S3R Sequential Switching Shunt Regulator
SAPV Stand-Alone Photovoltaic
SAT Single Axis Tracking
SBEMS Smart Building Energy Management Systems
SCPB Short-Current Pulse-Based
SCRIMP™ Seemann Composites Resin Infusion Molding Process
SDSHE Solar Dish Stirling Heat Engine
SEB Single Event Burnout
SEBP Single Event Burnout Phenomenon
SECE Synchronous Electric Charge Extraction
SEE Single Event Phenomenon
SEGR Single Event Gate Rupture
SEL Single Event Latch-up
SEPIC Single-Ended Primary Inductor Converter
SEU Single Event Upset
S-FSK Spread Frequency Shift Keying
SGEMP System-Generated EMP
SGI Silhouette Global Index
Si Silicon
SiC Silicon Carbide
SIDO Single İnductor Dual Output
SM Smart Meter
SME Shape Memory Effect
SMPS Switched Mode Power Supply
SoC State of Charge
SOFC Solid Oxide Fuel Cell
SoH State of Health
SS Spread Spectrum
SSPB Single-Supply Pre-Biasing
SST Shear Stress Transport
STAR Space Technology and Advanced Research
STC Standard Test Conditions
TDD Time Division Duplexing
TDMA Time Division Multiple Access
TE Trailing Edge
TEG Thermoelectric Generator
TENG Triboelectric Nanogenerator
THD Total Harmonic Distortion
TID Total Ionizing Doze
TSAT Tilted Single Axis Tracking
xxvi Abbreviations and Acronyms
TT&C Telemetry, Tracking, and Telecommand

TVAC-PSO Time Varying Acceleration Coefficients Particle Swarm
Optimization
UC Ultracapacitor
UDDS Urban Dynamometer Driving Schedule
VARIM Vacuum-Assisted Resin Infusion Molding
VAWT Vertical Axis Wind Turbine
VBRTM Vacuum Bag Resin Transfer Molding
VDRG Velocity-Damped Resonant Generator
VG Vortex Generator
VRM Voltage-Regulated Module
VSAT Vertical Single Axis Tracking
W Daily Electrical Energy Consumption
WAN Wide Area Network
WAST Warm Air Storage Tank
WBF Weighted Bit Flipping
WBG Wide Band-Gap
WF Daily Water Flow
WiSH Wind-Solar Hybrid
WLAN Wide Local Area Network
WM-Bus Wireless M-Bus
WPAN Wireless Personal Area Network
WSN Wireless Sensor Network
WT Wind Turbine
ZCS Zero Current Switching
ZVS Zero Voltage Switching
Chapter 1
Energy Saving and Efficient Energy Use
By Power Electronic Systems
Frede Blaabjerg, Huai Wang, Pooya Davari, Xiaohui Qu

and Firuz Zare
Abstract In the development of the modern society, one of the key factors is to
save energy in order to become more independent of other energy resources. Two
important approaches can be taken—one is to change behavior and thereby save
energy—the second one is to develop new technology which is able to save energy
in different applications. This chapter will give an overview of challenges and
possibilities in terms of energy saving and also energy efficient use. This includes a
discussion on high efficiency power electronics devices and the systems they are
used for energy loss reduction. The key enabling technologies are power elec-
tronics, Information and Communication Technology (ICT) as well as systems to
carry the electrical energy through power transmission, conversion and distribution.
A couple of examples will be given to demonstrate the energy saving possibilities
by power electronics systems, such as in the applications of adjustable speed drives
and solid-state lighting systems. Power electronics gives also a high flexibility when
renewable power production is introduced to electrical energy systems. Future
research opportunities and challenges will finally be discussed.
F. Blaabjerg H. Wang P. Davari (&)

Department of Energy Technology, Center of Reliable Power Electronics (CORPE),
Aalborg University, Aalborg, Denmark
e-mail: pda@et.aau.dk
F. Blaabjerg
e-mail: fbl@et.aau.dk
H. Wang
e-mail: hwa@et.aau.dk
X. Qu
School of Electrical Engineering, Southeast University, Nanjing, China
e-mail: xhqu@seu.edu.cn
F. Zare
Power and Energy Group, The University of Queensland, Brisbane, QLD, Australia
e-mail: f.zare@uq.edu.au
© Springer International Publishing AG 2017 1

N. Bizon et al. (eds.), Energy Harvesting and Energy Efficiency,
Lecture Notes in Energy 37, DOI 10.1007/978-3-319-49875-1_1
2 F. Blaabjerg et al.

Keywords Electrical energy Power electronics Adjustable speed drive Wide

band-gap power devices Intelligent lighting
Abbreviation and Acronyms

AF Active Filtering
AFE Active Front-End
AM Amplitude Modulation
ASD Adjustable Speed Drive
CFL Compact Fluorescent Lamp
GaN Gallium Nitride
HID High Intensity Discharge
ICT Information and Communication Technology
IEA International Energy Agency
LED Light-Emitting Diode
PF Passive Filtering
PFC Power Factor Correction
PWM Pulse-Width Modulation
Si Silicon
SiC Silicon Carbide
WBG Wide Band-Gap
1.1 Introduction
Among the technologies that support the growth of clean energy and the
improvement of energy efficiency, power electronics has been representing a major
enabler. It provides efficient conversion and flexible control of electrical energy.
The power electronics converters make possible the connection of renewable
energy generators to the legacy power systems, as schematically shown in Fig. 1.1,
and efficiency improvement of energy harvesting through dedicated controls.
Furthermore, power electronics is extensively used on the consumer side and is a
core technology for the new smart grid [1]. It unlocks the potentials of energy
savings in motor drives, buildings (e.g., lighting, energy-efficient appliances),
transportation (e.g., electric vehicles, high speed trains), consumer electronics (e.g.,
computers, cell phones), etc.
1 Energy Saving and Efficient Energy Use …
Fig. 1.1 The role of power electronics in electrical energy generation, transmission, distribution, and consumption [1]
3
1.2 Electrical Energy Generation and Consumption
Today more than 40% of all energy consumption is in the form of electrical energy,
which is expected to grow to 60% by 2040 [2]. The generation of the electrical
energy is becoming more renewable-based as shown in Fig. 1.2, which is according
to the projection by the International Energy Agency (IEA) [3]. The power gen-
eration capacity worldwide is expected to increase by 80%, from 5,952 GW in 2014
to about 10,700 GW in 2040. During this period, about 2,250 GW power capacity
currently in operation and about 200 GW capacity from newly commissioned
renewables is expected to be retired due to the end of service life. The cost gap
between electricity generated from renewable sources and fossil fuels is narrowing
as discussed in [1]. Connection of the electrical energy generated from renewable
energy sources (e.g., wind power and photovoltaics) to power grids is not possible
without power electronics based energy conversion systems.
The generated electrical energy is consumed by different sectors, such as
industry, residential, and transportation. The transportation sector sees a clear
electrification both in cars and ships, and also in air-planes which is made possible
by means of power electronics. In terms of applications, motor drives, and lighting
Fig. 1.2 Power generation capacity flow by source from 2014 to 2040 projected by the
International Energy Agency (IEA) (*over the projection period, a portion of renewable additions
is retired, with the assumption of 25 years of lifetime for wind farms and solar photovoltaics
plants) [3] © OECD/IEA 2014, World Energy Outlook 2014, IEA Publishing. Licence: www.iea.
org/t&c
1 Energy Saving and Efficient Energy Use … 5
are two of the dominant ones, which will be discussed in the next two sections,
respectively. The end-use energy efficiency of these applications plays a key role in
reducing CO2 emissions as discussed in [4, 5].
1.3 Power Electronics for Adjustable Speed Drives
Electric motors consume more than 40% of global electrical energy, as shown in
Fig. 1.3. This share is even higher in industrial applications, where electric motors
are utilized for approximately 64% of industrial electricity consumption [6, 7].
In fact, the high demand for electric motors in different applications such as
pumps, fans and conveyer systems has made them as the major source of electricity
consumption. Therefore, developing energy efficient motor drive systems holds a
great potential for reducing the worldwide energy consumption.
Introducing Adjustable Speed Drive (ASD) based on power electronics tech-
nology leads to more energy efficient motor drive systems. An ASD improves
energy efficiency of a system by controlling the speed of motor at an optimal speed
and/or torque. Hence, the energy consumption of the motor is reduced from full
power to a partial power for the same performance (i.e., speed and/or torque).
However, in order to achieve the maximum possible energy savings of a system,
understanding the application demand and operating mode of an ASD, its system
architecture and components are mandatory.
1.3.1 Motor Drive Applications
ASDs only contribute to energy saving when different speeds and/or torques are
required over time. This means that adding an ASD unit to a motor which has to
operate continuously at full speed and full load only result in excessive losses.
Fig. 1.3 Estimated share of electricity consumption for all electric motors [6]
Fig. 1.4 Block diagrams of: a direct connected AC motor and b ASD
Fig. 1.5 Relation between

the out power and speed in
variable torque applications
(for instance a reduction in
speed of 20% results in an
energy reduction of 50% [8])
Figure 1.4a exemplifies a direct connected AC motor to a main where a mechanical

valve is used to regulate the machine’s working point while in Fig. 1.4b an ASD
motor drive system is utilized to adjust the speed based on the applied voltage and
frequency.
In reality, in most of ASD applications, their load profiles vary over time and the
systems are commonly operated in partial load conditions [8]. However, the energy
saving potential is different in every application and it can be exploited by
understanding the application demand (Fig. 1.5).
1.3.2 System Architecture
Electric motors can operate in generative mode when decelerating from one speed
to another or during a braking mode. As illustrated in Fig. 1.6a, standard ASDs are
equipped with the conventional line-commutated front-end rectifiers, which provide
unidirectional power flow [9]. Therefore, the regenerated power will be dissipated
in braking resistors as heat. Feeding the regenerated power back to the grid may
improve the system energy efficiency to some extent [8, 10]. This requires bidi-
rectional power flow in ASDs which is achievable by replacing the conventional
front-end rectifier with a regenerative rectifier such as Active Front-End (AFE). The
resulting system is known as back-to-back topology shown in Fig. 1.6b.
Although most applications generate energy during deceleration and braking, but
feeding back this energy to the grid not necessarily justify the use of regenerative
rectifiers in an ASD system. This is due to the fact that a regenerative rectifier
accounts for more active components comparing with a standard rectifier unit.
Fig. 1.6 Typical single-phase and three-phase front-end rectifier configurations in an ASD system
with possible Passive Filtering (PF) and Active Filtering (AF) with a unidirectional rectifier and
b regenerative rectifier system
However, despite the losses in an ASD system, to select proper system architecture
the grid side efficiency needs to be accounted as well.
From power quality point of view the main concern with ASD systems is the
generation of current harmonics which may lead to high losses and stability issues in
the grid [11]. To cancel out the harmonics, additional passive or active filtering needs
to be employed at the front-end stage of the ASD [12–14]. While such a technique
has been matured in single-phase systems, known as Power Factor Correction
(PFC), it still remains a big challenge in three-phase systems [7, 9, 12–14].
Notably, regenerative rectifiers such as AFE extensively improve input current
quality which leads to a better power quality and lower losses from grid side point
of view [7, 10]. Therefore, employing a back-to-back system can only be validated
by considering the motor regenerative load cycle versus the additional costs and
losses affecting both the ASD and the grid.
Another alternative approach in improving energy efficiency is employing a
multi-drive or cascaded configuration (Fig. 1.7). In many applications it is a com-
mon practice to employ parallel connected drive units (e.g., multi-pump arrange-
ment) [7, 8, 15]. In this situation the application demand is met using multiple
modestly sized motor units rather than one single large unit. Hence, depending on
the applied control strategy such as multi-follower or multi-master the load can be
shared across the drive units evenly or unevenly, respectively [15]. This configu-
ration can satisfy partial load conditions more efficiently by keeping majority of units
in standby mode. Moreover, it allows easier system diagnosis and leads to better
redundancy. More importantly, by implementing a proper interaction among the
parallel connected units the input current harmonics can be significantly reduced
leading to a better power quality and losses from grid side point of view [7].
Fig. 1.7 Block diagram of a multi-drive configuration of ASD units

1.3.3 System Components
Every component contributes to system losses. In power electronics systems

switching losses of power semiconductors dominate significant share of total sys-
tem losses. Figure 1.8 illustrates typical structures for unidirectional and bidirec-
tional ASD systems. It is obvious that the bidirectional topology having AFE
includes larger number of active power switches. Notably, the chopper switch used
for braking operation is not in continuous operation in the unidirectional topology.
In fact, the introduction of additional losses, cost and size cannot justify the
investment in more complex topologies such as AFE in most of the three-phase
ASDs.
However, the advent of new power semiconductor devices, known as Wide
Band-Gap (WBG) devices, and their continuous decreasing price due to market
demand can break through many barriers and provide more degrees of freedom.
Employing WBG devices at front-end stage not only significantly improve per-
formance of single-phase systems, but also offer an opportunity to revolutionize the
next generation of three-phase ASD systems.
To sum up, although each three aforementioned sections can affect energy
efficiency, but the ultimate efficiency are to be made by optimizing the entire
system, which can be obtained by considering a trade-off between different sections.
Fig. 1.8 Block diagram of typical ASD systems: a unidirectional structure with diode rectifier,
b bidirectional structure with AFE
1.4 Power Electronics for Lighting
1.4.1 Evolution of Lighting Technology
Since the first incandescent light bulb was invented in 1879, lighting began to
consume electrical energy. Nowadays electric lighting has accounted for 22% of the
total electricity energy used in the world [16], and about 10% of the average
household’s energy use goes into lighting costs. Hence, energy-efficient lighting is
potential to improve the global energy saving and cut users’ energy cost.
In the last century, technology evolution has occurred in the lighting sources,
which brings various efficient and durable lighting lamps suitable for different
indoor and outdoor applications. According to the technology development, they
can be classified as heated filaments (e.g., traditional incandescent and
energy-saving incandescent (tungsten halogen) lamps), gas dischargers (e.g.,
tubular and compact fluorescent lamps (CFL), and high intensity discharge
(HID) lamps), and currently most promising solid state lighting, i.e., Light-Emitting
Diode (LED) lamps. Figure 1.9 compares the luminous efficacy among these dif-
ferent types of lamps, where the LED lighting has incomparable advantages on high
luminous efficiency and also extremely long lifetime to become the dominant
lighting source in the near future.
To promote more energy saving, some government in the world, such as
European Union, United States, and China have introduced measures to phase out
traditional incandescent lamps gradually [17]. LED lamps are the prospective
successors.
Fig. 1.9 Comparison of luminous efficacy among different types of lighting lamps
1.4.2 Energy Efficient and Intelligent Lighting Power

Electronics
Due to different light emitting principles, some lamps may not operate properly or
even fail without power electronic drivers. Efficient lighting power electronics can
improve the efficiency by a minimum of 20%. Intelligent power electronic drivers
with dimming can further save on average an additional 30%.
Filament incandescent lamps can be heated directly by AC mains. TRIAC
incandescent dimmer can change the turn-on phase of input voltage to dim light.
Compared to incandescent lamps, gas discharge lamps offer higher efficiency, but
are more complicated to manufacture and require auxiliary electronic equipment
such as high voltage initiation circuit to excite the gas atoms and driving ballast to
control current flow through the gas. Thus, the frequent turning on and off for these
lamps will affect the operating life and there are some rules for CFL blub dimming.
As the current-driven semiconductor, LEDs have fast dynamic response and the
operating life is unaffected by turning them on and off. A small increment in driving
voltage will cause the current increasing dramatically. Therefore, an LED driver
should provide a constant DC current for a stable light color and lumen output.
Figure 1.10 shows the typical LED driving configuration with accurate current
control for each LED lamp. Generally, the LED dimming have three main methods:
amplitude modulation (AM), pulse-width modulation (PWM) and TRIAC dim-
ming. AM and PWM dimming can be integrated in LED driver with smart and
simple dimming control for user requirement. TRIAC dimmers are suitable for the
interim lighting fixtures, which are expected to have the ability of accommodating
both incandescent bulbs and LED retrofit lamps to save the cost.
For the potential solid state lighting, advanced power electronics not only
improves the energy saving in the energy transformation, but also ensures a reliable
and durable operation to reduce the running and maintenance cost. For example, the
least reliable components in the LED driver are energy-buffering electrolytic
capacitors, i.e., the main storage capacitor as shown in Fig. 1.10, most of which last
below 5,000 h, far from the LED lifetime. LEDs emit heat, and 70–80% of
Fig. 1.10 Typical LED

driving configuration for
multiple lamps
electrical power are radiated as heat. Without good thermal control, heat from LEDs
and driver will significantly reduce the LED output lumen efficiency and lifetime.
By taking into account the electrical, thermal and optical issues, efficient and
intelligent lighting power electronics should integrate sensors, actuators and dim-
mer control schemes to offer more functions, and develop more compact, efficient,
and reliable drivers and control, leading to optimal LED lighting systems.
The future of both domestic and professional lighting is in the application of
LEDs, by taking advantages of the technology’s intrinsic qualities. This will lead to
products that inspire designers and enable lighting to develop in ways that con-
ventional lighting systems could never do.
1.5 Emerging Opportunities and Challenges
Power electronics technology is highly evolving and becoming more and more
multi-disciplinary based, bringing new opportunities and challenges in energy
saving and electrical energy conditioning. One of the promising opportunities is the
commercialization of WBG semiconductor devices (e.g., SiC and GaN) in the last
decade. These new devices enable further power loss reductions and extended
operation ranges, which are otherwise limited by Si based devices in terms of
voltage rating and maximum allowable junction temperature. Figure 1.11 shows the
figure of merit of SiC, GaN, and Si devices [18]. One of the important challenges is
Fig. 1.11 Figure of merit of SiC, GaN and Si based power semiconductor devices [18]
Table 1.1 Typical lifetime target in different power electronics applications

Applications Typical design target of lifetime
Aircraft 24 years (100,000 h flight operation)
Automotive 15 years (10,000 operating hours, 300,000 km)
Industry motor drives 5–20 years (60,000 h in at full load)
Railway 20–30 years (73,000–110,000 h)
Wind turbines 20 years (120,000 h)
Photovoltaic plants 5–30 years (90,000–130,000 h)
the reliability of power electronic systems, especially due to the increasing relia-
bility requirements and more stringent cost constraints as discussed in [19–21].
Table 1.1 lists the typical lifetime targets in different applications.
1.6 Conclusions
The key role of power electronics in energy saving and highly-efficient electrical
energy conditioning is discussed in this chapter. More and more power electronics
systems will be used, especially as renewables are becoming the mainstreamed
electrical energy sources, and the transportation sector sees a clear trend of elec-
trification. The power electronics applications in ASDs and solid-state lighting
systems are discussed. The future opportunities in WBG power semiconductor
devices and the challenges in reliability aspect performance are also addressed.
References
1. Blaabjerg F, Ionel DM (2015) Renewable energy devices and systems—state-of-the-art

technology, research and development, challenges and future trends. Electr Power Compon
Syst 43(12):1319–1328
2. European Center for Power Electronics (ECPE) (2007) Position paper on energy efficiency—
the role of power electronics, March 2007
3. IEA report (2014) world energy outlook 2014
4. IEA report (2015) Tracking clean energy progress 2015—energy technology perspectives
5. Gerber JP, Ferreira JA, van Wyk JD (2011) Quantifying the value of power electronics in
sustainable electrical energy systems. IEEE Trans Power Electron 26(12):3534–3544
6. Waide P, Brunner C (2011) Energy-efficiency policy opportunities for electric motor-driven
systems. In: International energy agency working paper, energy efficiency series
7. Davari P, Yang Y, Zare F, Blaabjerg F (2016) A multi-pulse pattern modulation scheme for
harmonic mitigation in three-phase multi-motor drives. IEEE J Emerg Sel Topics Power
Electron 4(1):174–185
8. Danfoss (2008) Facts worth knowing about frequency converters. Danfoss, Nordborg
9. Kolar JW, Friedli T (2013) The essence of three-phase PFC rectifier systems—Part I. IEEE
Trans Power Electron 28(1):176–198
10. Rodriguez JR, Dixon JW, Espinoza JR, Pontt J, Lezana P (2005) PWM regenerative rectifiers:
state of the art. IEEE Trans Ind Electron 52(1):5–22
11. Kumar D, Zare F (2016) Harmonic analysis of grid connected power electronic systems in
low voltage distribution networks. IEEE J Emerg Sel Topics Power Electron 4(1):70–79
12. Davari P, Yang Y, Zare F, Blaabjerg F (2016) Predictive pulse pattern current modulation
scheme for harmonic reduction in three-phase multi-drive systems. IEEE Trans Ind Electron
62(9):5932–5942
13. Davari P, Zare F, Blaabjerg F (2016) Pulse pattern modulated strategy for harmonic current
components reduction in three-phase AC-DC converters. IEEE Trans Ind Appl 52(4):3182–
3192
14. Klumpner C, Blaabjerg F, Thogersen P (2006) Alternate ASDs: evaluation of the converter
topologies suited for integrated motor drives. IEEE Ind Appl Mag 2(2):71–83
15. Vacon: Multiple pumps—simple control. http://www.vacon.com/fr-FR/Vacon/media/
Articles/Multiple-pumps—simple-control
16. Department of Energy. http://www.energy.gov/public-services/homes/saving-electricity/
lighting
17. https://en.wikipedia.org/wiki/Phase-out_of_incandescent_light_bulbs
18. Yole Development (2015) Market and technology trends in WBG power module packaging,
presentation at IEEE APEC
19. Wang H, Liserre M, Blaabjerg F (2013) Toward reliable power electronics—challenges,
design tools and opportunities. IEEE Ind Electron Mag 7(2):17–26
20. Wang H, Liserre M, Blaabjerg F, de Rimmen PP, Jacobsen JB, Kvisgaard T, Landkildehus J
(2014) Transitioning to physics of failure as a reliability driver in power electronics. IEEE J
Emerg Sel Top Power Electron 2(1):97–114
21. Chung H, Wang H, Blaabjerg F, Pecht M (2015) Reliability of power electronic converter
systems, IET. ISBN: 978-1-84919-901-8
Part I
Energy Harvesting
Chapter 2
Hybrid Energy Harvesters (HEHs)—A
Review
Nazenin Gure, Abdulkerim Kar, Erturul Tacgin, Alper Sisman

and Naser Mahdavi Tabatabaei
Abstract In this millennium, the methodologies to harvest existing dissipated

powers not only supply input energy to our sophisticated devices, but also con-
tribute the current technological researches and developments. Single harvester
generator or harvesting single power source may remain insufficient for the energy
feed into the systems like electronic devices, biosensors, human, structural and
machine health monitoring, and wireless sensor nodes. To overcome this problem,
hybridization of energy harvesters (EHs) takes place to increase the limited energy
generation of stand-alone EHs. In this chapter, piezoelectric and electromagnetic
generators are compared and classic as well as novel hybrid energy harvester (HEH)
designs are reviewed by considering fixed-frequency; broadband including linear,
nonlinear and tunable HEHs; multimode; and multisource powered configurations.
This review covers two-, three-, four-multi source powered HEHs in micro-, meso-
and large-scales. Overall comparisons of classic and novel HEHs are tabulated and
discussed in detail in order to guide potential researchers. In the scope of this
N. Gure (&) A. Kar E. Tacgin

Faculty of Engineering, Department of Mechanical Engineering, Marmara University,
Goztepe Campus - Kadikoy, Istanbul, Turkey
e-mail: nazeningure@marun.edu.tr; nazeningure@hotmail.com
A. Kar
e-mail: akar@marmara.edu.tr
E. Tacgin
e-mail: ertugrul.tacgin@marmara.edu.tr
A. Sisman
Faculty of Engineering, Department of Electrical and Electronics Engineering, Marmara
University, Goztepe Campus - Kadikoy, Istanbul, Turkey
e-mail: alper.sisman@marmara.edu.tr
N.M. Tabatabaei
Faculty of Engineering, Department of Electrical Engineering, Seraj Higher Education
Institute, Tabriz, Iran
e-mail: n.m.tabatabaei@gmail.com
N. Gure
Enhas R&D Energy Systems Ind. Co. Ltd, Maltepe, 34844 Istanbul, Turkey

18 N. Gure et al.
chapter review, it is seen that HEHs generate greater power outputs than its single
harvester components. The most promising power and energy generations are 315
mW by four-source powered novel HEH in meso-scale, 215 lW by tunable
broadband classic HEH in microscale and 440 kW h/day by partially three-source
powered HEH in large scale. This chapter indicates that HEHs not only increase the
output powers and power densities, but also enables endless configurations to
maximize harnessing existing power sources.
Keywords Hybrid energy harvesting

Multimode energy harvesting

Multisource energy harvesting Novel energy harvesters Microscale energy

harvesting Large-scale energy harvesting

A-Si Amorphous Silicon
AGS Automatic Generating System
DPE Direct Piezoelectric Effect
EM Electromagnetic
EMHs Electromagnetic Energy Harvesters
EHing Energy Harvesting
HAWT Horizontal Axis Wind Turbine
HEHs Hybrid Energy Harvesters
HRTHs Hybrid Rotary-Translational Harvesters
MEMs Microelectromechanical Systems
PAGV Power Augmentation Guide Vane
PE Piezoelectric
PEHs Piezoelectric Energy Harvesters
PET Polyethylene Terephthalate
PV Photo-voltaic
PZT Lead Zirconate Titanate
RF Radio Frequency
SME Shape Memory Effect
TENG Triboelectric Nanogenerator
VAWT Vertical Axis Wind Turbine
WiSH Wind-Solar Hybrid
2.1 Introduction
Ever since the beginning of the industrial age, being independent from man and
animal power sources, especially at greater energy levels, was the greatest inno-
vation. As the time passes by, the more technological improvements occur along
2 Hybrid Energy Harvesters (HEHs)—A Review 19
with wireless networks, the more devices are in our lives thus, elevating the quality
of life, production and work. In spite of the efforts to decrease the energy input of
the electronic devices, this ever increasing demand on energy surprisingly takes us
to seek using existing power sources like human movement, known as kinesiology,
similar to the energy source before industrial age [1–3]. In this millennium, the
methodologies to harvest existing dissipated powers not only supply input energy to
our sophisticated devices, but also contribute the current technological researches
and developments. Among energy harvesting (EHing) systems, one of the inno-
vative research trend is on hybrid energy harvesters [4, 5].
Obeying the first law of thermodynamics, conservation of energy implies that the
existing and dissipated power sources can be scavenged and transduce into usable
electrical energy [1]. Up until recently, energy harvester (EH) need is arose by the
dominant use of electronic devices, biosensors, human, structural and machine
health monitoring, and wireless sensor nodes [6–9]. Single harvester generator or
harvesting single power source, also known as stand-alone EH, may produce low
output powers to supply energy to the system. For sufficient energy feed to these
vast varieties of applications, hybridization of EHs takes place to increase the
limited energy generation of stand-alone EHs [10–23].
Harnessing multiple power sources or combining multiple generators for energy
extraction in a single unit is called “hybrid energy harvesting or multimodal energy
harvesting” [24–27].
In this chapter, piezoelectric (PE) and electromagnetic (EM) generators are
compared in Sect. 2.2. Classic HEH designs and novel configurations are reviewed
in Sects. 2.3 and 2.4, respectively. Sections 2.3 and 2.4 covers fixed-frequency,
broadband HEHs, and their comparisons. Furthermore, Sect. 2.4 includes multi-
mode vibration HEHs and multisource powered energy harvesters along with
meso-, micro- and large-scale applications. In the final part, HEH performances and
evaluations are compared and concluded.
2.2 Comparison of Piezoelectric and Electromagnetic

Generators
EM energy harvesters (EMHs) generate power based on faraday’s law of induction,

which equates the time derivative of flux to the electromotive force. As the scale
shrinks to microscopic level, the decreased coil area results smaller magnetic flux.
On the other hand, quasi-static (ultra-low-frequency) movements increase the time
intervals. Thus, both factors lead electromotive force to approach to zero. Apart
from the fabrication boundaries of coil diameters and turns, theoretically EM har-
vesters are bound to be limited at low speeds [28]. Thus, EM harvesters perform
better at high frequencies and PE harvesters outperform at low frequencies [8].
Additionally, at microscale level, EMH output voltage generally stays lower than
the need to power devices [17]. As a result, piezoelectric and electrostatic harvesters
20 N. Gure et al.
are more suitable for microscale applications, while electrostatic systems hold
greater advantage due to the ease of integration to microelectromechanical systems
(MEMs) [27].
Similar to EMHs, PE harvesters (PEHs) do not require voltage source while
electrostatic generators require separate voltage source and more difficult in prac-
tice, and in contrast to EMHs, PEHs produce sufficient output voltage but at low
current level [8, 19, 27]. Among these three types of transducers, piezoelectric
generators are the simplest ones in terms of required components, transducer
geometries and directly converting mechanical energy to voltage output [9]. In
addition to PEHs, at macroscopic level, EMHs also provide simplicity in geometry,
design and production [19].
In conclusion, PEHs are applicable for micro-, meso- and large-scales, while
EMHs are easily manufactured and although they perform better at mesoscale, they
are integrable to MEMs. As a result, abundant PE and EM HEH are reviewed and
compared in the following sections.
2.3 Classic HEH Systems
While tremendous amounts of multimode energy harvesters (EHs) are possible,

there exist such PE and EM combination that takes the greatest research and
development interest and turns out to be classic. As listed in Fig. 2.1, these are
composed of piezoelectric plate or patch attachment on Euler-Bernoulli beam and
either magnetic or coil tip mass is surrounded by respective coil or magnets to
achieve faraday law of induction [29].
In this chapter, PE unimorph and bimorph structures; comparison of rectangular
and trapezoidal beams; four and two poles magnet configurations and comparisons;
serial, parallel and isolated connections of PE and EM transducers in HEHs;
fixed-frequency applications; and rarely studied broadband HEH designs are
reviewed.
2.3.1 Fixed-Frequency Classic HEHs
Figures 2.1 and 2.2 shows the classic HEH designs and fabrications, respectively.
As a brief summary, in 2008, Wischke and Woias researched PE layers on uni-
morph and bimorph cantilevers with rectangular and trapezoidal layouts in HEH. It
is seen that trapezoidal shape is not superior in terms of power generation and its
fabrication is more complicated. Since unimorph HEH has greater tip velocity, EM
transducer generates greater power. In contrast to unimorph design, bimorph PE
part produces greater output than EM part. Upon this contrary output, authors
suggest using greater tip mass (magnet) to reduce EM coupling [11]. Becker et al.
Fig. 2.1 Classic energy HEH configurations are categorized and illustrated. a [20, 25, 26, 33–36]
and c [30, 32, 37] are the most common ones and the tip mass can either be two magnets as in (a),
or single magnet surrounded by coils as in (b) [4, 15] and (c), or vertical coil surrounded by
magnets as in (d) [20, 33, 38] as well as modified horizontal arrangement [21]. Along with varying
PE length on beam, PE unimorph (d) and bimorph (a, b and c) cantilever configurations are also
possible [11]. Interface circuit is schematically indicated on (d)
begin to test HEH prototype in Fig. 2.1c and research further for its adaptation into
synchronized switch harvesting interface [30].
Xu et al., both theoretically and experimentally analyzed PE and EM HEH.
Theoretical optimum output power is 1.02 mW at 77.8 Hz and experimental value
is 0.845 mW at the resonance frequency of 66 Hz under the vibration acceleration
of 9.8 m/s2. Respective to single PE and EM transducer, output powers are 667 and
22 N. Gure et al.
Fig. 2.2 Classic HEH

prototypes in test setup [25,
39, 40]
0.32 mW at 9.8 m/s2 and 66 Hz. Xu et al. proves that presented HEH generates
greater power than single EHs [31].
Ali et al., investigated total power outputs of PE and EM harvesters in serial,
parallel connection and separately. As seen in Fig. 2.1c, the HEH has approxi-
mately 1000 turns and 4 magnets (25 10 5 mm3) with opposite polarization at
cantilever tip. At the fixed input frequency of 76.2 Hz, total generated power is the
highest when PE and EM transducers are isolated. While PEH generates 27.56 mW,
EMH generates the lowest power output. Besides, single PEH output power is 3
times greater than serial connected HEH and parallel connected HEH generates 3
times more than single EM transducer [32].
As a recent study, Xia et al. not only investigates the classic HEH but also
compared the performances of HEH and EMH (Fig. 2.1c). Throughout the
experiments best HEH case generated the output power of 2.26 mW with 41%
efficiency at 23.3 Hz and 0.4 g input excitation and thus, greater performance
compared to EMH alone. HEH not only owns greater output power and efficiency,
but also enables broadband operation [33]. In contrast to these findings, Sang and
Shan et al. experiments result that HEH has the almost the same resonance fre-
quency with the PEH. Sang et al., considered the valuable Classic HEH
Fig. 2.3 a Two-pole and

b four-pole magnet
arrangements on classic HEH
designs [20, 34]
configurations and yet, similar to Fig. 2.1c with only difference of having vertical
coil placement on both sides of magnet is researched for four different cantilever
lengths. HEH with cantilever and PE layer respective sizes of 50 15 1 mm3
and 30 15 0.5 mm3, generated 10.7 mW while EM alone was 5.9 mW at
50 Hz with the acceleration of 0.4 g [21]. Supportively, Shan et al. also reported
that HEH produced greater output power of 4.25 mW than single PEH of 3.75 mW
at 40.5 Hz and optimum loads. Their design is slightly modified version of
Fig. 2.1c, d. The U-shaped magnet cage is fixed at the beam end and coil was fixed
filling the gap in U-shape magnets during vibration [16].
Ab Rahman et al. studied two and four pole magnet arrangements on classic
HEH in Fig. 2.1a, d, respectively (see Fig. 2.3). It is experimentally proved that
each HEH transducer with four-pole magnets produce greater output voltage than
two pole HEH. When the input excitation is 1 g, generated output powers of four
pole type PE and EM parts were 2.3 and 3.5 mW at 15 Hz, whereas those outputs
were 0.5 and 1 mW at 49 Hz for two pole HEH [20, 34]. More detailed comparison
of four-, two- and single magnet novel HEH performances are studied by
Castagnetti and covered in Sect. 2.4.2 [35].
2.3.2 Broadband Classic HEHs
The ever-demanded ideal EH efficiently performs in wide bandwidth. In order to

satisfy this demand, many researches have been conducted and broadband EHing is
still one of the popular EHing research subjects. These researches include passive
tuning either manually or with linear, non-linear, multi-stable and band-pass har-
vester structures; and active tuning which may result in negative power outputs due
24 N. Gure et al.
Fig. 2.4 a Shan et al. [15] and b Ping Li et al. [4] broadband HEH. NdFe35 magnet and PZT-5H
ceramics are used as EM and PE transducer components on Shan et al. prototype. Mahmoudi et al.
used oppositely aligned magnets and unimorph PE layers in (b) configuration [29]
to consumed energy by active parts and advanced electronic networks [26, 36, 37].
Among these efforts, there are some classic HEH design approaches exist.
Operation of an EH within wider frequency band is a very significant advantage in
EHs. A few studied novel micro scale HEHs are also mentioned in this section for
being broadband.
Linear Classic HEH: In 2013, Ping Li et al. designed similar to Fig. 2.4b without
fixed magnets. They analyzed their linear HEH performance under white noise
excitement to model random vibration. Their sensitivity analysis indicated that the
power generation of HEH dominantly affected by vibration frequency, damping
ratio, coupling coefficients, which widens the bandwidth and increase power output
as increases, and load resistances to achieve HEH impedance matching and max-
imize power generation. While PE load directly proportional to resonance
frequency of harvester, EM load has almost no effect. The most efficient perfor-
mances have the mean power levels of 0.44, 1.93, 4.2 mW at 77.5 Hz [17, 38].
Tunable Classic HEH: Wischke et al. focused on frequency tuning method for
HEH exactly shown in Fig. 2.1a and the picture of the prototype is on the in
Fig. 2.2a. In this technique, voltage is introduced to the PE layer. Applied voltage
changes the stiffness of the generator and relatedly, resonance frequency. By
matching input frequency with tuned generator’s resonance frequency, broadband
operation is achieved. Electrode length’s effect on tunability is also investigated and
found that greater than 10 mm PE beam length, tunable range almost saturates
between *50 to 60 Hz. To investigate widest tunable range, fabricated HEH
cantilever length was 20 mm with the width of 5 mm. Extractable output powers
from EM part is 60 lW, for PE part with parallel connection is around 200 lW and
with serial connection it is 215 lW. EM transducer generated minimum of 50 lW
at 56 Hz wide operation band width around the range of 267–323 Hz [25, 39].
Non-linear Classic HEHs: One method to achieve broadband operation is that
harvester to be nonlinear so that the frequency response range between half power
outputs can be widen [35]. Li et al. summarized nonlinear broadband mechanism
such that as nonlinearity increases, resonance frequency decreases and as the
acceleration increases, half power bandwidth broadens, whereas resonance fre-
quency decreases [4].
As inspired from classic HEHs in Fig. 2.1b and c, Shan et al. design and
materials are illustrated in Fig. 2.4a. In magnet configuration, poles are oppositely
aligned so that indirectly exerted force on the suspended magnet can yield nonlinear
mono-stable HEH [15]. Two peak powers and modes of HEH are 11.4 mW at
8.373 Hz by EM and 21.6 mW at 14.83 Hz by PE transducers. At half peak power,
the device band is around 7–17 Hz [15].
In Addition to HEH in Fig. 2.1c, Xu et al. used same pole magnet aligned in
front of the tip magnet (Fig. 2.2a) so that HEH can be nonlinear and operate at
larger frequency band. Their nonlinear HEH prototype achieved 5.66 mW power
output at 1 g, this result is 247% greater than at 0.5 g and at half power level
(3 dB), the frequency band is 83.3% wider than PE transducer alone [40].
Recently, HEH illustrated in Fig. 2.4b is researched by Mahmoudi et al. [29] and
Ping Li et al. [4]. Their device mirrors the configuration in Fig. 2.1b; opposite pole
magnet arrangement is used by Mahmoudi et al. and same pole magnet arrangement
is used by Ping Li et al. While the moving magnet is shared with both symmet-
rically placed beams having bimorph PE layers by Mahmoudi et al. and lead
zirconate titanate (PZT) patches by Ping Li et al. For Mahmoudi et al. HEH, EM
and PE parts respectively produce 39 and 61% of the power output. This EM
transducer can increase power density by 60% up to 1035 mW/cm3 and bandwidth
by 29% (155 to 220 Hz) at 0.9 g with respect to single EMHs [29]. Ping Li et al.
deeply studied modeling, tests, effects of nonlinear factors, loads, input frequency
and acceleration on amplitude, and found that their HEH design both enhance as
wider band with low resonance frequency and greater power output compared to
linear HEH designs. In contrast to linear EHs, optimal loads differ with excitation
26 N. Gure et al.
acceleration. Apart from Ping Li et al. statement, their theoretical and experimental
frequency responses show no significant band widening other than shifting the
linear resonance frequency from 119 down to 113.5 Hz. Experimental analysis
optimum results with respect to input accelerations of 0.2 and 0.45 g are 0.14 and
1.19 mW for EM generator and 0.085 and 0.5 mW for PE generator. The HEH peak
power output is 3.6 mW at 0.6 g and *110 Hz having half-power frequency range
of *107.5 to 112.5 Hz [4].
As a different vibration source and application on airflow harvesting system
example, hybrid aeroelastic vibration EH is modeled by Dias et al. Their system
includes an airfoil that is connected to fixed spring and damper at around mid-plane
and starting that point, cantilever beam as seen in Fig. 2.1c is connected. Dias et al.
propose 2 and 3 degree of freedom system dynamic modeling [41, 42]. Relatedly,
novel aeroelastic HEH harvesting incident sunlight is proposed by Chatterjee and
Bryant (Fig. 2.16), and their research is covered in Sect. 2.4.3.2. under ‘Two-Multi
Source Powered HEHs’ title.
2.3.3 Overall Classic HEHs Comparision
Up to this end, classic HEHs are classified and their performances are reviewed in
terms of generated powers by EM and PE parts as well as the whole HEH system.
The important factors of classic HEHs are peak power generations, HEH volumes,
magnet masses, input excitations, input frequencies and half power band width
ranges. Comparison of reviewed performances in Sects. 2.3.1 and 2.3.2, regarding
mentioned factors are listed in Table 2.1.
As seen in Table 2.1, among broadband and the whole classic HEHs, the
greatest power output of total of 33 mW (21.6 mW by PEH and 11.4 mW by EMH)
is achieved by Shan et al. with HEH configuration as illustrated in Fig. 2.4a [15].
Among fixed-frequency HEHs, Ali et al. reached the greatest power output of 27.56
mW [32]. It is then followed by 10.7 mW via Sang et al. HEH [21], both of HEHs
share the similar configuration as in Fig. 2.1c, Sang et al. HEH has almost the half
device volume of Ali et al. HEH as well as applying below half of the input
excitation.
Shan et al. [16] and Ab Rahman et al. [20, 34] HEHs are around similar power
generation levels of 4.25, 5.7 and 5.8 mW, respectively. For broadband HEHs to
achieve similar power levels, the device volume expands approximately ten times
with and exception of Xu et al. nonlinear HEH (Fig. 2.1a) having almost the same
device volume with Ab Rahman et al. HEH [40]. Table 2.1 also indicates that
four-pole arrangements [20, 34] are always superior to Ab Rahman et al. two-pole
arrangement [20].
Finally, the lowest power outputs belongs to Xu et al. [31] and Wischke et al.
(Fig. 2.2b) [25, 39]. However, Xu et al. fixed-band HEH volume is also the smallest
device volume in Table 2.1, whereas Wischke et al. tunable HEH owns the second
biggest device volume.
Table 2.1 The overall comparison of the reviewed classic HEH systems in Sects. 2.3.1 and 2.3.2
Type References Input Input Volume Peak Mass
acc. frequency/bandwidth (mm3) power (g)
range (Hz) (mW)
Fixed-frequency classic HEHs
[30] – 130 – – –
[31] 1g 66 187.2 0.845 –
[32] 1g 76.2 5992 PEH:27.56 –
[33] 0.4 g 23.3 2.26 11
[21] 0.4 g 50 2975 10.7 *15
[16] – 40.5 760 4.25 *21.5
Four-pole [34] 1g 15 2280 PEH: 2.2 –
EMH: 3.5
Four-pole [20] 1g 15 2280 PEH: 2.3 –
EMH: 3.5
Two-pole [20] 1g 49 1181 PEH: 1 –
EMH: 0.5
Broadband classic HEHs
Linear [17, 38] A* 77.5/*70–80 29,810 4.2 –
Tunable [25, 39] 1g 299/*267–323 31,201 215 lW –
Nonlinear [15] 0.5 g 8.373 for EM 20,726 PEH: 21.6 100
14.83 for PE/7–17 EMH: 11.4
Nonlinear [40] 1g *45.5/*43–47 2257 5.66 9.8
Nonlinear [29] 0.9 g 93/155–220 40,000 B** –
Nonlinear [4] 0.6 g 110/*107.5–112.5 18,437 3.6 –
(*) A: random acceleration with (0.1 g)2/Hz spectral density of acceleration
(**) B: Peak power density of 1035 mW/cm3
Note: Device volumes represent the minimum volume occupied by the harvester components and
do not include and remaining device parts and the air gaps in HEHs, and Mass generally stands for
the only stated magnet mass in references
2.4 Novel HEH Systems
While stand-alone systems are generally bound to be limited by one power source,
multimode designs offer never-ending possibilities. Especially, when multi-source
powered EHing concept is included as well as multimode HEHs, designs turn out to
be novel. In this section, vibrational HEH novel designs are covered along with
multiple vibration source harvesting, two and three multi-source powered har-
vesting systems in meso, micro and large scales.
28 N. Gure et al.
2.4.1 Fixed-Frequency Single-Source Powered HEHs
As a preliminary study, Reuschel et al. proposed axial flux and radial flux
arrangements, where set of opposite pole magnets aligned radially in a radial coil
house, designing EMH and PE cantilever modeling for the proposed arrangements.
They announce to combine both transducers and analyze HEH as a whole system
[43].
Harvesting from human motion is a demanded research subject especially to
power personal electronics. Wei and Ramasamy studied harvesting kinesiology and
it is shown that the HEH is suitable to feed personal electronics and charged mobile
phone in experiments. The mechanical harvesting part composed of flywheel and in
each footfall, it runs the shaft connected to one-third of diameter of the actual
wheel. Two piezoelectric configurations researched for shoe insole and it is seen
that rolled piezoelectric plate is placed in shoe sole. It is seen that though this HEH
is slow to charge mobile phone for being able to charge about 10% in 30 min, it is
also found that starting from half-fully charged phone. The user can end up with
70% of charge with HEH, whereas without any harvester, charge level would be
16%. Authors assume the potential over one-million personal usages of their har-
vester. In this case, they foreseen the total power generation of 60,000 kW h [1].
Halim et al. unique components turns classic HEH into novel one. The main
harvester body is almost same with the illustration in Fig. 2.4 on the right. Novel
parts are the parabolic top of the tip mass, which is intended to move vertically by
the non-magnetic ball action during horizontal input excitations (see Fig. 2.5). This
mechanism also leads vertical PE bimorph displacement at center and EM induction
with magnet attachment. This design is aimed to harvest human motion, thus
shaken manually by hand at around 5 Hz during experiments. Resulted frequency
responses of EM and PE transducers show that the first mode is at 816 Hz for both
parts. Optimum power generation performances are 0.64 mW for EM part and 0.98
mW for PE part of HEH system [14].
2.4.2 Broadband Single-Source Powered HEHs
Linear Novel HEHs: Castagnetti’s novel HEH is one of the most innovative one
as well possessing 60 Hz-bandwidth. The design concept is composed of Belleville
springs (B1 and B2 in Fig. 2.6b and c) and three different case of EM part for
having a single magnet as in Fig. 2.6a, 2 magnets (Fig. 2.6c) and four magnets
configurations. The vertical frame in (a) is shown horizontally in (c), denoted by
“F”. Experiments conducted for three cases of HEHs at 1 g and 19.62 m/s2. Input
acceleration of 19.62 m/s2 yield greater power outputs at resonance frequency.
Among HEHs, the generated power of four-magnet HEH is 2 times of two-pole
magnet HEH and 8 times of single magnet HEH. The four-magnet HEH
Fig. 2.5 a Halim et al. novel HEH design schematic representation, b listed fabricated
components, and c assembled prototype [14]
configuration produced the greater power of 15.31 mW and enables broadband

operation from 120 to 180 Hz at 2 g. Castagnetti also reported their HEH is
superior than commercial products like Perpetuum in terms of power generation
and band-width [35].
Non-linear Novel HEHs: Karami and Inman presented mono and bistable non-
linear thus, broadband novel HEH, as seen in Fig. 2.7. As the horizontal input
excitation is applied, then, the tip magnet moves harmonically. This oscillation
yields EM and PE energy generation. HEH magnets are aligned with opposite
polarization and these magnets’ distance is arranged such that the HEH can perform
as mono or bistable but nonlinear unless the gap is set to 50 mm in order to see
linear system performance. Additionally, the system behaves as linear at low input
base excitation and nonlinear at greater acceleration inputs. The best power output
results are close to 35 lW for EM transducer and 1.5 mW for PE part at 1.7 m/s2. It
is worth to mentioned that linear dynamics of these HEH systems at low excitations
[44, 45] are overcome by Leadenham and Erturk’s M-shaped PEH design [46].
30 N. Gure et al.
Fig. 2.6 a The technical

drawing of the Castagnetti’s
HEH prototype, b Belleville
spring scheme, and c HEH
scheme with two magnets
[35]
Fig. 2.7 Karami and Inman’s design illustration and prototype [44, 45]
2.4.3 Multimode and Multisource Powered HEHs
The term “multimode EH” refers larger literature than “hybrid EH”. Further,
multimode and multisource power terms are used in title to imply that novel
researches of the extreme levels of hybridization of harvesters, multi-vibrational
power source types (Sect. 2.4.3.1) and multi-power source inputs (Sects. 2.4.3.2,
2.4.3.3 and 2.4.3.4) are reviewed in this part.
2.4.3.1 Hybrid Rotary-Translational Harvesters (HRTHs)
The characteristics of motion highly affect the harvester design. Unlike the
excitement types so far, two axes translational and rotary motions are the realistic
cases of vibration inputs as in kinesiology and rotary machines. All these and
chaotic or random and varying excitations lead broadband methods, among these,
manual tuning is preferably employed for simplicity. However, as an alternative,
different types of random motions can be scavenged without active physical input
and it is known as “automatic generating system (AGS)” under the Kinetic brand.
One classic commonly known commercial example is SEIKO wristwatches.
Harmonic movement of the rotor drives the gear train and eventually almost 100
times of the eccentric mass rotational speed is converted to run the EM generator (at
5–15 krpm for 50 ms). In 1988, the device average power output range was 5–10
lW and in 1996 it was 10 lW when worn, 1 mW when shaken with a maximum
capacity of 10 mW [47–51]. Sasaki et al. researched this HRTH and analyzed
damping, self excited rotation and swinging motions as indicated in Fig. 2.8b.
Electric load connection on generator causes rotor axis damping to increase up to
some limit for self-excited rotation mode. This electromechanical damping limit is a
32 N. Gure et al.
Fig. 2.8 Schematic AGS mechanism and harvested multi-motions [48, 49]
bit greater than 10−4 Nms/rad. Self excited rotation reaches its maximum value of
10 mW at this point and as damping increases further, sudden decline in power
generation occurs down to 1 mW and the mode turns into swinging motion. This
indicated that power generation in self-excited rotation is 10 times greater than
swinging mode. In the light of these findings, AGS is better at low frequency and
large amplitude inputs like walking or running [48].
For medical implants, this HRTH is experimentally tested on the right ventric-
ular wall of the dog’s heart. Test results are 13 µJ per heartbeat, and 80 mJ in
30 min (44 lW), yet never further tested to feed aimed application of heart-pace
since less power is produced than needed 200 mJ per half an hour. This design’s
feasibility on pacemakers further studied: This harvester mechanism is placed on
the surface of a human chest and performance tested in office activity. In 8 h,
multi-mode harvester generated 0.5 lW, which is much less than peacemaker
battery recharge feed [50]. Related to Sasaki et al. conclusion on AGS, high
amplitude is necessary for greater power outputs. Hence, office environment does
not provide this vibration input as in running. However, generated power in office is
sufficient enough to run wristwatches.
Fig. 2.9 HRTH with

magnetic pendulum and DC
motor components [51]
Karami and Inman, investigated, modeled and simulated nonlinear HRTH in

order to enlarge operation bandwidth. This design has magnetic pendulum as an
eccentric mass connected to DC motor shaft and generates power from DC motor as
the magnet oscillates (Fig. 2.9). In case of full rotation is passed, the resulted
vibration is either two period vibration or chaotic. As the vertical base acceleration
increases from 1 g up to 1000 g, power output rises from 10 mW to 1 W. While the
base frequency increases up to 10 Hz, the output power is almost the same around 1
mW, after 20 Hz, logarithmically increases maximum of 0.5 mW, which is the
saturation value at 90 Hz. The multiplicity of these findings gave maximum power
generation of about 70 mW. The analytic calculations of power generations are
much higher than the alternative proposed HRTHs, which is highly favorable in
applications like tires and wind turbine blades [51].
Jung Kim Min et al. are especially designed HRTH for low frequency and
structural applications. The complete HRTH design is shown in Figs. 2.10 and
2.11c as Case 3. The whole system is separately analyzed and tested for three cases
of harvester components. The first case (Fig. 2.11a) is EM transducer where same
pole magnet movement results in EM induction. In case 2 (Fig. 2.11b), as the beam
oscillates, it is transformed into rotational motion and similar to Karami and Inman
design, this rotation input yields power generation by DC motor. When all the
harvesting mechanisms are gathered, there exists another implementation for the
permanent magnets to oscillate in boundary limits so that the efficiency can be
increased and since the beam is not fixed or not able to complete full rotation.
Therefore, In case 3, the mechanical stoppers are added to both oscillation radial
clockwise and counter-clockwise directions as seen in Fig. 2.11. In experiments, all
cases are tested at constant input excitation of 0.03 g and the frequency is shifted
34 N. Gure et al.
Fig. 2.10 Jung Kim Min et al. test HRTH prototype, it is analyzed in three cases [52]
from 0.5 to 5 Hz. Frequency response showed that the resonance frequency of all
cases is almost at 1 Hz and output power of the system in Case 2 that is power
generated by the motor with the gear ratio of 10:1, overwhelms the generation in
Case 1. As a result, the harvester power generation of 37 mW is mostly contributed
by the second case [52].
Larkin and Tadesse’s novel design is composed of 3 HEH cantilevers in
Fig. 2.1a mounted onto rotation center and a circular coil house (Fig. 2.12). The 3
HEH cantilever acts as eccentric mass and in addition to operation modes are shown
at the bottom of Fig. 2.8, PE beams allow harvesting perpendicular translational
forces. To sum up, the operation modes are rotational, and translational in both
vertical and horizontal axes in 1–10 Hz with a maximum of 30 Hz operation range.
Combined PE generator output power was 332 µW at 5 Hz and 0.8 g input exci-
tation, and 1.25 mW at 20.6 Hz and 0.6 g vibration input. Other experiments cover
testing at two body levels (wrist and ankle) when the device is placed horizontally
(normal to gravitational axis, Fig. 2.12d and f) and vertically (in line with gravi-
tational axis for walking, walk with spinning and jogging, Fig. 2.12e). The best
performances respective to EM and PE generators are 17 at mW vertical placement
on the wrist, and 759 µW at ankle while jogging. For both transduction mecha-
nisms, lowest performances are observed for horizontal placement on the wrist with
walking alone. The reason for vertical placement to produce greater powers is that
the gravitational force effect on the eccentric mass set which increases its rotational
Fig. 2.11 The tested three cases of HRTH. a Case 1 represents EMH part, b Case 2 illustrates
transformed beam oscillation as rotational input yields power generation by DC motor, and c Case
3 shows the schematic Fig. 2.2 of the whole HRTH system [52]
speed [24, 54]. Similar to AGS mechanism and Larkin and Tadesse’s novel HRTH,
Zhong et al. multi-source powered HEH (Figs. 2.17 and 2.18) is covered in
Sect. 2.4.3.2.
In addition to combination of multiple transducers for EHing in HEHs concept,
the other phenomenon is to harvest multiple power sources with either single or
multiple generators and the former is known as “hybrid power systems or multi
source power systems” [55]. In the following parts, multi source systems will be
investigated in three parts so that the number of different power sources can be
classified.
2.4.3.2 Two-Multi Source Powered HEHs
Gambier et al. presented combined solar and PE transducers as a multi-source HEH.

In addition to HEH, they also investigate thermoelectric generator and suggest
36 N. Gure et al.
Fig. 2.12 Larkin and Tadesse’s novel HRTH schematic design in (a), (b) and (c) [53], and the
tested prototype vertical (e) and horizontal (d and f) arrangements [24, 54]
combining these three generators as a HEH. HEH design is composed of flexible

solar (PowerFilm, Inc.), piezoceramic (QuickPack QP10n, Mide TC) generators
and thin-film battery (MEC102, Infinite Power Solutions, Inc.) supported by
metallic substrates, Kapton layers with implanted flexible copper electrodes for
electrical isolation and epoxy layers having high-shear strength, as seen in
Fig. 2.13. Single layer solar transducer dimensions are 93 25 0.178 mm3 and
cantilever volume including piezoceramic is 93 25 1.5 mm3. Tests are con-
ducted independently and together with solar and vibrational EHs. Three levels of
illumination intensities are used as 124, 223, 311 and 437 W/m2. As expected,
highest irradiance of 437 W/m2 resulted the maximum output power of 30 mW,
while it is 12.5 mW at 223 W/m2. Frequency swept from 0 to 500 Hz to catch the
device resonance frequency and after that, measured output powers at 56.4 Hz
are 0.4 and 0.49 mW with respect to the base excitations of 0.1 and 0.5 g.
Fig. 2.13 Gambier et al. novel HEH combined with flexible thin-film batteries. a Aluminum
structure layer (1), piezoceramic in Kapton material (2), flexible battery (3) and flexible solar layer
(4), and b HEH scheme after integration of the layers [56]
At illumination exposure at 223 W/m2 single solar generator charges 1 mA h of a

thin-film battery in 20 min and 1.3 mA h in 26 min, whereas it takes 8 h for
1 mA h and 3 h for 0.38 mA h of capacity with PEH at base excitation of 0.5 g and
56.4 Hz [56].
Novelty of Hehr et al. research is that both vibrational and radio frequency
(RF) inputs are harvested upon the same solenoid coil in EMH. As seen in
Fig. 2.14, EMH dimensions are 50.8 mm diameter and 95.3 mm length with
neodymium magnet of 38.1 mm diameter and 12.5 mm length. Designed harvester
has the first resonance frequency of 12.5 Hz, which is experimentally validated at
3.5 and 5 g acceleration amplitudes. Coil diameter of 40.5 mm (coil set 1 and 2) is
selected in order to keep the center frequency of band at 2.45 GHz and its
equivalent wavelength is 122.4 mm so that HEH can function for industrial, sci-
entific and medical radio band of 2.4 GHz. Throughout experiments, electromag-
netic resonance was measured as 2.46 GHz. For the helical antenna, grounding
copper ring diameter is 39 mm. For maximum voltage output, coil sets are
38 N. Gure et al.
Fig. 2.14 a Kinetic and RF EM generator and b RF EHing test set-up with the same multi source
powered harvester upon coil part alone [57]
connected in series for vibrational EHing and in parallel for RF EHing in the axial
orientation to the source antenna as seen in the bottom in Fig. 2.14. The interesting
finding was that the voltage differences between 1.52 V for coil 1 only and 1.32 V
for coil 1 and 2 connected in parallel. Hehr et al. explains the reason as the
indication of RF interface between the coils. At 1 g and 12.5 Hz, EM generator
output energy is 1 J in *90 s namely, *11 mW. At 2.46 GHz, RF generator
produces 70 mJ of energy in 100 s in other words, 0.7 mW [57].
Collado and Georgiadis also comprised unusual two energy sources of solar and
electromagnetic waves in a single device. Novelty of their research is not limited to
these harvesting sources but also include the HEH design in Fig. 2.15, developed
low cost and efficient solar cells, maximized power point tracking and low power
DC/DC converter to fix DC voltage. Collado and Georgiadis’s novel HEH design is
based on broadband monopole rectifying antenna on flexible polyethylene
terephthalate (PET) substrate structure and on top of the antenna, flexible hydro-
genated amorphous silicon (A-Si) solar panel is thin solar cell arrays are imple-
mented (Fig. 2.15). The antenna alone is the copper layer of 35 lm covering a PET
substrate of 75 lm. In contrast to Hehr et al., authors aimed to harvest broadband
range of 800 MHz to 6 GHz so that dissipated wide range of standard communi-
cation bands can be harvested. Solar array integration was chosen such that the RF
harvesting performance will not be affected. Thus the experiments conducted at a
distance of 3.45 m from the transmitter, which transmits RF power less than
Fig. 2.15 a HEH design with dimensions and implemented dual band rectifier and b broadband
rectenna on PET substrate [58, 59]
20 dBm for the rectenna of radius (R) of 70 and 100 mm and with and without the
solar cell integration. For solar and RF HEH, the illumination was set almost to
standard global solar irradiance of 100 and 15 mW/cm2 for shade simulation. With
respect to the stated illuminations, the voltage outputs of the solar array are 4.06 V
and 3.90 mV. It is seen that EH with and without solar cells generates similar
voltage trend and amount just as the radius of 10 and 7 mm prototypes resulted, yet
10 mm radii prototype generates slightly higher voltage at 850 and 1850 MHz.
Collado and Georgiadis concluded that the best performance of the HEH is able to
generate 56 mW [58, 59].
Resembling Gambier et al. novel multi source powered HEH [56], Chatterjee
and Bryant’s HEH harnesses solar and wind flow or vibration type kinetic powers.
Only Chatterjee and Bryant state possible application of HEHs on mobile robots.
As shown in Fig. 2.16, the PE patches are laminated underneath flexible thin film
solar ribbon with very low aspect ratio. HEH is fixed in both ends and during
longitudinal tension; transverse wind flow is induced to form aeroelastic vibration.
In spite of setting the test set up, in Chatterjee and Bryant’s paper, authors only
proposed theoretical modeling and tuning via changing the longitudinal tension.
They concluded that the transverse matrix methods is superior over Bokaikan
method, where as the stress is reduced, the power generation of PE patches also
reduces while being almost ineffective on mode shapes, yet effective on the mode
shape of the structure without the PE patches [60].
Similar to AGS mechanism and Larkin and Tadesse’s novel HRTH, Zhong et al.
combined planar rotary disk EMH with triboelectric nanogenerator (TENG)
in order to harvest hand induced rotating kinesiology (see Fig. 2.17).
40 N. Gure et al.
Fig. 2.16 Test setup of the solar and vibrational HEH [60]
Fig. 2.17 Assembled prototype used in tests (a and b). Hand motion induced to HEH and lights
40 LEDs (c), and a globe light (d) [61]
During experiments, hand induced rotation supplied to HEH and successfully lit 40
LEDs and a globe light with a maximum illumination up to 1700 lux (Figs. 2.17
and 2.18). At a rate of 200 rpm, HEH generates *8.4 mW by EM part and 8.4 mW
by TENG part. Under optimum loads of 1 and 2 kX, respective power generations
by EM and TENG transducers are 50 and 17 mW [61].
Microscale Examples: Yu et al. investigated the combination of solar panels for
low illumination and MEM vibrational EH. Similar to Collado and Georgiadis, Yu
et al. also selected A-Si solar panel with a surface area of 9.6 cm2 since it is suitable
for indoor light (fluorescent or LED). When authors’ five-parameter modeling
simulations of solar power generations are compared with the experiments, their
modeling is closer to the experimental measurements and thus, more accurate than
traditional models. The maximum power output is 110 lW at 530 lux. Microscopic
scale vibrational EH is composed of five PZT beam array connected in series with a
silicon proof mass (Fig. 2.19a). The peak power output is 66.75 lW at 234.5 Hz
and 5 m/s2 [62].
Fig. 2.18 a and b the first layer consists 6 magnets on a 5 mm thick circular acrylic disk with
140 mm in diameter. Likewise, 6 coil sets are aligned to the magnets on last layer. 2–4 layers are
TENG parts: second and fourth layers are the combination of the copper strips that are integrated
on a flexible substrate. The third polyamide layer is sandwiched in between these layers. c The
prototype picture in use and HEH runs the globe light [61]
Fig. 2.19 a Yu et al. vibrational HEH structure without solar panels [62] and b Jeon et al. self
cleaning rain drop and solar powered HEH [63]
42 N. Gure et al.
Jeon et al. not only propose a novel multi source powered HEH but also solves
the cleaning need of the solar cells and thus, keep its efficiency stable by the second
power source of raindrops. In addition contact electrification by mechanic input,
water solid interactions are also enough to harness energy via TENGs. Transparent
and superhydrophobic TENGs with small contact angle allows water droplets to
remove particles on the surface as well providing light to transmit and reach to the
solar cells. In Fig. 2.19b, transparent superhydrophobic PDMS and ITO-PEN
substrate layers on solar cells, and the prototype (below) are demonstrated.
Throughout the experiments conducted with distilled water, raw tap water, actual
rain, 0.01 and 1 M NaCl solutions. The resulted power generation is in the same
order of the tested solutions regarding the highest to lowest power generation.
While in the real rain case the output power is decreased, it is almost negligible
especially the standard deviations are considered. Jeon et al. multi-function HEH
generated maximum power output of 0.27 lW [63].
Novelty of Zakharov et al. study is to harness thermal energy via PE transducer
due to the resulted mechanical deformation. Their design couples shape memory
effect (SME) and direct piezoelectric effect (DPE). However, proposed novel har-
vester is partially HEH since the direct main source is thermal and the second
source is indirect mechanical deformation, caused by the thermal input (Fig. 2.20)
[64]. NiTi wire connected to fixed support and the tip of the cantilever shrinks
during heating and bend the bulk PZT ceramic plate, and also reach the initial state
as it is cooled down as shown in Fig. 2.20a. SME and DPE structure with 0.2 cm3
of active materials energy generations are 90 µJ over a temperature increase of
35 °C and 60 µJ while cooling [64].
Large Scale Examples: Dr. Tong’s research team installed HEH consists of sur-
rounding novel power augmentation guide vane (PAGV), vertical axis wind turbine
(VAWT) and solar panel at the top so that the LED outdoor light can be driven.
Fig. 2.20 a Zakharov et al. partially hybrid EH working principle scheme and b prototype picture
along with the heat source (for the video record during operation, please visit www.youtube.com/
watch?v=uS0z4ZEBeYw) [64]
Fig. 2.21 Eco-GreenergyTM Wind-Solar Outdoor Lighting System in Kuala Lumpur Campus,
University of Mala [65, 66]
The developed system runs against low wind speeds due to PAGV integration
(Fig. 2.21) [65, 66].
De et al. also studied wind-solar hybrid EH. Their so-called “WiSH” system
consists horizontal axis wind turbine (HAWT) having 500 W rotor and solar PV
panels with 250 W capacity so that the energy production can be increased during
low wind speed regimes in India. HAWTs have two types; having 3 blades and 4
blades. Each blade weighs about 600 g and to achieve this low weight and high
strength, they are made of carbon fiber composite. Performance curves gained from
mobile testing indicated that fabricated 4-blade HAWT is more efficient than
3-bladed one. Prototypes of WiSH systems are installed in Kodihalli Campus, India
for field tests (Fig. 2.22). Both 4 bladed HAWTs are parallel and 30 m-high. With
the help of industrial partnership of ARES, prototypes developed up to 1–5 kW
capacity and further produced 50 units of these WiSH systems [67].
2.4.3.3 Three-Multi Source Powered HEHs
In edition to Porcarelli et al. novel device comprising solar, airflow and hydrogen
micro fuel cell EHs (Fig. 2.23), their [68] and Chung et al. researches [69] lie in the
early studies about powering wireless sensor nodes by fuel cells.
Simulations for power generation of solar and wind EHs are compared with the
experimental results. Solar cells are tested for the irradiation levels of 8000; 24,000;
40,000; and 80,000 lux. In the same order with light intensities, the photo-voltaic
(PV) cell maximum output powers are 0.075, 0.13, 0.28, and 0.45 mW. Airflow
energy is harvested by meso-scale plastic four bladed HAWT, 6.3 cm in diameter
and 7.5 cm in length. Flow EH is tested for the flow speeds of 8.5, 15 and 16 km/h
44 N. Gure et al.
Fig. 2.22 De et al. WiSH prototypes under field test. For controlled conditions, HAWTs are 30 m
above the ground level and parallel to each other [67]
Fig. 2.23 Porcarelli et al. three-multi source powered HEH: solar, airflow and hydrogen micro
fuel cell harvesters, battery and harvester circuitry are shown [68]
and the respective wind generator maximum output powers are 3.1, 5, and 7.7 mW.
Solar cell and wind harvester experimental results are very close to simulation
findings. As a final component, fuel cell EH with 3.61 cm2 area, power generation
is around 1 mW [68].
Zheng et al. are researched a pretty novel application of single improved dual
mode TENG in such a way to harness raindrop and wind energies, and combined it
with solar EHing. The very similar structure is in Jean et al. HEH as demonstrated in
Fig. 2.19b. Zheng et al. also used the transparent superhydrophobic TENG with a
40 cm2 surface area on silicon-based solar cells. TENG’s transparency is validated by
spectra transmittance test as being more transparent even than a 3 mm-thick com-
mercial glass. Moreover, separation of polytetrafluoroethylene and nylon layers by
PET spacer border enhances dual mode of TENG (fabrication and experiments of
HEH components are given in detail in Zheng et al. article). These dual modes are:
water contact TENG as rain EH and the wind contact TENG as wind EH. The tests are
set as such: Rainy day conditions are considered (13.6 and 20 mL/s dripping rates)
and the incident angle between raindrops and TENG surface is 45°, with a distance of
40 cm from the rain shower. Common daily wind speeds of 1.7; 2.7; 4.1 and 4.9 m/s
are flowed from a faucet and the solar irradiation is set to 100 mW/cm2. At 20 mL/s
dripping rate, water TENG, water contact TENG and HEH drive 10, 20 and 50 LEDs,
respectively. Experiments have shown that the dual-mode TENG (water and water
contact TENGs are in series) voltage and current densities are greater than water
contact TENG that is greater than water TENG alone. Generated power of the pro-
totype is limited but since the size can be expanded, generated powers in per square
meter of the HEH surface area are more representative as in current densities, thus, for
HEH it is 86 mW/m2 at a dripping rate of 13.6 mL/s and in the absence of solar and
rain sources, it is 8 mW/m2 from wind at a speed of 2.7 m/s [70].
Microscale Examples: Chung et al. combined thermal, mechanical and magnetic
powers in such a novel way. As mentioned in two-multi source powered HEHs title
for Zakharov et al. 2015-design, in 2012, Chung et al. also used indirect source of
magnetism to harness direct thermal power source. So, Chung et al. device is
partially three-source powered HEH. As seen in the prototype picture in Fig. 2.24a,
main body is copper-beryllium spring and the PE cantilever is both fixed to glass
frame and attached to the spring. Design has two types of magnets: Moving
gadolinium soft-magnet on top of spring and fixed neodymium-iron-boron
hard-magnet on frame. The working principle of this novel HEH demonstrated in
Fig. 2.24b and c: As the soft magnet is cooled below Curie temperature, it gains
ferromagnetic property and almost attaches to the fixed hard magnet (b), which
bends the spring and PEH. Worth to mention that magnets never really touch due to
the distance between them in y direction. As the topside is cooled down (15 °C) and
heated (27 °C), magnetic attraction incidence and withdrawal of the magnetic
attraction leads spring, thus, PE beam oscillation and power generation. Second
working mode is classic vibrational spring-proof mass energy generation on PEHs.
Final creative additional working mode is the oscillation of the soft magnet on
spring due to the induced AC magnetic field in z direction (c) [69].
46 N. Gure et al.
Fig. 2.24 a Temperature-difference-driven thermal and b vibrational and magnetic field induced
schematic working principles. c Chung et al. fabricated prototype [69]
Chung et al. experiments resulted as: For temperature-difference-driven thermal

EHing, a 25 °C difference lead minimum and maximum of 15 and 70 peak-to-peak
voltages, respectively. Generated peak-to-peak voltage outputs are 175 and 20 mV
with respect to the input excitation of 1 mm at 46 Hz, and AC magnetic field of ±3.5
Oe at 43 Hz. The preliminary results clearly indicate that the vibrational energy
generation is undoubtedly greater than other working modes [69]. Almost the same
research team; Chen et al. took further to analyze their novel design (Fig. 2.25) [71].
Chen et al. simplified their previous study by using gadolinium soft-magnet as a
fixed beam on PZT sheet and silicon clamps and keeping the same fixed
neodymium-iron-boron hard-magnet on frame. Experimental measurements are
(a) (b) PZT Sheet Gd Cantilever Beam
Glas
Contai
Unit: mm
Fig. 2.25 Chen et al. simplified recent magnetic-piezoelectric thermal HEH prototype with
dimensions (a) and during experiments with ice water (b) [71]
37 mV of peak-to-peak voltage and 1.98 mV RMS voltage at 20 °C temperature

difference for the temperature range of 6.7 to 26.7 °C. The maximum output power
is and average power density is *1.37 nW with an average of *3.96 pW with the
breakthrough HEH volume of about 0.063 cm3 [71].
Large Scale Examples: Chong et al. presented VAWT, solar energy converter and
rainwater collection systems integration with PAGV. Since the rainwater collection
system is not an EH, it is only for storage purpose and indirectly saving energy from
the reduced pumping need [72, 73] so their design is partially three-source powered
HEH. Nevertheless, as Mithra et al. suggested in their research, it is also possible to
harness energy from rainwater by increasing its kinetic energy from falling and the
use of turbines [55]. As seen in their patent scheme in Fig. 2.26a–c, the wind
turbine ‘A’ is placed in the middle of the structure with rudder ‘C’, PAGV ‘B’ has
sloped upper ‘D’ and lower ‘E’ wall ducts, and surrounded by the protective mesh
‘M’. The upper wall duct also acts as a collector of the rain shower and at the same
time, having solar energy converter (PV and/or solar thermal panel or solar con-
centrator system) ‘F’ on the upper surface. The captured rainwater after filter ‘N’
flows through the passage ‘G’ and reach to the storage ‘K’, which has thermal
insulation at the base ‘L’ so that heat transfer into building can be blocked. VAWT
power drive shaft ‘H’ is connected to the generator ‘I’ via mechanical drive box ‘J’.
Their design is open to many modifications and one is illustrated in Fig. 2.26b.
Tests conducted with and without 3-blade Sistan rotor having 0.5 m diameter and
0.25 m height, surrounded by the PAGV having 30 m diameter. Pilot-scale test set
up height from the PAGV to base is 12 m. It is reported that PAGV amplifies the
wind speed by *1.8 times (46 rpm for pilot tests) and VAWT gains direct flow
angle. It is seen that PAVG integration increases energy generation by 1.25 times. It
is estimated that for the system on a 220 m high skyscraper, energy generations of
the wind energy harvester system is approximately 157 and 58.4 MW h/year, solar
panels with a 650 m2 active area is 280 kW h/day, and monthly the energy saving
for 220 m pumped domestic water is 79 kW h. Ultimately, Chong et al. research
suggests the annual total saved and generated energy is 160 MW h and by
48 N. Gure et al.
Fig. 2.26 a Patented and proposed design’s referred figure and b proposed modified design, and
c artistic demonstration of HEH on the high rise building in Tokyo, Japan [72, 73]
considering 2.8 MW h of energy consumption per house, one skyscraper applica-

tion can help to supply energy for 57 houses [72, 73].
Relatedly, Donnie holds a patent of the similar HEH application. Patented HEH
has solar panels on the outer rotating large blades, harvesting wind energy and the
inner part has either acoustic piezoelectric or electrostatic energy harvester in order
to harness entered wind’s remaining kinetic energy [74].
Sathiyamoorthya and Bharathib proposed the idea of combining PEH integration
roads, rotational solar panel and two-blade VAWT between two-way roads to gain
turbulence, and thus, rotational motion input. This configuration aims to feed
streetlights but they also suggested HEH implementation on rooftop to supply
electricity for households. Proposed VAWT is applicable at low levels for energy
generation, in other words, suitable for residential structures. The novelty lies on the
sun-tracking module that does not require any sensor or controller. It works as a
pendulum clock and as the hanging wheel oscillates, it rotates the escapement
wheel and the center shaft. Gear mechanism, reduce the clockwise rotation of the
shaft and reduced shaft is connected to solar panel and tracking sun’s direction from
east to west. PEH is selected as PZT (lead zirconium titanate) and connected in
series. Measured maximum output power and peak-to-peak voltage of solar panel
and PE are 10 W and 12 V, respectively. Authors propose wind turbine analytic
modeling but do not mention the value of the potential power generation [75].
Mithra et al. combined the classic rainwater, solar and wind harvesting systems.
Rainwater harvesting starts from the collection of the rainfall on the roof of a
three-floor, *9 m high building to the header tank. The water falls from the conduits,
which increase pressure thus, velocity. At the base the falling water meets the turbine
and its shaft is connected to alternator’s rotor. After the rainwater hits the turbine
blades it is collected and stored for further use, while the generated power is used to
feed first of all power electronics and then, power systems. Based on this plot scale
findings, the desired real-life application’s evaluated electricity production is stated as
1.53 kW h for 7 m high normal house and 290 kWh for 98 m skyscrapers. Classic
Solar PV arrays and wind turbine to harvest wind energy are suggested to meet
domestic power consumption. Nonetheless, Mithra et al. have not stated pilot scale
tests along with rainwater system. The combination of mentioned generation systems
are concluded as: The rainwater and wind power generations are rectified by AC-DC
converters and the PV power alone is converted to fix DC voltage by chopper. The
control unit combines these three different generated energies as a hybrid system [55].
2.4.3.4 Four-Multi Source Powered HEH
After Porcarelli et al. another attempt to research solar, wind and fuel cell EHs are
studied by Saini et al. with an additional novel segment of electrolyzer (Fig. 2.27).
Saini et al. uses solar, airflow, water, and produced hydrogen and oxygen as a result
of electrolysis. HEH solar module is 95 135 30 mm3 in dimension, 89 g of
weight and the maximum PV power generation is *190 mW when exposed to single
75 W lamp at 90°. Six blade HAWT and the maximum output power is 400 mW.
50 N. Gure et al.
Fig. 2.27 Four-multi source powered HEH system diagram by Saini et al. [2]
Electrolyzer is the basis of fuel cell for decomposing water into hydrogen and oxy-
gen. Afterwards, produced hydrogen and oxygen is used by fuel cell and increases
the total source number up to four. Electrolyzer dimensions are 50 40 57 mm3
and the generator is 54 g in weight with a decomposition voltage of 1.5 V in practice.
Final harvester component of fuel cell is selected as proton exchange membrane and
as a 5-fuel cell set, it is 60 70 175 mm3 in volume and 430 g in weight.
Four-multi source powered HEH maximum output power is 315 mW [2].
2.4.4 Overall Novel HEHs Comparison
The reviewed novel HEH systems peak power generations, total harvester com-
ponent volumes and/or active surface areas device masses, input excitations, input
frequencies and half power band width ranges are compared in Tables 2.2 and 2.3.
Table 2.2 The overall comparison of fixed-frequency and broadband single-source powered HEHs and multimode HRTHs in Sects. 2.4.1 and 2.4.2
Type References Input power Working modes Input Input properties Volume/surface Peak output power Mass
sources/types acc. A. voltage, power
density
Fixed-frequency single-source powered HEHs
[1] Mechanical: Rotary Generator, PEH – *5 Hz* – 0.06 kW h, 10 V –
foot fall
[14] Mechanical: Ball driven PEH, EMH – *5 Hz** 19 cm3 PEH: 0.98 mW, 20 V –
hand motion EMH: 0.64 mW,
0.65 V
HEH: 84.4 µW/cm3
Broadband single-source powered HEHs
Linear [35] Mechanical PE and EM 2g *150/120–180 Hz 884 cm3 4 magnet HEH: 15.31 –
mW, 1.75 V
Nonlinear [44, 45] Mechanical PE and EM 1.7 m/s2 *12 Hz 819 mm3 PEH: 1.5 mW 48 g
2 Hybrid Energy Harvesters (HEHs)—A Review
EMH: 35 mW
Multimode: hybrid rotary-translational harvesters
[47–51] Mechanical: Mechanical: – 5–15 krpm in the 1–10 mW –
wrist motion rotary-translational wristwatch
[50] Mechanical: AGS – – – 44 lW –
dog’s heart
Mechanical: on – – – 0.5 lW –
chest
(continued)
51
Table 2.2 (continued)
52
Type References Input power Working modes Input Input properties Volume/surface Peak output power Mass
sources/types acc. A. voltage, power
density
[51] Mechanical Rotary-translational 1– 10–90 Hz 10 mW to 1 W, (avg: –
generator, DC motor 1000 g 70 mW)
[52] Mechanical Rotary-translational DC 0.03 g 1 Hz *1000 cm3 37 mW, 0.27 V –
motor, EMH
[24, 54] Mechanical: Mechanical: – Jogging: PEH on 958 cm3 PEH: 759 µW, EMH: –
walking, rotary-translational ankle, EMH on 17 mW
jogging PEH and EM wrist
(*) Standard walking and running human motion frequency value [14]
(**) Input frequency is 5 Hz in horizontal axis, but the HEH resonance frequency is 816 Hz
N. Gure et al.
Table 2.3 The overall comparison of multisource powered HEHs. Classified according to micro-, meso- and large scales
# of source References Input power Working modes Input Input properties Volume/ Peak output power Mass
sources acc. surface A. voltage, power density (g)
Microscale
Two [62] Solar, PV, PE 0.5 g Solar: 530 lux, PE: 1096 mm2 Solar: 110 lW, PE: –
vibration 234.5 Hz 66.75 lW
Two [63] Raindrop, TENG, PV – Solar: 150 mW/cm2 13.4 mm2 0.27 lW –
solar
Two [64] Thermal, Thermally-driven PE, – Temp increase: 35 ° 200 mm3 Temp rise: 90 µJ –
(mechanical) (SME, DPE) C Cooling: 60 µJ
Three [69] Thermal, Thermally-driven PE, – T: 25 °C, 49 cm3 Thermally-driven –
magnetic Magnetic field induced PE: 1 mm at 46 Hz, PE: 35 V
field, PE, AC magnetic field:
vibration
Vibration induced PE ±3.5 Oe at 43 Hz
Three [71] Thermal, Thermally driven PE, – T: 20 °C 63 mm3 *1.37 nW, 1.98 V, 21.7 –
(magnetic (Magnetic Field driven nW/cm3

field, PE, Vibration)
vibration)
Mesoscale
Two [56] Solar, PV, PE 0.5 g Solar: 437 W/m2, 3488 mm3 Solar: 30 mW, PE: 0.49 –
vibration PE: 56.4 Hz mW
Two [57] RF, vibration RF, EM 5g RF: 2.46 GHz, 773 cm3 RF: 0.7 mW, 1.52 V, EM: –
RF wavelength: 11 mW, 4.5 V
122 mm, EM:
12.5 Hz
Two [58, 59] RF, solar RF, PV – RF: 4084 mm3 HEH: 56 mW, –
800 MHz-6 GHz Solar: 4.06 V
Solar: 100 mW/cm2
(continued)
53
54
Two [61] Mechanical: TENG, EM – 200 rpm 77 cm3 TENG: 17 mW, –
hand-induced EM: 50 mW
Three [68] Solar, PV, HAWT, hydrogen – PV: 80,000 lux, HAWT: PV: 0.45 mW –
airflow, fuel fuel cell Wind speed: 234 cm3, Fuel
cell 16 km/h cell: 361 mm3
HAWT: 7.7 mW
Fuel cell: *1 mW,
1.23 V, 2.5 A,
282 mW/cm2
Three [70] Solar, A-Si Solar cells, TENG – Solar: 100 mW/cm2 40 cm2 HEH: 344 lW, 86 –
raindrop, Rain: 13.6 mL/s mW/m2
airflow Wind speed: 2.7 m/s
Four [2] Solar, PV, wind mill, – 75 W lamp at 90° 1234 cm3 Solar: 190 mW 573
airflow, electrolyzer, hydrogen
water, fuel & Oxygen fuel cell
cell
Wind: 400 mW,
HEH: 315 mW, 3 V,
105 mA
Large-scale
Two [65, 66] Solar, airflow PV, HAWT – – Outdoor-light Enough to drive LED –
Size street-light
Two [67] Solar, wind PV, HAWT – 8.5 m/s 30 m-high up to 1–5 kW –
Three [72, 73] Solar, PV or Solar Thermal – – Pilot scale: VAWT: 157 kW h/day –
airflow, Panel or Solar 10,603 m3 PV: 280 kW h/day Saved
(rainwater Concentrator System, Pumping: 2.6 kW h/day
N. Gure et al.
collection) VAWT
(continued)
Actual: on
220 m high
skyscraper
Three [75] Solar, PV, VAWT, PE – – Actual on Solar: 10 W, PE: 12 V –
airflow, road two-way
traffic roads and
houses
Three [55] Solar, PV, Wind turbine, – – Pilot: 290 kWh –
airflow, Turbine three-floor,
rainwater *9 m high
building
Actual: 98 m
skyscrapers
55
56 N. Gure et al.
Table 2.2 covers fixed-frequency and broadband single-source powered HEHs and
multimode HRTHs in Sects. 2.4.1, 2.4.2 and 2.4.3.1. Table 2.3 emphasizes micro-,
meso-, and large-scale based comparisons of two-, three- and four-multisource
powered HEHs in Sects. 2.4.3.2, 2.4.3.3 and 2.4.3.4. The given device volume and
surface areas are the conservative values of the HEHs, and generally belong to
transducer parts and active surface areas. The overall novel HEHs comparisons are
evaluated in Sect. 2.5.
2.5 Conclusions
The need to achieve greater power generation leads researchers to investigate

hybridization of EHing transduction mechanisms, working modes and source
powers. As indicated in literature [10–23], in the scope of this chapter review, it is
seen that HEHs generate greater power outputs than their single harvester
components.
Among classic HEHs, Table 2.1 shows that Shan et al. HEH holds the greatest
power output of 33 mW along with broadband performance in the range of 7 to
17 Hz [16]. Regarding the device volume, Ali et al. holds greater power density
than Shan et al. [32]. Once PE and EM power generations are compared, generally,
EM power generations are lower than PE parts with an only exception of Ab
Rahman et al. HEH with four pole magnet arrangement [20, 34]. Apart from
Wischke et al. microscale HEH, Xu et al. device [31] has the lowest energy gen-
eration with the minimum volume of 187.2 [31].
In novel HEH class, the most promising power generations are proposed by Yu
et al. [62] in microscale, Saini et al. [2] and Karami and Inman [51] in meso-scale,
and Chang et al. [72, 73] in large scale HEHs. Broadband technique significantly
increases output power and density in nonlinear designs [44, 45]. In case of linear
HEHs, Halim et al. novel design suppresses power density of Castagnetti’s HEH
[35] about 5 times, but not the bandwidth [14]. Within HRTHs, best performance is
achieved by Karami and Inman with an average power output of 70 mW and
maximum of 1 W when exposed to 1000 g [51]. However, Jung et al. managed to
produce 37 mW at low excitation as much as 0.03 g [52]. At almost the same
device size with Jung et al. Larkin and Tadesse gained more than 17 mW during
jogging activity on wrist placement [24, 54] (see Table 2.2). In micro-scale HEH
class, Wischke et al. managed to gain the highest output power of 215 lW via
tuning in broadband operation, yet the total harvester volume is much greater [25,
39] than remaining prototypes. The closest result of 177 lW is reached by Yu et al.
by solar and vibration powered HEH with a surface area of *1 cm2 [62]. The
smallest size prototype with 13.4 mm2 surface area is fabricated and tested by Jeon
et al. and it produces 0.27 lW of output power [63]. Comparison of mesoscale
novel HEHs emphasize that four-source powered HEH produces the greatest power
output of 315 mW, whereas occupying the greatest volume [2]. It is seen that
two-multisource powered HEHs produce greater power than three-multisource
powered ones without an exception. Collado and Georgiadis proved that RF and
solar HEH are able to harness 56 mW of power with a small volume of 4 cm3 [58,
59]. Regardless of the device size, Zhong et al. hand induced hybrid TENG and EM
harvester generates total of 67 mW of power [61]. Lastly, large-scale HEHs result
the best when implemented on high buildings such as skyscrapers. The greatest
energy generation is estimated for solar and airflow harvesting along with rainwater
storage, which saves pumping cost and eventually serves a total of 440 kW h of
surplus energy daily [72, 73] (Table 2.3).
Among overall reviewed classic and novel HEHs, the peak power of 1 W is
achieved by Karami and Inman at an extreme excitation of 1000 g, yet reasonable
daily conditions are enough for the generation of 315 mW by Saini et al.
four-source powered novel HEH. Regarding the size of HEHs, Collado and
Georgiadis’ two-source powered HEH generates the highest output power of 56
mW. The most promising power generations are achieved by Saini et al.
four-source powered novel HEH in meso-scale, Wischke et al. tunable broadband
classic HEH in microscale and Chong et al. proposed partially three-source pow-
ered HEH arrangement in large scale. In conclusion, HEHs not only increase the
output powers and power densities, but also enable endless configurations to
maximize harnessing existing power sources.
Acknowledgements Provided supports of Scientific and Technological Research Council of

Turkey (TUBITAK) 2210-C Master Scholarship Program and Technological Entrepreneurship
Industry Support (TGSD) by Republic of Turkey’s Ministry of Science, Industry, and Technology
(MoSIT) are greatly acknowledged.
References
1. Wei CK, Ramasamy G (2011) A hybrid energy harvesting system for small battery powered
applications. In: 2011 IEEE conference on sustainable utilization and development in
engineering and technology (student), October, pp 165–170. doi:10.1109/STUDENT.2011.
6089346
2. Saini PK, Biswas A, Bhanja D (2015) Performance evaluation and simulation of solar panel,
wind mill, fuel cell hybrid system for small scale energy harvesting. J Clean Energy Technol 3
(6):417–421. doi:10.7763/JOCET.2015.V3.234
3. Weddell AS, Magno M, Merrett GV, Brunelli D, Al-Hashimi BM, Benini L (2013) A survey
of multi-source energy harvesting systems. In: Design, automation & test in Europe
conference & exhibition (date). IEEE Conference Publications, New Jersey, pp 905–908.
doi:10.7873/DATE.2013.190
4. Li P, Gao S, Cai H, Wu L (2015) Theoretical analysis and experimental study for nonlinear
hybrid piezoelectric and electromagnetic energy harvester. Microsyst Technol 10(8). doi:10.
1007/s00542-015-2440-8
5. Wu X, Khaligh A, Xu Y (2008) Modeling, design and optimization of hybrid electromagnetic
and piezoelectric MEMS energy scavengers. In: Custom integrated circuits conference
(CICC). IEEE, San Jose, CA, USA, pp 177–180
6. Energy Research Center (2015) Hybrid adaptive ambient vibration energy harvesting. University
of Maryland. http://www.umerc.umd.edu/projects/harvest01. Accessed 10 Oct 2015
58 N. Gure et al.
7. IDTechEx (2015) Multi-mode energy harvesting. Energy Harv J http://www.energyharvesting

journal.com/articles/8294/multi-mode-energy-harvesting. Accessed 10 Oct 2015
8. Yang B (2010) Hybrid energy harvester based on piezoelectric and electromagnetic
mechanisms. J Micro/Nanolithogr MEMS MOEMS 9(2):023002. doi:10.1117/1.3373516
9. Cook-Chennault KA, Thambi N, Sastry AM (2008) Powering MEMS portable devices—a
review of non-regenerative and regenerative power supply systems with special emphasis on
piezoelectric energy harvesting systems. Smart Mater Struct 17(4):043001. doi:10.1088/0964-
1726/17/4/043001
10. Challa VR, Prasad MG, Fisher FT (2009) A coupled piezoelectric–electromagnetic energy
harvesting technique for achieving increased power output through damping matching. Smart
Mater Struct 18(9):095029. doi:10.1088/0964-1726/18/9/095029
11. Wischke M, Woias P (2008) A multi-functional cantilever for energy scavenging from
vibrations. In: Proceedings of PowerMEMS 2008. IEEE, Sendai, pp 73–76. http://cap.ee.ic.ac.
uk/*pdm97/powermems/2008/pdfs/073-76%20Wischke,%20M.pdf
12. Khbeis MT (2010) Development of a simplified, mass producible hybridized ambient, low
frequency, low intensity vibration energy scavenger (half-lives). University of Maryland,
College Park. http://drum.lib.umd.edu/handle/1903/10778
13. Eun Y, Kwon D, Kim M, Yoo I, Sim J, Ko H-J, Cho K-H, Kim J (2015) A flexible hybrid
strain energy harvester using piezoelectric and electrostatic conversion. Smart Mater Struct 23
(4):045040. doi:10.1088/0964-1726/23/4/045040
14. Halim MA, Cho HO, Park JY (2014) A handy-motion driven, frequency up-converted hybrid
vibration energy harvester using PZT bimorph and nonmagnetic ball. J Phys Conf Ser 557
(1):012042. doi:10.1088/1742-6596/557/1/012042
15. Shan X, Guan S, Liu Z, Xu Z, Xie T (2013) A new energy harvester using a piezoelectric and
suspension electromagnetic mechanism. Appl Phys Eng 14(12):890–897. doi:10.1631/jzus.
A1300210
16. Shan X, Xu Z, Song R, Xie T (2013) A new mathematical model for a piezoelectric-
electromagnetic hybrid energy harvester. Ferroelectrics 450(1):57–65. doi:10.1080/00150193.
2013.838490
17. Li P, Gao S, Cai H (2013) Modeling and analysis of hybrid piezoelectric and electromagnetic
energy harvesting from random vibrations. Microsyst Technol 21(2):401–414. doi:10.1007/
s00542-013-2030-6
18. Chen S, Sun J, Hu J (2013) A vibration energy harvester with internal impact and hybrid
transduction mechanisms. In: 13th international conference on fracture. IEEE, Beijing, pp 1–10
19. Karthik KS, Ali SF, Adhikari S, Friswell MI (2013) Base excited hybrid energy harvesting.
In: 2013 IEEE international conference on control applications (CCA). IEEE, Hyderabad,
pp 978–982. doi:10.1109/CCA.2013.6662878
20. Ab Rahman MF, Kok SL, Ruslan E, Dahalan AH, Salam S (2013) Comparison study between
four poles and two poles magnets structure in the hybrid vibration energy harvester. In: IEEE
(ed) 2013 IEEE student conference on research and development (SCOReD), Putrajaya,
Malaysia Comparison, pp 16–17
21. Sang Y, Huang X, Liu H, Jin P (2012) A vibration-based hybrid energy harvester for wireless
sensor systems. IEEE Trans Magn 48(11):4495–4498
22. Tadesse Y, Priya S (2008) Multimodal energy harvesting system: piezoelectric and
electromagnetic. J Intell Mater Syst Struct 20(5):625–632. doi:10.1177/1045389X08099965
23. Matiko JW, Grabham NJ, Beeby SP, Tudor MJ (2014) Review of the application of energy
harvesting in buildings. Meas Sci Technol 25(1):012002. doi:10.1088/0957-0233/25/1/
012002
24. Larkin M, Tadesse Y (2014) HM-EH-RT: hybrid multimodal energy harvesting from
rotational and translational motions. Int J Smart Nano Mater 4(4):257–285
25. Wischke M, Masur M, Goldschmidtboeing F, Woias P (2010) Electromagnetic vibration
harvester with piezoelectrically tunable resonance frequency. J Micromech Microeng 20
(3):035025. doi:10.1088/0960-1317/20/3/035025
26. Tang L, Yang Y, Soh CK (2010) Toward broadband vibration-based energy harvesting.
J Intell Mater Syst Struct 21(18):1867–1897. doi:10.1177/1045389X10390249
27. Khameneifar F (2011) Vibration-based piezoelectric energy harvesting system for rotary
motion applications. Simon Fraser University. http://summit.sfu.ca/item/11906
28. Lafont T, Gimeno L, Delamare J, Lebedev GA, Zakharov DI, Viala B, Cugat O, Galopin N,
Garbuio L, Geoffroy O (2015) Magnetostrictive—piezoelectric composite structures for
energy harvesting. J Micromech Microeng 22(9):094009. doi:10.1088/0960-1317/22/9/
094009
29. Mahmoudi S, Kacem N, Bouhaddi N (2014) Enhancement of the performance of a hybrid
nonlinear vibration energy harvester based on piezoelectric and electromagnetic transduc-
tions. Smart Mater Struct 23(7):075024. doi:10.1088/0964-1726/23/7/075024
30. Becker P, Folkmer B, Manoli Y (2009) The hybrid vibration generator, a new approach for a
high efficiency energy scavenger. In: IEEE (ed) International workshop on micro and
nanotechnology for power generation and energy conversion applications. PowerMEMS,
Washington, DC, pp 439–442
31. Xu ZL, Wang XX, Shan XB, Xie T (2012) Modeling and experimental verification of a
hybrid energy harvester using piezoelectric and electromagnetic technologies. Adv Mater Res
569:529–532. doi:10.4028/www.scientific.net/AMR.569.529
32. Ali NM, Mustapha AA, Leong KS (2013) Investigation of hybrid energy harvesting circuits
using piezoelectric and electromagnetic mechanisms. In: IEEE student conference on research
and development, December, pp 16–17
33. Xia H, Chen R, Ren L (2015) Analysis of piezoelectric–electromagnetic hybrid vibration
energy harvester under different electrical boundary conditions. Sens Actuators A 234:87–98.
doi:10.1016/j.sna.2015.08.014
34. Ab Rahman MF, Kok SL, Ali, NM, Hamzah RA, Aziz KA (2013). Hybrid vibration energy
harvester based on piezoelectric and electromagnetic transduction mechanism. In: 2013 IEEE
conference on clean energy and technology (CEAT). IEEE, Langkawi TBD, pp 243–247.
doi:10.1109/CEAT.2013.6775634
35. Castagnetti D (2015) A belleville-spring-based electromagnetic energy harvester. Smart Mater
Struct 24(9):94009. doi:10.1088/0964-1726/24/9/094009
36. Twiefel J, Westermann H (2013) Survey on broadband techniques for vibration energy
harvesting. J Intell Mater Syst Struct 24(11):1291–1302. doi:10.1177/1045389X13476149
37. Zhou S, Cao J, Inman DJ, Lin J, Liu S, Wang Z (2014) Broadband tristable energy harvester:
modeling and experiment verification. Appl Energy 133:33–39. doi:10.1016/j.apenergy.2014.
07.077
38. Li P, Gao S, Niu S, Liu H, Cai H (2014) An analysis of the coupling effect for a hybrid
piezoelectric and electromagnetic energy harvester. Smart Mater Struct 23(6):065016. doi:10.
1088/0964-1726/23/6/065016
39. Wischke M, Masur M, Goldschmidtboeing F, Woias P (2010) Piezoelectrically tunable
electromagnetic vibration harvester. In: 2010 IEEE 23rd international conference on micro
electro mechanical systems (MEMS). IEEE, pp 1199–1202. doi:10.1109/MEMSYS.2010.
5442427
40. Xu ZL, Shan XB, Song RJ, Xie T (2014) Electromechanical modeling and experimental
verification of nonlinear hybrid vibration energy harvester. In: Applications of ferroelectrics
(ed) 2014 joint IEEE international symposium on the applications of ferroelectric,
international workshop on acoustic transduction materials and devices & workshop on
piezoresponse force microscopy. IEEE, State College, PA, USA, pp 1–4. doi:10.1109/ISAF.
2014.6923018
41. Dias JAC, Marqui C De, Erturk A (2013) Hybrid piezoelectric-inductive flow energy
harvesting and dimensionless electroaeroelastic analysis for scaling. Am Inst Phys 044101
(102):1–6
42. Dias JAC, De Marqui C, Erturk A (2015) Three-degree-of-freedom hybrid
piezoelectric-inductive aeroelastic energy harvester exploiting a control surface. AIAA J 53
(2):394–404. doi:10.2514/1.J053108
60 N. Gure et al.
43. Reuschel T, Salehian A, Kotsireas I, Melnik R, West B (2011) Analysis and modelling
towards hybrid piezo-electromagnetic vibrating energy harvesting devices. doi:10.1063/1.
3663465
44. Karami MA, Inman DJ (2011) Equivalent damping and frequency change for linear and
nonlinear hybrid vibrational energy harvesting systems. J Sound Vib 330(23):5583–5597.
doi:10.1016/j.jsv.2011.06.021
45. Karami M.A, Inman, DJ (2010) Nonlinear hybrid energy harvesting utilizing a
piezo-magneto-elastic spring. In: Ghasemi-Nejhad MN (ed) SPIE smart structures and
materials + nondestructive evaluation and health monitoring, vol 7643, p 76430U–76430U–
11. doi:10.1117/12.847566
46. Leadenham S, Erturk A (2015) Nonlinear M-shaped broadband piezoelectric energy harvester
for very low base accelerations: primary and secondary resonances. Smart Mater Struct 24
(5):055021. doi:10.1088/0964-1726/24/5/055021
47. Ishikuro H (2011) Energy harvesting technology, system LSI design. Tokyo, Japan
48. Sasaki K, Osaki Y, Okazaki J, Hosaka H, Itao K (2005) Vibration-based automatic
power-generation system. Microsyst Technol 11(8–10):965–969. doi:10.1007/s00542-005-
0506-8
49. Khaligh A, Zeng P, Zheng C (2010) Kinetic energy harvesting using piezoelectric and
electromagnetic technologies—state of the art. IEEE Transduct Ind Electron 57(3):850–860
50. Romero E, Warrington RO, Neuman MR (2009) Energy scavenging sources for biomedical
sensors. Physiol Meas 30(9):35–62. doi:10.1088/0967-3334/30/9/R01
51. Karami MA, Inman DJ (2013) Nonlinear dynamics of the hybrid rotary-translational energy
harvester. In: ASME conference on smart materials, adaptive structures and intelligent
systems (SMASIS). doi:10.1115/SMASIS2013-3110
52. Jung H-J, Kim I-H, Min DY, Sim S-H, Koo J-H (2013) A hybrid electromagnetic energy
harvesting device for low frequency vibration. In: Sodano H (ed) Active and passive
structures and intelligent systems 2013, SPIE, vol 8688, p 86881I. doi:10.1117/12.2010014
53 Taylor, Francis (2013) Int J Smart Nano Mater 4(4):257–285. www.tandfonline.com/doi/abs/
10.1080/19475411.2014.902870
54. Larkin MR, Tadesse Y (2014) Characterization of a rotary hybrid multimodal energy
harvester. In: Proceedings of SPIE, October 2015, vol 9057, p 90570U. doi:10.1117/12.
2045271
55. Mithra VTV, Sharmila R, Giridhar MVSS (2014) Development of hybrid power generation
model using rain water, solar and wind. In: Centre for water resources, IST, JNTUH. JNTUH,
Hyderabad
56. Gambier P, Anton SR, Kong N, Erturk A, Inman DJ (2011) Piezoelectric, solar and thermal
energy harvesting for hybrid low-power generator systems with thin-film batteries. Measur
Sci Technol 23(1):015101. doi:10.1088/0957-0233/23/1/015101
57. Hehr A, Park G, Farinholt K (2012) Hybrid energy harvesting/transmission system for
embedded devices. Proc SPIE Ind Commer Appl Smart Struct Technol 8343(07):1–6. doi:10.
1117/12.915461
58. Collado A, Georgiadis A (2013) Conformal hybrid solar and electromagnetic (EM) energy
harvesting rectenna. IEEE Trans Circuits Syst I Regul Pap 60(8):2225–2234. doi:10.1109/
TCSI.2013.2239154
59. Georgiadis A, Collado A, Via S, Meneses C (2011) Flexible hybrid solar/EM energy
harvester for autonomous sensors. In: IEEE MTT-S international microwave symposium
(IMS), Baltimore, MD, USA
60. Chatterjee P, Bryant M (2015) Transfer matrix modeling of a tensioned piezo-solar hybrid
energy harvesting ribbon. In: Liao W-H (ed) SPIE, vol 9431, p 94310D. doi:10.1117/12.
2086138
61. Zhong X, Yang Y, Wang X, Wang ZL (2015) Rotating-disk-based hybridized
electromagnetic-triboelectric nanogenerator for scavenging biomechanical energy as a mobile
power source. Nano Energy 13:771–780. doi:10.1016/j.nanoen.2015.03.012
62. Yu H, Yue Q, Zhou J, Wang W (2014) A hybrid indoor ambient light and vibration energy
harvester for wireless sensor nodes. Sensors 14(5):8740–8755. doi:10.3390/s140508740
63. Jeon S, Kim D, Yoon G-W, Yoon J, Choi Y-K (2015) Self-cleaning hybrid energy harvester
to generate power from raindrop and sunlight. Nano Energy 12:636–645. doi:10.1016/j.
nanoen.2015.01.039
64. Zakharov D, Lebedev G, Cugat O, Delamare J, Viala B, Lafont T, Gimeno L, Shelyakov A
(2012) Thermal energy conversion by coupled shape memory and piezoelectric effects.
J Micromech Microeng 22(9):094005. doi:10.1088/0960-1317/22/9/094005
65. Tong CW (2014) Lighting up University of Malaya—an innovative idea to provide outdoor
lighting using wind-solar hybrid renewable energy sources. University of Malaya. http://
www.researchsea.com/html/article.php/aid/8212/cid/1/research/science/university_of_malaya/
lighting_up_university_of_malaya_____an_innovative_idea_to_provide_outdoor_lighting_
using_wind-solar_hybrid_renewable_energy_sources.html. Accessed 12 Oct 2015
66. Malaya U (2014) Outdoor lighting using wind-solar hybrid renewable energy sources.
ScienceDaily. http://www.sciencedaily.com/releases/2014/05/140525204734.htm. Accessed
12 Oct 2015
67. De MM, Reddy KJ, Srikanth L, Jayasankar S, Muthu K, Samiullah TH, Dayananda GN (2014)
Research, design, development & demonstration (RD 3) of a wind solar hybrid (WiSH)
system. In: Indian technology congress, ITC2014. NIMHANS Convention Centre, The
Institution of Engineers (India), NDRF & Indian Technology Congress, Bangalore, pp 1–10
68. Porcarelli D, Brunelli D, Magno M, Benini L (2012) A multi-harvester architecture with
hybrid storage devices and smart capabilities for low power systems. In: International
symposium on power electronics, electrical drives, automation and motion. IEEE, Sorrento
doi:10.1109/SPEEDAM.2012.6264533
69. Chung T-K, Tseng C-Y, Chen C, Wang C (2012) Design, fabrication, and testing of a
thermal/mechanical/magnetic hybrid energy micro-harvester. In: ASME 2012 conference on
smart materials, adaptive structures and intelligent systems, SMASIS2012-8192, vol 2.
ASME, Stone Mountain, Georgia, USA, pp 1–6. doi:10.1115/SMASIS2012-8192
70. Zheng L, Cheng G, Chen J, Lin L, Wang J, Liu Y, Li H, Wang ZL (2015) A hybridized power
panel to simultaneously generate electricity from sunlight, raindrops, and wind around the
clock. Adv Energy Mater. doi:10.1002/aenm.201501152
71. Chen C-C, Chung T-K, Tseng C-Y, Hung C-F, Yeh P-C, Cheng C-C (2015) A miniature
magnetic-piezoelectric thermal energy harvester. IEEE Trans Magn 51(7):1–9. doi:10.1109/
TMAG.2015.2395385
72. Chong WT, Fazlizan A, Poh SC, Pan KC, Ping HW (2012) Early development of an
innovative building integrated wind, solar and rain water harvester for Urban high rise
application. Energy Build 47:201–207. doi:10.1016/j.enbuild.2011.11.041
73. Chong WT, Pan KC, Poh SC, Fazlizan A, Oon CS, Badarudin A, Nik-Ghazali N (2013)
Performance investigation of a power augmented vertical axis wind turbine for urban
high-rise application. Renew Energy 51:388–397. doi:10.1016/j.renene.2012.09.033
74. Donnie EJ (2014) Hybrid energy harvesting device and fixed threshold power production.
United States Patent, NV (US). USOO8847425B2 (12)
75. Sathiyamoorthy S, Bharathi N (2012) Hybrid energy harvesting using piezoelectric materials,
automatic rotational solar panel, vertical axis wind turbine. Procedia Eng 38:843–852. doi:10.
1016/j.proeng.2012.06.106
Chapter 3
Micro-scale Energy Harvesting
for Batteryless Information Technologies
Ali Muhtaroğlu
Abstract Continually growing integration levels and miniaturization in electronics

has led to the enrichment of features to enable a variety of new applications with
simultaneous reduction in system power consumption. The start of the 21st century
is hallmarked by the emergence of small information engines (microsystems) as
part of “internet of things” thrust to provide the intelligence behind the building
blocks of the increasingly automated, digitized, and connected eco-system around
residential, industrial, health, education, transportation, communication, and other
sectors of our civilization. These engines require small amount of power to work in
an embedded environment where frequent access for maintenance, and power
delivery or battery refurbishment is not desirable. The focus of this chapter is
batteryless operation, which is at the center-stage of microscale harvesting research
efforts to enable such applications. Energy and power budgeting, and system design
concerns are reviewed for batteryless operation, and system examples are provided.
First part of Sect. 3.1 discusses demand for micro-scale information technologies.
Section 3.1.3 provides typical power and energy budgets for sensor nodes. The
motivations for batteryless operation are examined in Sect. 3.1.4. Power generation
is provided as a recent power management thrust in such systems in Sect. 3.1.5.
System design issues for realizing batteryless information technologies are covered
in Sect. 3.2 with the review of energy harvesting techniques such as small-scale
solar irradiation, environmental vibration, thermal differences, and ambient
radio-frequency electromagnetic waves. Finally, the last part of the section briefly
discusses upcoming trends and further efforts in enabling batteryless information
technologies at a disruptive level.

Keywords Microsystems Batteryless operation Energy harvesting Energy

scavenging Information technologies
A. Muhtaroğlu (&)
Department of Electrical and Electronics Engineering, Middle East Technical University,
Northern Cyprus Campus, Kalkanlı, Güzelyurt, Mersin 10, Turkey
e-mail: amuhtar@metu.edu.tr

64 A. Muhtaroğlu

AC Alternating-current
ADC Analog-to-digital conversion
ASIC Application specific integrated circuit
BSN Body sensor nod
COTS Commercial off-the-shelf
DC Direct-current
DVS Dynamic voltage scaling (), 20
ICT Information and communication technologies
M2M Machine-to-machine (), 4
MCU Microcontroller unit
MEMS Micro-electromechanical system
MPPT Maximum Power Point Tracking
PC Personal Computer
PV Photovoltaic
PZT Piezoelectric
PCB Printed circuit board
SECE Synchronous electric charge extraction
TEG Thermoelectric generator
WSN Wireless Sensor Networks
3.1 Micro-scale Energy Harvesting
3.1.1 Introduction
The development of integrated microprocessor chip in 1971 was a key enabler for
the Personal Computer (PC). Since then, Moore’s Law has correctly predicted the
seamless technology advancement of doubled transistor count on a chip about every
two years, while the transistors themselves shrunk with every generation.
Additional transistors with better power consumption characteristics were utilized
by component and system architects to develop advanced features in electronic
platforms, and by circuit designers to build faster circuits. More integration was
achieved with every technology generation, which not only pulled down system
cost through reduced number of components, but also lowered overall power dis-
sipation due to the elimination of glue-logic between components, and the fact that
each transistor could be switched on/off with less power. Computer quickly stopped
being an expensive machine in a laboratory reserved for privileged scientists, and
by 1990s marketing segmentation drove differentiation of computer systems to
address emerging usage models, making computers available to virtually everyone.
3 Micro-scale Energy Harvesting for Batteryless Information Technologies 65
Some computers evolved into mobile productivity tools that travel along with us.
Some turned into entertainment centers with emphasis on multi-media performance.
Some continued their evolution on traditional high performance computing with the
goal to achieve more scientific computations per unit time. Another important
segment was enabled through simple-task dedicated ‘small’ information engines
shaping the building blocks of the progressively automated, digitized, and con-
nected eco-system around all major application sectors such as residential, indus-
trial, health, education, transportation, and communication. It was commonly
accepted by the end of the 20th century that wireless machines and networks would
have a fundamental role in intelligent system design. It has become clear ever since
that 21st century is the era of internet of things (IoT) or “internet of everything”,
meaning everything from small sensors to large machines get connected. These
“things” make our life easier, and increase our productivity. The information,
available at our finger tips, contributes to our health, comfort, safety, education,
professional development. This realization quickly leads to the need for embedded
applications without batteries i.e. longevity with minimum maintenance.
Contemporary microsystems for handling information require features to
dynamically trade off performance and power dissipation. Other simultaneously
critical ingredients are long battery life, low energy operation, wireless communi-
cation, compact, low cost, light platforms. Due to the recent emphasis on intelligent
environment, small embedded systems with very low power consumption and
longevity have become vital to the exponentially growing number of ‘miniaturized
and ubiquitous’ devices in the 21st century. Since good power and energy con-
sumption characteristics have been the determinant for the success of new infor-
mation engines, much engineering effort has been spent on various aspects of power
management. Batteryless operation, however, continues to be a great challenge.
Energy harvesting is thus introduced in this section as the means to achieve zero or
close-to-zero average input power for green micro-scale IT (Information
Technologies) of the future. Energy and power budgeting for such applications, the
need for batteryless operation, consumption management and generation are cov-
ered in the following discussion.
3.1.2 Demand for Micro-scale Information Technologies
Use of IT has been proven to increase productivity in organizations for the last few
decades [1]. Tendency for decentralized or distributed operations, mobile lifestyles
has naturally increased the demand for IT. In a recent white paper by Cisco,
tremendous growth in demand for smartphones, tablets, and other mobile com-
puting devices is projected. Average number of connected devices per capita is
projected to grow from 2 to 3.2 in the period between 2014 and 2019. Particularly,
machine-to-machine (M2M) connection is the fastest growing category as depicted
in Fig. 3.1 [2].
66 A. Muhtaroğlu
Fig. 3.1 Global projected

growth of connected devices
by type [2]
The reported trends further emphasize the forecasted growth in demand for
smaller or micro-scale IT devices for applications with M2M connectivity to
improve quality of daily services. The largest expected growth is in connected
health sector, with compound annual growth rate close to 50% between 2014 and
2019 [2]. Similarly, agriculture, construction, emergency services, and transporta-
tion (cars) sectors are expected to experience a fast growth rate in the adoption of
M2M connected devices. Sensing, digitization, processing, and communication are
ubiquitously embedded into everyday objects, turning them into the IoT [3].
Individual devices thus collaborate in Wireless Sensor Networks (WSNs) across a
variety of contexts to provide invaluable data and information on physical phe-
nomena. The applications of such WSNs are widespread as summarized in Fig. 3.2.
The generic WSN system architecture and requirements are outlined in Fig. 3.3.
Many embedded sensors attached to the system have the most stringent power
dissipation and energy requirements for sensing and transmission of data to the
local HUB. Microelectromechanical system (MEMS) components may be used for
miniaturization of mechanical aspects of the sensor nodes. These nodes also contain
some processing features implemented in application specific integrated circuit
(ASIC) components in a subset of the applications of interest. Adding to these the
size and cost constraints, and potentially dynamic, unfavorably out-of-touch
physical environment, the sensor node system design becomes a challenging
problem. The local processing and control hub follows a pre-programmed algorithm
to collect data from a variety of sensors in the neighborhood (few meters to few
kilometers as typical range), and may respond by activating various actuators in the
intelligent system after processing the received information. A desirable feature of
the WSN is the ability to communicate with cloud or servers over the internet for
remote monitoring and control. This aspect further allows remote policies and user
interfaces to be devised and implemented through software for a particular WSN.
Considering transportation, for example, use of WSNs in aeronautical vehicles is
an emerging trend. More than one thousand sensor nodes may potentially be uti-
lized in an aircraft, for automation of flight attendant control panels, passenger
services, cabin illumination, surveillance cameras, safety sensors (smoke, temper-
ature, humidity), passenger entertainment, signs and various other controls [7]. It
has become increasingly difficult to reliably wire signals and power lines to and
from the sensors and actuators through the tight space available within the structure
Fig. 3.2 Application of WSNs (based on [4–6])
Fig. 3.3 A generic WSN system architecture and component requirements [8]
of the plane. As in the case of aircrafts, many uses of WSNs require accurate
processing and delivery of data from one location to another, while at the same time
satisfying strict system requirements related to power source, power consumption
and size.
68 A. Muhtaroğlu
3.1.3 Power and Energy Budgets
Sensor nodes are either battery driven or are required to operate on energy scav-
enged (harvested) from the environment. Hence, the energy budget is tight. The
importance of optimizing design based on application needs cannot be stressed
enough. Designs using commercial off-the-shelf (COTS) components are often
disadvantaged in power consumption due to redundancy of features, and inter-
connect. Custom integrated designs and system-on-chip (SOC) components on the
other hand can potentially deliver minimum power for the application, but typically
are subject to longer design, fabrication and test lead times, and higher unit costs for
low volume operations.
A typical architecture for wireless sensor nodes is depicted in Fig. 3.4. The
voltage from the battery is converted to one or more voltage levels by DC/DC
converter(s) to supply power to sensor(s), one or more analog-to-digital converters
(ADCs), microcontroller unit (MCU), a memory if storage requirement is higher
than what is available in the MCU, and a radio for communication of data. Local
algorithm for capturing and processing the data is programmed into the MCU.
Although transceiver and/or receiver circuits in the radio module are generally
accepted to be the most power-hungry of all blocks, the rest of the system com-
ponents, software and hardware power management features, algorithms and pro-
tocols may potentially have far greater cumulative impact on the average power
dissipation and energy consumption of the node.
Table 3.1 illustrates the power dissipation values for fundamental components
and modes of the two sensor node architectures captured by a case study in 2002.
Although the technology scaling in integrated circuit industry results in significant
(potentially up to 2–3 orders of magnitude) reduction today in all of the provided
power figures, the table is indicative of the relative spread of power consumption
across modes of operation, and across different architectures. For calculation of
accurate energy consumption, it is necessary to characterize the portion of the total
operation time that is statistically spent in each valid sensor node power state i.e.
percent residence per power state. Then the average power per sensor node can be
computed for battery life estimation through weighted summation of all power
levels. A comprehensive model was recently developed [10] for accurately
Fig. 3.4 System architecture of a typical wireless sensor node [9]

Table 3.1 Power analysis of (a) Rockwell’s WINS, (b) Medusa II nodes [9]
(a) Rockwell’s WINS (b) Medusa II
MCU Sensor Radio mode Power MCU Sensor Radio mode Power
mode mode (mW) mode mode (mW)
ACTIVE ON Tx (Power: 1080.5 ACTIVE ON Tx (Power: 24.58
36.3 mW) 0.7368 mW)
Tx (Power: 986.0 Tx (Power: 19.24
19.1 mW) 0.0979 mW)
13.8 mW) 0.7368 mW)
3.47 mW) 0.0979 mW)
2.51 mW) 0.7368 mW)
0.96 mW) 0.0979 mW)
0.30 mW) 0.7368 mW)
0.12 mW) 0.0979 mW)
ACTIVE ON Rx 751.6 ACTIVE ON Rx 22.20
ACTIVE ON IDLE 727.5 ACTIVE ON IDLE 22.06
ACTIVE ON SLEEP 416.3 ACTIVE ON OFF 9.72
ACTIVE ON REMOVED 383.3 IDLE ON OFF 5.92
SLEEP ON REMOVED 64.0 SLEEP OFF OFF 0.02
ACTIVE REMOVED REMOVED 360.0 – – – –
predicting the power consumption of wireless sensor nodes across a variety of

usage scenarios. It can be inferred from the table that, the variation with various
power modes aside, transmitter and receiver (radio) circuits in sensor nodes con-
sume a significant portion of the total power budget. Low-power radios typically
have a minimum power consumption of around 50 µW, for example, in order to
provide sufficient sensitivity in body sensor networks [11]. In a recently reported
bladder-pressure monitoring application [12], the concern was battery life, since the
recharge frequency was not high. A custom application specific integrated circuit
(ASIC) implant in this application achieved around 33 µW average power when
enhanced with a special power control unit (PCU). When the unit was excluded, the
average power exceeded 1 mW.
3.1.4 The Case for “Batteryless”
A viable energy source continues to be a challenge in wireless sensors that make up

the IoT infrastructure. Although batteries have traditionally been used to power
70 A. Muhtaroğlu
small microelectronic systems, much of recent research is focused on eliminating

them from future sensor networks. Historically three parallel motivations have
driven this thrust:
i. Green Computing: Information and communication technologies (ICT) gen-
erate 2% of worldwide CO2 emissions, which roughly corresponds to a quarter
of the emissions generated by cars [13]. With the unprecedented growth
forecasted in ICT, as outlined in Sect. 3.1.2, this trend brings on significant
concerns about the contribution of computing to global warming. There have
been many emerging initiatives on energy efficiency around the globe, such as
the Energystar [14], which is clearly a significant component in green com-
puting approach. The other important component is achieving a sustainable
life-cycle cost in microelectronics. An undisputed problem in this regard has
been the disposal of batteries or disposal of microelectronic components that
contain batteries.
ii. Avoiding Invasive and Bulky: Implantable sensors are of high interest in
biomedical field for monitoring various body functions, such as blood pressure
[15], electrocardiogram (ECG), electroencephalogram (EEG), and elec-
tromyogram (EMG) [16]. However, their utility reduces significantly if they
become invasive to the part of the body they are integrated in, or even cause
additional patient anxiety due to bulkiness. Significant portion of the volume
and intrusiveness of non-batteryless body sensor networks can be traced to
on-board batteries.
iii. Cost Reduction: Due to the cost and inconvenience or in some cases unfea-
sibility of regular refurbishing of batteries, the operational expenses of the
WSNs increase significantly. In many embedded applications, it may even be
cheaper to install a new sensor node instead of accessing the old one for
maintenance and battery replacement.
3.1.5 Consumption Management and Generation
Due to the concomitant power resource constraints, as discussed in the previous

section, management of power and energy at sensor nodes takes high priority. There
are many energy saving methods to improve the working model or algorithms
employed by a given sensor at the data manipulation or communication layer in
order to ensure different components are activated on demand whenever there is a
significant “event”. These techniques significantly reduce average power con-
sumption, and hence the energy. A comprehensive tree of approaches that can be
applied to the energy conservation problem for a particular system is outlined in
Fig. 3.5.
Duty cycling solutions emphasize management of the radio transceiver subsys-
tem, which traditionally consumes much more energy than the processing sub-
system. Distributed sleep/wake-up scheduling algorithms are applied to the radio
Fig. 3.5 Taxonomy of approaches for energy consumption management in sensor networks [17]
circuits in order to turn them on when there is data to send or receive. The
present-day subsystem components also support multiple power management
modes, depending on “how active” they need to be in any given instant. Such
component power management features will be further discussed later. Excessive
data sampling unnecessarily keeps both the processor(s) and the transceivers awake
causing them to burn active power. Thus, data driven techniques shown as one of
the main branches in Fig. 3.5 target diminution in the amount of sampled or
transmitted data, while maintaining the minimum required sensing accuracy. When
the sensor nodes are static, some nodes may be on congested data paths while
others may hardly receive data ‘hops’ from the surroundings. Such asymmetry in
WSN may require location based management schemes. In cases when sensor
nodes are dynamic, data ‘hops’ may first detect proximity of the receiving sensors
to the transmitting sensors in order to optimize power dissipated in transmission.
Such algorithms are categorized under mobility-based approaches in the figure.
Details of the individual energy conservation schemes in the tree will not be dis-
cussed further here, but can be accessed at the original publication [17].
Various hardware design and fabrication methods are utilized to reduce power
consumption at component level, which all directly contribute to the power man-
agement of the sensor node. Different facets of power management in microelec-
tronic computing systems are presented in Fig. 3.6. The balloon on the lower right
represents the software design features described in the above paragraphs. As the rest
of the picture exhaustively presents, the hardware power management design task is
shared across different expertise areas. Process technology based features combine
physical enhancements to semiconductors with fabrication methods to reduce static
power consumption (leakage) per device for a targeted switching performance.
Circuit designers focus on effective use of transistors in order to simultaneously
achieve goals related to circuit delay, dynamic and static power dissipation through
various circuit design techniques. Architects use their knowledge of hardware/
software interface and how different hardware blocks are expected to “behave” to
optimize power at a higher abstraction level. For example, memory blocks can be put
in low-leakage (e.g. low supply voltage) mode, when not in use. Clocks may be
72 A. Muhtaroğlu
removed from blocks of processing units (clock gating) to reduce dynamic power
dissipation, or power may be shut down to inactive blocks selectively to lower total
power. Dynamic speed/power tradeoffs can be realized in real-time through the
simultaneous modulation of supply voltage and operating system clock frequency.
Integration of power electronics, such as DC/DC regulators, into the SOC reduces
system cost, and allows elimination of losses in large power delivery networks. More
importantly, processing engines can control supply voltage with better resolution to
support advanced power management. Temperature, power, motion sensors incor-
porated into the electronic components or SOCs further provide tools for system
level power/performance optimizations, while keeping within reliability constraints.
Energy scavenging (a.k.a. energy harvesting), depicted at the bottom of Fig. 3.6,
has recently been a fast growing area of research and development to enhance
system power management characteristics, and address the issues associated with
batteries covered in Sect. 3.1.4. This trend is encouraged by the ever-diminishing
power and energy dissipation requirements of IoT devices with technology scaling
trends. Energy harvesting is the ultimate sustainable or green solution to powering
micro-scale IT systems. It involves use of constant or continually available energy
Fig. 3.6 Different facets of power management [18]

sources in the vicinity of the sensor node to generate electrical power. Since such
resources practically never run out, they are referred to as “renewable”. Although
traditional renewable energy sources such as sunlight and wind can and should be
utilized in a variety of IoT contexts when applicable, they are often non-existent in
the environment of the embedded sensors. When they do get utilized, the associated
transducers need to be miniaturized to satisfy the requirements of the micro-scale IT
nodes. Therefore, there are multiple system design challenges involved. Examples
from appealing solutions will be discussed in the next section in the context of
applications.
3.2 Systems and Applications
3.2.1 Introduction
System design issues associated with incorporating and utilizing available energy
sources around intelligent embedded systems are reviewed in this section.
Particularly micro-scale (solar) photovoltaic, piezoelectric, vibration based elec-
tromagnetic, thermoelectric and ambient radio-frequency (RF) wave energy will
receive focus. The final section of the chapter highlights examples from emerging
applications to demonstrate successful technology integration, and upcoming
trends.
3.2.2 System Design
The goal of an ideal energy harvester, which is often referred to as a micro-power

generator, is to replace the battery previously depicted in Fig. 3.4. Availability of
sufficient energy at the source is an important but an insufficient condition for
energy harvesting. This is due to the fact that useful power available at the
micro-scale power generators typically varies between micro-Watts to tens of
milli-Watts, with voltage output between tens of milli-Volts to hundreds of
milli-Volts, at most, under realistic scenarios. It may take one or more power
electronics stages before the power can be safely delivered to the sensor node with
acceptable voltage and current characteristics.
Each stage bears losses, and inadvertently reduces overall system efficiency by
some amount. For example, if the generated voltage from the harvester is in AC
(alternating-current) form, it requires rectification because typical electronic loads
consume DC (direct-current) power. If the harvester output impedance or resistance
characteristic is far from that of an ideal supply or varies significantly with envi-
ronmental conditions, then impedance matching circuit may be required to extract
maximum energy. Finally, the voltage needs to be stepped-up to suitable levels for
74 A. Muhtaroğlu
the load circuits in the target system. Integrated design approaches often combine a
number of such circuits in one design to simultaneously reduce the cost and size of
the interface electronics, and improve power conversion efficiency. It is challenging
to design interface electronics with high efficiency, especially in cost and area
constrained systems. Therefore, one needs to carefully analyze the target applica-
tion, and evaluate solution space for energy harvester integration at minimum cost.
3.2.3 Micro-scale Harvesters
Micro-scale IT systems vary widely in volume between few cm3 to few tens of cm3,
which sets the range for acceptable sizes for the energy harvesters. The energy
sources in the vicinity also alter in magnitude and availability. Depending on use
environment, some sensor nodes may experience motion and vibration, while others
may be subject to temperature variations. A number of embedded applications may,
by design, shield RF radiation, while others may be exposed to strong RF waves.
Therefore, it is necessary to carefully analyze application environment and
requirements before incorporating the correct energy harvesting mode. Common
ambient energy sources for embedded sensors are outlined in Table 3.2, along with
available power density and voltage levels at their output. Pros and cons have also
been listed.
3.2.3.1 Photovoltaic (Solar) Energy Harvesting
In a typical photovoltaic (PV) system shown in Fig. 3.7, MPPT (Maximum Power
Point Tracking) sub-system matches the input impedance of the power conditioning
circuitry to the output impedance of the PV source, and the output impedance of the
power conditioning circuitry to the input impedance of the load. The function of
maximizing power extraction across varying temperature and irradiance conditions
is also often incorporated into the MPPT. All of these features bring along the extra
burden of power consumption, and only make sense when this consumption is
much lower than the harvested average power. System analysis is performed on
feasibility of incorporating such circuits, which often results in bare minimum
power conditioning design in sensor systems without a digital controller.
One such simplified design based on COTS is presented in [19], with robust
functionality, bare-bone maximum power point (MPP) support, and up to 10%
divergence from the truly optimal operating point (PV current and voltage). The
energy harvesting interface circuit consumes less than 1 mW. Since no MCU or
DSP (Digital Signal Processor) is utilized for MPP regulation, the system can be
shut down when unused for energy savings. A small (9 mm2) pilot cell with the
same characteristics as the main harvester array (112 cm2) is used to track maxi-
mum power point based on fractional open circuit voltage method as depicted in
Fig. 3.8. The governing relationship in Eq. 3.1 justifies this simple approach to
Table 3.2 Ambient energy sources and characteristics of micro-power generators
PV Solar Thermoelectric Piezoelectric Electromagnetic Ambient RF
vibration vibration
Power Outdoor: 50–100 10–200 µW/cm3 1–2 µW/cm3 0.0002–1 µW/cm2
density 100 mW/cm2 µW/cm2 per 0C
Indoor: <100 µW/cm2
Output 0.5 V max 10–100 mV 10–20 V few 100 mV 3–4 V (open ckt)
voltage (open ckt)
Availability Lighted Surfaces with DT Hz–kHz Hz Vibration Vicinity to radiation source
condition environment Vibration
Pros High power Non-intermittent/less intermittent High voltage Well developed Antenna can be integrated
density than alternatives Well developed Widely available
Well developed technology
technology
Cons Intermittent Low voltage Highly variable Bulky Very sensitive to distance of
Highly dependent on Need DT output Low power the RF source
light Large area density
High output Low output
3 Micro-scale Energy Harvesting for Batteryless Information Technologies
impedance voltage
75
76 A. Muhtaroğlu
Fig. 3.7 PV harvesting

system [18]
Fig. 3.8 PV harvester platform with simple pilot-cell based MPPT [19]
MPPT implementation to optimize extracted power across temperature and irradi-

ation variations:
VMPP KFOC VOC KFOC ðKPILOT VPilotCell Þ ð3:1Þ
where, VMPP is the target voltage at maximum power point, KFOC is the fractional
open circuit voltage constant, and KPilotCell is the scaling constant associated with
the pilot cell reference. A supercapacitor is deployed for storage. Power conversion
efficiency, using commercial LTC3401 step-up voltage regulator in the last stage
(DC-DC block in Fig. 3.8), varies between 50–80% with 50 mW of power deliv-
ered to the commercial embedded sensor platform called Tmotesky.
3.2.3.2 Thermoelectric (Temperature Difference) Energy Harvesting
The available voltage level from thermoelectric micro-modules is low, as noted

previously in Table 3.2. Recent work on integrating TE modules into WSN nodes
have thus focused on developing efficient voltage step-up circuits. An integrated
interface circuit in 0.35 µm CMOS technology reported by [20] boosts 35 mV
input voltage from a TE module up to regulated 1.8 V output to drive a sensor
node. A motion activated switch is used to mechanically assist the starting of power
generation, as depicted in Fig. 3.9. This approach avoids batteries for start-up. The
interface achieves 58% efficiency using external inductors for the buck converter
and start-up. The existing large inductors (not shown in the figure) cannot be
integrated on-chip in this design, adding incrementally to the system cost (and size).
The printed circuit board (PCB) for the thermal harvesting system only occupies
7.5 cm2.
A 130 nm CMOS ultra-low power batteryless energy harvesting body sensor
node (BSN) system-on-chip (SoC) was presented in [16], which acquires, pro-
cesses, and transmits electrocardiogram (ECG), electromyogram (EMG), and
electroencephalogram (EEG) data. Dynamic power management concepts, low
voltage and subthreshold circuit techniques for low power design are applied to
constrain average power dissipation to 19 µW, which is supplied by thermoelectric
module and/or RF harvester. It operates at a minimum thermoelectric supply
voltage of 30 mV. The design, depicted at high level in Fig. 3.10, is a first of its
kind in multiple aspects, but certainly in the small power dissipation achieved for
the number of features it supports. A 4 4 cm2 COTS thermoelectric generator
(TEG) module successfully delivers the required sensor node power at room tem-
perature when attached to the back of the neck or chest of a human being. First, the
boost converter and voltage regulator together step up harvested TEG input up to a
regulated 1.35 V using an off-chip capacitor and an inductor for efficient
Fig. 3.9 Thermoelectric energy harvesting system block diagram [20]

78 A. Muhtaroğlu
Fig. 3.10 High level block diagram of the BSN design [16]
conversion. The regulator in addition provides five regulated voltage supplies and a
bandgap reference to the rest of the chip. Monitored signal is processed through an
analog front-end (AFE) with programmable gain and sampling range, and con-
verted to digital domain using the analog-to-digital conversion (ADC) block.
Signals as low as a few µVs are amplified with less than 4 µW power consumption
per analog channel. The digital processing system that follows is implemented
using subthreshold circuit design techniques, and incorporates power and clock
gating as well as dynamic voltage scaling (DVS) for optimization of power dissi-
pation. Internal voltage regulation control circuit requires 600 mV to start up,
which is not available from the TEG. Wireless RF power is therefore used for
kick-starting. A short RF burst of −10 dBm completes a one-time precharge of
VBOOST node within few seconds after TEG output settles through a six-stage
charge pump circuit. The system is able to communicate data wirelessly in
433 MHz ISM band with low (280 µW) instantaneous (burst) power consumption.
3.2.3.3 Piezoelectric (Vibration) Energy Harvesting
Piezoelectric (PZT) harvesters have high output impedance, and need an impedance
matching interface in order to supply maximum power. Recent improvements to
conventional full-bridge rectifier and voltage doubler interface circuits have
increased the piezoelectric harvester efficiency to above 85% [21]. This is achieved
through a number of circuit enhancements for efficiency. External inductor arbiter
component (with value in tens of µHs), shown in Fig. 3.11, cannot be integrated
on-chip in the design. On the other hand special features, such as analog multi-
plexers, are embodied to enable various power converters to share the same
inductor for cost reduction.
A different approach for power conversion, named synchronous electric charge
extraction (SECE) is currently researched by several groups. The Multi-shot SECE
system (MS-SECE), depicted in Fig. 3.12 [22], can self-start and handle a wide
range of piezoelectric power and voltage levels with the help of small off-chip
components. It only consumes 1 µW at 5 Hz vibration frequency, and is suitable
for low frequency motion available in daily human activity. The efficiency is
Fig. 3.11 Bias-flip rectifier based interface system for PZT harvesting with high efficiency [21]
Fig. 3.12 Autonomous piezoelectric energy harvester interface schematic [22]
reported at 61% using a 125 mm3 coupled-inductor. Storage capacitors Casic and
Cwsn in Fig. 3.12 are dedicated to the power supplies of the MS-SECE integrated
circuit (IC) component and WSN device respectively. There is an “optimized path”
controlled through the KP and KS switch transistors to implement the SECE tech-
nique. The PZT harvester is optimally discharged through the flyback circuit to
Cwsn. The SECE technique is based on storing charge to the primary inductor (Lp)
when PZT harvester voltage reaches maximum value, and then transferring charge
to secondary inductor (Ls) after a predetermined period by closing KS and opening
KP. The further optimized MS-SECE technique transfers the energy from the
harvester in N successive transfers instead of a single transfer (hence the name
“multi-shot”). As the maximum value of the voltage vp is detected, multi-shot
80 A. Muhtaroğlu
generator and ring-oscillator turn on to precisely control KP and KS for charge

transfer. When VDDwsn reaches a predetermined threshold, Pgood_wsn signal trig-
gers WSN block to perform a measurement or a RF transmission. The
“non-optimized path”, which is used to charge Casic directly from the PZT har-
vester, is activated during the cold-start, and periodically when Casic voltage drops
below a minimum. FSM_slow block utilizes the Pgood_asic signal from the
Pgood-ASIC block and clock from the relaxation-oscillator to manage transition
between optimized and non-optimized paths.
An application oriented study [23] on potential of typing as energy source
couples electromagnetic and piezoelectric energy harvesting from the computer
keyboard strikes. A prototype demonstrates that 40.8 µW can be harvested from a
single key using a micro-machined piezoelectric component, and 1.15 µW from an
electromagnetic component. Another similar application oriented study [24] focuses
on the potential to power a body sensor node (BSN) through piezoelectric and
thermoelectric harvesters. Real time experiments on human body demonstrate
generation of up to 20 µW at room temperature in a 9 cm2 area of thermoelectric
generator. In addition, a 0.5 cm3 piezoelectric harvester used in this work generates
3.7 µW while running at 7 miles/h.
3.2.3.4 Electromagnetic (Vibration) Energy Harvesting
Another practical kinetic energy harvesting type relates to the electromagnetic

principle of Faraday’s Law. A magnet is typically actuated with low frequency
environmental motion within a coil to generate current. Since the coil has low
output impedance, power extraction does not typically require the type of complex
interfaces described in the previous section.
A novel batteryless interface was presented in a recent study [25] to be utilized
in a system with low voltage electromagnetic harvesters to supply WSN nodes. The
system, shown in Fig. 3.13, has an in-house 2 cm 3.5 cm cylindrical harvester
that feeds a novel rectification circuit with a passive (less efficient) first stage that
powers an active AC-DC doubler with high efficiency. The first stage AC-DC
converter consists of diode-connected MOSFET transistors, and doubles the input
peak voltage.
The active doubler gets powered by the passive circuit to deliver DC voltage to
the final stage with high efficiency. The final stage consists of an integrated step-up
converter built using a charge-pump. The minimum AC input peak voltage that can
be rectified by the system is 125 mV. The output voltage range is 1–2.48 V. The
compact and low cost solution is suitable for a number of low frequency applica-
tions. The combined rectifier stage can deliver 25 µW and 0.61 V with 67% effi-
ciency, while the second stage can deliver higher voltages and few µWs with 13%
efficiency. Charge-pumps in general are commonly used for fully integrated voltage
step-up, but are low in efficiency as demonstrated in this work.
Fig. 3.13 Electromagnetic energy harvester (top), and integrated interface circuit block diagram
(bottom) [25]
3.2.3.5 Ambient RF Harvesting
Wireless digital television signals are constantly broadcasted over the air in many
geographies. A wireless energy harvesting prototype, named E-WEHP [26],
demonstrated harvesting of sufficient power from a TV broadcast source that is
6.3 km away to sustain a 16-bit embedded microcontroller for batteryless M2M
applications. The system, depicted in Fig. 3.14, uses a log-periodic antenna
designed for empirically characterized 512–566 MHz frequency spectrum. A mixed
L-section matching network was designed and built using discrete components and
distributed transmission line elements to match the mainly capacitive impedance of
the RF-to-DC charge-pump circuit to the 50 X impedance of the antenna.
Energy efficiency of the system was characterized in lab by emulating a
collect-first (charge/sleep) and use-later (discharge/active) mechanism through 1- and
18-MX resistors that represented PIC24F 16-bit embedded microcontroller power
82 A. Muhtaroğlu
Fig. 3.14 E-WEHP system block diagram [26]
management modes. In these modes, wireless power was rectified and transferred to
the charge-tank capacitor. With total channel power levels of 8.99 dBm (126 mW)
that represented a location at 6.3 km from the TV source, laboratory tests on the
E-WEHP yielded efficiencies between 55% and 15% for 1- and 18-M output loads
across the charge-tank capacitor respectively.
3.3 Conclusion
Managing power consumption through traditional and new technologies enables

smaller batteries and more compact microelectronic systems. In parallel, energy
harvesting continues to be a pivotal research and development area in efforts to
achieve batteryless operation in increased number of platforms. Micro-scale energy
harvesting will enable many more aspects of intelligent living and working spaces.
The reasons for the current trends have been presented in Sect. 3.1, and few selected
examples have been provided in Sect. 3.2 from recent literature to demonstrate the
state of the art. The upcoming developments are projected in the next paragraphs to
conclude the chapter.
3.3.1 Upcoming Trends
A number of critical issues around sensor platforms with energy harvesting will
continue to receive attention from researchers in order to enable new batteryless
applications. These areas are outlined below in no particular order.
i. Hybrid mode harvesting: Wireless sensor nodes need to support multiple
harvesting modes at the same time in a hybrid configuration as depicted in
Fig. 3.15 in order to extend on-board features and communication range.
Examples of these have recently been noted, combining thermoelectric and
RF, thermoelectric and piezoelectric, piezoelectric and RF harvesting.
However, seamless implementation of a single power supply with multiple
sources requires further study.
ii. True self-starting operation: Efforts will continue to provide self-starting
operation in all systems without batteries. Many of the more efficient systems
today take advantage of small batteries, or depend on a kick-starting RF signal
or motion to self-start. The system goal is to start on detection of any type and
amount of energy in the surroundings without requiring any special
arrangement.
iii. Integration and Efficiency: SoC integration will strive to pull in harvesters,
interface circuits, and sensors into the same package, and eventually into the
same semiconductor component. Elimination of all off-chip discrete compo-
nents (capacitors, inductors) will receive focus. Research on integrated CMOS
Fig. 3.15 Wireless sensor platform enabled by multiple harvesters simultaneously (based on [27])
84 A. Muhtaroğlu
compatible interface circuits will hence maximize power conversion efficiency

without the use of discrete components. This is required not only for minia-
turization of sensor nodes, but also for reduction of cost to levels that will
enable massive use of sensors in disposable or semi-disposable applications.
iv. Improved materials: Material improvements will deliver transducers with
higher intrinsic efficiency. This includes efficiency improvements in solar
cells, thermoelectric modules, piezoelectric cantilevers, magnetic components,
and RF antennas. Storage capacitors and super-capacitors with lower leakage
will also be of major interest to reduce losses. New harvesters and harvesting
modes will be researched in order to extend the harnessed frequency band for
solar irradiation, vibrations, and RF waves without increasing the system size
and cost.
References
1. Brynjolfsson E, Hitt LM (1998) Information technology and organizational design: evidence

from micro data
2. Cisco Visual Networking Index: Forecast and Methodology 2014–2019 (2015) Cisco, White
Paper
3. Chen YK, Wu AY, Bayoumi MA, Koushanfar F (2013) Low-power, intelligent, and secure
solutions for realization of internet of things. IEEE J Emerg Sel Top Circ Syst 3:1–4
4. Culler D, Estrin D, Srivastava M (2004) Overview of sensor networks. Computer 37:41–49
5. Heppner A (2007) Sensornetzwerke - Beispiele aus der Praxis. Sensornetzwerke
6. Verdone R, Dardari D, Mazzini G, Conti A (2008) Wireless sensor and actuator networks.
Academic Press/Elsevier, London
7. Blanckenstein J, Klaue J, Karl H (2015) A survey of low-power transceivers and their
applications. IEEE Circ Syst Mag 15:6–17
8. Muhtaroglu A, Kulah H (2013) Micro-power generation for autonomous systems (Invited
Talk). In: The 2nd Saudi international electronics, communications and photonics conference
(SIECPC’13)
9. Raghunathan V, Schurgers C, Park S, Srivastava MB (2002) Energy-aware wireless
microsensor networks. IEEE Sig Process Mag 19:40–50
10. Martinez MMB, Vilajosana I, Prades JD (2015) The power of models: modeling power
consumption for IoT devices. IEEE Sens J 15:5777–5789
11. Calhoun BH, Lach J, Stankovic J, Wentzloff DD, Whitehouse K, Barth AT et al (2012) Body
sensor networks: a holistic approach from silicon to users. Proc IEEE 100:91–106
12. Majerus SJA, Fletter PC, Damaser MS, Garverick SL (2011) Low-power wireless
micromanometer system for acute and chronic bladder-pressure monitoring. IEEE Trans
Biomed Eng 58:763–767
13. Karmokar A, Anpalagan A (2013) Green computing and communication techniques for future
wireless systems and networks. IEEE Potentials 32:38–42
14. E. P. A. (EPA) (2005) Energy star program requirements for computers, eligibility criteria
preliminary draft
15. Cong P, Ko WH, Young DJ (2010) Integrated electronic system design for an implantable
wireless batteryless blood pressure sensing microsystem. IEEE Commun Mag 48:98–104
16. Zhang Y, Zhang F, Shakhsheer Y, Silver JD, Klinefelter A, Nagaraju M et al (2013) A
batteryless 19 lW MICS/ISM-band energy harvesting body sensor node SoC for ExG
applications. IEEE J Solid-State Circ 48:199–213
17. Anastasi G, Conti M, Di Francesco M, Passarella A (2009) Energy conservation in wireless

sensor networks: a survey. Ad Hoc Netw 7:537–568
18. Muhtaroglu A (2012) Power management and energy scavenging. In: Kaabouch N, Hu W-C
(eds) Energy-aware systems and networking for sustainable initiatives. IGI Global, Hershey
19. Brunelli D, Moser C, Thiele L, Benini L (2009) Design of a solar-harvesting circuit for
batteryless embedded systems. IEEE Trans Circ Syst I 56:2519–2528
20. Ramadass YK, Chandrakasan AP (2011) A battery-less thermoelectric energy harvesting
interface circuit with 35 mV startup voltage. IEEE J Solid State Circ 46:333–341
21. Ramadass YK, Chandrakasan AP (2010) An efficient piezoelectric energy harvesting interface
circuit using a Bias-flip rectifier and shared inductor. IEEE J Solid-State Circ 45:189–204
22. Gasnier P, Willemin J, Boisseau S, Despesse G, Condemine C, Gouvernet G, Chaillout J
(2014) An autonomous piezoelectric energy harvesting IC based on a synchronous multi-shot
technique. IEEE J Solid-State Circ 49:1561–1570
23. Wacharasindhu T, Kwon JW (2008) A micromachined energy harvester from a keyboard
using combined electromagnetic and piezoelectric conversion. J Micromech Microeng
18:055017
24. Wahbah M, Alhawari M, Mohammad B, Saleh H, Ismail M (2014) Characterization of human
body-based thermal and vibration energy harvesting for wearable devices. IEEE J Emerg Sel
Top Circ Syst 4:354–363
25. Ulusan H, Gharehbaghi K, Zorlu O, Muhtaroglu A, Kulah H (2015) A fully integrated and
battery-free interface for low-voltage electromagnetic energy harvesters. IEEE Trans Power
Electron 30:3712–3719
26. Vyas RJ, Cook BB, Kawahara Y, Tentzeris MM (2013) E-WEHP: a batteryless embedded
sensor-platform wirelessly powered from ambient digital-TV signals. IEEE Trans Microw
Theory Tech 61(June):2491–2505
27. Kim S, Vyas R, Bito J, Niotaki K, Collado A, Georgiadis A, Tentzeris MM (2014) Ambient rf
energy-harvesting technologies for self-sustainable standalone wireless sensor platforms.
Proc IEEE 102:1649–1666
Chapter 4
Efficient Energy Harvesting Systems
for Vibration and Wireless Sensor
Applications
Mustafa Doğan, Sıtkı Çağdaş İnam and Ö. Orkun Sürel
Abstract In the first part of the research, we present the design of a vibration-based
energy harvesting system. Robotic flexible arm having variable cross-section is
investigated to overcome serious problems, e.g. insufficient bandwidth and model
inaccuracies. Most of the energy harvesting systems are linear with unimodal
characteristics. On the other hand, real vibrations can be modeled as random,
multi-modal and time varying systems. Hence, unimodal linear systems can give
highly unsatisfactory results under certain circumstances. However, non-linear
systems can have multi-modal character with increased performance in real and
practical situations. In this work, tapered links are preferred with nonlinear coupling
setup to provide sufficient bandwidth and output power requirements for modern
applications. Thus, the proposed scheme has been proven by simulated and
experimental results successfully. In the second part of the research, we present
design and experimental results of an electromagnetic harvester, energy source of
which is single-phase household AC power with a nominal voltage of 220 V and a
frequency of 50 Hz. In this case, energy harvesting is based on the induced elec-
tromotive force (EMF) as a result of the periodic variations of the magnetic field
around the AC power cord. In this part, we also discuss basic principles of a
wireless sensor network design powered by electromagnetically harvested energy
obtained from household alternating current.

Keywords Energy harvesting Multi-modal vibrations Piezoelectric PDE

modelling Tapered beam Flexible arm Wireless sensors Electro-magnetic

radiation Self-powered
M. Doğan (&) S.Ç. İnam Ö. Orkun Sürel

Department of Electrical and Electronics Engineering, Baskent University,
Bağlıca Kampüsü Eskişehir Yolu 20. Km Bağlıca, 06810 Ankara, Turkey
e-mail: mudogan@baskent.edu.tr
S.Ç. İnam
e-mail: inam@baskent.edu.tr
Ö. Orkun Sürel
e-mail: orkunsurel@gmail.com

88 M. Doğan et al.
Nomenclature
a1 Linear slope for tapering
bi(x) Variable height of the links due to tapering
E Young’s modulus
Ii(xi) Variable beam cross-section moment about the z-axis at the location xi
Ihi Inertia of ith hub
Iti Tip inertia of ith beam
li Length of ith link
mhi Mass of ith hub
mti Tip mass of ith beam
t Time (t 0)
wi(xi, t) Flexural deflection of point i at the location xi of ith beam
wix(xi, t) Flexural slope of point i at the location xi where the subscript in wix
denotes spatial derivative w.r.t. x
wixx(xi, t) Bending strain of point i at the location xi
xi Coordinate along the axial centre of the ith beam (0 xi li)
h1 Angular position of the first link
h2 Angular position of the second link
qi(xi) Variable density of the ith link depends on the cross-sectional area
si Input torque at ith motor
Abbreviation and Acronym

FEM Finite element method
MEMS Micro-electro-mechanical systems
ODE Ordinary differential equations
PDE Partial differential equations
FFT Fast Fourier transform
EMF Electromotive force
4.1 Introduction
Portable devices cannot be maintained steadily by limited power sources, e.g.

batteries. This idea improved and motivated the research in energy harvesting area.
Harvesting by vibrating resources, e.g. trains, buildings, bridges, cars, aircraft,
machinery, is well-studied. One of the harvesting methods is based on piezoelectric
materials that can transform electrical energy into mechanical strain or force and
vice versa [1–4]. Recently, remote sensors/actuators are usually hard to reach as
they are hidden in inaccessible locations such as bridges, civil buildings, even
human body. Thus, energy harvesting is the only way to keep them alive [5–7].
4 Efficient Energy Harvesting Systems … 89
As the first case study, e.g. [8], we consider a vibration-based harvesting system
including beams with piezoelectric layers and proof mass. Most of the energy
harvesting systems are linear with unimodal characteristics. On the other hand, real
vibrations can be modeled as random, multi-modal and time varying systems.
Hence, unimodal linear systems can give highly unsatisfactory results under certain
circumstances. However, non-linear systems can have multi-modal character with
increased performance in real and practical situations. In this work, tapered links are
preferred with nonlinear coupling setup to provide sufficiently bandwidth and
output power requirements for modern applications. Thus, the proposed scheme has
been proven by simulated and experimental results successfully.
The flexible beams with piezoelectric layers can capture the environmental
vibrations. The partial differential equations (PDE) are used to model such a
challenging system due to nonlinear coupling for producing accurate and consistent
results, compare with finite-dimensional models, e.g. ordinary differential equations
(ODE) or finite element method (FEM), approaches, refer to [9, 10]. Thus, we will
use the infinite-dimensional model in [9] for the energy harvesting system.
Optimal geometry of flexible links can be investigated to improve some features,
e.g. low mass, low moments of inertia and high natural frequencies, e.g. see [11].
Besides, large bandwidth is required for fast and stable motion, can be achieved
with high fundamental frequency, e.g. refer to [12]. Therefore, non-uniform
cross-section is implemented for the links in this research.
The tapered beam structure and inherent non-linearity of the flexible arm model
are indispensable to improve the bandwidth and output power. Thus, different from
the literature, e.g. cantilever beams with permanent magnets, see [13–18], we do not
use magnets, multiple beam-mass systems and lumped parameter models to
increase the bandwidth and/or to introduce the non-linearity. One of main contri-
butions is that the proposed structure is simple enough to catch multi-modality and
nonlinearity.
As the second case study, we will deal with an energy harvesting case for
wireless network sensors. Wireless sensor nodes typically have power requirements
as low as microwatts, and as the batteries suitable for wireless sensor have limited
lifetimes of the order of a few years, powering such systems by energy scavenging
is an interesting and practical alternative since sensor nodes will no longer require
the occasional battery replacement [19].
We consider a household power consumption detector for real time monitoring.
Real time power consumption monitoring helps customers be more informed about
their energy consumption and have greater control of their monthly costs. Our
detector is based on a sensor with a current sense transformer which is powered by
electromagnetically harvested energy originated from the AC source and is capable
of wireless communication [20]. The detector transmits its readings wirelessly to a
computer or a mobile terminal.
90 M. Doğan et al.
4.2 Vibration-Based Energy Scavenger
The power spectrum of the proposed system is significantly improved by

non-uniform geometry of the links, especially for low frequency band that covers
the most of practical ambient vibrations. The flexible robot arm configuration and
equations of motion for the system are obtained with the extended Hamilton’s
principle, see extensive work in [8, 9].
4.2.1 Energy Harvester Design
Usual vibration sources have random, multi-modal and time varying characteristics.
Hence, unimodal linear systems could be rather inefficient compared to non-linear
systems that are capable of responding over a broad frequency range, thus per-
forming better in realistic vibration spectrum.
The proposed novel energy harvesting system with tapered geometry of the
flexible arm, covered with thin piezoelectric materials. The analytic model which is
developed in [8, 9], are modified with damping terms and external excitation. Two
kinds of damping mechanisms are added to the model in [9]: viscous damping and
Kelvin–Voigt (or strain-rate) damping. This will be helpful for modelling due to its
extreme sensitivity to mechanical damping for the power output under resonance
excitation. Therefore designing the energy harvester beam to have less damping can
be more important than choosing the right piezoelectric material [21]. Modified
equations (PDE) of motion that includes damping terms are given in [8]. The
vibration modes will be distributed over rich spectrum content due to non-linear
coupling as rigorously proven in [8]. On the other hand, tapered beam geometry
provides uniform strain field, e.g., trapezoidal geometry can supply efficiently more
than twice the energy per unit volume than the rectangular one [22]. Besides,
increasing the slope of tapering improves bandwidth [23, 24]. The proposed energy
harvesting system can be easily implemented due to its simple geometry and
structure.
4.2.2 Simulation Results
The proposed harvesting system is simulated using a MATLAB code. The PDEs
are discretized in both spatial and time domains using finite difference method. In
order to avoid complexity of the fourth-order derivative approximation, the
second-order derivative approximation has been managed by using intermediate
states that corresponds to physical variables such as deflections, velocity and
bending moments, e.g. refer to [25]. The same method was also applied success-
fully and proven in [26]. The model parameters for the system, partially given in
Table 4.1 Simulation parameters of the flexible arm

Parameter Value
Length of links l1 = 0.1 m, l2 = 0.024 m
Time step Dt = 10−5 s
Spatial steps Dx1 = l1/100, Dx2 = l2/100
Young’s modulus, E 70 GPa
Density, q 2742 kg/m3
Thickness of links, m c1 = 0.003175, c2 = 0.00238
Maximum height for tapering at the root of the link bo = 0.0654 m
Linear slope for tapering a1 = 0.12, and 0.28
Hub inertias, kg m2 Ih1 = 0.0055, Ih2 = 0.0068
Tip inertias, kg m2 It1 = 0.0139, It2 = 0.00024
Hub mass, kg mh2 = 0.678
Tip mass, kg mt1 = 0.800, mt2 = 0.0204
Structural damping coef., Sf 2 105
Viscous damping coef., Vf 1.5
Attenuation factor, att 0.5
[9], are listed in Table 4.1. A white noise of certain power amplitudes of which
peak near ±2 is applied as external base excitation [8].
The simulations are summarized in Figs. 4.1 and 4.2. The tapered beam structure
and inherent non-linearity of the flexible arm model are indispensable to improve
the bandwidth and output power. Increasing the slope of the beam result in robust
vibrations with increased amplitudes, see Fig. 4.2.
Fig. 4.1 Multi-modal character, slope = 0.12, duration = 0.1 s

92 M. Doğan et al.
Fig. 4.2 Multi-modal character, slope = 0.28, duration = 0.1 s
Although, any kind of nonuniform cross-section can be implemented; in this

particular simulation, rectangular cross-sections of given uniform thickness are
examined. For small values of tip mass and tip inertia moment relative to the ones
for beam, the optimum shape is approximately a linearly ‘tapered’ beam, see for
example [12]. Consequently, the height bi(x), density, qi(x) and cross-section area
moment Ii(x) at any point can be calculated using parameters listed in Table 4.1,
e.g. see [8].
4.2.3 Experimental Results
Having simple structure, stable performance and high reliability; common piezo-
electric diaphragms (buzzers) are utilized as basic electronic sound components.
Besides, these buzzers are also used in many sensitive equipment as shock sensors.
The main benefits of piezoelectric materials are their large power densities and
ease of use. Two-link flexible arm, made of aluminum alloy, are supported by PZT
buzzers. For efficient configuration, these piezoelectric materials are resided on two
opposite sides of the links. The experimental setup used for measurements is shown
in Fig. 4.3. Random vibration is obtained by using adjustable electronic shaker
(MIKROTEK Dental Shaker RC—402) and measured by digital storage oscillo-
scope (RIGOL DS 5202MA, Two Channel, 200 MHz, 1 GS/s). Two-link arm are
clamped to the shaker frame for robust base excitation. The second link is adjusted
to freely rotate with tip mass (0.02 kg). Unamplified oscilloscope signals are
Fig. 4.3 Experiment, 1 oscilloscope, 2 data acquisition system, 3 flexible arm, 4 electronic shaker
transferred to a personal computer (PC). Built-in fast Fourier transform (FFT) utility
of the digital oscilloscope are used to obtain power spectrum. All figures in this
section are oscilloscope images.
Time response of the single link has expected unimodal spectrum, see Fig. 4.4
and this linear system behavior is observed for different levels of external vibra-
tions. However, time responses of the flexible arm in Figs. 4.5, 4.6 and 4.7 are
similar to that of white noise. This broadband multi-peak behavior of the harvested
signal can also be observed at frequency spectrum in Figs. 4.8, 4.9, 4.10, 4.11 and
4.12 as well. The frequency value at upper right of all related figures is just an
estimation of dominant frequency by oscilloscope. Change in this value throughout
the all figures and also rich content of the frequency spectrum in Figs. 4.8, 4.9,
4.10, 4.11 and 4.12 show that wide bandwidth with multiple peaks have been
obtained by non-linear-coupled design. Note that estimated frequency value in
Fig. 4.12 is quite high as expected due to removed proof mass [8].
Fig. 4.4 Temporal response

of the single link with tip
mass
94 M. Doğan et al.

of the two links with tip mass
with rms voltage of 3.54 V

with rms voltage of 2.95 V

with rms voltage 2.29 V
Fig. 4.8 Frequency response

with an estimated dominant
frequency of 246 Hz
Fig. 4.9 Frequency response

with an estimated dominant
frequency of 147 Hz
Fig. 4.10 Frequency

response of the two links with
tip mass with an estimated
dominant frequency of 225
Hz
96 M. Doğan et al.
Fig. 4.11 Frequency

response of the two links
without tip mass with an
estimated dominant frequency
of183 Hz
Fig. 4.12 Frequency

response of the two links
without tip mass with an
estimated dominant frequency
of 490 Hz
4.3 Wireless Sensor Network Powered

by Electromagnetically Harvested Energy
In this section, we present design and the experimental results of an electromagnetic

energy harvester. The source of harvested energy is single-phase household AC
power with a nominal voltage of 220 V and a frequency of 50 Hz. We also discuss
the use of harvested energy to power a wireless sensor network.
4.3.1 Energy Harvester Design
Energy harvesting is based on the induced electromotive force (EMF) due to change
in the magnetic flux as a result of the periodic variations of the magnetic field
around the AC power cord. Using Faraday’s Law of Induction, it is seen that
amplitude of induced EMF is proportional to four quantities: Frequency of the
alternating current (AC), amplitude of the alternating current, number of turns of the
coil and the area enclosed by the coil. We are interested in harvesting power from a
sub-kHz frequency and our typical current amplitude are at most of the order of a
few 10 Amperes. Moreover, we have to restrict the area enclosed by the coil due to
two reasons: Firstly, both magnetic field strength falls dramatically around a wire
with distance. It typically changes with 1/r for very long straight wires, but in real
life problems where the length of the straight part of the wire is of the order the
geometric dimensions of the coil this decrease might be more substantial. Secondly,
we seek for a harvesting solution that is as compact as possible. Thus, we try to
increase the number of turns of the coil as much as possible without significantly
extending the size of the coil. We also choose the type of the wire to ensure that it is
not too thin to limit the total resistance and it is not too thick to limit the overall size
of the coin.
Properties of the coil used as the energy harvester are listed in Table 4.2.
Figure 4.13 shows geometry of the coil.
To harvest energy from the AC wire, we work on three different experimental
setups. For the first and second setups (see top left and top right panels of
Fig. 4.14), we separated the phase wire from the neutral wire so that they are more
than 30 cm away. Then, as for the first setup, we place the phase wire close to the
coil to enhance the magnetic flux and thus the amplitude of induced EMF. Similarly
to test the amount of decrease in the harvested energy with increasing distance from
the phase wire, we increase the distance between the phase wire and the coil.
Thirdly, to increase the magnetic flux even further, we place both phase and neutral
wires in the opposite edges of the coil (see bottom panel of Fig. 4.14). For all of
these setups, the wire (or wires) are carefully held on the same plane with the area
enclosing the coil.
Table 4.2 Properties of the coil used as the energy harvester

External dimensions 315 mm(L) 75 mm(W) 25 mm(D)
Internal dimensions 260 mm(L) 35 mm(W) 25 mm(D)
Number of turns 1,260 turns
Wire type 25AWG (cross-sectional area 0.45 mm2)
Total resistance 94 X
Fig. 4.13 Geometry of the coil used as the energy harvester

98 M. Doğan et al.
37,5mm 87,5mm
37,5mm
Fig. 4.14 Edge-on views of the coil corresponding to three different experimental setups. For all
setups, the wires and the area enclosed by the coil lie on the same plane to maximize magnetic flux
passing through the area (red arrows denote magnetic field line directions). (Top left) phase wire
(separated from neutral wire) is placed close to the coil. (Top right) phase wire (separated from
neutral wire) is placed farther away from the coil. (Bottom) phase and neutral wires are placed
close to the coil to further enhance the magnetic flux
4.3.2 Experimental Results
In Fig. 4.15, a sample oscilloscope view obtained by measuring the voltage across
the ends of the coil when it is placed close the phase wire is shown.
For the three setups shown in Fig. 4.14, we study two cases: Either AC power
cord is connected to a 1000 W heater or a 2000 W heater. We measure current as a
function of voltage (see left panels of Figs. 4.16, 4.17, 4.18, 4.19, 4.20 and 4.21).
Fig. 4.15 A sample oscilloscope view obtained by measuring the voltage across the ends of the
coil when it is placed close to the phase wire
Fig. 4.16 (Left) current versus voltage and (right) output voltage versus load resistance plots for
the first experimental setup and a cord connected to a 1000 W heater
We also alter the load resistance and measure corresponding voltages (see right
panels of Figs. 4.16, 4.17, 4.18, 4.19, 4.20 and 4.21).
We also measure harvested and dissipated power as a function of load resistance
for three experimental setups for the cases of AC power cord connected to a 1000
and a 2000 W heater (see Figs. 4.22, 4.23, 4.24, 4.25, 4.26 and 4.27).
100 M. Doğan et al.
the first experimental setup and a cord connected to a 2000 W heater
the second experimental setup and a cord connected to a 1000 W heater
the second experimental setup and a cord connected to a 2000 W heater
the third experimental setup and a cord connected to a 1000 W heater
the third experimental setup and a cord connected to a 2000 W heater
Fig. 4.22 Harvested and

dissipated power as a function
of load resistance plot for the
first experimental setup and a
cord connected to a 1000 W
heater

first experimental setup and a
heater

second experimental setup
and a cord connected to a
1000 W heater

second experimental setup
and a cord connected to a
2000 W heater

third experimental setup and a
heater

third experimental setup and a
heater
4.3.3 Design of a Wireless Sensor Network Powered

by Electromagnetically Harvested Energy Obtained
from Household Alternating Current
Energy harvesting from household alternating current requires more diligent designs
compared to energy harvesting from an RF field due to the fact that frequencies
involved are orders of smaller compared to RF frequencies, see e.g. [27–29]. As seen
from Figs. 4.16, 4.17, 4.18, 4.19, 4.20, 4.21, 4.22, 4.23, 4.24, 4.25, 4.26 and 4.27,
harvested energy input from an AC power cord can be obtained very efficiently using
a harvester design presented as the third experimental setup as shown in bottom
panel of Fig. 4.14. For such a design, harvested powers can reach up to a few Watts
for loads up to 1 kX in case of household currents of 5–10 A.
Harvested
Input 2
Computer
or Base Station
Harvested Boost Converter Wireless Sensor

Input 1 DC-DC or Transmitter
Storage Circuit Step-up Circuit

and Backup and Regulation
Fig. 4.28 Block diagram of a wireless sensor network powered by electromagnetically harvested
energy. A single node and a computer or base station are shown only
This harvested energy can efficiently be used by a sensor network illustrated in

Fig. 4.28. Such a sensor network might be used to track detailed household energy
usage.
4.4 Conclusion
In this chapter, firstly we investigated the design of energy harvester by a flexible

arm with non-uniform cross-section. It is proven by the simulation results that the
new design method can provide efficient energy harvesting without complex
structures, e.g. magnets and separated links. Coupled non-linear systems can have
multi-modal character with increased performance in real and practical situations.
The shape of the power spectrum is retained effectively even with some geometry
and mass changes. This proves the robustness of the proposed scheme.
The tapered beam structure and inherent non-linearity of the flexible robot arm
are indispensable to improve the frequency response and maximum output power.
Thus, different from the literature, e.g. cantilever beams with permanent magnets,
we do not use magnets, multiple beam-mass systems and lumped parameter models
to increase the bandwidth and/or to introduce the non-linearity. One of main con-
tributions is that the proposed structure is simple enough to catch multi-modality
and nonlinearity.
In this work, variable cross-section links are preferred with rigorous nonlinear
coupling setup to provide sufficiently bandwidth and output power requirements for
modern applications. Thus, the proposed scheme has been proven by simulated and
experimental results successfully.
Secondly, we design and realize an efficient energy harvester which makes use
of both phase and neutral wires of a household alternating current. Having
presented our experimental results, we show that harvested powers can reach up to
a few Watts for loads up to 1 kX for three different configurations.
We also discuss design of a wireless sensor network powered by electromag-
netically harvested energy obtained from household alternating current. To power
such a wireless network, it is discussed that storage/back-up and step-up/regulating
circuits can be developed to feed the wireless sensors or transmitter.
References
1. Shahruz SM (2006) Limits of performance of mechanical band-pass filters used in energy

scavenging. J Sound Vib 293(1–2):449–461
2. Shahruz SM (2006) Design of mechanical band-pass filters with large frequency bands for
energy scavenging. Mechatronics 16(9):523–531
3. Shahruz SM (2008) Design of mechanical band-pass filters for energy scavenging:
multi-degree-of-freedom models. J Vib Control 14(5):753–768
4. Shahruz SM (2008) Performance of mechanical bandpass filters used in energy scavenging in
the presence of fabrication errors and coupling. J Vib Acoust 130(5):054505(9 pp). doi:10.
1115/1.2950056
5. Chennault KAC, Thambi N, Sastry AM (2008) Topical review powering MEMS portable
devices: a review of non-regenerative and regenerative power supply systems with special
emphasis on piezoelectric energy harvesting systems. Smart Mater Struct 17(4):043001(33
pp). doi:10.1088/0964-1726/17/4/043001
6. Beeby SP, Tudor MJ, White NM (2006) Review article energy harvesting vibration sources
for microsystems applications. Meas Sci Technol 17:R175–R195
7. Paradiso JA, Starner T (2005) Energy scavenging for mobile and wireless electronics.
Pervasive Comput 1:18–27
8. Dogan M (2012) Efficient energy scavengers by flexible robot arm with non-uniform
cross-section. IET Control Theor Appl 6(7):935–942. doi:10.1049/iet-cta.2011.0173
9. Doğan M, Morgül Ö (2010) On the control of two-link flexible Robot arm with nonuniform
cross section. J Vib Control 16(5):619–646. doi:10.1177/1077546309340994
10. Luo ZH (1993) Direct strain feedback control of flexible Robot arms: new theoretical and
experimental results. IEEE Trans Autom Control 38(11):1610–1622
11. Moallem M, Patel RV, Khorasani K (2000) Optimization of an actuated flexible arm for
improved control properties. In: Proceedings of IEEE international conference control
application, pp 709–714, Anchorage
12. Wang FY, Russell JL (1992) Optimum shape construction of flexible manipulators with tip
loads. In: Proceedings of the 31st IEEE conference decision and control, Tucson, pp 311–316
13. Shahruz SM (2008) Increasing the efficiency of energy scavengers by magnets. J Comput
Nonlinear Dyn 3(4):041001(12 pp). doi:10.1115/1.2960486
14. Stanton SC, McGehee CC, Mann BP (2010) Nonlinear dynamics for broadband energy
harvesting: investigation of a bistable piezoelectric inertial generator. Phys D Nonlinear
Phenom 239(10):640–653
15. Külah H, Najafi K (2008) Energy scavenging from low-frequency vibrations by using
frequency up-conversion for wireless sensor applications. IEEE Sens J 8(3):261–268. doi:10.
1109/JSEN.2008.917125
16. Sarı I, Balkan T, Külah H (2010) An electromagnetic micro power generator for
low-frequency environmental vibrations based on the frequency upconversion technique.
J Microelectromech Syst 19(1):14–27. doi:10.1109/JMEMS.2009.2037245
17. Ferrari M, Ferrari V, Guizzetti M, Andò B, Baglio S, Trigona C (2009) Improved energy
harvesting from wideband vibrations by nonlinear piezoelectric converters. Proc Chem
1:1203–1206. doi:10.1016/j.proche.2009.07.300
18. Lin J-T, Lee B, Alphenaar B (2010) The magnetic coupling of a piezoelectric cantilever for
enhanced energy harvesting efficiency. Smart Mater Struct 19:045012(7 pp). doi:10.1088/
0964-1726/19/4/045012
19. Yeatman, EM (2004) Advances in power sources for wireless sensor nodes. In: Proceedings
of international workshop wearable and implantable body sensor networks, pp 20–21
20. Lee V (2012) Energy harvesting for wireless sensor networks. M.S. project in Engineering—
Electrical Engineering and Computer Sciences, UC, Berkeley
21. Erturk A (2009) Electromechanical modeling of piezoelectric energy harvesters. Ph D thesis,
Virginia Polytechnic Institute and State University, Blacksburg, November 2009
22. Roundy S, Leland ES, Baker J et al (2005) Improving power output for vibration-based
energy scavengers. Pervas Comput 1:28–36
23. Goldschmidtboeing F, Woias P (2008) Characterization of different beam shapes for
piezoelectric energy harvesting. J Micromech Microeng 18(10):104013. doi:10.1088/0960-
1317/18/10/104013
24. Özdemir Ö, Kaya MO (2006) Flapwise bending vibration analysis of a rotating tapered
cantilever Bernoulli-Euler beam by differential transform method. J Sound Vib 289:413–420
25. Abhyankar NS, Hall EK II, Hanagud SV (1993) Chaotic vibrations of beams: numerical
solution of partial differential equations. ASME J Appl Mech 60:167–174
26. Dogan M, Morgül Ö (2011) Boundary control of a rotating shear beam with observer
feedback. J Vib Control 18(14):2257–2265. doi:10.1177/1077546311429145
27. Chang K-S et al (2012) Electric field energy harvesting powered wireless sensors for smart
grid. J Electr Eng Technol 7(1):75–80
28. Tsunoda Y et al (2015) A small-size energy-harvesting electric power sensor for
implementing existing electrical appliances into HEMS. IEEE Sens J 16(2):457–463
29. Moghe R et al (2014) A low-cost electric field energy harvester for an MV/HV
asset-monitoring smart sensor. IEEE Trans Ind Appl 51(2):1828–1836
Chapter 5
Nonlinear Problems in Piezoelectric
Harvesters Under Magnetic Field
Erol Kurt and Yunus Uzun
Abstract This chapter focuses on the nonlinear problems in the piezoelectric har-
vester systems under the magnetic field. In this manner, the chapter initially men-
tions an introductory section on the studies of piezoelectric harvester dynamics.
After the introductory section, the basic properties of the piezoelectric systems and
their energy harvester applications will be presented. Since the harvesters have a
complicated structure under the magnetic field, the electromagnetic design, mod-
eling and algebraic studies of a novel harvester study will be pointed out. After the
presentation of a theoretical outline on the harvester systems, the experimental
setups will be explained in detail. Thus, a complete picture of the problem will be
produced in order to sustain a comparable study on the theory and experiment. The
main dynamic quantities such as displacement and velocity of the vibrating piezo-
electric layer as function of the system parameters will be explored. According to
results, the effect of periodic magnetic flux can give varieties of responses from
regular dynamics to chaotic one. Phase space constructions, Poincare sections and
FFTs are evaluated depending on the parameter sets including the excitation fre-
quency f, amplitude Uc of electromagnet and the distance d. It is proven that the
periodic magnetic flux can exert high frequency velocity fluctuations nearby the
minimal and maximal values of the velocity, whereas the situation differs for
the position. Therefore it will be pointed out that the magnetic field mostly governs
the velocity by yielding complicated vibrations. According to the detailed analyses,
the FFTs prove the high frequency responses in addition to the main frequency.
When f differs from the natural frequency of the system f0, the responses become
chaotic. It is proven that lower and higher frequency fluctuations in displacement and
velocity, which are different from f0 decrease the electrical power harvested by the
E. Kurt (&)
Technology Faculty, Department of Electrical and Electronics Engineering,
Gazi University, 06500 Teknikokullar, Ankara, Turkey
e-mail: ekurt52tr@yahoo.com
Y. Uzun
Faculty of Engineering, Department of Electrical and Electronics Engineering,
Aksaray University, Aksaray, Turkey
e-mail: yunusuzun@aksaray.edu.tr

108 E. Kurt and Y. Uzun
piezoelectric pendulum. Indeed, it is remarkable to get a relation between the rms

values of displacement/velocity and the harvested power according to the mea-
surements. Thus this relation can be used to estimate the power output in harvester
systems. The piezoelectric harvester generates much energy when f is closed to f0
and the distance to the magnetic device should be closer in order to decrease the
nonlinearities in displacement and velocity. The pendulum-like harvesters as the
most preferable ones can be applied to many devices or units as a power source. The
maximal power for these magnetically-excited structures can be estimated by the
system parameters. At the end of the chapter, the recent techniques of maximal
power point tracking (MPPT) and proposed controller units are explained for the
piezoelectric harvester systems in order to optimize the harvested power.
Keywords Piezoelectric harvester Magnetic field Nonlinearity Attractor

Maximal power point tracking

MEMS Micro-electro-mechanical system
CDRG Coulomb-damped resonant generator
DAQ Data acquisition
FEM Finite element method
FFT Fast Fourier transform
MPPT Maximal power point tracking
RMS Root mean square
SIDO Single inductor dual output
SSPB Single-supply pre-biasing
THD Total harmonic distortion
VDRG Velocity-damped resonant generator
5.1 Introduction
In parallel with the improvements on the fields of smart material and energy
material technologies, there exists a growing interest on the harvester systems,
world-widely [1–3]. The energy harvesting studies mainly focus on three types of
harvester systems: Piezoelectric, electromagnetic, and electrostatic [4–7]. Although
the piezoelectric materials have a complicated material production procedure
(which makes the piezo-structures much expensive than the others in any induction
system), higher energy densities from the piezo-systems can be available [8–10].
Table 5.1 gives brief information on the energy densities of three harvesting types.
Piezoelectric systems are more convenient than electromagnetic ones for MEMS
applications, because of the problems in magnets miniaturization with current
state-of-the-art microfabrication processes [12]. Piezoelectric materials can
5 Nonlinear Problems in Piezoelectric Harvesters … 109
Table 5.1 Maximum energy density of three types of energy harvesting [11]
Type Power Practical maximum Theoretical maximum
equation (mJ/cm3) (mJ/cm3)
Electromagnetic E = B2/2l0 4 400
Electrostatic E = 1/2e E2 4 44
Piezoelectric E = r2y k2/2Y 17.7 335
efficiently convert vibrations to electrical energy with relatively high output voltage
but without any other voltage source [5]. In addition to above-mentioned superi-
orities of piezoelectrics, the advances in solid material behavior shed a light to
combine the dynamical aspects of solids and the piezoelectric features in order to
explain the energy harvesting mechanism behind the solid/piezo structures. At this
point, there is very thin separation between a piezoelectric dynamical system and a
piezoelectric harvester system: Mainly if a piezoelectric system can harvest energy
from any ambient effect such as wind, human movement, seismic forces, water
flow, and traffic, etc. [13–16], it is classified as a piezoelectric harvester, otherwise
any other systems related to the piezo-structures can be acceptable as piezoelectric
dynamic system. From this point, any piezoelectric harvester system is also a
piezoelectric dynamic system, thereby the dynamical tools to explain the physics
and engineering of energy production and fluctuation phenomena can be used to
determine the amplitude/power or velocity/power relations [17, 18].
According to the dynamics of a fluctuating beam, there exists a wide literature,
which has been mostly studied by physicists and mechanical engineers [19–21].
According to their approach, a spring-mass model can be used to understand the
dynamics of a fluctuating beam under an external stress and damping effect.
Equivalently, a basic pendulum model has also been used to identify the dynamical
features of the fluctuating beam in some other researches [22–24]. According to
these two models, the basic harmonic motion is described by the restoring force of
spring or gravitation.
The piezoelectric beams have certain force-voltage relation. Indeed, when the
beam is buckled by an external stress (i.e. mechanical, magnetic, electrical, mag-
netic, etc.), the beam terminals induce a voltage. Note that if the buckled beam then
becomes static in any position, it does not induce any potential difference since there
is no certain electrical dipole moment inside the crystal structure of piezo-material.
Note also that the polarization direction also plays an important role to have an
efficient voltage production. Strictly speaking the force should be in the polarization
direction of the electrical dipoles. In the case of magnetic stress, the magnetic flux
should be oriented in such a way that the permanent magnet(s) should buckle the
piezoelectric beam in the polarization direction. Therefore the poles of the magnet(s)
and the external magnetic flux direction should be well-simulated [25].
In the following section, initially some basic information on the electromagnetic
design, modeling and simulation will be given. Next section will handle the
experimental techniques on the piezoelectric systems. Section 5.4 explains the
methods of nonlinear analyses, which are used in the fluctuation studies of
piezoelectric systems. Section 5.5 introduces some piezoelectric systems and

explains their nonlinear analyzes. While the nonlinear analyzes are discussed in the
next section, Sect. 5.7 explains the dependence of dynamics and power to the
system parameters. Section 5.8 presents MPPT techniques for piezoelectric har-
vester systems. Finally the conclusions give the main results of the chapter and
some clues on the future perspectives.
5.2 Electromagnetic Design, Modeling and Simulation
The electromagnetic (EM) design is very important for a piezoelectric system in

order to estimate the experimental results and the optimization of the system. If the
EM design is made in a proper way, it helps to decrease the time consumption in the
construction of the setup and the experimental researches. According to the liter-
ature there exists some tools to make electromagnetic design and simulation.
However, all of them use the finite element method (FEM) because of its reliability
and solution speed. Before starting any solution, the geometry and material
assignment should be made correctly. Since the magnetic force depends on the
distance between the magnetic sources and ferromagnetic bodies strictly, especially
the magnetic bodies in the model should be positioned in the correct position in
order to estimate the correct magnetic force fm. In Fig. 5.1, a representative mesh
structure is shown. Note that the number of mesh elements can be useful for the
FEM simulation. Extremely large elements yield to long runs of computer and the
results cannot be obtained in the proper time duration, whereas an optimized
number of elements yield to both correct and faster solution.
In order to determine the number of mesh elements, one can assign different
numbers to FEM programme and look for the simulation results. If there is a valley
in the graph of a quantity (i.e. magnetic flux U, magnetic flux density B, field
Fig. 5.1 A representative mesh structure

Fig. 5.2 A sample magnetostatic solution of a magnetic piezoelectric harvester
strength H, etc.) as function of number of mesh, then the left edge of this valley can
be used for a proper mesh number [26].
The modeling of piezoelectric systems can be made as function of time by using
Runge-Kutta time integration technique. It is a proper technique, since the fourth
order term addiction gives good sensitivity for the estimation of quantities such as
position, velocity, voltage, current, etc. [18]. In addition, the findings of
Runge-Kutta technique can be directly used for the dynamic analyzes as in [18, 19].
In the following chapters, the applications of this technique will be given.
After the initial drawings of the model harvester system, magnetostatic analyses
can be carried out. From magnetostatic solutions, the effects of magnetic sources,
magnetic fields can be found in detail in 3D media (see for instance Fig. 5.2).
Note that the fields may increase at the vicinities of magnetic sources such as
magnets up to B = 1.1 T. The advantage of the 3D magnetostatic solution is that it
can give the field values at every solution region. Therefore the model can be
re-drawn easily for the best result. While the distance to a magnetic source is
adjusted, one should keep in mind that the piezoelectric beam should not be put
closer to the source. Since the magnetic force extremely increases, it can break the
beam. Thus the simulations can also give the resulting magnetic force in order to
estimate the optimal distance and save the laboratory materials with that respect.
5.3 Experimental Techniques
The setups in the piezoelectric energy harvester labs can include different equip-
ments from various research fields. Therefore experimental techniques may require
scientists from different fields such as acoustics, physics, electrical and computer
engineering. The equipments, which are frequently used in the harvester studies can
be summarized as follows: Shaker, amplifier, laser displacement sensor and con-
troller, data acquation card, PC, oscilloscope, signal generator, magnets.
Figure 5.3a–c shows the setup of piezoelectric shaker test. The shakers operate
with an amplifier in order to boost the amplitude of the vibration as in Fig. 5.3a.
The amplifier and the signal generator are used to excite the shaker for piezoelectric
tests. In some setups, a special package programme is used to excite the shaker. It
can be operated under a laptop and a specific frequency and amplitude can easily be
Fig. 5.3 a The signal generator (top) and connected amplifier (bottom) for the excitation of
shaker. b Side view of a shaker. c Top view of test equipment including a piezoelectric beam and
the laser displacement sensor
adjusted via the package and it does not require any signal generator. However it
can increase the expenditures for such a setup. In Fig. 5.3b, c, the piezoelectric
mounted shaker and the laser displacement sensor (black coloured) are seen on top
of the system.
Indeed the excitation signal which comes from a signal generator with a certain
frequency (generally 1–50 Hz) and amplitude (lower than 1 V peak to peak) is
increased in amplitude and directed to the shaker to produce certain displacements
at the tip of the piezoelectric beam. The vibration displacement can be adjusted by
the amplifier just by increasing the output amplitude. However, if a certain fre-
quency region will be sweeped, the amplitude should be constant for the experi-
mental sensitivity. Because any small increase or decrease in displacement at the
beam tip yields to different results, since the piezoelectrics are very sensitive to the
buckling. The shaker setup can be used for the testing of frequency responses of the
piezoelectric beams. When it vibrates the beam tip at certain frequencies f, the beam
would produce fluctuations between the negative and positive sides of the equi-
librium point. Figure 5.4 gives such a fluctuation on the laptop screen, if the
fluctuations are read by the laser displacement sensor and send via a DAQ card.
For the interface of the laptop a LabView or other DAQ softwares can be used
efficiently. It is possible with recent equipments that the laser displacement sensor
and DAQ card can measure the vibrations with the sampling rate of 1 ms. This is a
good accuracy in order to measure the higher frequencies of these kind of systems.
Figure 5.5 represents the results of a shaker test for three piezoelectric beams
with various lengths. While the vibration amplitude of the long beam becomes
large, the short one has small amplitude. In addition, the maximal amplitude, which
Fig. 5.4 Fluctuations of the piezoelectric beam at the shaker test

Fig. 5.5 Representative

vibration amplitudes for three
piezoelectric beams with
various lengths: Long (circle),
moderate (square) and short
(triangle) from a shaker test
Fig. 5.6 The experimental diagram for the magnetic test system of a piezoelectric beam (from
[27])
refers to the maximum power, is obtained for different frequencies. The reason is
the difference in the natural frequencies of the beams due to their lengths. Similar to
pendulums, the length will increase the period of the vibration and the f is
decreased.
In order to test the harvesters in a magnetic media, the following setup has been
proposed (Fig. 5.6). This system is different than the previous one, since there
exists magnetic excitation in place of a mechanical one.
In this system, the response of the piezoelectric beam is tested for different
magnetic excitation. The setup can give opportunity to adjust the strenght and the
frequency of the magnetic field B. The electromagnet is responsible to generate the
Table 5.2 The experimental Parameter Symbol Value

parameters
Force/voltage ratio a 0.0001 N/V
Pendulum mass m 27.4 g
Piezoelastic layer mass mp 10 g
Magnetic force coefficient F0 2750
Magnetic force coefficient F1 1500
Elastic stiffness coefficient k 33.435 N/m
Piezoelastic layer capacitance C 232 nF
Damping ratio c 1.48
Inductance of electromagnet coil L 125 mH
Resistance of electromagnet coil Rc 5.9 X
Fig. 5.7 The test setup under periodic field. From left to right, piezoelectric and
non-ferromagnetic beam, laser displacement sensor, electromagnet, DAQ card, signal generator,
electromagnet and LDS feeding sources, laptop and variable resistive loads
periodic B. The distance of electromagnet to the tip of the beam is important,

because the magnetic force strictly depends on the distance. The laser displacement
sensor, DAQ Card, rectifier and storage circuit and signal generator are the main
test equipments. Table 5.2 gives the experimental parameters of the equipments. In
order to create a mathematical expression on the magnetic field effect, the waveform
of the applied signal to the electromagnet is important. Such a problem will be
introduced in Sect. 5.5. The vibration amplitude and the distance are given by x
(t) and d, respectively. Note also that a non-ferromagnetic beam is adhered to the
piezoelectric beam tip in order to construct the desired length. The overall setup is
shown in Fig. 5.7.
More than 500 thousand data points can be captured per second by the card in
this setup. All experimental outputs are gathered by the LabView software. To
determine the responses of harvester at different distances d to electromagnet, an
adjustable slide mechanism is utilized, thereby the electromagnet can be moved
forward and backward with respect to the beam tip. Throught the tests, input
voltages for the electromagnet can be sweeped between 6 and 10 V in order to
change the magnetic force Fm(t). Note that the voltage directly influences the
current on the electromagnet. Since the electromagnet is an inductive load (XL), the
resulting waveform can change (see Sect. 5.5). The magnetic periodic force is
characterized as function of current Ic and the position of the piezoelectric layer
tip u.
According to the test system, piezoelectric beam (sizes of 70 32 1.5 mm
and the weight of 10 g) produced by Piezo System Inc is used. The capacitance and
stiffness values of piezo-material are 232 nF and 188 N/m, respectively. The laser
displacement sensor has a IL-065 type head and a IL-1000 control unit made by
Keyence Inc. A rectifier and a storage circuit, a square pulse signal generating
circuit, and an NI USB-6250 DAQ unit are also included. The magnetic excitation
unit- the electromagnet has 1050 turns with 0.7 cm diameter copper wire.
A ferromagnetic core (sizes of 120 20 20 mm and the relative permeability of
µr = 10,000) is used to increase the field density. The laser displacement sensor
which measures the vibration amplitude of the tip has a sensitivity of 4 µm. It is
possible to get multiple records of different physical parameters such as vibration
amplitude and harvested voltages in the experiments, synchronously.
5.4 Basic Methods of Nonlinear Analyses
The nonlinearity in engineering systems is important in the sense that the

time-dependent systems can be out of stability by slight change in system param-
eters. Thus the estimated behavior cannot be observed and even it may affect the
overall results. Even in many systems such as converters [28, 29], RL-diode circuits
[30, 31], beams [19] the system can come out the controlled state and the behavior
of the physical quantities such as position, velocity, voltage, amplitude, current, etc.
becomes of randomized. Indeed the time-dependent change in these quantities may
lead to unpredictable values or periodic states with higher harmonics. At that point,
one requires a brief explanation on what the nonlinearity is. Figure 5.8 shows some
quantities as function of a variable.
Fig. 5.8 Linearity and

nonlinearity of the quantities
While the line gives a linear behavior with respect to the variable (i.e. x-axis),
other two curves show complicated responses. Strictly speaking, the increase of
x-axis does not yield to simple linear formula (i.e. f(x) = ax), indeed it may have
higher order dependences to variable x such as x2, −x3, ex, Sin(x), etc. A small
increase can yield to different results in f(x) function, thereby all these functions are
called as nonlinear [19, 32]. In many harvester systems, the quantities such as force,
voltage, magnetic field indicate such a behavior.
There exist many tools to examine the results of a dynamic piezoelectric system.
The following parts give explanations on these tools:
(a) Phase Space Construction (Attractor Formation): In any higher dimensional
system (i.e. higher than 1 variable), one can create a graph by putting one of the
variables to the x-axis and the other to the y-axis. This representation simply
gives the relation between these two variables (i.e. quantities) simultaneously.
Thus, if these variables have a closed well-defined curve in the phase space
representation, then one can identify the dynamics as “periodic”. Indeed a
periodic variable can have 2, 3 or more closed curves, then the periodicity is
called such as 2P, 3P, etc. by giving a few different frequency in the variable.
Figure 5.9 shows such an example.
If the phase space has no well-defined trajectory, indeed the output is assigned
as “irregular”. In such an attractor, the phase space fulls of many points and
their position always changes by time by filling the certain region of the phase
space. Beyond the quasi-periodic regime, these kinds of attractors are
encountered frequently and those are the indicator of strong unpredictability.
On the other hand, this irregular change in the variable is called as “chaotic”.
A representative example of such an attractor is shown in Fig. 5.10.
At first glance, the response of a dynamical system can be interpreted by
forming the attractors, easily. It is easy since there exists a fundamental dif-
ference between Figs. 5.9 and 5.10 and it really helps to get an idea on the
dynamic regimes.
Fig. 5.9 Periodic phase

space attractor from the
regular motion of a
piezoelectric beam
Fig. 5.10 Chaotic phase

space attractor from the
irregular motion of a
piezoelectric beam
(b) FFT Analysis: Fast Fourier Transform (FFT) is very-well known process on
the signal or image processing problems. It also finds broad application area in
power electronics, since the harmonics, sub-harmonics and super-harmonics
have importance on the design of a power electronics device. Indeed the
dynamics of an output from any system variable can be analyzed by FFT and
the frequency spectra can be identified in detail. From the analyses on the
piezoelectric beams, periodic, quasi-periodic or chaotic spectrum can be easily
understood. Mainly a chaotic output yields to a wide frequency spectra. It
means that the output has the combination of many individual frequencies.
Some of them can be larger than the main frequency f0, some others can be
smaller than f0. Many peaks in the FFT indicate the chaotic output of the
studied variables. On the other hand, the periodic output gives a basic FFT.
A main peak and a few frequencies (they can be sub- or super- harmonics).
Note that the peaks in a periodic FFT form can easily be identified, indeed it
does not become complicated. Figure 5.11a, b shows two experimental
examples on the FFTs from a piezoelectric beam.
(c) Estimation of Lyapunov Exponents: Lyapunov exponents determine the
distance of the successive trajectories for a certain time evaluation in the phase
space [33]. It mainly considers the exponential difference from an initial state
for the system variables. Therefore the exponential change may increase the
distances of future successive trajectories or it may decrease it. Therefore the
logarithm of the ratio of the distances gives an idea how the variables improve
in the phase space. The expression,
1 X n
dðti Þ
k¼ log2 ð5:1Þ
tn t0 i¼1 dðti1 Þ
gives the fastest growing exponential from the previous state. Indeed, there
exist same numbers of exponential with the system variables. If the harvester
Fig. 5.11 Representative FFT analysis of a regular (a) and a chaotic (b) output
system has for instance, a coupled model of position, velocity and current as
function of time, it means that it has 3 exponentials for each variable. However,
the maximal exponential, which drives the overall system, is accepted as the
maximal Lyapunov exponent and it is responsible for the future of the system
trajectories. Note that if k becomes negative, the trajectories get closer by time,
however if it is positive, the trajectories move away.
5.5 Nonlinear Analysis of the Piezoelectric Harvester

Systems
In this section, the nonlinear modeling procedure of some piezoelectric pendulum

system under the changeable magnetic field is explained. In our previous study (see
[19]), it was found that the vibrations of pendulum structures yield to many com-
plicated responses. Indeed, the elasticity, the magnetic field strength, distance to the
field source play important role to identify the dynamics. Initially, the pendulum
model under the periodic external field is considered. The model of the setup in is
given in Fig. 5.12.
In this model, it is considered that the electromagnet generates a pulsed magnetic
field near the tip of beam. It has been known [19, 32] that the electromagnet can
create a static and inhomogeneous magnetic field near the tip of the beam.
Therefore, dynamics of the beam tip would include the elastic, gravitational and
magnetic forces. In the considered system, the elastic force is much higher than the
gravitational force since the mass of the system is small. With that respect, the
effects of elastic and magnetic forces can be considered dominantly [19, 32, 34]. All
forces which affect the tip can be summarized as in Fig. 5.13.
Following our studies [19, 27, 32] the forces in Fig. 5.13 gives the system
equations. When the displacement x(t) occurs at the tip of the pendulum, it produces
Fig. 5.12 The model of the

piezoelectric pendulum in
Fig. 5.7 (from [27])
Fig. 5.13 The forces acting on the beam near the electromagnet
and internal mass displacement u(t) inside the piezoelectric material. Here,
Figs. 5.12 and 5.13 describe the system as a mass-spring, a damper and a capacitor.
In this system, the rigid mass mp and the stiffness constant k defines the mechanical
property under a damper c which denotes the mechanical losses. When the mass
displacement u occurs in the layer, current I and voltage V are generated. In this
case, the system equations is given by,
Ftotal ðtÞ ¼ kuðtÞ þ aVðtÞ þ Fm ðtÞ

duðtÞ dVðtÞ ð5:2Þ
IðtÞ ¼ a C
dt dt
Note that the equations include the mechanical and electrical parts. Here, the
backward piezoelectric coupling is also included in addition to the mechanical
effects. While C is the clamped capacitance, a denotes the force factor. If the
displacement of mass inside the piezoelectrics u′(t) and the pendulum tip dis-
placement x′(t) are used before the dimensionless form,
d 2 x0 ðtÞ du0 ðtÞ d 2 u0 ðtÞ

ðm þ mp Þ ¼ c m p ku0 ðtÞ aVðtÞ Fm ðtÞ ð5:3Þ
dt02 dt0 dt02
Here, the pendulum and piezoelectric beam masses are denoted by m and mp,
respectively. Considering the periodic magnetic force Fm(t), the equations can be
written explicitly as follows:
d 2 x0 ðtÞ c du0 ðtÞ mp d 2 u0 ðtÞ ku0 ðtÞ aVðtÞ Fm ðtÞ

02
¼ 0
02

dt ðm þ mp Þ dt ðm þ mp Þ dt ðm þ mp Þ ðm þ mp Þ ðm þ mp Þ
dV 0 ðtÞ a du0 ðtÞ IðtÞ
¼
dt0 C dt0 C
ð5:4Þ
Since x′ is measured as 3 times of the displacement u′, experimentally (see in

Fig. 5.8), u′ = 0.33x′ can be written. Then one arrives at,
d 2 u0 ðtÞ c du0 ðtÞ k u0 ðtÞ a VðtÞ Fm ðtÞ

02
¼ 0

dt 3m þ 4mp dt 3m þ 4mp 3m þ 4mp 3m þ 4mp
ð5:5Þ
dV 0 ðtÞ a du0 ðtÞ IðtÞ
¼
dt0 C dt0 C
The dimensionless form can be obtained when t′ = st, y′ = yd/s, u′ = ud and

V′ = V0V are introduced for time, velocity, position and voltage scaling, respec-
tively. Note that s determines the natural period of the pendulum and cannot be
confused by the excitation period of the magnetic field T. Besides, to avoid the
misunderstandings, d in later two expressions refers to the distance between the
equilibrium point of pendulum and the tip of electromagnet. Then the dimension-
less form can be written as;
du
¼y
dt
dy cs ks2 ðm þ mp Þs2 Fm ðtÞ as2 V0
¼ y u V ð5:6Þ
dt 3m þ 4mp 3m þ 4mp ð3m þ 4mp Þd ð3m þ 4mp Þd
dV ad du sV
¼
dt CV0 dt V0 CRL
In above expression, d transfers the coordinate system into the equilibrium point
of the beam. The vibration amplitude of the tip is represented by u.
The main problem is to find out the expression of magnetic force. The best way
to find out the magnetic force expression, a magnetostatic analysis as similar to the
case in Ref. [27] should be realized. The field lines are shown in Fig. 5.14 at the
vicinity of the tip. Here, the Poisson SuperFish simulation package is used with
FEM. The magnetic force Fm increases near the electromagnet and affects the
ferromagnetic tip, dominantly. Flux density B decreases with the distance, thereby
the 2D simulation can find the effect of the distance.
Figure 5.15 presents the results of the magnetostatic simulations. It has been
found that these variations can be fitted to a combined function depending on the
variables of u, Uc (i.e. Ic) and t. By considering Fig. 5.16, the overall nonlinear
expression can be considered as the functions of electromagnet voltage Uc and the
distance u,
Fm ðtÞ ¼ fm ðuÞ fm ðUc ; tÞ ð5:7Þ
Fig. 5.14 The field lines at

the vicinity of beam tip
Fig. 5.15 Results of magnetostatic simulations: a Force, b magnetic flux density B, c force as
function of the distance between the pendulum tip and electromagnet and d force on the plane of
current and distance
Fig. 5.16 Applied voltage to

electromagnet terminals and
the electromagnet current
waveform which exerts the
magnetic force
Note that the third-order polynomial function can fit better as in Ref. [18]. Since
fm(Uc,t) is the function of time due to the periodic excitation as seen in Fig. 5.16,
the exponential functions appear as follows:
8n o 9
>
> ð1 0:7056=dÞ þ 0:0623ð1 3uÞ þ 28026dð1 3uÞ2 106 d 2 ð1 3uÞ3 >
>
>
> >
>
>
> n o >
>
>
> >
>
< a ðURcc ð1 eRc t=L ÞÞ2 b URcc ð1 eRc t=L Þ 0\t T
2
=
Fm ðxÞ ¼ n o
>
> 2 3 >
>
> ð1 0:7056=dÞ þ 0:0623ð1 3uÞ þ 28026dð1 3uÞ 106 d 2 ð1 3uÞ
> >
>
>
> n o >
>
>
> >
>
: a ðURcc eRc t=L Þ2 b URcc eRc t=L T
\t T ;
2
ð5:8Þ
Here, d, Uc, Rc, L indicate the distance from the electromagnet to the pendulum
equilibrium, voltage, resistance and inductance of the electromagnet coil, respec-
tively. In the expression, the electromagnet force constants are defined as
a = 8 10−8 and b = 10−9 from the electromagnetic simulation.
In order to simulate Eq. 5.6, the fourth order Runge-Kutta method can be used
efficiently in the MatLab media. In that case, the time dependent results for position,
velocity and voltage can be obtained. The main task is to fit the parameters of model
and experiment. For this reason, the responses of simulation should be compared by
the experimental findings. Therefore the damped results (without the magnetic
excitation) can be compared at first step to estimate the best parameter fit.
From these graphs (Fig. 5.17), the natural period and frequency are found as
s = 0.210 s and f0 = 4.76 Hz. In addition, the damping constant is c = 1.48 and the
ratio of elastic constant to mass is k/m = 894. With the help of these parameters, the
simulations and experiments fit well. Moreover, this model can be used for finite
field values with the excitation frequencies.
Fig. 5.17 a Experimental and b Theoretical vibration amplitudes u for non-magnetic case
In order to perform the averaged power output <P>, Eq. 5.6 is considered in
frequency domain. While the second equation in Eq. 5.6 is written as,
jRL axm u
V¼ ð5:9Þ
1 þ jCRL xm

uRL ajxm ðm þ mp Þ x2m F0 \Ic2 [ þ F1 \Ic [
V¼ þ h:o:t: ð5:10Þ
ðcjxm mp x2m kÞ ð1 þ RL Cjxm Þ a2 RL jxm
Here the frequency xm is determined by the field. Note that the linear terms of
the piezoelectric amplitude u is used in order to have an expression on power.
Here <Ic> and <I2c > indicates the time-averaged values as below:
Vc Vc Lxm LxRc p

2e m 1 eLxm
Rc 2p
\Ic [ ¼ þ 2
2Rc 2pRc
ð5:11Þ
V 2
V 2 Lxm Lx Rc p
\Ic2 [ ¼ c2 þ c 3 1 eLxm þ eLxm 1
Rc 4p m
2Rc 4pRc pRc
Here, the electrical parameters belong to the electromagnet coil. A certain

voltage Vc with an excitation frequency of xm = 2pf is applied to the coil. Then, the
power relation is found as follows:
2
u2 RL a2 x2m ðm þ mp Þx2m F0 \Ic2 [ þ F1 \Ic [
\P [ ¼ n o
k 2 ð1 þ C 2 R2L x2m Þ þ 2kx2m ðmp þ a2 CR2L þ C mp RL xm Þ þ xm 2a cRL þ a RL þ 2a Cmp RL xm þ c2 ð1 þ C 2 R2L x2m Þ þ m2p x2m ð1 þ C 2 R2L x2m Þ
2 2 2 2 2 4 2 2 2 2
ð5:12Þ
The linear part of Eq. 5.8 as function of u and its complex conjugate divided by
load resistance RL, one arrives at \P [ ¼ \VV [ =RL . Equation 5.12 gives the
relation between the output power and the electrical/mechanical parameters of the
system. By using this formula, it is also possible to find out the optimal load
resistance. If one gets the derivative of this equation with respect to RL, the optimal
Fig. 5.18 The load resistance (a) and vibration amplitude (b) dependences of power. d = 2 cm
and the voltage applied to the electromagnet terminals is 8 V [27]
load resistance is found as (xC)−1. Here, x and C give the frequency of magnetic
field and the capacitance of the piezoelectric material, respectively.
Figure 5.18 gives the experimental and theoretical results of the averaged power.
The resistive load about 1 MX gives the optimum power of the piezoelectric beam.
As in other electrical systems, the piezoelectric beam gives the maximal power
output when the load resistance equals to its internal empedance.
It is obvious that the theoretical and experimental explorations give the same
value. In addition, power depends on the vibration amplitude with a higher order
polynomial function (see in Fig. 5.18b). With higher deflactions, much power is
obtained.
5.6 Experimental and Theoretical Results of Nonlinearity
In this part, the test results with respect to vibration amplitude, tip velocity and
output voltage are discussed. Figure 5.19a–c presents the dynamic responses of the
beam tip. It is understood that velocity has some ripples at the maximal values when
the tip is at the equilibrium point. The output voltage is not ideal sinusoidal,
however it has the same phase with the velocity. In Fig. 5.19d the phase plane
trajectories give ripples at minimal and maximal points of velocity as stated before.
This phase space representation shows these ripples much clear. Since an ideal
periodic motion yields to a simple circle or ellipse in the phase space, the ripples
here point out a non-regular dynamics even at the natural frequency of the system.
It has been proven that these ripples are resulted by the magnetic field [27]. The
ripples cannot be related to any noisy data, because the external magnetic excitation
is very strong for the adjusted parameter set. Even in the theoretical formulation, it
yields to similar ripples. The spectral analysis shows that there exist some high
frequencies but low amplitude effects in the velocity data. It can have 88 Hz
Fig. 5.19 Observation of a harvester tip displacement, b tip velocity, c voltage, d phase space
portrait and e power spectrum of the velocity in the case d = 2 cm, xm = 4.76 Hz, RL = 500 kX
component and that explains the ripples occurring in a very short time scale. Strictly
speaking, the main frequency 4.7 Hz and other high frequency components such as
14.6, 26.3, 34.9 and 56.3 Hz are clearly seen in the spectrum (Fig. 5.19e).
When the field frequency xm is increased to 6.09 Hz, the dynamics becomes
much complicated (Fig. 5.20a). Both the velocity and voltage shows ripples as in
Fig. 5.20b, c, however the vibration amplitude has still a sinusoidal character. In
addition to the main frequency 6.09 Hz, there exist infinite numbers of frequencies
in the spectrum as shown the inset of Fig. 5.20e. Some frequencies with larger
amplitudes can be summarized as follows: 24, 36, 54, 156, 256, 425, 452 Hz etc..
These frequencies produce many ripples as in Fig. 5.20d and dominate the phase
space. Thus the nonlinearity governs the dynamics for non-natural frequencies. On
the other hand, these high frequencies such as 425 and 452 Hz in the velocity data
are observed first time, experimentally in such a large-scale piezoelectric system to
our knowledge. This result proves how the nonlinearity is important for such
harvester systems and encourages us to obtain better power solutions for magnet-
ically excited systems. Although the vibration amplitude becomes sinusoidal, the
wide-band structure in velocity causes smooth power decay for different excitation
frequencies as will be shown later.
Fig. 5.20 The experimental results of a harvester tip displacement, b tip velocity, c voltage,
d phase space portrait and e power spectrum of the velocity in the case d = 2 cm, xm = 6.09 Hz,
RL = 500 kX
Fig. 5.21 Experimental (a) and theoretical (b) attractors for xm = 4.76 Hz and d = 2 cm. The
points represent Poincare sections
Figure 5.21 represents the experimental and theoretical attractors for another
experimental set. Note that the only change in this parameter set is the current over
the electromagnet (i.e. lower magnetic force). The ripples still exist, however their
Fig. 5.22 Experimental (a) and theoretical (b) attractors for xm = 6.02 Hz. The points represent
Poincare sections [18]
lengths are low compared to the case in Fig. 5.19. The open circles on the
experimental data gives the Poincare section data, which is separated on the phase
space after the same time interval. The theoretical attractor, which was calculated
from Eq. 5.6 is shown in Fig. 5.21b and gives a good similarity to the experimental
one.
The vibration amplitude of the beam tip becomes larger, when the field frequency
approaches to the natural frequency of the system (i.e. xm = x0 = 4.76 Hz). Thus
the frequencies near the natural frequency of the system are better to harvest much
energy from the system, since the voltage generated by the piezoelectric layer is
found to be dependent on the vibration amplitude. In Fig. 5.22, the attractors of
another parameter (i.e. xm = 6.02 Hz) are compared. The changeable character of
the magnetic flux can be seen in velocity fluctuations dominantly, whereas the
vibration amplitude gets smaller values compared to the earlier cases. The similarity
persists again between the theoretical and experimental findings. Note also the
distributed Poincare section points on the attractor in Fig. 5.22a.
In addition to the similarities between the model system and the experimental
one, the harvester systems show another interesting phenomena called hardening
and softening effects. In Fig. 5.23, an example of this strange hysteresis bahaviour
is presented. Here, while the blue plot gives the maximal voltage when the exci-
tation field frequency is increased up step by step, the red plot gives the maximal
voltage when the frequency is decreased gradually.
The frequency increase causes a harvesting effect with small amplitudes around
5 V (as in sweep up case) as in Fig. 5.23. But the frequency lowering beyond the
natural frequency causes three-fold voltage harvest in the softening effect. This
behavior is the result of the piezoelectric layer type. While some layers indicate
softening effect, others give hardening effect, where increasing frequency yields to
high amplitudes. As a result of these hysteresis effects, the excitation frequency
Fig. 5.23 Peak voltage as

function of the excitation
frequency xm. Softening
effect is observed for the
sweep up/down cases. The
parameters are d = 2 cm, the
load resistance RL = 820 kX
and the maximum voltage
over electromagnet Uc = 8 V
should always change at the vicinity of natural frequency, where jumps in ampli-
tudes can be observed.
5.7 The Dependence of Dynamics and Power to the System

Parameters
In this section, the parametrical dependence of power in pendulum-like harvesters

will be explained. For this aim, the dynamic feature of such a system can be
summarized by an intuitive bifurcation diagram (Fig. 5.24). Indeed, such diagram
can be produced by adjusting a parameter or two parameters on a plane. If one
parameter dependency is studied, it is called one-dimensional bifurcation diagram,
whereas two parameters are studied together on a parameter plane, it would be a
two-dimensional bifurcation diagram. The bifurcation data can be produced by
displacement, velocity or harvested voltage for successive parameter changes.
Therefore it is nothing else than the collection of many Poincare sections.
According to Fig. 5.24, the frequency values less than 4.8 Hz give regular dis-
placement values and a periodic behavior is observed. However beyond this value,
displacement values fluctuate between 0.4 and 1.4 cm till the excitation frequency
becomes 5.1 Hz. The displacement vibrations become regular (i.e. periodic) for
larger values. In the case of velocity (i.e. filled circles), the motion is much com-
plicated since the fluctuations does not decay to a certain velocity value for any
excitation frequency.
Fig. 5.24 The experimental

bifurcation diagram
(d = 2 cm, Uc = 8 V). The
bifurcation parameter is
magnetic field excitation
frequency
The parametrical dependencies can also be found by root mean square

(RMS) data, which is given by [35]:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2
xrms ¼ x1 þ x22 þ þ x2n ð5:13Þ
n
Displacement and velocity data are represented by xn, and n gives the total data
point. It is interesting that there exist a good relation between RMS values of data
and the corresponding output power as in Fig. 5.25. Indeed, this relation can be
found for both experimental and theoretical data. All data can be fitted to a second
order formula, which states that power increases parabolically as function of dis-
placement and velocity rms data.
This invention of this relation is new for magnetically-excited systems and it
gives a very clear result, which enables one to estimate output power just by
looking at rms data of displacement and velocity [18]. In Fig. 5.26, another
interesting result is given. The RMS values of displacement and velocity enables us
to determine the natural frequency of the harvester system. Both RMS displacement
and velocity data jumps at natural frequency. In this experiment, the natural fre-
quency is 4.76 Hz.
Fig. 5.25 Experimental (a, b) and theoretical (c, d) relation between displacement/velocity and
output power. (d = 2 cm, Uc = 8 V and the frequency changes between 4.25 and 6.53 Hz.)
One of the important concepts to measure the nonlinearity of time-dependent

data is the fluctuation amount. The nonlinearity of any fluctuation increases, when
different frequencies are included in the data. When one gets the amplitudes of
individual frequencies from the FFT result, a quantitative scaling can be obtained
for the data. This process is mainly described with the concept of total harmonic
distortion (THD) [17, 36]:
sffiffiffiffiffiffiffiffiffiffiffi
PNf
A2i
i¼1
THD ¼ ð5:14Þ
A0
where A0 denotes the main frequency of the displacement or velocity data. Ai and Nf
denote other frequencies and the number of total frequencies in the FFT,
respectively.
In Fig. 5.27a, b, the THD results of the data are presented as function of exci-
tation frequency. It is obvious that displacement fluctuations are smoother than
velocity fluctuations in Fig. 5.27b. The nonlinearity becomes 5 times powerful in
velocity fluctuations for some f values compared to the displacement fluctuations.
Fig. 5.26 The dependence of

rms position and velocity to
the excitation frequency
While the maximal distortions are obtained at lower f for both displacement and
velocity, a contradiction occurs at the vicinity of natural frequency.
Because the distortion in displacement increases up to 6 at the vicinity of f0, the
distortion in velocity stays lower near 15 at the same frequency. Note also that there
is a slightly decreasing character in THD of displacement (Fig. 5.27a). However, in
the case of velocity, THD increases for higher frequencies as seen in Fig. 5.27b.
Thus, these results prove that while the nonlinearity of displacement is generally
higher for the frequencies lower than natural frequency of the system, the nonlin-
earity of the velocity stays higher for lower and higher frequencies and becomes
lower at relatively medium frequencies. In order to find a relation between
displacement/velocity THD values and output power, Fig. 5.28 is depicted. While
the output power maximizes at f = 4.8 Hz, THD of velocity minimizes. Thus, it
proves that the power decreases when irregularity increase in the velocity data.
Whereas there is no significant correlation or anti-correlation between displacement
and power.
From the electrical point of view, the resistive load RL and excitation frequency
f are important. In order to show this combined effect, Fig. 5.29 is presented.
There exists a certain maximum for specific resistance and magnetic excitation
frequency. The output power of harvester becomes maximal, when f gets closer to
the natural frequency. While the dependence on the frequency is much strict in
Fig. 5.27 The total harmonic

distortion percentages (THD
%) of a displacements and
b velocities for excitation
frequency
Fig. 5.28 The corresponding

output power as function of
excitation frequency
order to get the maximal output, the resistance dependence affects the output power
relatively low. Note that the resistance axis is in the logarithmic scale. Another
important result is that the harvester power shows the wider frequency region at its
same value for the resistive loads of 1 MX. Thus if one requires much stable output
for a wider range of excitation frequencies from the harvester, the load should be
adjusted at that value otherwise the maximal output can be generated at certain
parameters.
5.8 Maximal Power Point Tracking Techniques

for Piezoelectric Harvesters
Parallel to the findings of previous section, a harvester gives the maximal power
generation for a specific vibration frequency and load (see in Fig. 5.29). Therefore
an efficient maximal power point tracking (MPPT) technique should be applied in
order to get the optimized energy. In the literature there exist various techniques for
the MPPT techniques [37–39]. Initially, it should be pointed out that surface
mounted devices (SMD) can be preferable to decrease the power loss in the circuit
Fig. 5.29 Harvested power on the plane of resistive load and excitation frequency xm
Fig. 5.30 The circuit diagram which enables to use of the harvested power: Rectification, storage
and MPPT
elements since they generate lW or mW scale powers. Figure 5.30 shows a general
circuit model with rectification, storage, MPPT and electrical load.
The harvested signal is initially rectified and transferred to an MPPT circuit just
before the storage unit. Two recent MPPT techniques are mentioned below:
1. DC/DC single inductor dual output (SIDO) control technique,
2. The single-supply pre-biasing (SSPB) technique.
The SIDO control equipments and MPPT circuit diagram which are introduced
by Ramond et al. [37] is shown in Fig. 5.31. After the harvester part in the left
hand-side an AC/DC recitifier is first used as also refered in Fig. 5.30. Later, the
MPPT and controller part is added to the system for an optimized power control.
In the MPPT part of Fig. 5.31, there is a certain voltage value called VMPP for
which the power PMPP is delivered maximal from the harvester. After this MPP,
the power decreases by the voltage increase until it reaches a value VOC, the Open
Circuit voltage of the harvester for which the harvested power is null. The main
Fig. 5.31 The detailed connection of MPPT and DC/DC SIDO control circuit (from [37])
Table 5.3 Harvester characterization summary with SIDO control technique (from [37])
Vibration level (g) |PMPP| [W] |VMPP| [V] |VOC| [V] |VOC|/|VMPP|
0.1 3.5 10−6 1.26 2.10 0.6
0.3 35.5 10−6 4.02 6.77 0.59
0.5 103 10−6 6.83 11.48 0.59
function of this MPPT technique is to sample the open circuit voltage of the
harvester, periodically in order to provide a reference voltage to the DC/DC con-
verter. Indeed as shown in Table 5.3, the ratio VMPP/VOC is constant over the
whole vibration level range (0.1–0.5 g). Thus the DC/DC converter regulates the
harvester polarization voltage around VMPP leading to a maximal power emission
with an average efficiency of 95%.
Consequently in order to achieve an optimal power, the micropower manage-
ment system should regulate the voltage of the harvester and place it close to the
MPP. According to literature, this kind of MPPT problem has been solved using the
charge pump or inductive DC/DC converters in order to isolate the piezoelectric
layer from heavy electrical load and enable to control the polarization of the source
independently of the load voltage (Fig. 5.32a). The solutions proposed by Yi et al.
[40] and Simjee and Chou [41] are interesting but all the power consumed by the
Fig. 5.32 Various MPPT management systems in the literature (from [37])
Fig. 5.33 The harvested

power with/without MPPT
from a piezoelectric layer
(from [42])
load passes through the storage element and two converters (see in Fig. 5.32b),
which decrease the overall efficiency of the system.
Figure 5.32c proposes another MPPT solution by providing a high efficiency
direct power path from the source to the load in addition to the MPPT scheme by
using a Single Inductor Dual Output (SIDO) DC/DC converter. The MPPT system
operates with the control circuit as in Fig. 5.31 as well. According to the findings of
Do et al. [42], the effect of an MPPT system can be seen in Fig. 5.33.
The bell-shaped dotted curve shows the harvested power after the rectifier for
various ohmic loads. For the high loads then 200 kX, there exists a drametical
decrease in power as also shown in Fig. 5.29 in the previous section. However, the
MPPT system produces a much rough output for high loads and the output power is
nearly fixed around 70 lW.
Other recently proposed MPPT technique is called as single-supply pre-biasing
(SSPB). According to Elliott and Mitcheson [39], their technique enables the
damping force to be set by a fixed voltage and by varying that voltage, real-time
adaptation to variations in the mechanical force can be implemented.
Every half cycle of vibration, charge is placed on the piezoelectric material
which will induce a force opposing the motion of the piezoelectric material. This
tries to dampen the beam’s motion, causing an increase in the harvested power.
SSPB changes the circuit from a velocity-damped resonant generator (VDRG) to a
Coulomb-damped resonant generator (CDRG) [43], enabling the level of damping
applied to be readily set by the pre-bias voltage, Vcc. At the piezoelectric beam’s
extreme points, charge from the pre-biasing capacitor, Cbias, is transferred on to the
beam. This generates a Coulomb force to oppose the beam’s vibration until it
reaches the opposite extreme position. The energy on the beam is then discharged
back into Cbias, and the process repeats. Similarly to the bridge rectifier, the
damping force can be optimally set by setting the voltage on the storage device in
order to achieve maximum power extraction,
8Q 2 c
Pmax ¼ V f0 Cp and it is applied when Vopt ¼ 2Vpo ð5:15Þ
p po 1 c2
Here Q is the Q-factor of the resonant current discharge path through the
inductor and is the fraction of the magnitude of the voltage conserved on the
capacitor of an RLC oscillator with Q-factor after a half-cycle of the oscillator.
The system must detect a change in the piezoelectric induced voltage, quantify the
magnitude of the change, and operate an appropriate response. Since both the
bridge rectifier and SSPB techniques use a voltage on a bias capacitor to apply the
optimal conduction angle and damping force respectively, adding a buck converter
with battery enables the bias capacitor voltage to be independently set. The MPPT
scheme operates by adjusting the off-time of the buck converter causing the energy
on the bias capacitor to be more or less frequently transferred, resulting in a
decrease or increase in voltage respectively. Figure 5.34 shows the suggested
topology to adjust the voltage applied for the SSPB circuit. The voltage on the
biasing capacitor, Cbias, is controlled by varying the power transferred through the
buck converter. If the time between energy transfers is increased, Cbias voltage
would rise and a greater damping force is applied to the piezoelectric material. For
the implementation of the SSPB technique, the system requires three elements [39]:
A circuit to measure the harvested power, a controller to decide whether the power
has increased or decreased object to the previous measurement and a buck converter
to transfer the energy. In the design of these circuits, power consumption should be
minimised for the highest efficiency. Figure 5.35 gives the implementation to
demonstrate the MPPT technique.
The measurement and control circuitry should operate at the lowest voltage
possible, but the voltage on the biasing capacitor maybe several times larger than
this. A potential divider can also be used by switching in across the biasing
capacitor during the operation (see in Fig. 5.35). The switch can be an n-type
MOSFET (BSS138), which can be driven by a low power FPGA (Igloo Nano). It
can also be used to implement the SSPB scheme [39]. A low power
Analog-to-Digital Converter (ADC AD7468) can be used to measure the voltage
and a Booth Multiplier algorithm [44] on the FPGA can square the value.
Fig. 5.34 The SSPB unit attached to a buck converter and battery (from [39])
Fig. 5.35 SSPB MPPT implementation and controller (from [39])
5.9 Conclusions
In this chapter, the outlines of the dynamic explorations of the harvester systems
have been explained. Mainly, the 3D system modeling, electromagnetic analyses
and time-dependent simulations have been defined for the widely worked pendulum
system. While the 3D modeling of the system gives a clue on the electromagnetic
effects, especially the magnetic force steming from the permanent magnets or a
stable curring carrying winding play an important role to determine the dynamics. It
has been also proven that the distance between the magnetic or electrical souce
contributes to the nonlinear force term as well as the current flowing the electro-
magnet. The other parameters such as resistive load, electromagnet excitation fre-
quency and strength (i.e. current) dominate the dynamics and affect the harvested
power in the system. In parallel with the other systems in literature, the natural
frequency of the harvester causes maximum power generation. In fact, the non-
linearities in the velocity data for such periodic magnetic excited system subject to
decrease the power generation obviously, whereas there exists no remarkable effect
of the vibration amplitude on the output power. Considering the methodology of
dynamics, many applications of the pendulum-like harvester systems should be
realized and both theoretical and experimental explorations should be carried out
together in order to draw a complete picture of the applications. The chapter ends
with the recent proposed MPPT techniques in order to optimize the harvested power
from the piezoelectric systems.
Acknowledgements The supports from Gazi University Research Development Unit under Grant
Nos. BAP 07/2010-01, BAP 07/2012-12, Turkey Ministry for EU Affairs—National Agency
under Grant No. 2015-1-TR01-KA203-021342 and The Scientific and Technological Research
Council of Turkey (TUBITAK) under grant EEEAG-114E017 are acknowledged.
References
1. Tufekcioglu E, Dogan A (2014) A flextensional piezo-composite structure for energy

harvesting applications. Sens Actuators A Phys 216:355–363
2. Saha CR, O’Donnell T, Wang N, McCloskey P (2008) Electromagnetic generator for
harvesting energy from human motion. Sens Actuators A Phys 147:248–253
3. Ferrari M, Ferrari V, Guizzetti M, Andò B, Baglio S, Trigona C (2010) Improved energy

harvesting from wideband vibrations by nonlinear piezoelectric converters. Sens Actuators A
Phys 162:425–431
4. Erturk A, Inman DJ (2008) Issues in mathematical modeling of piezoelectric energy
harvesters. Smart Mater Struct 17:065016 (14 pp)
5. Lin JH, Wu XM, Ren TL, Liu LT (2007) Modeling and simulation of piezoelectric MEMS
energy harvesting device. Integr Ferroelectr 95:128–141
6. Wang L, Yuan FG (2008) Vibration energy harvesting by magnetostrictive material. Smart
Mater Struct 17:045009 (14 pp)
7. Poulin G, Sarraute E, Costa F (2004) Generation of electrical energy for portable devices:
comparative study of an electromagnetic and a piezoelectric system. Sens Actuators A Phys
116:461–471
8. Rocha JG, Gonçalves LM, Rocha PF, Silva MP, Lanceros-Méndez S (2010) Energy
harvesting from piezoelectric materials fully integrated in footwear. IEEE Trans Ind Electron
57:813–819
9. Al-Ashtari W, Hunstig M, Hemsel T, Sextro W (2013) Enhanced energy harvesting using
multiple piezoelectric elements: theory and experiments. Sens Actuators A Phys 200:138–146
10. Muralt P, Marzencki M, Belgacema B, Calamea F, Basrourb S (2009) Vibration energy
harvesting with PZT micro device. In: Proceedings of the eurosensors XXIII conference,
pp 1191–1194
11. Roundy S, Wright PK, Rabaey JM (2004) Energy scavenging for wireless sensor networks.
Springer, New York, pp 49–50
12. Caliò R, Rongala UB, Camboni D, Milazzo M, Stefanini C, Petris G, Oddo CM (2014)
Piezoelectric energy harvesting solutions. Sensors 14:4755–4790
13. Ward JK, Behrens S (2008) Adaptive learning algorithms for vibration energy harvesting.
Smart Mater Struct 17:035025 (9 pp)
14. Priya S (2007) Advances in energy harvesting using low profile piezoelectric transducers.
J Electroceram 19:165–182
15. Cho JH, Richards RF, Bahr DF, Richards CD (2006) Efficiency of energy conversion by
piezoelectrics. Appl Phys Lett 89:104107 (3 pp)
16. Cottone F, Vocca H, Gammaitoni L (2009) Nonlinear energy harvesting. Phys Rev Lett
102:080601 (4pp)
17. Uzun Y, Kurt E (2013) Power-vibration relation for a piezoelectric harvester under magnetic
excitation. In: 13th international conference on electric power system, high voltages, electric
mach, Chania, Crete Island, Greece, pp 59–64
18. Uzun Y, Kurt E, Kurt HH (2015) Explorations of displacement and velocity nonlinearities and
their effects to power of a magnetically-excited piezoelectric pendulum. Sens Actuators A
Phys 224:119–130
19. Kurt E (2006) Nonlinear responses of a magnetoelastic beam in a step-pulsed magnetic field.
Nonlinear Dynam 45:171–182
20. Tseng WY, Dugundji J (1970) Nonlinear vibrations of a beam under harmonic excitation.
J Appl Mech 37:292–297
21. Tseng WY, Dugundji J (1971) Nonlinear vibrations of a buckled beam under harmonicex-
citation. J Appl Mech 38:467–476
22. Poddar B, Moon FC, Mukherjee S (1988) Chaotic motion of an elastic–plastic beam. J Appl
Mech 55:185–189
23. Emam SA, Nayfeh AH (2004) Nonlinear responses of buckled beams
tosubharmonic-resonance excitations. Nonlinear Dyn 35:105–122
24. Saymonds PS, Yu TX (1985) Counterintuitive behavior in a problem of elastic–plasticbeam
dynamics. J Appl Mech 52:517–522
25. Uzun Y, Demirbas S, Kurt E (2014) Implementation of a new contactless piezoelectric wind
energy harvester to a wireless weather station. Elektron Elektrotech 20:35–39
26. Uzun Y, Kurt E (2012) Implementation and modeling of a piezoelastic pendulum under a
harmonic magnetic excitation. In: 11th international conference on applications of electrical
engineering, Athens, Greece, pp 1–6
27. Uzun Y, Kurt E (2013) The effect of periodic magnetic force on a piezoelectric energy
harvester. Sens Actuators A Phys 192:58–68
28. Bizon N, Oproescu M (2007) Power converters for energy generation systems (Convertoare
de Putere utilizate in Sistemele de Generare a Energiei). Publishing House of the University of
Piteşti, Piteşti
29. Bouzelata Y, Kurt E, Altın N, Chenni R (2015) Design and simulation of a solar supplied
multifunctional active power filter and a comparative study on the current-detection
algorithms. Renew Sust Energ Rev 43:1114–1126
30. Kurt E, Ciylan B, Taskan OO, Kurt HH (2014) Bifurcation analysis of a resistor-double
inductor and double diode circuit and a comparison with a resistor-inductor-diode circuit in
phase space and parametrical responses. Sci Iran 21:935–944
31. Linsay PS (1981) Period doubling and chaotic behaviour in a driven anharmonic oscillator.
Phys Rev Lett 47:1349–1352
32. Kurt E, Kasap R, Acar S (2003) Effects of periodic magnetic field to the dynamics of vibrating
beam. J Math Comput Appl 9:275–284
33. Kurt E, Böyükata M, Güvenç ZB (2006) Lyapunov exponent as an indicator of phase
transition in melting Pd13 clusters. Phys Scripta 74:353–361
34. Kurt E, Uzun Y (2010) Design and bifurcation analysis of a piezoelectric energy harvester
under a changeable magnetic field. In: 2nd international conference on nuclear and renewable
energy resources, Ankara, Turkey, pp 877–884
35. Moser RD, Kim J, Mansour NN (1999) Direct numerical simulation of turbulentchannel flow
up to Re = 590. Phys Fluids 11:943–945
36. Morris DJ, Youngsman JM, Anderson MJ, Bahr DF (2008) A resonant frequency tunable
extensional mode piezoelectric vibration harvesting mechanism. Smart Mater Struct
17:065021 (9 pp)
37. Ramond A, Ardila Rodríguez GA, Durou H, Jammes B, Rossi CA (2009) SIDO buck
converter with ultra low power MPPT scheme for optimized vibrational energy harvesting and
management. In: PowerMEMS, Washington DC, USA, pp 415–418
38. Kong N, Cochran T, Ha DS, Lin HC, Inman DJ (2010) A self-powered power management
circuit for energy harvested by a piezoelectric cantilever. In: 25th applied power electronics
conference and exposition (APEC), CA, USA, pp 2154–2160
39. Elliott ADT, Mitcheson PD (2014) Piezoelectric energy harvester interface with real-time
MPPT. J Phys Conf Ser 557:012125(5 pp)
40. Yi J, Su F, Lam YH, Ki WH, Tsui CY (2008) An energy-adaptive MPPT power management
unit for micro-power vibration energy harvesting. In: IEEE international symposium on
circuits and systems, Seattle, WA, USA, pp 2570–2573
41. Simjee F, Chou PH (2006) Everlast: long-life, supercapacitor-operated wireless sensor node.
In: Proceedings of the ISLPED, pp 197–202
42. Do XD, Han SK, Lee SG (2014) Optimization of piezoelectric energy harvesting systems by
using a MPPT method. In: IEEE 5th international conference on communications and
electronics, Danang, Vietnam, pp 309–312
43. Miller LM, Mitcheson PD, Halvorsen E, Wright PK (2012) Coulomb-damped resonant
generators using piezoelectric transduction. Appl Phys Lett 100:233901(4 pp)
44. Booth AD (1951) A signed binary multiplication technique. Q J Mech Appl Math 4:236–240
Chapter 6
Energy Harvesting from the Photovoltaic
Hybrid Power Source Based on Extremum
Seeking Control Schemes
Nicu Bizon, Marian Raducu, Luminita-Mirela Constantinescu

and Mihai Oproescu
Abstract The energy harvesting is known as the conversion process of ambient

energy into usable electrical energy. The energy of the renewable and green Energy
Sources (ES) is free and available without territorial restrictions. In this chapter the
possibility to use the Extremum Seeking Control schemes for harvesting the solar
energy via a Photovoltaic Hybrid Power Source is presented. The new ESC
schemes based on a band-pass filter instead of the series combination of high-pass
and low-pass filters are analyzed in order to evaluate their performance. The per-
formance indicators used are the search speed and the tracking accuracy. The
simulations performed highlight the advantages of the Extremum Seeking Control
schemes based on a band-pass filter in comparison with the classical Extremum
Seeking Control schemes. A Maximum Power Point tracking technique based on a
modified Extremum Seeking Control slightly improves the energy efficiency of the
Photovoltaic Hybrid Power Source. The advanced Extremum Seeking Control
scheme reduces the power ripple, so the energy efficiency of the Photovoltaic
Hybrid Power Source increases as well. The analysis of the dither persistence in the
Extremum Seeking Control loop scheme shows the relations between the search
speed and the derivatives of the Photovoltaic power. The ratio of these search
speeds is also used as the performance indicator. Finally, the dynamical operation
of the Photovoltaic Hybrid Power Source under variable irradiance profile is shown.
N. Bizon (&) M. Raducu L.-M. Constantinescu M. Oproescu

University of Pitesti, Targu din Vale Street no. 1, 110040 Pitesti, Romania
e-mail: nicu.bizon@upit.ro; nicubizon@yahoo.com
M. Raducu
e-mail: marian.raducu@upit.ro
L.-M. Constantinescu
e-mail: luminita.constantinescu@upit.ro
M. Oproescu
e-mail: mihai.oproescu@upit.ro
N. Bizon
University Politehnica of Bucharest, Splaiul Independentei Street no. 313, 060042
Bucharest, Romania

144 N. Bizon et al.

Keywords Photovoltaic Hybrid Power Source (PVHPS) Extremum Seeking

Control (ESC) Maximum Power Point (MPP) Search speed Tracking accu-

racy Energy harvesting Harmonics analysis

aESC Advanced ESC
BPF Band-pass filter
CS Charge-sustaining
CI Charge-increasing
CD Charge-depletion
DG Distributed generation
EMS Energy management strategy
ESC Extremum seeking control
ESS Energy storage system
ES Energy source
EQ Equivalence
FW Freewheeling diode
FC Fuel cell
GMPP Global MPP
HPS Hybrid power source
HT Hydro-turbines
HF High frequency
hoESC High-order ESC
HPF High-pass filter
IC Incremental conductance
IGBT Insulated-gate bipolar transistor
LF Low frequency
LPF Low-pass filter
MEP Maximum efficiency point
MPP Maximum power point
MPPT MPP tracking
mESC Modified ESC
P&O Perturb & observe
PV Photovoltaic
PVHPS Photovoltaic hybrid power source
WT Wind turbine
6 Energy Harvesting from the Photovoltaic Hybrid Power Source … 145
6.1 Introduction
In the last decades the solar energy is widely used as ambient energy that can be
efficiently harvested and converted into electrical energy. The solar, wind and water
energy are all ambient green energy resources without territorial restrictions, which
can generate electrical energy via the Photovoltaic (PV) panels’ array, Wind
Turbines’ (WT) farm and Hydro-Turbines (HT). The Energy Source (ES) shown
in Fig. 6.1 can represent a single source or a mix of renewable energy sources (such
as PV array, WT farm, and HTs). The Fuel Cell (FC) stack can be used as an
auxiliary energy source of the Hybrid Power Source (HPS) in order to minimize the
Energy Storage System (ESS) if the HPS operates under high dynamic load.
The DC loads will be connected on the ESS bus (directly or via a DC-DC con-
verter) and the AC load via a DC-AC converter.
The HPS architecture which will be analyzed in this chapter (see Sect. 6.2) is
based on the generic HPS architecture shown in Fig. 6.1. Note that each ES will be
connected to the ESS bus by an unidirectional DC-DC converter that is controlled
using a Maximum Power Point (MPP) tracking algorithm [1]. The MPP tracking
(MPPT) controller can control sequentially all ESs acquiring the samples of each
ES current (iES) and the common ES voltage (vES) [2]. The MPP of each ES will be
tracked using the Extremum Seeking Control (ESC) scheme that obtains the
searching gradient based on the dithering action of the control loop [3].
Both searching gradient and dither signals are parts of the reference current, iref.
The reference current (iref) is compared with the ES current (iES) using an appro-
priate current-mode control strategy to generate the switching pulses [4].
If the HPS supplies a load that has a sharp and dynamic profile, then the ESS is
necessary to assure the power flow balance on the ESS bus:
Fig. 6.1 Architecture of the Hybrid Power Source [26]

146 N. Bizon et al.
pL ¼ g pES þ pESS ð6:1Þ
The ESS power is given by (6.2):
pESS ¼ pL g pES ð6:2Þ
The ES power depends on the environmental and weather conditions, so it is

difficult to dynamically compensate the power flow balance by controlling the fuel
flow rates of the FC stack [5, 6]. Consequently, a hybrid ESS (of batteries and
ultracapacitors connected with DC-DC converters to the ESS bus [7]) must be
designed to minimize the cost of the HPS. The ESS will be charged and discharged
during a load cycle and the average ESS power, PESS(AV), can be zero, positive or
negative if the ESS operates in charge-sustaining (CS), charge-increasing (CI) or
charge-depletion (CD) mode, respectively [8, 9].
The Energy Management Strategies (EMSs) proposed in the literature are
dependent to HPS application. For example, the EMS is applied to the HPS used as
the Distributed Generation (DG) unit in the smart grids [10] or as the energy source
for the power-train of the hybrid vehicles [8]. The EMS of the load-following type
[9] will operate the ESS in CS mode based on the FC power controlled to assure the
averaged power flow balance:
PFCðAVÞ ¼ PLðAVÞ g PESðAVÞ ð6:3Þ
The FC stack is efficiently operated based on the efficiency map [11] or the
Maximum Efficiency Point (MEP) tracking algorithm [12, 13].
If the EMS is not of load-following control type, then the ESS will operate in CI
or CD mode by controlling the ESS charger (the bidirectional DC-DC converter)
based on the charging ES factor (kES) and the regenerative load factor (kL) defined
as following [9]:
kES ¼ PES!ESSðAV Þ = PESðAV Þ ð6:4Þ
kL ¼ PL!ESSðAVÞ = PLðAV Þ ð6:5Þ
where
PES!ESSðAVÞ is the part of the ES power flow that charge the ESS;
PL!ESSðAVÞ is the reverse power flow from the regenerative load that charge the
ESS.
The ESS charger usually has the plug-in feature, so choosing between the CI
mode or CD mode will depend by the price of the energy offered at the plug-in units
during the route [9].
The energy must be harvested from each ES and the FC stack must be operated
at MEP in order to optimize the HPS operation. In this chapter the PV HPS opti-
mization based on an ESC scheme is analyzed. The perturbed ESC scheme will
Fig. 6.2 Searching of the MPP on the power characteristic [26]
track the MPP of the PV panel (see Fig. 6.2) or the Global MPP (GMPP) of the PV
array.
Besides the ESC scheme, different MPPT techniques have been proposed during
the last decades. These techniques can be categorized as direct and indirect methods
to maximize the power generated by an energy source that has a power charac-
teristic with one maximum [1, 14]. The power characteristic of the solar cells is
modeled using mathematical equations [15] or numerical approximations [16].
The MPP tracking performance is dependent on the resolution of the
analog-to-digital converter [17] and the cut-off frequencies of the low-pass filter
(LPF) [18]. The Perturb & Observe (P&O) and Incremental Conductance (IC)
methods are the most used MPPT algorithms implemented in the commercial
controllers [19]. The P&O method uses a perturbation of the PV voltage or PV
current to search the MPP by observing the sense of the variation of the PV power.
The step size of the perturbation must be carefully designed to balance the tradeoffs
between the performance indicators (the search speed and tracking accuracy),
assuring in meantime the robustness to variability of the irradiance [20, 21]. The
irradiance profile during a sunny-cloudy day has a dynamic shape in reality, so a
model of this profile is usually used in simulation [21].
This chapter analyzes three ESC schemes to compare the performance of the
MPPT process. The modified ESC (mESC) scheme (see Fig. 6.3) has improved
performance in comparison with the high-order ESC (hoESC) scheme (see
Fig. 6.4) related to the search speed if high cut-off frequency of the band-pass filter
(BPF) is higher than that of the LPF. The low cut-off frequency of the BPF may be
the same with that of the high-pass filter (HPF). It can be noted that the tracking
148 N. Bizon et al.
Fig. 6.3 The mESC scheme [26]
Fig. 6.4 The hoESC scheme (or equivalent (EQ) mESC scheme) [26]
accuracy remains almost the same [22], where the MPP tracking accuracy is defined
as 100 (PPV / PMPP) [%], PMPP being the power of the PV panel operating at the
MPP and PPV the power harvested from the PV panel via DC-DC converter under
the MPPT controller. The 99.8% tracking accuracy could be obtained with
advanced MPPT algorithms (which decrease the level of perturbation close to
MPP). An adaptive decrease of the perturbation is used in the advanced ESC
(aESC) scheme proposed in [21]. The 99.99% tracking accuracy could be obtained
without reducing the search speed and response time to irradiance changes.
The generic topologies of the mESC and hoESC schemes will be detailed in the
next sections for the Photovoltaic Hybrid Power Source (PVHPS) application. It is
obvious that the performance inherited from the basic hoESC scheme (such as a
guaranteed convergence and an internal robustness) is taken up as well [23].
The ESC schemes can be used to search the unknown maximum of the PV power
characteristic because it is a control method of adaptive type [24, 25].
Also, it is easy to show that the mESC and hoESC schemes are functionally
equivalent if the cut-off frequencies of the BPF are the same with those of the HPF
and LPF (see Fig. 6.5). This will be briefly shown in next section based on the
specific ESC schemes for the PVHPS [5, 26].
The chapter is organized as follows. Section 6.2 briefly presents the architecture
of the PVHPS. The modeling of the PVHPS is presented in Sect. 6.3, where the
models of the solar cell, PV panel, boost converter, and ESC schemes used in
simulation are shown. Section 6.4 deals with the performance analysis of the ESC
schemes. The analysis in frequency domain of the ESC loop is performed here in
order to evaluate the performance of the ESC schemes under step-up in the irra-
diance profile. The dither persistence on the ESC loop is also approached here.
The PVHPS behavior under variable irradiance profile is shown in Sect. 6.4 using
the mESC and aESC schemes in Sect. 6.6.
Fig. 6.5 Bode diagrams for the LPF (●), HPF (○), and BPF (■) [26]
6.2 Architecture of the Photovoltaic Hybrid Power Source
The Photovoltaic Hybrid Power Source (PVHPS) architecture is shown in Fig. 6.6,
where the main subsystems are detailed by the diagram implemented in
Matlab-Simulink®. The dynamic load is connected on the ESS DC bus and the ESS
is implemented using a passive topology [7]. A simple current-mode hysteretic
controller is used here to generate the switching command for the boost converter.
The boost converter is usually used as a DC-DC power interface between the
low DC voltage of the ES (here the PV voltage) and the high DC voltage (VDC) of
the common DC bus for ESS and inverter system [14, 27]. If the voltage inverter
system is grid connected, then the switching command must be synchronized with
the grid frequency and the harmonics of the inverter and the grid voltages must
have the same magnitude [28]. Thus, the low frequency (LF) and high frequency
(HF) ripple having harmonics at multiples of the grid frequency and switching
frequency will appear on the DC bus. The LF ripple by three harmonics is con-
sidered in the load model because only these are important to analyze the ESC
control loop. It is known that the ripple correlation method can use both LF and HF
ripple of the PV power signal (PPV) [29], but the natural inverter ripple is usually
utilized by the perturbed MPPT algorithms based on the P&O methods [30] or the
ESS schemes [31].
150 N. Bizon et al.
Fig. 6.6 The diagram of PV/ESS HPS operating at MPP under aESC [21]
6.3 Modeling the Photovoltaic Hybrid Power Source
6.3.1 Solar Cell and PV Panel Models
6.3.1.1 Solar Cell Model
The one-diode model of the solar cell contains a current source (IL), a diode, a shunt
resistor (Rp) and a series resistor (Rs) with the output current (IPV(cell)) given by the
(6.6) [1]:

qðVPVðcellÞ þ Rs IPVðcellÞ Þ
IPVðcellÞ ¼ ILðcellÞ I0S exp 1
nkB T
ð6:6Þ
VPVðcellÞ þ Rs IPVðcellÞ

Rp
where VPV(cell) represents the solar cell voltage and T is the temperature.
The light-generated current, IL(cell), and the cell reverse saturation current, I0S,
are given by (6.7) and (6.8):
G
ILðcellÞ ¼ IscðcellÞ ½1 þ a ðT TR Þ ð6:7Þ
GR
3
T VG TR
I0S ¼ I0R exp 1 ð6:8Þ
TR nVT T
where VT = kBTR/q.
If the parallel resistance Rp will be neglected and the temperature is assumed to
be constant, T = TR = 298 K, then relations (6.6)–(6.8) can be further simplified as:

qðVPVðcellÞ þ Rs IPVðcellÞ Þ
IPVðcellÞ ¼ ILðcellÞ I0S exp 1 ð6:6’Þ
nkB T
G
ILðcellÞ ¼ IscðcellÞ ¼ KIGðcellÞ G ð6:7’Þ
GR
I0S ¼ I0R ð6:8’Þ
where KIG(cell) = Isc(cell)/GR is irradiation Irradiation to short-circuit current gain.

Consequently, the simplified model of the solar cell is given by (6.9):

VPVðcellÞ þ Rs IPV ðcellÞ
IPV ðcellÞ ¼ KIGðcellÞ G I0R exp 1 ð6:9Þ
nVT
where:
VPV(cell)—represents the solar cell voltage;
IPV(cell)—the solar cell current;
Rs—the series resistance of the solar cell;
KIG(cell) = Isc(cell)/GR—the irradiation to short-circuit current gain;
IL(cell) = Isc(cell)G/GR—the light-generated current;
G—the level of instantaneous irradiation
Other parameters are mentioned in Table 6.1.
6.3.1.2 PV Panel
The PV panel used 34 cells connected in series. Thus, the open-circuit voltage, Voc,
and the short-circuit current, Isc, of the PV panel will be about 21 V and 3.8 A
under standard test conditions: irradiation intensity of 500 W/m2, AM1.5 G, and
temperature of 25 °C. Also, the MPP of about 51.2 W is obtained for 3.5 A and
14.63 V under the same test conditions.
152 N. Bizon et al.
Table 6.1 Parameters of solar cell model [21]

Parameter Description Value (unit)
GR Reference irradiation 1,000 (W/m2)
TR Reference temperature 298 (K)
q Electron charge 1.6e−19 (C)
kB Boltzmann’s constant 1.38e−23 (JK−1)
n Diode ideality factor 1.3 (-)
VG Silicon band-gap energy 1.12 (eV)
VT = kBTR/q Thermic voltage 26 (mV)
IOR Reverse saturation current at T = TR 2e−9 (A)
a Short-circuit current temperature coefficient 0.0025 (AK−1)
Rs Cell series resistor 3 (mX)
Rp Cell shunt resistor 10 (X)
Voc(cell) Cell open-circuit voltage 0.61 (V)
Isc(cell) Cell short-circuit current (A)
6.3.2 Boost Converter Model
The basic boost converter is implemented using power devices from the
SimPowerSystem® toolbox.
If the both insulated-gate bipolar transistor (IGBT) and freewheeling (FW)
diode are in on-state, then the operating relationships are [10]:
diPV
vPV ¼ ðrL þ RDSðonÞ Þ iPV þ L þ VDSðonÞ
dt ð6:10Þ
diPV
vPV ¼ ðrL þ RDðonÞ Þ iPV þ L þ VDðonÞ þ vDC
dt
where the on-state parameters, (RDS(on), VDS(on)) and (RD(on) and VD(on)), use the
preset values.
The value of boost inductance is designed based on the magnitude of the HF
components of the PV current, DI, and hysteretic band:
L ¼ DVFCðMPPÞ = ð2fsw DIÞ ð6:11Þ
where VFC(MPP) is the maximum PV voltage that can be obtained during a sunny
day, fsw = 1/Tsw is the switching frequency and D is the duty cycle estimated with
(6.12):
D ¼ 1 VFCðMPPÞ = VDC ð6:12Þ

The switching frequency may be chosen to be about 100-times higher than the
dither frequency (100 Hz) to be better tracked by the LF shape of the PV current by
the reference current. Thus:
irefðLFÞ ffi iPVðLFÞ ð6:13Þ
6.3.3 The ESC Schemes
The ESC schemes used in this chapter will be briefly presented here.
6.3.3.1 The hoESC Scheme
The hoESC scheme shown in Fig. 6.1 is adapted for PVHPS as in Fig. 6.7. The
HPF is necessary to extract the useful LF harmonics from the PV power signal and
the LPF to reduce the HF noise [32].
The information about the searching gradient is obtained by using a dither
signal, kAsin(2pfdt) (kA being the amplitude, fd the dither frequency and t the time),
injected in the ESC control loop. A compromise must to be made between the
choosing of the design parameters (k1 and k1 gains) to obtain good performance for
both indicators: the search speed and the tracking accuracy. The search speed can
be increased if the cut-off frequency of the LPF is set about 5-times higher than the
dither frequency. If the series connection of the HPF and LPF is replaced with an
equivalent BPF, then the modified ESC (mESC) scheme will be obtained (see
Fig. 6.8).
Note that the cut-off frequencies of the LPF and HPF used in hoESC scheme are
usually set as xl(ho) = bl(ho)x, 0 < bl(ho) < 1, and xh(ho) = bh(ho)x, 0 < bh(ho) < 1.
6.3.3.2 The Modified ESC Scheme
The modified ESC scheme is shown in Fig. 6.8. The search speed of the mESC
scheme increases about 2-times in comparison with the hoESC scheme if the same
Fig. 6.7 The hoESC scheme [22]

154 N. Bizon et al.
Fig. 6.8 The mESC scheme [22]
Fig. 6.9 Advanced ESC (aESC) scheme [21]
k1 and k1 gains are used for both mESC and hoESC schemes, but the cut-off
frequency of BPF high is set bl(m)-times higher than dither frequency (xl(m) = bl
(m)x, 3 < bl(m) < 6) [26].
The HPF removes the average value of the PV power, so the low cut-off fre-
quency of the HPF and BPF can be set the same for the both ESC schemes:
xh = bhx, 0 < bh < 1.
It is known that a large value of the k2 gain will increase the search speed, but the
tracking accuracy is depreciated. A high power ripple will appear even during the
stationary phase (constant irradiance). Thus, the k1 gain must be higher during the
transitory phase (change of the irradiance level) and very small during the stationary
phase. This can be obtained by using the advanced ESC (aESC) scheme, where the
dither injected has the amplitude modulated with the first harmonic (H1) of the PV
power: kA = Ak2H1 (see Fig. 6.9).
6.3.3.3 The Advanced ESC Scheme
The advanced ESC (aESC) scheme is shown in Fig. 6.9, and its equivalent
(EQaESC) topology based on HPF and LPF in series is shown in Fig. 6.10.
It is obvious that all the ESC schemes (hoESC, mESC, and aESC schemes) will
have the same equivalent operating equations (see for example [21, 26]):
Fig. 6.10 Equivalent aESC (EQaESC) scheme [21]
pPV ¼ pPV ðiPV Þ; i1 ¼ iPV = VMPP ð6:14Þ

iHPF ¼ xh iHPF þ xh i1 ;

ð6:15Þ
i2 ¼ i1 iHPF ; iBPF ¼ xl iBPF þ xl i2 ; i3 ¼ iBPF A sinðxtÞ

iHPF ¼ xh iHPF þ xh i1 ; i2 ¼ i1 iHPF ; i3 ¼ xl i3 þ xl i2 A sinðxtÞ ð6:16Þ

i4 ¼ k1 i3 ð6:17Þ
iPV ffi iref ; iref ¼ i4 þ kA sinðxtÞ þ IPV0 ð6:18Þ
where Eqs. (6.14), (6.17), and (6.18) represent the PV power characteristic, inte-
grator, current—mode controller based on the current reference, iref, and Eqs. (6.15)
and (6.16) represent the signal processing in aESC and EQaESC schemes based on
BPF and series connection of HPF and LPF, respectively.
The following notations are used for all ESC schemes (see Figs. 6.7, 6.8, 6.9 and
6.10):
• k1 is the loop gain;
• kA is the dither’s amplitude (kA = Ak2 for the hoESC and mESC schemes, and
kA = H1Ak2 for the aESC scheme);
• H1 is the magnitude of the first harmonic of the PV power;
• xl = blx is the cut-off frequency of the LPF (0 < bl(ho) < 1 for hoESC scheme,
and 1 < bl(m) < 6 for the mESC and aESC schemes);
• xh = bhx, 0 < bh < 1, is the cut-off frequency of the HPF for all ESC schemes;
• A and fd = x/2p are the amplitude and frequency of the dither signal;
• i1, i2, and i3 are the signals after normalization block, HPF, and LPF blocks (see
Fig. 6.9 and the Eqs. (6.14)–(6.16));
• iHPF is an intermediate variable related to the HPF operating;
• iBPF is the output signal from the BPF;
156 N. Bizon et al.
• i4 is the signal that estimates the unknown parameter;

• VMPP is voltage on the MPP;
• IPV0 is the initial value for the PV current.
Note that A = 1 and IPV0 = 0 for all simulations presented in this chapter.
So, the k1 gain can be set higher (up to the stability limit) in order to have search
speed up to the save value of HPS operating, without affecting the tracking accu-
racy. It was shown in [26] that the performance of the aESC scheme is better than
that obtained with the extended Kalman-filter used for gradient estimation [32].
Thus, the main feature of the aESC scheme is related to its capability to set
independently the performance indicators for each HPS application. The ESC
schemes will be analyzed in next section.
6.4 Performance Analysis of the ESC Schemes
The comparative performance is performed in this section based on the analysis in

the frequency domain of the ESC loop. The analytical results will be validate by a
simulation using the testing diagram shown in Fig. 6.11.
6.4.1 The Analysis in Frequency Domain of the ESC Loop
The searching of the MPP on the power characteristic is mainly the same for all the
ESC schemes considered here, being based on dither-pertubed search (see Fig. 6.2).
Fig. 6.11 The diagram for testing the ESC scheme [22]
The PV power can be approximated by the Taylor series:

pPV ¼ PPV þ pPV iref þ pPV i2ref = 2! þ pPV i3ref = 3! þ . . . ð6:19Þ
where iref = IPV + kAsin(xt) is the reference current, (IPV, PPV) is one point on static

power map, and pPV ; pPV ; pPV , and so on are derivatives of the PV power to iref
current. The iref current is linear variable in time (see Fig. 6.2), so the derivatives of
the PV power will be related to time as well.
The PV power signal (pPV) will become a current signal (i1) after normalization
to VMPP:
i1 ¼ pPV = VMPP ¼ I1 þ iLF1 þ iHF1 ð6:20Þ
where I1, iLF1 and iHF1 are the DC, LF and HF components of i1 signal.
The first order HPF (with transfer function GHPF(s) = s/(s + xh), where xh =
hx, 0 < bh < 1) will remove the DC component of i1 signal, so the i2 signal will be
given by (6.21):
i2 ¼ ðiLF1 þ iHF1 Þ L1 fGHPF ðsÞg ¼ iLF2 þ iHF2 ¼ i2ð1Þ þ iLF2 þ iHF2 ð6:21Þ
where L−1 is the inverse Laplace transform, * is the convolution operator, iLF2 and
iHF2 are the LF and HF components of i2, i2(1) is the fundamental harmonic of i2,
and iLF2 is remaining LF component of i2 − i2(1) signal.
The first order LPF (with transfer function GLPF(s) = xl/(s + xl), where xl =
l x, 0 < bl < 6) will remove the HF component of the i2 signal, so the i3 signal will
be given by the (6.22):
i3 ¼ ½ðiLF2 þ iHF2 Þ sinðxtÞ L1 fGLPF ðsÞg ffi ½iLF2 sinðxtÞ L1 fGLPF ðsÞg
¼ I3 þ iLF3
ð6:22Þ
where I3 and iLF3 are the DC components

and
LF component of i3 signal.
The information about the gradient pPV is obtained by multiplying the fun-

damental harmonic of the i2 signal, i2ð1Þ ¼ pPV jGHPF j sinðxt þ /HPF Þ; with the
dither, Asin(xt), i.e., performing a demodulation:

i2ð1Þ A sinðxtÞ ¼ pPV jGHPF jA sinðxt þ /HPF Þ sinðxtÞ

ð23Þ
¼ ðpPV jGHPF jA = 2Þ ½cosð/HPF Þ cosð2xt þ /HPF Þ
where
158 N. Bizon et al.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jGHPF j ¼ 1 = 1 þ ðxh = xÞ2 ; /HPF ¼ argðGHPF Þ ¼ arctanðxh = xÞ ð6:24Þ
The following values of the searching gradient (I3) are obtained:

I3 ¼ pPV AjGHPF jjGLPF jcosð/HPF Þ = ð2VMPP Þ ð6:25Þ
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jGLPF j ¼ 1 = ð x = x l Þ 2 þ 1; /LPF ¼ argðGLPF Þ ¼ arctanðx = xl Þ ð6:26Þ
If the mESC and aESC schemes are based on BPF, then the searching gradient
(I3(BPF)) are obtained as:

I3ðBPFÞ ¼ pPV AjGBPF jcosð/BPF Þ = ð2VMPP Þ ð6:27Þ
where
jGBPF j ¼ jGHPF jjGLPF j; /BPF ¼ argðGBPF Þ ¼ /HPF þ /LPF ð6:28Þ
Consequently, if the same parameters are set for the HPF and LPF in series, and
equivalent BPF, then the searching gradient will almost be obtained. The simula-
tions shown below will validate this result.
The next signal processing block for all ESC schemes is the integration block,
which contains the k1 gain as well:
Z Z Z
i 4 ¼ k1 i3 dt ¼ k1 ðI3 þ iLF3 Þdt ¼ k1 I3 t þ k1 iLF3 dt ¼ KSS t þ iLF4 ð6:29Þ
where iLF4 represents the LF components of i4 and KSS [A/s] is the search speed that
is specifically computed for all ESC schemes.
Finally, the reference current, iref, is given by (6.30):
iref ¼ IPV 0 þ KSS t þ iLF4 þ kA sinðxtÞ ¼ IPV0 þ KSS t þ iLFðref Þ ð6:30Þ
where IPV0 and iLF(ref) are the initial value and the LF component of iref.
The MPP tracking process on the P-I phases plane is shown in Fig. 6.12 using
the aESC scheme. The PV power spectrum is shown in three points of the PV
power characteristic versus PV current.
A zoom of the tracking accuracy is shown in the lower right corner. The high
power ripple during the searching phase and negligible ripple during the stationary
phase (constant irradiance) can be observed.
Fig. 6.12 The MPP tracking process using aESC [21]
6.4.2 On the Equivalence of the mESC and EQmESC

Schemes
The LF equivalence (EQ) of search speed for the EQmESC and mESC schemes
computed based on (6.29) is given by (6.31) and (6.32):

KSSðEQmÞ ¼ pPV k1 A GHPFðEQmÞ GLPFðEQmÞ cosð/HPFðEQmÞ Þ = ð2VMPP Þ ð6:31Þ

KSSðmÞ ¼ pPV k1 A GBPFðmÞ cosð/BPFðmÞ Þ = ð2VMPP Þ ð6:32Þ
Thus, the ratio of the searching speeds (KSS(m) and KSS(EQm), respectively) is:
RSS ¼ KSSðmÞ = KSSðEQmÞ ¼ cosð/HPFðmÞ þ /LPFðmÞ Þ = cosð/HPFðEQmÞ Þ ð6:33Þ
If
0\bhðmÞ \1\3\blðmÞ ð6:34Þ

160 N. Bizon et al.
Fig. 6.13 The diagram for comparative tests using as reference the PV panel without MPP control
[21]
Then
1\RSS 1 = cosð/HPFðhoÞ Þ 1:0541 ð6:35Þ
the equality being obtained when:
/HPFðhoÞ ¼ /LPFðhoÞ , blðhoÞ bhðhoÞ ¼ 1 ð6:36Þ
The diagram from Fig. 6.13 will be used to validate the equivalence of mESC
and EQmESC schemes, and aESC and EQaESC schemes.
In conclusion, based on (6.35), the mESC and EQmESC schemes have almost
the same search speed if GBPF(m)(s) = GHPF(EQm)(s)GLPF(EQm)(s). The simulation
results are shown in Fig. 6.14 for the following parameters set for the both mESC
and EQmESC schemes: k1 = 400, kA = 0.1 (k2 = 1, and A = 0.1), fd = 100 Hz, bl
(m) = 5.5, and bh(m) = 0.5 (so bl(m) bh(m) = 2.75).
Some differences in the searching process are shown in zooms for bl(m) bh
(m) = 1 case, when bl(m) = 5.5 and bh(m) = 1/bl(m)≅0.18.
Fig. 6.14 Simulation results for mESC and EQmESC schemes [22]
The simulation results validate the analytical results presented above:

• Both the mESC and EQmESC schemes almost have the same search speed. For
example, at 0.05 s it can be observed that the DC components are about
IPV = 1.5A, VPV = 18 V, so PPV = 27 W, resulting on this point the search

rates: iPV ¼ 1A=0:04 s ¼ 25 As1 , and pPV ffi VPV iPV ffi 450 Ws1 . This power
rate is verified considering the result shown on the top plot of Fig. 6.14.
• The LF harmonics content of the i4 signal is almost the same for both mESC and
EQmESC schemes.
• The power ripple is the same for both mESC and EQmESC schemes, being
about 0.2Wp-p Thus, the normalized power ripple is 0.2/51.4 ≅ 0.004 = 0.4%.
• The tracking accuracy is the same for both mESC and EQmESC schemes, being
about 51.3/51.4 ≅ 0.998 = 99.8% (see the zoom on right of top plot of the
Fig. 6.14).
6.4.3 On the Performance of the mESC and hoESC

Schemes
The search speed for mESC and hoESC schemes computed based on (6.29) is given
by the (6.37) and (6.38):
162 N. Bizon et al.

KSSðmÞ ¼ pPV k1 A GBPFðmÞ cosð/BPFðmÞ Þ = ð2VMPP Þ ð6:37Þ

KSSðhoÞ ¼ pPV k1 A GHPFðhoÞ GLPFðhoÞ cosð/HPFðhoÞ Þ = ð2VMPP Þ ð6:38Þ
Thus, the ratio of the searching speeds (KSS(m) and KSS(ho), respectively) is:
RSS ¼ KSSðmÞ = KSSðEQmÞ ¼ cosð/HPFðmÞ þ /LPFðmÞ Þ = cosð/HPFðhoÞ Þ ð6:39Þ
This ratio can be approximated by (6.40):

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
RSS ffi 1 þ 1 = b2lðhoÞ ð6:40Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
because blðmÞ = b2lðmÞ þ 1 0:95 for 0\blðhoÞ \1\3\blðmÞ .
The simulation results are shown in Fig. 6.15 for the following parameters set
for the both mESC and hoESC schemes: k1 = 400, kA = 0.1 (k2 = 1, and A = 0.1),
fd = 100 Hz, bl(m) = 5.5, bl(ho) = 0.5, and bh(m) = bh(ho) = 0.5.
The simulation results validate the analytical results presented above:
• The search speed at t = 0.05 s is about 450 and 300 Ws−1 for the mES and
hoESC schemes, so the ratio is 450/300 = 1.5. The theoretical value given by
(6.40) is about 2.2. This difference could be explained based on simple
assumptions considered in obtaining the relationship (6.40), which estimates this
Fig. 6.15 Simulation results for the mESC and hoESC schemes [22]
rate at the dither frequency. It can be observed that the LF spectrum of the i4
signal is different for the mESC and hoESC scheme (see plot 3 in Fig. 6.15).
The dither persistence is improved for the mESC scheme [26]. This aspect will
be briefly analyzed in Sect. 6.4.5 to clearly highlight the role of harmonics in the
ESC loop.
• The difference between the shapes of the PV power during the searching phase
(see Fig. 6.15, top plot) is about 5 J, which means an additional power har-
vested of about 5 J/0.2 s = 25 W from the PV module (SX60 type having PMPP
≅ 51.2 W) if the mESC scheme is used.
6.4.4 On the Performance of the aESC and mESC Schemes
The main difference between the mESC scheme and the aESC scheme is related to
the dither gain, which is constant, kA(m) = Ak2, and time variable, kA(a) = Ak2H1,
respectively, where H1 is the magnitude of first harmonic of the PV power. So, the
searching speed for aESC and mESC schemes are given by (6.41) and (6.42) [21]:

KSSðmÞ ¼ pPV k1ðmÞ A GBPFðmÞ cosð/BPFðmÞ Þ = ð2VMPP Þ ð6:41Þ

KSSðaÞ ¼ pPV k1ðaÞ H1 A GBPFðaÞ cosð/BPFðaÞ Þ = ð2VMPP Þ ð6:42Þ
If the same design parameters will be used for the aESC and mESC schemes (k1
(a) = k2(m) and GBPF(a) = GBPF(m)), then the ratio of searching speeds will be:
RSS ¼ KSSðaESCÞ = KSSðmESCÞ ffi H1 ð6:43Þ
If the S1 saturation block has the limits 0 and 1, then kA(a) = kA(m) during the
searching phase (when H1 > 1) and kA(a) = H1kA(m) if the MPP was located and
H1 < 1. The H1 magnitude is very small close to MPP, so the PV power ripple is
negligible during the stationary phase and the MPP is found accurately (see
Fig. 6.16, where the magnitudes of the first three harmonics (H1, H2, and H3) of the
PV power are shown).
Note that the searching speed will be H1-times higher for the aESC scheme in
comparison with the mES scheme if the S1 saturation block has the upper limit set
to infinit.
The magnitude of H1 and its mean value are shown in the first plot of Fig. 6.16.
The magnitude of H2 and H3 is shown in next plots. The PV power during the
MPP searching process is shown in plot 4 of Fig. 6.16 and the zoom shows the
tracking accuracy. The search speed is about 200 W/s (see plot 4 of Fig. 6.16) if the
values used for the k1 and k2 gains are set to 400 and 0.8, and A = 0.2.
The effect of amplitude limitation of the dither used for the aESC scheme is
shown in Fig. 6.17.
164 N. Bizon et al.
Fig. 6.16 The PV power signal and magnitude of its harmonics [21]
Fig. 6.17 The dither signal for the aESC scheme [21]
Fig. 6.18 The MPP searching for 500 W/m2 irradiance level using the aESC (■) and mESC (●)
schemes; the reference is set by the PV panel without MPP control (▲) [21]
If this type of limited dither is used for the aESC scheme, then the ratio of the
searching speeds (6.43) will be 1 during the searching phase. This result is validated
by simulation (see first plot in Fig. 6.18, where almost the same search speed is
obtained for both mESC and aESC schemes). The search speed is about 800 W/s
because k1 = 1,600 is 4-times higher than the previous value used in Fig. 6.16, and
the saturation block (S1) is used to limit the H1 magnitude of the iBPF signal at 0.2
(which is also the amplitude of the dither used in both ESC schemes), then the
search speed will be the same for the both ESC techniques (about 800 W/s; see
Fig. 6.10).
The MPP tracking accuracy for the both control schemes is highlighted in the
zooms from the right side of Fig. 6.18. The power ripple has a decreasing and
constant magnitude of ripple for the aESC and mESC schemes. This validates the
ratio of the searching speeds (6.43) that is H1 during the stationary phase. Note that
the H1 magnitude decreases to a very small value during the stationary phase (see
the last plot in Fig. 6.16).
The PV panel without control (see Fig. 6.13) is used as reference to compute the
MPP value for 500 W/m2 irradiance level, the zooms are used to evaluate the
performance indicators for both ESC schemes. The performance evaluated based on
Fig. 6.18 are the following:
• The search speed is almost the same for both ESC techniques if H1 magnitude is
limited at 1; the search speed using the aESC scheme is H1—times higer than
that using the mESC scheme (see simulations shown in Sect. 6.5).
166 N. Bizon et al.
• The MPP tracking accuracy using the mESC scheme is about 99% (because 1%
≅100% (51.4 − 50.9)/2 51.4) and lower than 0.1% if the aESC scheme is
used;
• The power ripple during stationary phase is about 0.5 Wp-p (= 51.4 − 50.9) and
negligible for mESC scheme and aESC scheme;
6.4.5 On the Dither Persistence on the ESC Loop
The dither persistence on the ESC loop will be analyzed using the following
simplifying assumptions:
• Ideal HPF and LPF;
• Unitary value for kN and kT gains;
• Initial value for the /k is set to zero;
• xd xh(ES), where xh(ES) is the cut-off frequency of the Energy Source (ES)
transfer function.
The dither will not interact with the ES dynamic based on last assumption, so the
ES power signal will result as harmonics’ superposition based on the ES static
characteristic (see Fig. 6.2) and, for example, the EQmESC scheme (Fig. 6.3). The
first three harmonics will be considered for the iref(LF) signal:
X
3
iref ðLFÞ ¼ Iref ðkÞ sinðkxd t þ /k Þ ð6:44Þ
k¼1
So, the reference current, iref, will be given by (6.45):
iref ðtÞ ¼ Iref þ iref ðLFÞ ðtÞ ffi Iref þ a1 sinðxd tÞ

ð6:45Þ
þ a2 sinð2xd tÞ þ a3 sinð3xd tÞ
The power response, pES, to one harmonic can be approximated by the Taylor
series:
pES ðiref Þ ¼ PES jIref þ p_ ES ðiref Þ Iref þ ak sinðk xd tÞ

€
pES ðiref Þ 2
þ Iref þ ak sinðk xd tÞ ð6:46Þ
2!
pvES ðiref Þ 3
þ Iref þ ak sinðk xd tÞ þ
3!
So, the small signal (SS) components of the pES signal are:
€
pES ðtÞ 2
pESðSSÞ ðtÞ ffi p_ ES ðtÞ ak sinðk xd tÞ þ ak sin2 ðk xd tÞ
2
ð6:47Þ
pv ðtÞ
þ ES a3k sin3 ðk xd tÞ þ
6
or
€
pES ðtÞ a2k
pESðLFÞ ðtÞ ffi p_ ES ðtÞ ak sinðk xd tÞ cosð2 k xd tÞ
2 2
ð6:48Þ
pv ðtÞ a3
þ ES k ½3 sinðk xd tÞ sinð3 k xd tÞ
6 4
If the HPF is ideal and kN=1, then the p1(LF) signal will be given by (6.49):
p1ðLFÞ ðtÞ ¼ pESðLFÞ ðtÞ ð6:49Þ
Thus, the signal after the demodulation is:
p2ðLFÞ ðtÞ ¼ p1ðLFÞ ðtÞ sinðxd tÞ ¼ p_ ES ðtÞ ak sinðxd tÞ sinðk xd tÞ

a2k
€
pES ðtÞ sinðxd tÞ cosð2 k xd tÞ
4
a3 ð6:50Þ
þ pvES ðtÞ k sinðk xd tÞ sinðxd tÞ
8
3
a
þ pvES ðtÞ k sinð3 k xd tÞ sinðxd tÞ
24
Considering all harmonics (k = 1, 2, 3), the p2(LF) signal is given based on the
superposition techniques as:
p2ðLFÞ ðtÞ ¼ p_ ES a1 sin2 ðxd tÞ þ a2 sinð2 xd tÞ sinðxd tÞ

þ a3 sinð3 xd tÞ sinðxd tÞ
€pES 2
a sinðxd tÞ cosð2 xd tÞ þ a22 sinðxd tÞ cosð4 xd tÞ
4 1
þ a23 sinðxd tÞ cosð6 xd tÞ
pvES 3 ð6:51Þ
þ a sin2 ðxd tÞ þ a32 sinð2 xd tÞ sinðxd tÞ
8 1
pvES 3
a sinð3 xd tÞ sinðxd tÞ þ a32 sinð6 xd tÞ sinðxd tÞ
24 1
168 N. Bizon et al.
or (after simple trigonometric manipulations)
p_ ES p_ ES
p2ðLFÞ ðtÞ ¼ a1 ½1 cosð2 xd tÞ þ a2 ½cosðxd tÞ cosð3 xd tÞ
2 2
p_ ES €
pES 2
þ a3 ½cosð2 xd tÞ cosð4 xd tÞ a1 ½sinðxd tÞ þ sinð3 xd tÞ
2 8
€
pES 2 €
pES 2
a2 ½sinð3 xd tÞ þ sinð5 xd tÞ a3 ½sinð5 xd tÞ þ sinð7 xd tÞ
8 8
pv pv
þ ES a31 ½1 cosð2 xd tÞ þ ES a32 ½cosðxd tÞ cosð3 xd tÞ
16 16
pvES 3 pv
þ a ½cosð2 xd tÞ cosð4 xd tÞ ES a31 ½cosð2 xd tÞ cosð4 xd tÞ
16 3 48
pvES 3 pvES 3
a ½cosð5 xd tÞ cosð7 xd tÞ a ½cosð8 xd tÞ cosð10 xd tÞ
48 2 48 3
ð6:52Þ
If the LPF is ideal, then the p3eq(LF) signal is given by (6.53):
p_ ES a1 pvES a31
p3eqðLFÞ ðtÞ ffi p2ðLFÞ ðtÞ ¼ þ þ sinðxd tÞ
2 "16 #

€pES a21 p_ ES a2 pvES a32
þ cosðxd tÞ þ þ sinð2 xd tÞ½0
8 2 16
" #
p_ ES a1 p_ ES a3 pvES a31 pvES a33 pvES a31 pvES a31
þ cosð2 xd tÞ þ þ þ
2 2 16 16 48 48
" #
€
pES a1 €
2
pES a2
2 v
p_ ES a2 pES a32
þ sinð3 xd tÞ þ þ cosð3 xd tÞ
8 8 2 16
ð6:53Þ
Thus, the p4eq(LF) signal (after the integrator block) is:

" #
1 p_ ES a1 pvES a31 1 €
pES a21
p4eqðLFÞ ðtÞ ¼ þ t cosðxd tÞ
k1 2 16 xd 8
" #
1 p_ ES a2 pvES a32 1
þ sinðxd tÞ þ cosð2 xd tÞ ½0
xd 2 16 2 xd
" #
1 p_ ES ða3 a1 Þ pvES 3 a33 4 a31
þ sinð2 xd tÞ þ
2 xd 2 48
2
1 €
pES a2 a1
2
cosð3 xd tÞ
3 xd 8
1 8 p_ ES a2 þ pvES a32
sinð3 xd tÞ
3 xd 16
ð6:54Þ
If kT = 1, then the i5eq(LF) signal is:

i5eqðLFÞ ðtÞ ffi kT p4eqðLFÞ ðtÞ ¼ KSSðEQmÞ t þ b1 sinðxd tÞ c1 cosðxd tÞ

þ b2 sinð2xd tÞ c2 cosð2xd tÞ b3 sinð3xd tÞ þ c3 cos ð3xd tÞ
ð6:55Þ
where:
" #
p_ ES a1 pvES a31
KSSðEQmÞ ¼ k1 þ ð6:56Þ
2 16
and
" #
1 p_ ES a2 pvES a32
b1 ¼ k 1 þ ;
xd 2 16

1 €
pES a21
c1 ¼ k1 ;
xd 8
" #
1 p_ ES ða3 a1 Þ pvES 3 a33 4 a31
b2 ¼ k 1 þ ;
2 xd 2 48 ð6:57Þ
c2 ¼ 0;
1 8 p_ ES a2 þ pvES a32
b3 ¼ k 1 ;
3 xd 16

1 €ES a22 a21
p
c3 ¼ k1
3 xd 8
Thus, the harmonics of the i5(LF) signal are:

qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
I5ð1Þ ¼ b21 þ c21 ;
I5ð2Þ ¼ b22 þ c22 ¼ b2 ; ð6:58Þ
I5ð1Þ ¼ b23 þ c23
Consequently, besides the gradient and the gained dither, the iref(eq) signal has
three harmonics:
X
3
iref ðeqÞ ¼ KSSðEQmÞ t þ I5ðkÞ sinðkxd t þ /5ðkÞ Þ þ k2 A sinðxd tÞ ð6:59Þ
k¼1
The better approximation of the gradient by (6.56) in comparison with (6.37),

where only first derivative appears can be observed.
170 N. Bizon et al.
Fig. 6.19 Harmonics of the i5 signal computed based on (6.58) [26]
The i5(LF) harmonics will be estimated based on (6.58) using the power
derivatives computed at a given Iref value:
p_ ES ðtÞjIES ¼Iref ; €
pES ðtÞjIES ¼Iref ; pvES ðtÞ IES ¼Iref
ð6:60Þ
The shape of the i5(LF) harmonics is shown in Fig. 6.19.
Fig. 6.20 Harmonics of the i5 signal computed based on the simulation of the signal processing
in the ESC loop [26]
Also, the i5(LF) harmonics will be evaluated based on signal processing related to
ESC loop considering the reference current given by (6.45) as dither. The shapes of
i5(LF) harmonics are shown in Fig. 6.20. The minor differences appear close to the
MPP due to high distortion of the dither (which means more harmonics in the
simulated ESC loop.
The conclusion is that the searching speed is proportional to a linear combination
of the first and third derivative of the pES power, besides the ESC loop gain (k1) and
the dither amplitude (A). Consequently, the search speed is dependent to the pass
band set for the BPF of the mESC and aESC schemes.
The performance of the PVHPS under variable irradiance profile and constant
load will be analyzed below for both mESC and aESC schemes.
6.5 The PVHPS Using the ESC Schemes
6.5.1 Testing the mESC Scheme
The variable irradiance profile (■) of step-up stairs type is shown in Fig. 6.21.
The MPP searching process is also shown in P-I phase plane (see Fig. 6.21). The
simulation diagram was that from Fig. 6.6. The MPPs are tracked with a small
ripple by using the mESC scheme (see the zooms on Fig. 6.22).
The MPPs tracking is also shown in the first plot of Fig. 6.21, where the PV
power (●) has the same MPPs as in Fig. 6.22. The ESS voltage (■) and the ESS
power (●) are shown in the second plot of Fig. 6.21. The ESS power is given by
(6.2). So for a constant load of 50 W and variable PV power, the ESS will com-
pensate the power flow balance on the DC bus.
The power ripple is about 2 Wp-p (see zooms on Fig. 6.22), which means a
normalized ripple of about 2/51 ≅ 4%, but this can be reduce further if the aESC
scheme will be used instead of mESC scheme.
Fig. 6.21 The PVHPS behavior under a irradiance profile [22]

172 N. Bizon et al.
Fig. 6.22 The MPP searching process [22]
Fig. 6.23 The PVHPS behavior under aESC scheme and step up-down profile of the irradiance
[21]
6.5.2 Testing the aESC Scheme
The irradiance profile is step-up and -down type (see Fig. 6.23, plot 1). The fol-
lowing plots are shown in Fig. 6.23:
• Plot 1: the irradiance sequence ( ) and the PV current, iPV ( );
Table 6.2 Performance of the mESC and aESC schemes

Performance mESC scheme aESC scheme
indicator
Search speed Maximum 0.8 kW/s; higher Higher than kW/s; H1—times
(kW/s) search speed means lower tracking higher that mESC search speed; can
accuracy be set independently to the tracking
accuracy
Tracking Lower than 99.8% Higher than 99.99%
accuracy (%)
Normalized About 4% Negligible
p-p power
ripple (%)
Robustness High High
Control Simple A bit complex than mESC
complexity
Global MPP No Yes (see [33–36])
feature
• Plot 2: the PV power ( ) and the H1 magnitude of the first harmonic of the iBPF
signal( );
• Plot 3: the ESS power ( ) and the voltage on DC bus ( ).
It can be observed that the search speed under aESC scheme is H1-times higher
than the search speed under mESC scheme. The H1 magnitude increases quickly to
any variation in the irradiance level (see Fig. 6.23, plot 2). The power ripple is also
negligible during the stationary phase, after the MPP is located (see Fig. 6.23, plot
2). Consequently, the performance of the aESC scheme is better than that of the
mESS scheme.
The ESS will operate in charge depletion mode to assure the power flow balance
on the DC bus when the PV power is lower than the load (see Fig. 6.23, plot 3—the
period when the irradiance level is lower that 500 W/m2). The DC bus remain
almost constant around 50 V set by the ESS battery voltage (see Fig. 6.23, plot 3).
The PVHPS behavior under the aESC scheme was tested in Matlab-Simulink®
for different irradiance and load profiles. The following performance may be high-
lighted (see Table 6.2): (1) the robust tracking of the MPP under irradiance profile as
in reality; (2) the search speed could be set higher than 1 kW/s; (3) the tracking
accuracy is higher than 99.99%; (4) a negligible ripple during the stationary phase.
6.6 Conclusion
In this chapter some analytical results related to the new ESC schemes are high-
lighted. The mESC scheme based on BPF was proposed to improve the search
speed: this is about twice in comparison with that obtained with the hoESC scheme.
174 N. Bizon et al.
Anyway, the PV power ripple remains the same for the both mESC and hoESC
schemes analyzed. The PV power ripple can be reduced by using the aESC scheme
that has the following performance (see Sect. 6.4.5): (1) the tracking accuracy is
higher than 99.99%, (2) the PV power ripple is negligible during the stationary
phase after the MPP is located; the aESC search speed is H1—times higher that
mESC search speed; (3) the aESC search speed can be set independently (up to a
safe value of PVHPS operation) in order to obtain high tracking accuracy during the
transitory phase as well; (4) the aESC-based MPPT control is robust to irradiance
profiles such as in reality; (5) the aESC-based MPPT control is simple to be
implemented.
In conclusion, the aESC scheme slightly outperforms the mESC scheme in total
power efficiency of the PVHPS.
Acknowledgements The research that led to the results shown here has received funding from
the project “Cost-Efficient Data Collection for Smart Grid and Revenue Assurance (CERA-SG)”,
ID: 77594, 2016-19, ERA-Net Smart Grids Plus. Some figures, tables and text are reproduced from
[21, 22, 26] here with kind permission from Elsevier Limited, UK, WCST, UK, and APPEL, CZ
[September 13, 2016].
References
1. Salas V, Olías E, Barrado A, Lázaro A (2016) Review of the maximum power point tracking
algorithms for stand-alone photovoltaic systems. Sol Energ Mat Sol C 90(11):1555–1578
2. Azar FE, Perrier M, Srinivasan B (2011) A global optimization method based on multi–unit
extremum–seeking for scalar nonlinear systems. Comput Chem Eng 35:456–463
3. Tan Y, Nesic D, Mareels I (2008) On the choice of dither in extremum seeking systems: a
case study. Automatica 44:1446–1450
4. Esram T, Chapman PL (2007) Comparison of photovoltaic array maximum power point
tracking techniques. IEEE Trans Energ Conv 22(2):439–449
5. Bizon N, Radut M, Oproescu M (2015) Energy control strategies for the Fuel Cell Hybrid
Power Source under unknown load profile. Energy 86:31–41
6. Bizon N (2014) Load-following mode control of a standalone renewable/fuel cell hybrid
power source. Energ Convers Manage 77:763–772
7. Kuperman A, Aharon I (2011) Battery–ultracapacitor hybrids for pulsed current loads: a
review. Renew Sust Energ Rev 15:981–992
8. Mehrdad E, Yimin G, Ali E (2010) Modern electric, hybrid electric, and fuel cell vehicles.
CRC Press (15)
9. Bizon N (2012) Energy efficiency of multiport power converters used in plug-in/v2g fuel cell
vehicles. Appl Energ 96:431–443
10. Bizon N (2012) Distributed generation systems integrating renewable energy resources. Nova
Science Publishers Inc., New York
11. Feroldi D, Serra M, Riera J (2009) Energy Management Strategies based on efficiency map
for Fuel Cell Hybrid Vehicles. J Pow Sour 190:387–401
12. Bizon N, Oproescu M, Raceanu M (2015) Efficient energy control strategies for a standalone
renewable/fuel cell hybrid power source. Energ Convers Manage 77:768–772
13. Bizon N (2014) Improving the PEMFC energy efficiency by optimizing the fuelling rates
based on extremum seeking algorithm. Int J Hydrogen Energ 39(20):10641–10654
14. Patel H, Agarwal V (2009) MPPT scheme for a PV-fed single-phase single-stage
grid-connected inverter operating in CCM with only one current sensor. IEEE Trans Energ
Convers 24:256–263
15. Wang J-C, Su Y-L, Shieh J-C, Jiang J-A (2011) High-accuracy maximum power point
estimation for photovoltaic arrays. Sol Energ Mat Sol C 95:843–851
16. Toledo FJ, Blanes JM, Garrigós A, Martínez JA (2012) Analytical resolution of the electrical
four-parameters model of a photovoltaic module using small perturbation around the
operating point. Renew Energ 43:83–89
17. Kimball JW, Krein PT (2008) Discrete-time ripple correlation control for maximum power
point tracking. IEEE Trans Power Electron 23:2353–2362
18. Leyva R, Artillan P, Cabal C, Estibals B, Alonso C (2011) Dynamic performance of
maximum power point tracking circuits using sinusoidal extremum seeking control for
photovoltaic generation. Int J Electron 98(4):529–542
19. Reisi AR, Moradi MH, Jamasb S (2013) Classification and comparison of maximum power
point tracking techniques for photovoltaic system: a review. Renew Sust Energ Rev 19:433–443
20. Qiang M, Mingwei S, Liying L, Guerrero JMA (2011) Novel improved variable step- size
incremental-resistance MPPT method for PV systems. IEEE Trans Ind Electron 58:2427–2434
21. Bizon N (2013) Energy harvesting from the PV Hybrid Power Source. Energy 52:297–307
22. Bizon N, Raducu M, Oproescu M, Constantinescu LM (2014) Energy efficiency of the PV
panels using a MPPT controller with improved search speed—Part I & II. In: APPEL’14—
international conference on applied electronics. http://ieeexplore.ieee.org/xpl/articleDetails.
jsp?arnumber=7011666&sortType%3Dasc_p_Sequence%26filter%3DAND%28p_IS_
Number%3A7011649%29
23. Krsti¢ M (2000) Performance improvement and limitations in extremum seeking. Syst
Control Lett 39(5):313–326
24. Krsti¢ M, Wang H-H (2000) Design and stability analysis of extremum seeking feedback for
general nonlinear systems. Automatica 36(2):595–601
25. Ariyur KB, Krsti¢ M (2003) Real-time optimization by extremum-seeking control.
Wiley-interscience edition, Wiley Inc.
26. Bizon N, Oproescu M, Raducu M (2012) On the dither persistence in the extremum seeking
control—Part I & II, WCST-2012—World Congress on Sustainable Technologies, pp 104–
114. http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6482927&url=http%3A%2F%
2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6482927
27. Harada K, Zhao G (1993) Controlled power interface between solar cells and AC source.
IEEE Trans Power Electron 8:654–662
28. Subudhi B, Pradhan R (2013) A comparative study on maximum power point tracking
techniques for photovoltaic power systems. IEEE Trans Sustain Energ 4:89–98
29. Brunton SL, Rowley CW, Kulkarni SR, Clarkson C (2010) Maximum power point tracking
for photovoltaic optimization using ripple-based extremum seeking control. IEEE T Power
Electr 25(10):2531–2540
30. Midya P, Krein PT, Turnbull R (1996) Dynamic maximum power point tracker for
photovoltaic applications. In: IEEE 27th annual power electronics specialists conference,
pp 1710–1716
31. Gelbert G, Moeck JP, Paschereit CO, King R (2012) Advanced algorithms for
gradient estimation in one- and two-parameter extremum seeking controllers. J Process
Contr 22:700–709
176 N. Bizon et al.
32. Bizon N (2010) On tracking robustness in adaptive extremum seeking control of the fuel cell
power plants. Appl Energ 87(10):3115–3130
33. Bizon N (2013) Energy harvesting from the FC stack that operates using the MPP tracking
based on modified extremum seeking control. Appl Energ 104:326–336
34. Lim YH, Hamill D (2000) Simple maximum power point tracker for photovoltaic arrays.
Electron Lett 36(11):997–999
35. Bizon N (2013) FC energy harvesting using the MPP tracking based on advanced extremum
seeking control. Int J Hydrogen Energ 38(4):1952–1966
36. Bizon N (2015) Global maximum power point tracking based on new extremum seeking
control scheme, progress in photovoltaics: research and applications. http://onlinelibrary.
wiley.com/doi/10.1002/pip.2700/full
Part II
Energy Efficiency of the Photovoltaic
Systems
Chapter 7
Improving Tracking Efficiency
of Two-Axis Sun Tracking Systems
Fevzi Kentli and Musa Yilmaz
Abstract Due to both reduction and insufficient of fossil fuel to supply current
growing energy needs, investigation and employing of renewable resources has
been accelerated. Besides, using fossil fuel affected the environment negatively.
Therefore, renewable energy resources in the most studies are solar, wind and
geothermal. In this study, electrical energy production methods from solar energy
have been examined, a fixed and a two axis tracking system have been designed.
Both systems are compared each other regarding to several factors by performing
annual measurements. Energy consumption of the system is minimized by
employing actuator motor in two axis solar tracking system. According to the
efficiency of two-axis tracking system, the annual average has been calculated as
31.67% more. This efficiency has been calculated as 70% in winter, 11% in
summer. As a result of these measurements several graphics of a year empirically
daily, monthly and annual data have been contributed to the literature for
Diyarbakir, one of the prominent cities of Southern east, having the most solar
energy of Turkey. In the first section, literature review will be indicated. In the
second section, solar angles, photovoltaic panels and systems, sample designs and
solar tracking systems are examined. In the third section, photovoltaic two-axis
solar tracking system and qualifications, work and advantages of fixed system
which we designed are stated. In following section, obtained results will be given
and in last section, financial analysis of fixed and tracking photovoltaic systems has
been performed. Also, recommendations for increasing their efficiency have been
noted.
Keywords Solar energy Photovoltaic Two-axis sun tracking

F. Kentli (&)
Department of Mechatronics Engineering, Technology Faculty,
Marmara University, 34722 Istanbul, Turkey
e-mail: fkentli@marmara.edu.tr
M. Yilmaz
Department of Electronics and Communications Engineering, Technology Faculty,
Batman University, Batman, Turkey
e-mail: musa.yilmaz@batman.edu.tr

180 F. Kentli and M. Yilmaz
7.1 Introduction
Importance of energy needed today in every area of life becomes increasing. Energy
is the main input for our civilization and its production and consumption is one of
the reliable merits to measure the improvement and development of the country.
Studies on finding and developing alternative energy resources have been
accelerated as a result of insufficiency of existing resources to provide constantly
increasing energy need. Yet, traditional energy production and consumption
methods, especially fossil fuel based ones, lead to unrecoverable environmental
damages. Saving the living environment on earth and decreasing the environmental
damage caused by energy production and consumption beneath the damage caused
by climate change give responsibility to all humankind. As a result of this
responsibility; preparing national and international legal regulations, prioritizing the
environmental effects in selection of energy production technology and resources
and giving maximum effort in efficiently using energy are gaining much more
importance by the time.
As a consequence, instead of using traditional and fossil fuel based energy
production systems which are causing worldwide environmental pollution and
climate change, finding energy resources and developing new technologies pro-
viding sustainability and renewability and causing less environmental damage and
climate change became necessary. Both of the facts that fossil fuel resources are
limited and their production and consumption systems cause environmental damage
led to investigate renewable, unlimited and environmental friendly energy resources
and technologies. So, renewable energy resources including solar energy gained
importance [1, 2].
Solar energy with the characteristics such as its potential, cleanliness, renewa-
bility, easy-to-use and being environmental friendly features has much more
probability to become widespread when compared with other renewable energy
resources. Overcoming it limitations such as having higher setup cost, having less
efficiency and capacity will make it more attractive at the future. Unfortunately,
Turkey could not use today all its potential even though it has a favorable position
from solar energy perspective.
Solar energy is vast and also does not damage environment. This makes it a
good choice in this consumption era as most known energy sources have becoming
exhausted and some complications are occurred such as corrosion in living world
and climate changes. However, there are different opportunities which allow us fix
the situation. One of these opportunities and also the most convenient one is to find
out the new and clean energy sources. Another opportunity is to reduce the amount
of consumed energy.
Nuclear energy has first considered as sharp cut solution but having high hazard
risk and changing nearby habitat made it less preferable. On the other hand, sun has
much more potential as it has not such kind of disadvantages. Available energy on
7 Improving Tracking Efficiency of Two-Axis Sun Tracking Systems 181
Fig. 7.1 World energy consumption and production status
the world produced by sun is in 152,424 1013 kWh while total energy produced
by power plants is 100 thousand times less and produced solar energy on earth is 15
thousand times less as seen in Fig. 7.1. Thus, today solar energy systems have
begun to be used in every part of life even though solar energy is intended to use
after the second half of our century [1–12].
At the present time, utilization of solar energy becomes significant in almost
every country as there are many different application areas. It is used in providing
hot water or heating house. Also, it is converted into electrical energy and then used
in lighting of buildings, for propulsion of cars and even for assisting to grow crops
by the help of solar water pumps. In last 20 years, new technologies are improved
on solar panels and studies are focused on making them cheaper [2, 3].
Optimal usage of photovoltaic (PV) panels for different cases is an important
problem because the efficiency of the panels is changing as the weather changes. To
solve this problem, panels are used in rural area or away from city centers where
there are plenty of empty spaces letting get much more sunlight onto panel surface.
But even in these conditions it is needed to maximize obtained solar energy by
determining appropriate declination angle. That is why so many studies on decli-
nation angle have been done [1, 2].
Output power of any photovoltaic system depends on amount of obtained solar
irradiance. To provide more sunlight as an input to system, it is required to track the
sun. Main concern of these systems is first cost of system and cost of consumed
energy during sun tracking. In this study, consumed energy of the system during
sun tracking is minimized and it is compared with the other studies [11–15].
Applications brought out different sun tracking systems [16–20].
7.2 Maximizing Sunlight on Panels
Easiest way to maximize sunlight in most system is to mount panels onto a hori-
zontal tracking system. Sun takes highest place in sky according to horizontal axis
at noon, so it is the time that sun has minimum distance in atmosphere and air has
the minimum density during day. At other times, incident radiation flux becomes
less since mass density is increasing and the angle between sunlight and normal of
panel surface is increasing. Consequently, tracking at two axes is needed to increase
the efficiency by decreasing this angle.
In summer, even with a dry weather, approximately 12% more energy could be
produced by using dynamic photovoltaic panels. During winter, 70% more energy
and as an annual average 31% more energy could be produced. Figure 7.2 shows the
comparison of stationary and dynamic photovoltaic panels at different months [2].
To select the type of PV system (single axis or two axes) is a difficult decision as
two-axis systems are much more expensive even though they are much more
efficient. Slope will be changed during the year if it is decided to use one-axis
tracking system. Inclination of panel changes depending one semester. For exam-
ple, inclination of the system used during summer is different than others (Fig. 7.2).
Fig. 7.2 Comparison of fixed and dynamic photovoltaic panels at different months for Diyarbakır
(Turkey) in 2012 [2] [Gain = (Powertracking – Powerfixed) * 100/Powertracking]
7.3 Sun Tracking Systems
Solar cells produce electrical energy using photon energy of sunlight. Output
electrical energy is directly related with the amount of sunlight hitting onto the
panel surface. Sunlight reaches earth with different angles during the day. So, to
obtain maximum electrical energy by solar cells, PV systems should track the sun.
This system aiming to maximize sunlight by tracking sun is called Sun Tracking
System. Sun tracking systems are designed to track through West-East direction
(one axis), West-East and North-South directions (two-axis) [3, 4]. Main aim of the
all studies on photovoltaics is to decrease payback time of the system. It becomes
possible only by increasing its efficiency. And efficiency could only be increased by
increasing amount of solar radiation hitting on to the panel surface (Fig. 7.3).
Amount of radiation hitting on to panel surface is related with cosine of the angle
between the sunlight and the normal of the surface. It means that efficiency is
maximized when sunlight goes through the normal of the panel surface. This sit-
uation brings the need of a system to provide sunlight always normal to the panel
surface by tracking the sun.
Finsten has developed the first tracker in 1962. Later Saavedra has added an
automatic control unit to this mechanical system in 1963. This unit inspired to build
Eppley pyrheliometer. Maldonado designed and constructed a sun tracker system at
Technical University Federico Santa Maria [5, 6].
Main function of all tracking systems is to provide 1 or 2 degree of freedom to
motion. In other words, main function is to supply sunlight at desired direction.
There are many applications of single (one) axis tracking systems (SAT) as they
provide power gain: Horizontal (HSAT), Vertical (VSAT), Tilted (TSAT) and Polar
Aligned (PASAT). But studies on these systems are seen rarely in literature except
one study which has concerned these systems. It is mentioned in this study that
TSAT with 5 has collected 10% more radiation than HSAT where HSAT 15% more
than VSAT. Also, it is found that PASAT collected 10% more than HSAT [3, 4].
Sun tracking systems are classified considering number of degree of freedom,
position of turning axis or the systematic for sun tracking.
Fig. 7.3 Radiation hitting on to the panel (To provide maximum sunlight, h should be equal to
zero)
Fig. 7.4 One axis tracker

system
7.3.1 Classification Considering Number of Degree

of Freedom
One Axis Sun Tracking, Two Axis Sun Tracking.

One axis sun tracking systems: One axis tracking systems performs movement in
either azimuth or vertical axis (Fig. 7.4). Reason to prefer single axis tracking
system is to have free area where it is mounted [11]. For example, parabolic trough
systems tracks using azimuth and most of the roof application of photovoltaic
systems track using only height angle due to limited space. SAT is direct and
cheaper way to get better efficiency. By using a simpler system, efficiency is
increased. RayTracker, a tracking system manufacturer, implied that they improved
the performance 23 and 38% when compared respectively with fixed angle modules
and outer flat modules.
Two axis tracking systems: Two-axis tracking systems let movement in two axes
(vertical and azimuth) which maximizes the efficiency of the tracking process.
But SAT developers are questioning the efficiency as two axis tracking systems
have additional installation, permitting and ongoing maintenance costs. Also higher
defect risk related to having more moving parts compared to SATs is another
concern. Moreover, two-axis systems need bigger spaces to place and consequently,
to receive planning permits is much more difficult for them [3, 4].
7.3.2 Classification Considering Position of Turning Axis
Sun tracking systems are also classified considering joint axis. One axis tracking
systems are classified as polar (Fig. 7.5), azimuth (Fig. 7.6) and horizontal
(Fig. 7.7). On the other hand, two axis tracking systems are divided into two
groups: azimuth and polar (Fig. 7.8).
Fig. 7.5 Polar one axis

tracking system
Fig. 7.6 Azimuth one axis

tracking system
Panel surface of polar one axis tracking systems is tilted through south direction
and azimuth angle is tracked by turning around the axis goes through weight center
of panel surface. While panel is tracking azimuth angle, it also tracks height angle
in the meantime due to its geometry. Thus, polar one axis tracking systems have
higher performance values when compared with other one axis tracking systems
from energy production perspective. Azimuth one axis tracking systems (Fig. 7.6),
turns on a structure with tilted panels around the axis normal to ground. This kind
of one axis tracking systems are preferred when there are big areas for the panels.
Horizontal one axis tracking systems turns on an axis parallel to ground
(Fig. 7.7). This kind of tracking systems is generally used where height angle is
approximately 90° like equatorial region (0°–±30°). Structure is compatible to
move as groups. One motor could run many photovoltaic arrays at the same time.
On the other hand, two axis systems are separated as polar and azimuth angles.
Polar two axis systems only completes turning around its axis in a day. During the
year it moves complying with sun movements and the issuing the commands comes
from control system or sensor by the time.
On the other hand, two axis azimuth moves both around its axis and along two
axis during a day and so its control is much more complex (Fig. 7.8).
Fig. 7.7 Horizontal one axis tracking system
Fig. 7.8 Polar and Azimuth

two axis tracking system
Setup and maintenance costs of sun tracking systems are higher than the fixed
systems having certain angle. Thus, this kind of systems needs being robust, less
maintenance required and high capacity to redeem their costs in long term. Also,
this kind of systems should be built at rural areas and at the places where sunlight
will be present from sunrise to sunset as the performance of sun tracking systems
will be lowered in city centers and in the places where intense building shadows
occurred (Fig. 7.9). At these places, costs related with sun tracking system and its
sensors will be decreased as sun tracking will be able to be accomplished from one
center for whole system. Any battery will not be needed as the system will be on
grid and its cost will be removed.
7.3.3 Classification Considering Sun Tracking Systematic
Generally control circuits consisting of electronic sensor systems, controller,

microprocessor and GPS are used to position the system to where it is required to be.
Fig. 7.9 A high capacity

solar farm tracking sun in two
axes
Passive sensor tracking: Some companies have tried to develop more energy
efficient and cheaper tracking systems. For example, Portasol tracking systems—
Australia—developed a hydraulic passive tracking system working according to
thermal changes during the day to track the sun. Company states that electrical
energy is not needed as tracking system does not include any electronic equipment,
motor or gearbox. Figure 7.10 shows (a) panel stopped at night through west, at
morning sun heats more the bottom part of tube and liquid expands through top,
(b) panel turns until sun heats both tube equally, (c) panel tracks the sun with a
certain angle by the time sun moves, (d) as a last step panel stopped at night through
west [3, 4].
First decision that should be taken is to choose whether passive or active
tracking system will be used. While low boiling point of compressed gas is used to
move in passive tracking systems, motor, gearbox or hydraulic components are
used in active tracking systems and they consume energy. Passive systems using
gas pressure produced by sun does not consume energy but they are more insuf-
ficient to track the sun [3, 4].
Active sensor tracking: Two photoresistances and photovoltaic cell are used as a
sensor for one axis tracking systems. These sensors place next to each other. So,
both sensors are at same conditions. Sensor are directly connected to motor, gen-
erally direct current (DC) motors are used. The angle between panel surface and
sunlight changes when sun moves. This leads to occur a difference between sensor
outputs. Control circuit runs the motor and turns the system considering where
more sunlight has come. Control circuit adjusts as panel surface is always normal to
sunlight using this difference. The example by Poulek and Libra claimed their
system was able to collect 95% of the energy with a ±5° tolerance. Figure 7.11
shows a sample structure [3, 4].
Chronological tracking: Systematic of chronological tracking is based on deter-
mining the movement positions before for every day during the year. Calculations
are done for the related geographical region. System includes a pressor
Fig. 7.10 Running of passive sensor tracking system. a Morning b Before noon c After noon
d Night
Fig. 7.11 Tracker sensor setups from left to right: divider, tilted mount, and collimator [4]
(microprocessor, PLC, computer etc.) making calculations. Latitude and longitude

values of the region are given as constant input to the processor. Variable time
value is taken from Real Time Clock. For the current date and time, position of sun
is calculated and system is turned as panel surface is normal to sunlight.

Disadvantage of this system is that system continues to track and consume energy
even during overcast sky.
7.4 Application
In this section, design and comparison of fixed and two axis photovoltaic panel
system applications are given. Energy consumption is an important factor in two
axis systems differing from one axis systems. It affects the system efficiency. To
minimize this effect, literature is investigated and a low energy consumed DC motor
having gearbox (Damper motor) is selected to use. Both used motors are identical.
To be able to compare rightfully, both systems are used in same circumstances and
panels having same trademark and model are used. Also, disturbing environmental
factors (shadows, reflections etc.) are minimized as much as possible. Thus, a
location where sunlight reached the panel all day during the whole year is chosen.
7.4.1 Fixed Photovoltaic Panel Systems
Widely used systems are fixed panel systems having low efficiency. Block scheme
of fixed system is shown in Fig. 7.12 and picture of the used system is shown in
Fig. 7.13. This system contains panel, measurement system, battery, inverter and
load. Photovoltaic panel is placed as fixed considering geographical and meteo-
rological data. Obtained power from panels are stored in batteries optimally.
Fig. 7.12 Block scheme of fixed photovoltaic panel system

Fig. 7.13 Picture of the fixed photovoltaic panel system
Fig. 7.14 Picture of the used two-axis dynamic system [2]
7.4.2 Two Axis Photovoltaic Panel System

(Sun Tracking System)
In this section, proposed two axis PV (sun tracking) system will be detailed.
Consumed energy by the system has a higher priority in designing dynamic systems
as it is related with the system performance. To minimize the consumed energy,
literature is investigated and a new motor type (DC motor with gear unit) is found
as a solution. Figure 7.14 shows the application of proposed system.
This dynamic system moves in two axes. East-West direction is defined as
horizontal axis and North-South direction is defined as vertical axis. System is able
to move in both axes at the same time as shown in Figs. 7.14 and 7.15. Two roller
bearings are used as supports in the system. Wind load data of the last 10 years for
Fig. 7.15 Improved two-axis

sun tracking system [2]
the application area is acquired from meteorology are so low that no need to be
taken into account in design stage eventhough system is capable of resisting the
loads over average [1].
Designed system consists of 2 Damper Motor (DM), 4 gel batteries (12 V–
100 Ah), 2 PV panel (120 W), power inverter, solar battery charger, Maximum
Power Point Tracker (MPPT), data acquisition card, Light-Dependent Resistors
(LDR)s and mechanical system (Fig. 7.16).
Required energy to track the sun is provided from batteries. System tracks sun
continuously at vertical and horizontal axis in sunny weather conditions. System is
stopped by a LDR in cloudy weather conditions and at nights. Produced energy just
after sun rise is enough to start the motor even if all energy in batteries is consumed.
Because total need power for panels is 240 W while it is only 12 W for motors.
Two axes photovoltaic system contains solar panel moving through North-South
and East-West directions. Solar panel is moved by the help of motor tracking sun by
photoresistances (LDRs) and control system. So, it is profited from sun efficiently.
Efficiency of photovoltaic panel is maximized when sunlight comes normal to panel
surface. For this reason, control circuits need to command motor at certain points to
provide panel angle as 90° to sunlight. This task is accomplished by control circuit
by the help of photoresistances. Control system has electronic and electrome-
chanical components. Suns movement is determined by 5 LDRs in control circuit to
provide being normal to surface (Fig. 7.17). Also, time delay relays arranged with
3 s time difference are used to prevent PCBs work at the same time. 2 LDRs are
used to measure movement in horizontal axis (East-West). 2 LDRs are used to
measure movement in vertical axis (North-South) and one is to know whether being
Fig. 7.16 Schematic view of two axis photovoltaic system (sun tracking system)
Fig. 7.17 Picture of the used control unit
sunny or not. Scheme of the sun checker PCB is shown in Fig. 7.18. So, damper
motors are able to move panel to provide being normal to surface [1]. To determine
the position of sun, several variables (irradiance, declination angle, azimuth angle,
zenith angle etc.) should be defined or calculated. Control unit calculates these
variables and then adjust the position of the system. Calculation of some important
variables will be given in next sections.
Fig. 7.18 Control circuit of sun checker PCB
Fig. 7.19 Mechanical model

of two-axis dynamic system
[2, 10]
7.4.3 Mathematical Modeling of Two Axis Photovoltaic

System
This system tracks in two axis. Even horizontal (East-west) and vertical
(North-south) axis seems together, they are moving independently. System freely
moves in horizontal and vertical as shown in Fig. 7.19. Two roller bearings ease
movement in both axes. Wind load is omitted as it is seen low in Diyarbakır’s last
10 year data (<3.5 m/s). But system is designed for overloads. Mathematical
modelling of system for horizontal and vertical axis is done as seen below [2, 10].
In modeling the system, Eqs. (7.1)–(7.5) (derived for system shown in Fig. 7.20)
are used for horizontal (East-West) and vertical (North-South) axis [2, 10].
Fig. 7.20 Electric circuit of

DC motor for horizontal and
vertical axis [2, 10]
For vertical (North-South) axis of the system;
diva
evertical
a ¼ iva R1 þ L1 þ evb ð7:1Þ
dt
dhvm
evertical
b ¼ Kbv ð7:2Þ
dt
Torque constant ! KvT

Electromotive force constant ! Kvb
diva dhv
eva ¼ iva R1 þ L1 þ Kbv m ð7:3Þ
dt dt
Mathematical modeling of vertical axis of mechanical systems is (Fig. 7.19);

Bvt ¼ Bvm1 þ Bvm2 ! Viscous damper coefficient of bearing
d 2 h2m v dhm
v
Tmv ¼ Jtv þ Bt ð7:4Þ
dt2 dt
Tmv ¼ KTv :iva ð7:5Þ
Jvt ¼ 1. Equivalent moment of inertia for torque of motors (kg mm2)
d 2 h2m v dhm
v
KTv iva ¼ Jtv þ B t ð7:6Þ
dt2 dt
Laplace transform of equation;
Eav ðsÞ ¼ Iav ðsÞR1 þ L1 sIav ðsÞ þ Kbv shvm ðsÞ ð7:7Þ
KTv Iav ðsÞ ¼ Jtv s2 hvm þ Bvt shvm ðsÞ ð7:8Þ

Equation (7.7) is shortened as;
Eav ðsÞ Kbv shvm ðsÞ

Iav ðsÞ ¼ ð7:9Þ
R1 þ L1 s
When Eq. (7.9) is substituted into Eq. (7.8);
Eav ðsÞ Kbv shvm ðsÞ

KTv ð Þ ¼ Jtv s2 hvm ðsÞ þ Bvt shvm ðsÞ ð7:10Þ
R 1 þ L1 s
KTv Eav ðsÞ KTv Kbv shvm ðsÞ ¼ ½Jtv s2 hvm ðsÞ þ Bvt shvm ðsÞðR1 þ L1 sÞ ð7:11Þ
KTv Eav ðsÞ ¼ Jtv s2 hvm ðsÞR1 þ Jtv s3 L1 hvm ðsÞ þ Bvt sR1 hvm ðsÞ
ð7:12Þ
þ Bvt s2 L1 hvm ðsÞ þ KTv Kbv shvm ðsÞ
KTv Eav ðsÞ ¼ hvm ðsÞ½Jtv s3 L1 þ ðJtv R1 þ Bvt L1 Þs2 þ ðBvt R1 þ KTv Kbv Þs ð7:13Þ
hvm ðsÞ KTv

¼ v 3 v ð7:14Þ
Ea ðsÞ Jt L1 s ðJt R1 þ Bt L1 Þs2 þ ðBvt R1 þ KTv Kbv Þs
v v
It is not needed to derive the equations for horizontal axis because both axes have
same situation [2–10].
7.5 Modeling of Sun Tracking
As same principles are valid for both axes, modelling for one axis will be enough in
designing system. Working principle of system is shown in Fig. 7.21. Four LDRs
helped to direct panel into the right direction by giving correct angle (between
normal and coming sunlight). Location of LDRs are adjusted as every LDR has the
same amount of sunlight when sunlight comes as normal to panel surface. The
current produced by LDRs are amplified by an operational amplifier and measured
change in current is accepted as error which shows that sunlight doesn’t come to
surface normal. Then control unit drives the motor to turn the system into position
here error value becomes zero.
Coordinate System: Center of coordinate system is assumed to be on output gear.
When reference axis is accepted as the stationary body of DC motor, all rotations
are measured around this axis. Sun axis (line connecting center of output gear with
sun) states hr(t) reference angle (Fig. 7.22). Purpose of control system is to mini-
mize the angle difference, a(t), between Bvt and h0(t) where h0(t) represents the
angle between reference axis and tracking axis [10].
Fig. 7.21 Working principle scheme of improved two-axis solar tracking system [2, 10]
aðtÞ ¼ hr ðtÞ h0 ðtÞ ð7:15Þ
Error Detector: When tracking axis overlaps with sun axis: a(t) = 0 and
ia(t) = ib(t) = I or ia(t) – ib(t) = 0. As shown in Fig. 7.21, sunlight having W width
reaches at certain a angle to A cell with oa width and to B cell with ob width [10].
Fig. 7.22 Coordinate system

of sun tracking system
Fig. 7.23 Nonlinear characteristic of error detector
W
oa ¼ L tan aðtÞ ð7:16Þ
2
W
ob ¼ L tan aðtÞ ð7:17Þ
2
On the other hand, when 2I is the maximum current of a LDR and 0 tan
a(t) W/2L as ia(t) current is related with oa and ib(t) current is related with ob;
2LI
ia ðtÞ ¼ I þ þ tan aðtÞ ð7:18Þ
W
2LI
ib ðtÞ ¼ I þ tan aðtÞ ð7:19Þ
W
Sunlight goes only onto A cell (LDR1) and ia(t) = 2I, ib(t) = 0 when W/2L tan
a(t) (C − W/2)L. When (C − W/2)L tan a(t) (C + W/2)L, ia(t) current
goes through zero from 2I. When tan a(t) (C + W/2)L, ia(t) − ib(t) = 0.
Nonlinear characteristic shown in Fig. 7.23 can be used to define error detector. As
a(t) angle is very small, a(t) is used instead of tan a(t) [10].
Operational Amplifier: There is following relation between ia(t) and ib(t) and
operational amplifier:
eo ðtÞ ¼ RF ½ia ðtÞ ib ðtÞ ð7:20Þ
Servo Amplifier: Gain of servo amplifier is –K. Output of servo amplifier as

shown in Fig. 7.21 could be written as;
ea ðtÞ ¼ Keo ðtÞ ð7:21Þ
Damper Motor: The systems using solar panels produce DC power and it is
preferred to use DC motors and devices working with DC power in the system. As a
matter of fact that an inverter will be needed for alternative current (AC) systems
and both cost of inverter and losses of inverter will decrease the efficiency of
system.
Damper motors are the AC motor types which can easily control under varying
loads. Easiness to control and to start running at desired speed makes DC motors
preferred in tracking systems. Moment could be changed by changing armature
current and speed could be changed by changing voltage in motors by the help of
electronic circuits. The motor used in the system is a brushed permanent magnet
DC motor. DM consists of DC motor and a coupled gear unit (having 15 gears).
Moment of DM is increased by the help of gearbox (16 Nm). For both axes, same
motor is used. Movements through North-South and East-West of panels in the
system are performed slowly considering movement of sun. Selected DC motor
with a gear unit is shown in Fig. 7.24 and its specifications are given in Table 7.1
[1, 2].
Damper motor consists of two components. First is gear unit and second is DC
motor. Angular position h0 of output gear is related with angular position of motor
hm and 1/n gear ratio.
1
ho ¼ hm ð7:22Þ
n
Fig. 7.24 Picture of used DC

motor with gearbox (Damper
motor)
Table 7.1 Specification of Parameter Unit Value

the motor used in the study
Voltage V 24 DA/220 AA
Current A 0, 5
Power VA 12
Moment Nm 16
°
Rotation angle 360
Rotation time s 320
Weight g 1200
Fig. 7.25 Block diagram of sun tracking system [2, 10]
While J and B represents inertia and friction coefficient of motor shaft of DC motor;
ea ðtÞ ¼ Ra ia ðtÞ þ eb ðtÞ ð7:23Þ
eb ðtÞ ¼ Kb xm ðtÞ ð7:24Þ
Tm ðtÞ ¼ Ki ia ðtÞ ð7:25Þ
dxm ðtÞ
Tm ðtÞ ¼ J þ Bxm ðtÞ ð7:26Þ
dt
Block diagram showing functional relations of system is given in Fig. 7.25. It is not
needed to derive equations for vertical axis as both axes has same situation.
7.6 Experimental Results
In this section, data obtained from fixed and two axis systems are compared.
Diyarbakır is chosen as the location where the area of highest solar radiation potential
in Turkey is and experiments are run on the roof of Engineering Faculty of Dicle
University in this city. Efficiency of the designed sun tracking system is observed
Fig. 7.26 Obtained output power of fixed and two axis systems during 2012 in Diyarbakır [2]
Fig. 7.27 Gain of two axis system during 2012 in Diyarbakır compared to fixed system [2]
[Gain = (Powertracking – Powerfixed) * 100/Powertracking]
under different weather conditions during the year [2]. To be able to compare the
designed tracking system with the fixed system placed at the same location, panels
having same trademark and model are used under same conditions. Comparison is
done for the whole year data. Also, measurements are compared with the meteoro-
logical data. These data belong to the interval between 01.01.2012 and 30.12.2012.
Measurements are done at each 10 min daily. Then, hourly and daily data are cal-
culated using minute measurement data and also monthly and annual data are derived
from daily data. Obtained data are shown in Figs. 7.26 and 7.27.
Energy consumption of two-axis sun tracking system is seen as low as not to
take into consideration to calculate. As seen in Fig. 7.28, the permanent magnet DC
motor of 12 W has a small structure and total consumed power during a day doesn’t
exceed 3 Wh. As sun tracking system stop at west position at sunset, at sunrise
system turns to east for 3 min and most of the energy consumption occurs at this
stage (0.6 Wh) as shown in Fig. 7.28. Then, until sunset energy is consumed at
small values (as an average 0.1 Wh) according to movement of sun at horizontal
and vertical axis.
Fig. 7.28 Daily energy consumption of the used motor
Fig. 7.29 Comparison of produced and consumed energies of tracking systems [2, 11–15]
Disadvantage of sun tracking systems is its cost and consumed energy during
tracking. Damper motor, used firstly in this system, has decreased both cost and
also the consumed energy during tracking. Power of used motor is equal to 12 W.
Daily power consumption even for the longest day (21 June) didn’t exceed 3 Wh.
DC motor is generally used in sun tracking system. Asynchronous motor is also
used in several studies [1–3]. Obtained power is converted to AC by the help of an
inverter in these studies. In this case, inverter losses decrease energy efficiency.
Comparison with previous studies has shown that the system using damper motor
consumed less energy as shown in Fig. 7.29 [1–4]. There are two reasons that
damper motor consumes less energy. First reason is to have less power (12 W) and
second reason is to have many gears (15 gears). To use these motors is recom-
mended in tracking systems as sun moves slowly.
7.7 Conclusions
Energy production from sun by using panels is related with their cost and efficiency.
Having high cost (1 W = 1$) and low efficiency (10–25%) of panels led to develop
sun tracking systems. The feasible way to reduce the cost and produce more energy
in a unit panel area is to use sun tracking systems. Even setup cost is higher than
fixed systems; they are more advantageous when produced energy and occupied
area is considered. Today it is inevitable to use these systems.
Efficiency of photovoltaic systems used to produce electrical energy directly
from sun energy is still low. Most difficult task for sun tracking system which is one
of the methods applied to increase efficiency is to minimize the consumed energy
and cost. Panel cost is much higher than design and consume energy costs of
tracking system. Also, need surface area will be low as power obtained from a unit
panel area is higher [2, 17].
Disadvantages of the sun tracking systems are setup cost and the consumed
energy during tracking. Using damper motor, contrary to general application, in
designed system and having more basic control circuit made possible to decrease
both cost and consumed energy during tracking. In the mentioned study 12 W
damper motor is used and daily consumed power is not reaching even 3 W/h during
the longest day (21 June). Design of control circuit is critical as it provides tracking
at both horizontal (West-East) and vertical (North-South) direction during the day.
Developed circuit stops the system at cloudy weather and night and so it helps
saving energy.
Obtained data showed that efficiency of newly designed two axis sun tracking
system is 70% higher during winter, 11% higher during summer and 31.67% higher
as an average during the year when compared with fixed system. Figure 7.2 shows
that fixed system is also producing high power during summer where sun is gen-
erous even two axis system is given better results. So, power difference is becoming
low during summer. But this difference changes during winter and efficiency of two
axis system becomes higher.
Another important factor for tracking system is how much cheaper and light
mechanical components are built. This factor affects directly payback time. New
studies on minimizing its effect will make tracking system cheaper and advanta-
geous and sun as a clean and renewable energy resource will continue to heat and
give energy to next generations.
Acknowledgements We would like to thank the Research Council of Scientific Projects

Coordinator—Dicle University.
References
1. Kentli F, Yilmaz M (2015) Mathematical modelling of two-axis photovoltaic system with

improved efficiency. Elektron Elektrotech 21:40–43
2. Yılmaz M (2013) Determination of methods deriving electrical energy from solar energy and
optimum efficiency by solar tracking system, PhD thesis, Department of Electrical Education,
Marmara University, Istanbul
3. Yilmaz M, Kentli F (2015) Increasing of electrical energy with solar tracking system at the
region which has Turkey’s most solar energy potential. J Clean Energy Technol 3:287–291
4. Catarius A, Christiner M (2010) Azimuth-altitude dual axis solar tracker, Graduate thesis, W.
P.I, Engineering, Massachusetts
5. Roth P, Georgiev A, Boudinov H (2005) Cheap two axis sun following device. Energy
Convers Manag 46:1179–1192
6. Samimi J, Soleimani EA, Zabihi MS (1997) Optimal sizing of photovoltaic systems in varied
climates. Sol Energy 60:97–107
7. Helwa NH, Bahgat ABG, El Shafee AMR, El Shenawy ET (2000) Maximum collectable solar
energy by different solar tracking systems. Energy Source 22:23–34
8. Akkilic K, Ocak YS, Yilmaz M (2015) Analysing enhancement of electricity generating
capacity with solar tracking system of the most sunning region of Turkey. J Clean Energy
Technol 3:291–295
9. Appleyard D (2012) Solar trackers: facing the sun, renewable energy world
10. Kho CB (2002) Otomatik Kontrol Sistemleri, Yedinci Baskı, Literatür Yayıncılık, İstanbul,
pp 207–209
11. Drury E, Lopez A, Denholm P, Margolis R (2014) Relative performance of tracking versus
fixed tilt photovoltaic systems in the USA. Prog Photovolt Res 22:1302–1315
12. Şenpinar A, Cebeci M (2012) Evaluation of power output for fixed and two-axis tracking
PVarrays. Appl Energy 92:677–685
13. Mecasolar Products Inc. (2014) Mecasolar two-axis 60P, Spain
14. Huang BJ, Ding WL, Huang YC (2011) Long-term field test of solar PV power generation
using one-axis 3-position sun tracker. Sol Energy 85:1935–1944
15. Salsabila A, Suhaidi S, Mohd ZAAK (2013) Power feasibility of a low power consumption
solar tracker. Procedia Environ Sci 17:494–502
16. Tarabsheh A, Etier I, Nimrat A (2012) Energy yield of tracking PV systems in Jordan. Int J
Photoenergy 2012:1–5
17. Burduhos BG, Visa I, Neagoe M, Badea M (2015) Modeling and optimization of the global
solar irradiance collecting efficiency. Int J Green Energy 12:743–755
18. Lee CY, Chou PC, Chiang CM, Lin CF (2009) Sun tracking systems: a review. Sensors
9:3875–3890
19. Reisi AR, Moradi MH, Jamas S (2013) Classification and comparison of maximum power
point tracking techniques for photovoltaic system: a review. Renew Sust Energy Rev
19:433–443
20. Mousazadeh H, Keyhani A, Javadi A, Mobli H, Abrinia K, Sharifi A (2009) A review of
principle and sun-tracking methods for maximizing solar systems output. Renew Sust Energy
Rev 13:1800–1818
Chapter 8
Maximum Power Point Tracking (MPPT)
Algorithms for Photovoltaic Systems
Ersan Kabalci
Abstract The solar energy have become a challenging area among other renewable
energy sources (RESs) since the photovoltaic (PV) systems have the advantages of
not causing pollution, having low maintenance, and long-lasting operation life.
Besides these advantages, a PV system has several drawbacks such as considerably
higher installation cost comparing some other RESs, and limited efficiency ranges
between 9–18%. The feasibility analyses have a great role in order to determine the
most appropriate plant site before installation. On the other hand, the operating
analyses and improvements based on maximum power point tracking (MPPT) are
quite important to increase the harvested total energy. The intermittent characteristic
and perturbing power curve of a PV module is one of the most important defects that
should be tackled to increase the generation efficiency. The power-voltage (P-V) and
current-voltage (I-V) curves are main efficiency indicators of a PV system that
exhibit nonlinear characteristics in its natural structure. Furthermore, the generated
maximum power with a PV panel depends on two main quantities of temperature
and irradiation. However, it is possible to increase the generated power up to
maximum rates by MPPT algorithms. This chapter introduces most widely used
algorithms respecting to their implementation and utilization properties. The indi-
rect, direct, and computational methods are presented considering their advantages
and disadvantages. The conventional and novel algorithms are explained with
flowcharts and analytical details in order to provide clear comparison. The artificial
methods are expressed in the last section where fuzzy logic, artificial intelligence,
and optimization-based approaches are discussed.
Keywords Maximum power point tracking (MPPT)

Perturb and observe

Incremental conductance Fuzzy logic controller Artificial neural network
Particle swarm optimization
E. Kabalci (&)
Faculty of Engineering and Architecture, Department of Electrical and Electronics
Engineering, Nevsehir HBV University, Nevsehir, Turkey
e-mail: kabalci@nevsehir.edu.tr

206 E. Kabalci

ABC Artificial Bee Colony
AF Activation Function
COA Centroid of Area
CV Constant Voltage
EMI Electromagnetic Interference
ESS Energy Storage System
FLC Fuzzy Logic Controller
GA Genetic Algorithm
HC Hill Climbing
IncCond Incremental Conductance
I-V Current-Voltage
MLI Multilevel Inverter
MLP Multilayer Perceptron
MPPT Maximum Power Point Tracking
OV Open Voltage
P&O Perturb and Observe
PSO Particle Swarm Optimization
PV Photovoltaic
P-V Power-Voltage
RES Renewable Energy Source
SCPB Short-Current Pulse-Based
8.1 Chapter Overview
The maximum power point tracking (MPPT) is an algorithm that is associated with
dc-dc power converters and inverters to track maximum power point during energy
conversion process. Thus, the generated energy is maximized in this way. Although
there are several methods proposed to implement an MPPT system, there are two
algorithms known as “perturb and observe” (P&O) and the “incremental conduc-
tance” (IncCond methods are widely used since they are commercially preferred.
However, simpler algorithms lack to provide the anticipated performance on the
output power. The recent researches on MPPT algorithms exhibited that more
sophisticated algorithms yield better outputs comparing to widely known basic
methods. Therefore, a wide variety of numerical methods including fuzzy logic,
neural networks and other computational methods are proposed. Although these
recent algorithms require increased complexity, they easily compete with malfunc-
tions of previous methods in terms of partial shading, misdirection during tracking,
power fluctuations around MPP, and inadequate performance at low irradiance.
8 Maximum Power Point Tracking (MPPT) Algorithms … 207
In this chapter, the MPPT algorithms have been analyzed in there groups where the
first group includes indirect algorithms while others are described as direct algo-
rithms, and numerical algorithms. The groups have been allocated considering the
progress on methods and approaches where the indirect and direct algorithms are
known as conventional MPPT techniques. The most recent computational algorithms
including genetic algorithms, particle swarm optimization (PSO) algorithms, and
artificial bee colony algorithms (ABC) are the hottest topics in MPPT algorithms.
The trade-off between conventional MPPT methods and recent algorithms are related
to complexity in algorithm and to cost in application.
This chapter is dedicated to these widely used MPPT algorithms of PV systems.
Therefore, the initial sections introduce analytical background of a PV panel at a glance,
and power curves of any PV panel in brief. Afterwards, the power conversion system of
a PV plant with power converters is described to emphasize the application of a generic
MPPT block in power conversion issues in the following section. The MPPT algorithms
have been presented in three sections including indirect algorithms, direct algorithms,
and numerical algorithms with their application examples.
8.2 Basic Principles of Photovoltaic Energy Conversion
Edmund Becquerel, a French physicist, firstly discovered the photovoltaic (PV)

energy by generating electricity when he illuminated an electrode in electrolyte
solution in 1839. However, Adams and Day made the first practical application of
PV by applying the PV energy to solid materials about 40 years later. The primary
PV cells made of selenium were performing almost 1 or 2% efficiency. Probably
Einstein was the most important contributor of PV by explaining theoretical PV
effect in 1904 that brought a Nobel Prize to him in 1923. The first generation silicon
PV cells were produced by 1940s and 1950s by Czochralski, a Polish scientist. The
solar panels constituted with numerous PV cells were commercially available in
1963 when Sharp Corporation succeeded in producing practical silicon photovoltaic
modules. Although there were several applications of solar cells seen in Vanguard I
space satellite in 1958, Explorer III, Vanguard II, and Sputnik-3 in 1959. World’s
first largest PV array was built in Japan with 242 Wp rated power in 1963 [1–4].
A PV energy conversion system is composed of a PV module, a dc-dc converter,
an inverter, and preferably an energy storage system (ESS). The PV module is
constituted by PV cells that are series and parallel connected to generate the desired
rated power. The cells are produced in monocrystalline or polycrystalline structure
depending to the purity of semiconductor [5–7]. The polycrystalline cells that
provide limited efficiency around 13–14% are less efficient comparing to the
monocrystalline that the efficiency increases up to 20%.
The analytical model of a PV cell is designed referring to one-diode electrical
equivalent as shown in Fig. 8.1. This circuit includes a photocurrent source, a
diode, and serial and shunt resistors that are known as one-diode or five-parameter
model [8]. The calculations of the one-diode model seen in Fig. 8.1 are depended to
the output current;
208 E. Kabalci
Fig. 8.1 Single-diode

electrical equivalent of a
mono or polycrystalline PV
cell
Io ¼ IPV ID ðVÞ Ish ðVÞ ð8:1Þ
where (V) depicts the dependency of diode and resistor currents to terminal voltage
while they are independent from irradiation value.
The Shockley equation is called to express the semiconductor diode behavior to
define ID as given in Eq. (8.2);
VD
ID ¼ IR ðegVT 1Þ ð8:2Þ
where ID is current and VD is voltage of the diode, IR is reverse blocking current, VD

is thermal voltage, and η is the ideality factor of the diode [9, 10]. The current of
shunt resistor Ish is rearranged regarding to its dependence to terminal voltage as
seen in (8.3);
VO þ IO RS
Ish ¼ ð8:3Þ
Rsh
where the equivalence shown in Eq. (8.1) is achieved as seen in Eq. (8.4) referring
to irradiation current IPV, thermal voltage VT, and other parameters denoted above
[8–10];
VO þ IO Rs V þI R
O O S
IO ¼ IPV IR e gI VT 1 ð8:4Þ
Rsh
There are several other equivalent circuit models presented in the literature to
increase the reliability of the PV cell design. The implementation structure of the
cell is expressed by additional electrical elements in the improved models. One of
these models that is being extensively studied one, amorphous silicon PV cell is
illustrated in Fig. 8.2. The polymer and amorphous silicon based PV cell is con-
stituted with an additional dependent current source controlled by photocurrent and
terminal voltage.
Fig. 8.2 Single-diode

electrical equivalent of an
amorphous silicon PV cell
This equivalent circuit is obtained by modifying the well-known five-parameter

model, and the calculations are rearranged considering the dependent current
source. The output current equation is rearranged by taking into account the
additional dependent current source as seen in Eq. (8.5);
IO ¼ IPV ID ðVÞ Ish ðVÞ Ir ðIPV ; VL Þ ð8:5Þ
The photocurrent IPV is proportional to the irradiation w where the dependent

source current can be converted to Ir ðIPV ; VL Þ ¼ Ir ðw; VL Þ. Since the output current
of amorphous silicon PV cell is calculated regarding to the Eq. (8.6) that depicts the
dependency of diode and shunt current to the irradiation and terminal voltage;
IO ¼ IPV ID ðw; VÞ Ish ðw; VÞ ð8:6Þ
A PV module is composed of a number of series cells NS as 36, 60, or 72 while

the cells are connected in series and parallel NP in PV arrays as shown in Fig. 8.3
where the arrangement defines the maximum output voltage VM and maximum
output current IM. The maximum output currents of a module IM and an array IA are
calculated regarding to Eq. (8.4) as follows [10–12];
VM þ IM NS Rs V þI N R
M M S S
IM ¼ IPV IO e gI NS VT 1 ð8:7Þ
NS Rsh
!
qðVA þ IA S Rs Þ
N
NP VA þ IA NNPS RS
IA ¼ NP IPV NP IO e gI Ns VT
1 NS ð8:8Þ
NP Rsh
The basic open-circuit and short-circuit currents of a PV cell can be derived

referring to Eq. (8.4) as given by respectively;
VOC V
OC
0 ¼ IPV IO egI NS VT 1 ð8:9Þ
Rsh
210 E. Kabalci
Fig. 8.3 Electrical

connection diagram of a PV
array
ISC Rs I R
SC S
ISC ¼ IPV IO egI NS VT 1 ð8:10Þ
Rsh
The verification of a PV model is performed by measuring the current-voltage I-

V and power-voltage P-V curves against various irradiation values. It is clear that
the basic maximum power current–voltage pair belongs to the I–V characteristic
that Eq. (8.4) is recalled to define the following maximum power relation;
VM þ IM Rs V þI R
M M S
IM ¼ IPV IO e gI NS VT 1 ð8:11Þ
Rsh
A PV module is modeled referring to the relations given above that define the
effect of Rs, Rsh, Io, IPV, and η. The curves shown in Fig. 8.4 are produced by
changing the irradiation value from 200 W/m2 to 1000 W/m2. The axis on the
left-hand side of figure represents the current variation I-V curve, while the
right-hand side illustrates the output power of PV panel in terms of P-V curve. The
effect of irradiation on current, power, and voltage can be easily seen from the
graphic. The curves generated by the modelled PV panel are verified by comparing
to datasheet of commercial PV modules at 240 Wp power [13].
The parameters of the modeled PV module are given in Table 8.1 where each
parameter is verified in the simulation studies. The values are provided for 25 °C
cell temperature and 1000 W/m2 irradiation level.
Fig. 8.4 I-V and P-V characteristics of a 240 Wp PV module at various irradiation conditions
Table 8.1 Electrical Quantity Value

parameters of the modeled PV
module [13] Rated power (Pmax) 240 W
Open circuit voltage (VOC) 52.4 V
Maximum power voltage (VMPP) 43.7 V
Short circuit current (ISC) 5.85 A
Maximum power current (IMPP) 5.51 A
Temperature coefficient (PMax) –0.30%/°C
Temperature coefficient (VOC) –0.126 V/°C
Temperature coefficient (VSC) 1.76 mA/°C
Module efficiency 19%
Nominal operating cell temperature 48.3 °C
8.3 Power Conversion and MPPT
The solar power generation systems have attracted extensive attention in several
application areas such as agricultural, residential, and even industrial sites.
Furthermore, the sensations related to greenhouse emissions and carbon footprints
are key factors to promote the utilization of solar power systems. Nowadays, the
installation costs are decreased and overall efficiency of a PV system is increased
212 E. Kabalci
comparing to a few decades ago. The decreased costs allow redeeming earlier by
ranging five years to seven years that accelerate the usage of solar energy in
distribution systems. The solar power generation systems are constructed in two
types; either standalone or grid-tied [14–17].
The standalone systems are attractive for remote sites where the distribution lines
are unavailable. The standalone plants are composed of solar arrays, dc-dc converter,
ESS, and inverter. On the other hand, the grid-tied systems eliminates ESS
requirement and requires a grid-tie inverter instead of classical inverters. The
complete block diagram of a solar power generation system with grid-connection is
depicted in Fig. 8.5. The solar array is a combination of PV modules in series and
parallel to generate the required power in various voltage and current ratings. The
power conversion stage consists of dc power interface and its ac conversion pairs.
The dc-dc converters are used to stabilize the intermittent characteristic of solar array
that is considerably depended to solar irradiation and ambient temperature. The
power conversion structure can be in single-stage or double-stage interface where
the single-stage includes just a dc-ac inverter while the double-stage is composed of
dc-dc converter and dc-ac inverter as seen in Fig. 8.5. The single-stage interface
lacks in the stabilizing the dc bus voltage against rapidly varying dc output of solar
array. However, the inverter requires operating an algorithm to track the maximum
power point in order to match dc bus voltage of solar array.
However, it is not possible to sustain optimum matching at all radiation levels
since the maximum power point rapidly fluctuates depending to the radiation and
temperature. This operation can be performed in large-scale solar plants where the
generated dc bus voltage excesses the required supply voltage of inverter. Therefore,
the most proper way is to use double-stage power interface in any case [4, 16, 18–20].
A dc-dc converter is connected between solar array and inverter to match the
required dc bus voltage in the double-stage power conversion system. The dc-dc
converter handles the MPPT operation and the dc bus voltage is matched at this first
stage. The MPPT algorithm of converter increases or decreases the dc bus voltage
Fig. 8.5 The detailed block diagram of a solar power generation system
according to varied climatic conditions and maintains the match between dc supply
limits of inverter. The MPPT algorithm controls the duty cycle of dc-dc converter
where it changes the converted dc voltage level by adjusting the operation period of
semiconductor switch in the topology. Thus, the maximum power is extracted from
solar array by appropriately matching I-V balance of solar modules. The widely
used dc-dc converter topologies are Buck-Boost and Cuk besides classical buck or
boost topologies in solar power generation systems. It is not placed in this section
since the converter topologies can be found in several textbooks [4, 20, 21].
The second interface used in double-stage is inverter that converts the dc bus
voltage to ac power where the converted ac power is either supplied to the stan-
dalone loads or is injected to the utility grid. The inverter topology to be used in this
interface depends to the power level where two-level or multilevel topologies can
be selected. The multilevel topologies that were firstly introduced by Nabae gen-
erate the third voltage level with the neutral point of dc bus. The most widely used
multilevel inverter (MLI) topologies are diode clamped, flying capacitor, and cas-
caded H-bridge in renewable energy. The diode clamped and cascaded H-bridge
topologies provide the most proper involvement in high power applications where
megavolt-ampere MVA power handling is desired. The flying capacitor arrange-
ment eliminates filtering requirement comparing to diode clamped topology.
However, this circuit is not accepted as robust as cascaded H-bridge to obviate the
harmonic contents and the cost issues. Furthermore, several other MLI topologies
such as grid-tied or string structures are implemented depending to the power
ratings in solar energy application [4, 20].
The efficiency of a power conversion system is related to several factors such as
conduction losses, switching losses, electromagnetic interference (EMI), and total
harmonic distortion (THD) rates. The power efficiency is primarily decreased by
power loss that are caused by active devices while the EMI and THD losses induce
to defects on power quality and efficiency. The power losses increase proportionally
to the switched voltage and current rates. The MLI topologies decrease the switched
voltage and current rates at each cycle owing to its staircase output waveforms.
Consequently, the power losses are also decreased comparing to regular two-level
inverter topologies. The harmonic contents, THD ratios, and EMI rates are reduced
by this way. The MLI output waveforms, either voltage or current depending to the
topology, are increased to several levels, are particularly presented in odd numbers
such as 3-level, 5-level, 7-level etc. In addition to these, the asymmetrical MLI
topologies allow generating more increased output voltage levels by using same
power stages as cascaded topologies [4, 19–21].
The rated power of a solar power generation system is increased by several string
connections of power modules where the series-connected PV modules comprise
the strings. Furthermore, several strings can be connected in parallel to achieve
higher power ranges. The grid connection of PV plants that are constructed by using
such strings is performed with various converter infrastructures. The most widely
used power conversion schemes are illustrated in Fig. 8.6 where the connection
diagrams of central converter, multi-string converter, string converter, and
micro-converter are exhibited from Fig. 8.6a–d, respectively. Central converters are
214 E. Kabalci
constituted of inverters that convert the dc voltage achieved from PV strings ac

voltage. They are widely used in residential, commercial, and utility scale systems
with a power level of 1 kW or higher. The MPPT operation of the PV panels is
performed centrally at the dc-ac inverter stage as shown in Fig. 8.6a. The main
advantages of this scheme are decreased cost of inverter and decreased switching
losses owing to unique converter structure. However, the central converter usage
causes to a main drawback in acquired energy level since it is controlled by a single
MPPT where the individual MPPT support is not available for each PV strings.
Thus, the acquired maximum power level is limited under shading, thermal vari-
ations, aging or disorientation of PV strings. The multi-string converter seen in
Fig. 8.6b is comprised of several individual dc-dc converters usually managing two
Fig. 8.6 Power conversion schemes in grid-tied PV systems, a central converter, b multi-string
converter, c string converter, d micro-converter
PV strings or three PV strings at most. In this case, the double-stage interface is

compulsory. The most important contribution of the double-stage interface is
additional dc-dc converters that allow increasing the dc input voltage in a wide
range. The PV plant can be freely constructed by arranging several dc-dc converter
connections of strings. The dc-ac conversion is configured by using a central
inverter that eliminates the drawbacks of the previous converter structure since the
MPPT operations are handled by dc-dc converters.
The string converter that is shown in Fig. 8.6c is configured with individual
MPPT controllers that do not allow parallel connection of strings. This option
provides to locate each PV string in any direction or orientation. Furthermore, it is
freely possible to care shading conditions and to determine PV module number in a
string. The prominent feature of string converter is relevant to its power level that is
widely designed at low power levels to be used in single-phase applications. In
addition to this, its common topology includes full-bridge inverter with lien fre-
quency transformer at the output stage. Thus, this configuration causes to a drawback
of higher cost per kW. The micro-inverter is constituted with a dc-ac converter for
each PV module where individual MPPT and power conversion features are
acquired by this way. The configuration seen in Fig. 8.6d enables each PV module to
connect to the ac grid without any dc wiring. Although this configuration provides
simpler usage and installation, the cost per kW is higher than string converters.
Also, the operation cycle of the converter is lower than of a PV module [20, 21].
Although the utilization of PV systems are increased day by day, two main draw-
backs are being extensively studied to overcome which are high installation costs,
and low conversion efficiency. The acquired output power of a PV system is
increased by using several mathematical algorithms. The MPPT is a fundamental
method that eliminates the mismatch along the load characteristic and maximum
power output of a PV module. MPPT algorithms ensure the optimal usage of PV
modules against several defective conditions such as disorientation, shading or
partial shading. Furthermore, the intermittent characteristic of solar power is sus-
tained in a reliable way by using MPPT methods. Many MPPT methods have been
proposed for single-stage and double-stage systems to cope with performance lim-
itations. The most prominent classifications of MPPT methods are indirect and direct
methods. The indirect methods include several algorithms such as open-circuit,
constant voltage, short-circuit, and pulse method where some of them requires
former knowledge on the PV array characteristics while some requires mathematical
calculations. These requirements limit the efficiency of the algorithm at any case of
irradiance and temperature variations.
The alternative MPPT methods are classified under direct methods where the most
widely known ones are perturb and observe (P&O), hill climbing (HC), and incre-
mental conductance (IncCon). These methods are competitor to indirect methods in
terms of simplicity, easy implementation with limited prior information, and effi-
ciency The commercial applications of MPPT methods are performed with P&O
algorithm where the algorithm instantly measures voltage value at the maximum
power point and tracks the variation. The MPPT algorithm is used to determine the
switching angle of power converters depicted in Fig. 8.6. The algorithm detects and
216 E. Kabalci
tracks the reference voltage to change the duty cycle, and consequently the output
voltage of converter [4, 22–25]. The most recent improvement of MPPT algorithms
are performed by using intelligence based methods such as fuzzy logic controllers
(FLCs), artificial neural networks (ANNs), particle swarm optimization (PSO)
algorithms, and several genetic algorithm (GA)methods [23, 26]. The soft computing
methods provide better efficiency and faster response comparing to indirect and
direct methods. However, they lack on simplicity. The MPPT methods and out-
standing properties are presented in the following sections.
8.4 Indirect MPPT Algorithms
The indirect MPPT algorithms differ from direct MPPT methods by calculations to
determine the PV cell voltage and current at MPP. However, the direct MPPT
methods are based on measurements instead of calculations. The indirect algorithms
are based on several calculation methods including system voltage adjustment,
temperature measurement, open-circuit voltage VOC calculation, and azimuth and
angle calculation sometimes. The system voltage or operating voltage adjustment is
performed to detect the MPP voltage according to seasonal variations where it
should be higher in winters. The operating voltage depends on the PV module
temperature. The MPP voltage is calculated regarding to VOC that is instantly
measured. The main advantage of indirect MPPT methods including constant
voltage, open-circuit voltage, and short-current pulse-based is their simplicity.
However, they lack against aging and pollution of PV cells [26–29].
8.4.1 Constant Voltage MPPT Method
The constant voltage (CV) method is the plainest MPPT algorithm that is based on
regulating the array voltage to track the MPP voltage VMPP as seen in Table 8.1.
The PV array voltage is adjusted around VMPP regarding to the reference voltage
VRef that is the regulated array voltage. It is intended to match the VRef to VMPP in
the most proper adjustment. If this is not the case, then the algorithm tries to acquire
the most adequate adjustment around the VMPP. The CV algorithm assumes that the
irradiance and temperature variations on the cells do not cause to meaningful
changes on VMPP and therefore, the VRef can be kept constant during the operation
of algorithm. The flowchart of the CV MPPT operation is illustrated in Fig. 8.7
where the PV array voltage is required to be measured at the beginning of the
algorithm. The algorithm does not involve any other input data. The measured array
voltage VPV is used to set up the duty-cycle of dc-dc converter. Then the duty cycle
is updated at each turn by comparing the array voltage to reference voltage [26, 28].
When the array voltage is greater than reference, the duty cycle is decreased in the
next step or vice versa.
Fig. 8.7 Flowchart of the

constant voltage MPPT
method
It should be noted that CV method is more efficient comparing to regular P&O

technique during the lower irradiance conditions. Therefore, it may be integrated
other MPPT methods.
8.4.2 Open Voltage MPPT Method
The open voltage (OV) method is based on the assumption of that VMPP is ever
around a fixed percentage of the open-circuit voltage VOV. It is also assumed that
the production variations of cells, temperature, and solar irradiance rates change the
MPP around 2% of its regular value. Therefore, this method defines a reference
voltage VRef at the 76% of VOV, which is quite near to VMPP.
The flowchart of OV method is depicted in Fig. 8.8 where open-circuit voltage
VOV measurement is required to determine the reference voltage. This measurement
is performed by opening the circuit, and a series static switch is used to carry out
this action. This operation causes to zero PV current that prevents the power
generation at this stage. In the next step, the measured PV voltage VPV is compared
to the reference voltage VRef to determine the exact duty-cycle of dc-dc converter.
The pre-defined duty-cycle Dn is increased in the next step, if VRef is greater than
VPV [26, 28, 30, 31].
218 E. Kabalci
Fig. 8.8 Flowchart of the open-voltage MPPT method
If the VPV is greater than VRef then the Dn is decreased by varied value for the
next duty-cycle calculation Dn+1. The regulated dc output voltage is controlled by
PI block at each cycle to track the VMPP.
8.4.3 Short-Current Pulse-Based MPPT Method
The short-current pulse-based (SCPB) MPPT method determines VRef that is

required to generate the optimum operating current IOP. In the ideal conditions, IOP
is proportional to the short-circuit current ISC by a coefficient k and irradiance level
of S as seen in Eq. (8.12);
IOP ðSÞ ¼ k ISC ðSÞ ð8:12Þ
The equation exhibits that the operating current is defined by irradiance level and
ISC where it is also temperature dependent that varies between 0 and 60 °C. On the
other hand, the coefficient is assumed to be around 92%. It is obvious that this
Fig. 8.9 Flowchart of the short-current pulse-based MPPT method
algorithm is based on short-circuit current measurement and that is performed by a

parallel static switch similar to previous method. The parallel switch provides the
short circuit conditions where the array voltage is zero and power generation is
prevented. The flowchart of this control method is shown in Fig. 8.9 that the
preliminary measurements and comparisons are depicted similar to OV MPPT
algorithm with a difference where the measured and compared parameters are
current in this method [27, 28, 30, 31].
The PI regulator requires the array voltage as in the OV method to generate the
operating current where the reference voltage is also acquired at this stage.
8.5 Direct MPPT Algorithms
The direct MPPT algorithms are based on measurement of PV module voltage,

current, or instant power where the performance variations are rapidly detected and
more accurate tracking is performed. The basic principles of direct MPPT
220 E. Kabalci
algorithms are related to periodic tuning of I-V and P-V curves, and adjusting the
operating voltage in small steps to increase the accuracy. The first principle is used
to determine the maximum output power of PV module that is tracked at each turn
of duty-cycle calculation of the dc-dc converter. The second principle is performed
by a method known as hill climbing that is based on detecting the power or current
direction either increasing or decreasing. In case the measured value increased
depending on the operating voltage, the tracking direction is set to forward,
otherwise it is set to backward. The MPP is detected by this principle and the
operating point track the real MPP with little oscillations regarding to step size [24,
25, 29, 32]. The characteristics of a typical PV cell under regular irradiance and
partial shading conditions are illustrated in Fig. 8.10 where the MPP dot depicts the
current source and voltage source operation of PV module. If the operating point of
the load is on the left-hand side of MPP, it means the module operates as a current
source and the error signal M required by Hill-Climbing HC MPPT is calculated as
given in (8.14);
If the operating point of the load is on the right-hand side of MPP, then PV
module acts as a voltage source, and M is calculated as follows;
@PPV
[ 0 ) M ¼ M þ DM ð8:13Þ
@VPV
@PPV
\0 ) M ¼ M DM ð8:14Þ
@VPV
Apparently, the error signal will be zero at the MPP that yield DM = 0. This
approach is applied at several direct MPPT methods where the most outstanding
algorithms are P&O and Incremental Conductance (IncCon) [29, 31, 32].
Fig. 8.10 Characteristics of a

typical PV cell under regular
irradiance and partial shading
8.5.1 Perturb and Observe MPPT Algorithm
The P&O algorithm is one of the most widely used and studied method owing to its
simplicity and practical implementation. It perturbs the terminal voltage, as its name
implies, and compares the actual PV power to previous value. The algorithm
decides to track the MPP in one direction or in the opposite direction by comparing
the terminal voltage change and output power increment. In case the output power
is increased against the voltage change, the algorithm keeps on tracking the MPP in
the same direction. Otherwise, it shifts the tracking direction [4, 26, 32, 33].
There are several textbooks and papers can be found on regular P&O algorithm
in the literature. However, the main attention should be directed to improve instead
of applying the regular P&O. The flowchart of an improved algorithm proposed by
Dolara et al. [26] is shown in Fig. 8.11 where the power and voltage perturbations
are detected in the progress, and then the voltage difference is detected to operate
the algorithm. In the proposed method, the array power is sampled to generate an
Fig. 8.11 Flowchart of optimized P&O MPPT method [26]

222 E. Kabalci
average value, and is used for arranging the magnitude of perturbation Dd at the
MPP operating point.
The arrangement is performed by calculating the a(n) regarding to voltage error
signal M and power to voltage ratios as shown on the right-hand side of flowchart.
The duty cycle is varied around 0.5% for each regular P&O methods. However, the
proposed algorithm in [26] tries to adjust the duty cycle variation between 0.5% and
2.7% in the next turn regarding to k coefficient, a (n), and Dd where it is equated as
follows;
dðn þ 1Þ ¼ dðnÞ + k aðnÞ Dd ð8:15Þ
An improved P&O MPPT method with PI control is proposed by Kabalci et al.

[34, 35] where the schematic diagram of combined controller is illustrated in
Fig. 8.12. The MPPT control of converters are improved with additional PI to
regular P&O algorithm where measurements of solar array, dc converter, and
control diagram of PI controller are shown on the left-hand side of Fig. 8.12. The
flowchart of the improved PI controlled MPPT algorithm is depicted in Fig. 8.13
where additional PI calculation is managed by the error calculation segment on the
flowchart.
The flowchart of MPPT algorithm is depended on voltage and power calcula-
tions that are based on current and voltage values acquired from sensors [34].
Once the actual power is calculated, then the next cycle of the measurement is
compared to previous value to track the reference voltage Vref. The implemented
control section that the upper and lower parts of detailed PI controller block exhibit
P&O detection part while the middle part on the left hand side of figure illustrates
the PI control added to MPPT algorithm considering Kp and Ki parameters of the
Fig. 8.12 Schematic diagram of the PI assisted MPPT [34]

Fig. 8.13 Flowchart of optimized PI assisted P&O MPPT method [34]
system. The actual inputs and reference outputs are switched as seen in the center of
PI controller block and modulating signal is conditioned by the zero-order sample
and hold block. The logical comparator compares the triangular carrier signal and
modulating signals to generate the PWM at each interval where the actual power
measurement is repeated [34].
In order to determine duty ratio the latest observation of the obtained power is
compared to the previous observation. Afterwards, ensured difference is used to
specify the next PWM duty cycle. The implemented P&O algorithm looks for the
increment of power perturbation and attempts to keep the duty cycle in the same
way in order to achieve the MPP of PV array, also PI control is activated to reduce
oscillation on the output voltage. The algorithm can reverse the perturbation when
there is a decrement occurs in the power observation. The algorithm repeats this
process for reach to MPP of PV array, if the system reaches to the MPP; the
algorithm produces steady duty cycle value. The oscillation at the MPP of PV array
is demanded to be reduced in order to increase the stability [35, 36]. The PI
function, which is used to reduce oscillation on the output voltage is calculated as
given in short code of Eq. (8.16) [35];
224 E. Kabalci
9
Void PIðÞ >
>
>
>
f >
>
>
>
>
=
Error ¼ Vref Vo;
ð8:16Þ
Error in ¼ Error in þ Error dt; >>
>
>
duty ¼ Error Kp þ Error in Ki; >
>
>
>
>
;
g
8.5.2 Incremental Conductance MPPT Algorithm
The incremental conductance InCon algorithm depends on the detection of the

slope occurred in the PV curve. The slope is zero when the MPP of the curve is
reached as shown in Fig. 8.10. The movement direction of operating point is
detected by tracking the change occurred in the I/V ratio comparing to the MPP
point. The operating point increases the slope when it is moving towards to the
MPP, while it turns negative on the right side of MPP. This situation can be
expressed as follows [25, 37–39];

dI I dP
¼ ; ¼ 0 ; at MPP
dV V dV

dI I dP
[ ; [ 0 ; at left of MPP ð8:17Þ
dV V dV

dI I dP
\ ; \0 ; at right of MPP
dV V dV
The Eq. (8.17) and flowchart given in Fig. 8.14 are based on the instant con-
ductance (I/V) and the incremental conductance (dI/dV) of PV array where the
reference voltage is calculated regarding to incremental value in an iteration that is
repeated until MPP is reached. Once the VRef acquired at the MPP, the algorithm
maintains to track this value in order to ensure maximum output power. In case a
change occurs in PV current dI, algorithm starts a new iteration at the beginning of
the flowchart to track VRef by increasing or decreasing the previous value.
When the irradiance increases on the PV array, the array voltage is increased and
the MPP moves to the right-hand side as can be seen on Fig. 8.10. The algorithm
increases the operating voltage of PV array for compensating this increment. On the
other hand, the decrement on irradiance triggers the algorithm to decrease the
operating voltage of array in order to compensate the change. That is the way that
InCon algorithm track the slope value to detect the change on irradiance and react
against the cases. The easier way to determine the change is comparing previous
and actual values of voltage and current that are depicted as dV and dI in the
Fig. 8.14 Flowchart of incremental conductance MPPT method
algorithm. The dV and dI are zero when the irradiance and temperature are stable,
and algorithm maintains to operate at MPP.
If the irradiance increases, it causes an increment on dI (dI > 0) while dV = 0 that
also means the MPP is being increased by irradiance. The algorithm decides to
increase the operating voltage at this stage. On the other hand, the decrement on
irradiance causes to dI < 0 that requires to decrease the operating voltage. The
rapidly changing cases are easily detected by InCon algorithm since it depends on
tracking the operating point towards MPP. This approach prevents misleading in
tracking as seen in P&O algorithm, and that makes InCon more reliable by decreasing
the perturbation around MPP. However, there are several drawbacks of this algorithm
related to MPP as others. Tey and Mekhilef [40] proposed an optimized alternative to
regular incremental conductance method by increasing the step size that is named as
variable step size InCon. The algorithm is based on defining a step size to compensate
the variations on MPP change and increasing or decreasing the duty cycle consid-
ering this step size additional to previous duty cycle. Thus, an error rate is permitted
by algorithm and the step size is adjusted to maintain the error control in the limit.
There are several methods are proposed to optimize regular MPPT methods.
However, the numerical methods including FLCs, Neural network controlled
226 E. Kabalci
algorithms, and some other soft computing methods are applied to improve existing
methods. The rapidly adapting structures, complexity, and sophisticated approaches
of these algorithms increase the reliability of regular MPPT methods.
8.6 Artificial MPPT Algorithms
The most widely used MPPT methods are introduced in 4th and 5th sections. There
are several other methods are also proposed in the literature including non-linear
and sliding mode control methods in order to improve the stability and efficiency of
regular MPPT algorithms such as perturb and observer or incremental conductance.
However, the main drawbacks seen in the proposed methods are caused by irra-
diation and cell temperature that are not robustly handled by existing algorithms.
The widely known methods are based on discontinuous control strategy where the
intersection tracking is performed around the MPP. The tracking approach of the
previously introduced algorithms lack in efficiency since they are not supported by
intelligent controls and are less efficient against the rapidly changing conditions
[41–44]. Furthermore, the existing methods do not pay attention to measurement
noises of voltage and current that has significant effect on the algorithm decisions.
These drawbacks require fast time response and increased stability comparing to
widely known methods. The oscillations occurred in the acquired power output
should be decreased in order to decrease power losses. The latest improvements in
MPPT methods are performed by using several numerical and artificial methods
including FLC, neural network, neuro-fuzzy, genetic algorithm, and some opti-
mization algorithms such as particle swarm optimization, artificial bee colony etc.
The FLC is one of the simplest algorithms among other artificial methods. Recently,
an increasing attention paid to FLC based MPPT algorithms owing to its robust
response against unpredictable air conditions and its simple structure not requiring
complex mathematical arrangements. The neural networks are also being exten-
sively studied owing to their high level control structure and sophisticated man-
agement features [42, 43]. The following subsections are dedicated to artificial
methods used in MPPT algorithms with analyses.
8.6.1 Fuzzy Logic Control MPPT
The combining artificial intelligence with regular MPPT algorithms is a new trend
of research and application in renewable energy sources. FLC is the most widely
researched artificial method in MPPT applications since it provides better perfor-
mance comparing the conventional methods, and involves simpler requirements
comparing to sophisticate neural or estimation algorithms. These advantages of
FLC are integrated to overall system by combining it to existing MPPT controller
that provides to acquire maximum power transfer to the loads.
Fig. 8.15 Block diagram of a FLC MPPT controller
The block diagram of a FLC based MPPT controller is depicted in Fig. 8.15
where the contents of a regular FLC are also shown in a significant block. The
reference input of FLC is used to determine the error signal E and change rate of the
error CE where it is shown on the left-hand side of figure. The output of FLC is the
duty cycle that is applied to converter where it is generated by the four sections of
FLC. These sections are fuzzification where the numerical inputs are converted to
lexical variables that are defined by rule base and depended to membership func-
tions. The levels of membership functions can be increased to present better res-
olution in the calculations. The analyzed inputs are weighted according to several
methods, namely modus ponens, Takagi-Sugeno or Mamdani, in the inference
section, and then the lexical outputs of FLC are converted to numbers in the
defuzzification section in order to generate the duty cycle. The most widely used
defuzzification method is the is the centroid of area (COA) owing to its successful
averaging properties and more accurate results [41, 42, 44]. The integration of FLC
to the regular P&O algorithm is illustrated in Fig. 8.16 where the decision making
against output power is performed regarding to the power and voltage perturbation
by FLC controller that generates the adaptive duty cycle.
8.6.2 Neural Network MPPT
The most recent researches on computational MPPT methods include ANN studies
that are significantly effective on wide disruptive conditions such as partial shading
[45, 46]. The ANN based MPPT provides rapid and reliable estimations against
quickly varying irradiance and temperature parameters. An ANN is composed of
several layers in multilayer perceptron (MLP) structure as shown in Fig. 8.17. In a
model as shown in figure, the neural-like organization is comprised of neurons seen
in the lower part of figure where the inputs are linearly weighted and summed in a
function. The sum of weighted input are transferred to a nonlinear function called
228 E. Kabalci
Fig. 8.16 Flowchart of a

FLC P&O MPPT algorithm
activation function (AF), and lastly are sent to the following neurons. The definition
of input function in AF is depicted as given in Eq. (8.18);
X
M
x¼ w m xm þ a ð8:18Þ
m¼1
where x1 ; x2 ; . . .; xm are input signals, and w1 ; w2 ; . . .; wm are the weights of each

input signal [45].
The weights define the significance of each input data and the learning operation
of the network is performed regarding to continuously varying weights. The
training set of the ANN is composed with several different input and output data
where they are voltage and current in inputs and MPP voltage at the output as seen
in Fig. 8.17. There are several ANN based MPPT operation approaches are pro-
posed. In one hand, some ANN algorithms integrated to regular MPPT methods use
irradiance and temperature measurement acquired from separate sensors and
Fig. 8.17 Block diagram of an ANN controlled MPPT algorithm
directly generates variable duty cycle. On the other hand, the measured irradiation
and temperature data are used to detect the voltage and current of PV module in
other ANN MPPT group. The second group is widely used rather than first one and
it is constructed with MLP structure [46–48].
8.6.3 Particle Swarm Optimization MPPT Algorithm
Particle swarm optimization PSO is a novel swarm optimization algorithm that is

firstly proposed by Kennedy as an evolutionary algorithm based on behavior of
birds [49, 50]. PSO uses a set of particles that each one suggests a solution to the
optimization problem. It is based on the success of all particles that emulates a
population where the position of each particle depends to the agent position to
230 E. Kabalci
detect the best solution Pbest by using current particles in the population G. The
position of any particle xi is adjusted by
xki þ 1 ¼ xki þ vi ð8:19Þ
where the velocity component vi represents the step size and is calculated by using
the Eq. (8.20);
vki þ 1 ¼ xvki þ c1 r1 ðPbest i xki Þ þ c2 r2 ðG xki Þ ð8:20Þ
where x is the inertial weight, c1 and c2 are the acceleration coefficients, r1 and r2
are random values that belong to the interval of [0, 1], Pbest_i is the best position of
particle i, and G is the best position in the entire population [50].
A typical MPPT method should be used to integrate PSO algorithm to controller.
Most widely used integration is based on hill-climbing or P&O algorithm to PSO.
The flowchart of a PSO MPPT algorithm is depicted in Fig. 8.18. The operation
given in flowchart can be analyzed in five steps that are initialization, fitness
evaluation, updating the individual and global best value, updating the velocity and
position of each particle, and convergence determination [51]. In the first step,
particles are randomly initialized in the distribution space, or are initialized on
Fig. 8.18 Flowchart of PSO MPPT algorithm

described grid nodes covering the search space. Similarly, the initial velocity values
are defined randomly. The fitness value of each particle is evaluated in the second
step where the fitness evaluation is leaded to provide candidate solution to the
objective function.
The individual and global best fitness values are determined in the third step
where pbest_i and gbest are respectively determined. Then the positions are updated
and replaced with better fitness values if they are found. The velocity and position
of each particle are updated in the fourth step respecting to Eqs. (8.19) and (8.20).
The last step of the flowchart checks the convergence criterion. If the criterion is
met, the process is finished. Otherwise, the iteration number is increased and
procedure returns to step 2 [51–53].
The application of PSO MPPT in a PV system is depended to the matching
definitions of both systems. The particle positions are used to define the duty cycle
of dc converter, and the fitness value evaluation function stands for the output
power of PV array. The success of algorithm is performed by increasing the number
of particles that provides more accurate MPP tracking operation, even for shading
problems. On the other hand, increased number of particles causes to longer
computation times. The particle number is usually selected as the number of series
connected cells in a PV array in order to obtain the most exact operation time.
The particles are placed either on fixed positions or on random places in the PSO
initialization step. It depends on the information about global MPP since the par-
ticles can be placed around the point, if there is any information about it. The PV
voltage and current are measured and digitally filtered by using finite impulse
response filters in order to calculate the fitness value, PPV of particle i. The cal-
culated fitness value is compare to the best fitness value in order to keep pbest_i or to
update it with the new pbest_i in case of better fitness value. The velocity and
positions of all particles are updated and convergence criteria are controlled to
generate the best control signal [51–53].
The dependency to pbest_i and gbest is the main problem of conventional PSO
algorithm under nonlinear shading conditions. Since there are several PSO methods
proposed in the literature. The outstanding alternative to classical PSO are deter-
ministic and dormant PSOs [52, 53].
8.7 Conclusions
This chapter deals with most widely used and recent trends in MPPT algorithms.
The literature reviews show that several methods are proposed to track the MPP in a
PV power system. The main concerns of all the proposed algorithms are fast
response, reliable tracking, and easy implementation. The power conversion devi-
ces and configurations are improved to increase acquired energy level of a solar
plant. Although there are various structures proposed in literature, double-stage
power conversion holds the great share of usage in residential and industrial plants.
The first stage is comprised of dc-dc converters while the second stage performs
232 E. Kabalci
dc-ac conversion by using inverters. The dc interface requires robust control

methods to increase converted energy under heavily or partially shaded conditions.
The MPPT algorithms are focused on coping with irradiance and temperature
variations to balance the acquired total output power.
The preliminary algorithms classified as indirect algorithms are based on several
calculation methods including system voltage adjustment, temperature measure-
ment, open-circuit voltage VOC calculation, and azimuth and angle calculation
sometimes. These methods lack on responding against rapidly changing weather
conditions. The direct methods that includes hill-climbing, P&O, and indirect
conductance are proposed to tackle this situation. The basic operation principles of
direct MPPT algorithms are based on periodic tuning of I-V and P-V curves, and
adjusting the operating voltage in small steps to increase the accuracy. The
decreased step size and effective measurement methods increased the reliability of
direct MPPT methods. However, the researches exhibited the perturbation issue
seen around the MPP can be overcome with computational methods. Their simple
analytical background led the main drawback of direct methods. Therefore, more
sophisticated and complex algorithms were researched to prevent the perturbations
occurred in power curve of PV arrays.
A wide variety of numerical methods including fuzzy logic, neural networks and
other computational methods are proposed. The most recent computational algo-
rithms including GAs, PSO algorithms, and ABC are prominent topics in MPPT
algorithms. The trade-off between conventional MPPT methods and recent algo-
rithms are related to complexity in algorithm and increased cost in application.
However, improved efficiency and fast response leads to extensive researches.
References
1. Twidell J, Weir T (2006) Renewable energy sources. Taylor & Francis, London ISBN
0-419-25330-0
2. Kabalcı E, Kabalcı Y, Develi I (2012) Int J Electr Power Energy Syst 34:19–28
3. Masters GM (2004) Renewable and efficient electric power systems. Wiley-IEEE Press.
ISBN: 13: 978-0471280606
4. Colak I, Kabalci E (2014) Chapter 7-Control methods applied in renewable energy, use,
operation and maintenance of renewable energy systems. Springer, pp 205–246. doi:10.1007/
978-3-319-03224-5_7. ISBN: 978-3-319-03224-5
5. Zeman M (2014) Chapter 9-Photovoltaic systems, Delft University of Technology
6. Hersch P, Zweibel K (1982) Basic photovoltaic principles and methods, SERI/SP-290-1448
solar information module 6213. Golden, Colorado
7. Gray JL (2003) In: Luque A, Hegedus S (eds) Handbook of photovoltaic science and
engineering. Wiley, West Sussex, pp 61–112
8. Kabalcı E, Gokkus G, Gorgun A (2015) 7th international conference electronics, computers
and artificial intelligence, Bucharest, Romania, 25–27 June 2015, pp SG23–SG28
9. Lineykin S, Averbukh M, Kuperman A (2014) IEEE Trans Ind Electron 61:6785–6793
10. Kadri R, Gaubert JP, Champenois G (2012) IEEE Trans Power Electron 27:1249–1258
11. d’Alessandro V, Guerriero P, Daliento S, Gargiulo M (2011) Solid State Electron 63:130–136
12. Tian H, David F, Ellis K, Muljadi E, Jenkins P (2012) Sol Energy 86:2695–2706
13. Panasonic HIT Photovoltaic Module, HIT Power 240S. http://www.panasonic.com/business/

pesna/includes/pdf/eco-construction-solution/HIT_Power_SA06_Series_(240W)_Data_sheet-
v2.pdf
14. Radjai T, Rahmani L, Mekhilef S, Gaubert JP (2014) Sol Energy 110:325–337
15. Pradhan R, Subudhi B (2015) Int J Electr Power Energy Syst 64:792–803
16. Mohanty P, Bhuvaneswari G, Balasubramanian R, Dhaliwal NK (2014) Renew Sustain
Energy Rev 38:581–593
17. Bendib B, Krim F, Belmili H, Almi MF, Boulouma S (2014) Energy Procedia 50:383–392
18. Wu JC, Wu K, Jou HL, Chang SK (2014) IET Power Electron 7:2717–2725
19. Jiang J, Su YL, Shieh J, Kuo K, Lin T, Lin T, Fang W, Chou J, Wang J (2014) Appl Energy
124:309–324
20. Zheng H (2013) Solar photovoltaic energy generation and conversion—from devices to grid
integration. PhD dissertation, Department of Electrical and Computer Engineering, Graduate
School of The University of Alabama, Tuscaloosa, Alabama
21. Lorenzani E, Franceschini G, Bellini A, Tassoni C (2014) Chapter 6-Single-phase grid
connected converters for photovoltaic plants, renewable energy. In: Hammons TJ (ed) InTech.
ISBN: 978-953-7619-52-7
22. Rezk H, Eltamaly A (2015) Sol Energy 112:1–11
23. Bendib B, Belmili H, Krim F (2015) Renew Sustain Energy Rev 45:637–648
24. Sivakumar P, Kader AA, Kaliavaradhan Y, Arutchelvi M (2015) Renew Energy 81:543–550
25. Eltawil MA, Zhao Z (2013) Renew Sustain Energy Rev 25:793–813
26. Dolara A, Faranda R, Leva SJ (2009) Electromagn Anal Appl 3:152–162
27. Faranda R, Leva S; Maugeri V (2008) Power and energy society general meeting—
conversion and delivery of electrical energy in the 21st century, 20–24 July 2008, pp 1–6
28. Faranda R, Leva S (2008) WSEAS Trans Power Syst 3:446–455
29. Onat N (2010) Recent developments in maximum power point tracking technologies for
photovoltaic systems. Int J Photoenergy 1:1–11
30. Park M, Yu K (2004) In: 30th annual conference of IEEE industrial electronics society,
pp 2040–2045
31. Coelho RF, Concer FM, Martins DC (2010) 2010 IEEE international conference on
sustainable energy technologies (ICSET), pp 1–6
32. Rújula AA, Abián JA (2014) Sol Energy 109:95–104
33. Ahmed J, Salam Z (2015) Appl Energy 150:97–108
34. Kabalci E, Kabalci Y, Canbaz R, Gokkus G (2015) 4th international conference on renewable
energy research and applications (ICRERA 2015), Palermo, Italy, 22–25 November 2015
35. Kabalci E, Gokkus G, Gorgun A (2015) ECAI 2015 7th international conference electronics,
computers and artificial intelligence, Bucharest, Romania, 25–27 June 2015, pp SG23–SG28
36. Sea TY, Ka JS, Lee, CU, Chung DH (2013) 13th international conference on control,
automation and systems (ICCAS 2013). IEEE
37. Hohm DP, Ropp ME (2003) Prog Photovolt Res Appl 11:47–62
38. Hsieh GC, Hsieh H, Tsai CY, Wang CH (2013) IEEE Trans Power Electron 28:2895–2911
39. Sera D, Mathe L, Kerekes T, Spataru SV, Teodorescu R (2013) IEEE J Photovoltaics
3:1070–1078
40. Tey KS, Mekhilef S (2014) Sol Energy 101:333–342
41. Mohd Zainuri MAA, Mohd Radzi MA, Soh AC, Rahim NA (2014) IET Renew Power Gener
8:183–194
42. Chiu C (2010) IEEE Trans Energy Convers 25:1123–1132
43. Chiu C, Ouyang Y (2011) IEEE Trans Control Syst Technol 19:1516–1526
44. El Khateb A, Abd Rahim N, Selvaraj J, Uddin MN (2014) IEEE Trans Ind Appl 50:2349–2358
45. Rizzo SA, Scelba G (2015) Appl Energy 145:124–132
46. Kofinas P, Dounis AI, Papadakis G, Assimakopoulos MN (2015) Energy Build 90:51–64
47. Khanaki R, Radzi MAM, Marhaban MH (2013) In: IEEE conference on clean energy and
technology (CEAT 2013), pp 287–292
234 E. Kabalci
48. Messalti S, Harrag AG, Loukriz AE (2015) In: 6th international renewable energy congress
(IREC 2015), pp 1–6
49. Shi J, Zhang W, Zhang Y, Xue F, Yang T (2015) Electr Power Syst Res 123:100–107
50. Mirhassani SM, Golroodbari SZM, Mekhilef S (2015) Int J Electr Power Energy Syst
64:761–770
51. Liu Y, Huang S, Huang J, Liang W (2012) IEEE Trans Energy Convers 27:1027–1035
52. Lian KL, Jhang JH, Tian IS (2014) IEEE J Photovoltaics 4:626–633
53. Ishaque K, Salam Z (2013) IEEE Trans Industr Electron 60:3195–3206
Chapter 9
Photovoltaic System: Case Studies
Ali Durusu, Ismail Nakir and Mugdesem Tanrioven
Abstract Solar energy is one of the most important energy, which is environ-
mentally friendly such as clean, inexhaustible and free, among the renewable
energy sources. Studies on solar photovoltaic (PV) energy generation system were
promoted in last two decades. The main application of PV systems are in
stand-alone (water pumping, lighting, electrical vehicle, etc.), hybrid and
grid-connected (PV power plants) configuration. Stand-alone PV power generation
system is considered as good alternative for places that are far from conventional
power generation/transmission/distribution system. PV generation systems have
two big problems; PV conversion efficiency is very low and PV electricity gen-
eration is effected from changing of weather condition. PV output varies periodi-
cally in a year and in a day, and is not stable due to environmental condition.
Accordingly, in order to increase PV output and PV efficiency, it is crucial to
analyze PV output considering solar radiation, temperature, wind speed, shadow,
etc. Maximum power point trackers (MPPTs) are employed for extracting power
from photovoltaic (PV) panels. MPPTs enforce the solar modules to operate at
maximum power point (MPP) under the fluctuations of ambient conditions.
Therefore, they take a vital role for increasing of PV system efficiency. In this part,
the case studies of MPPT system, which includes stand-alone and hybrid PV
systems, will be briefly reviewed, followed by discussion of the MPPT modeling,
design, etc. Several stand-alone and hybrid MPPT application will be presented.
Latest developments in MPPT methods will be summarized. Finally some of the
present challenges facing the MPPT techniques will be explored.
Keywords Photovoltaic systems Case studies MPPT algorithms MPPT

algorithm modeling
A. Durusu (&) I. Nakir M. Tanrioven

Yildiz Technical University, Davutpasa Campus, 34220 Istanbul, Turkey
e-mail: adurusu@yildiz.edu.tr
I. Nakir
e-mail: inakir@yildiz.edu.tr
M. Tanrioven
e-mail: mtanrioven@gmail.com

236 A. Durusu et al.

IC Incremental Conductance
MPP Maximum Power Point
OC Only Current Photovoltaic
P&O Perturbation and Observe
PV Photovoltaic
SC Short Circuit Current
9.1 Introduction
Solar energy is one of the most important energy, which is environmentally friendly
such as clean, inexhaustible and free, among the renewable energy sources [1].
Studies on solar photovoltaic (PV) energy generation system were promoted in last
two decades. The main application of PV systems are in stand-alone (water
pumping, lighting, electrical vehicle, etc.), hybrid and grid-connected (PV power
plants) configuration. Stand-alone PV power generation system is considered as
good alternative for places that are far from conventional power generation/
transmission/distribution system. Such systems applications presented in two scale:
application at a small scale from 1 to 10 kW and stand-alone PV system in size,
from 10 to 100 kW [2]. In hybrid PV systems, more than one type of electricity
generator, such as wind turbine, fuel cell, diesel generator, etc., is employed. In
grid-connected PV systems, it is usual practice to connect PV system to the elec-
tricity grid [3]. A grid-connected PV system provides parallel work with the already
established electricity grid and number of PV system features are determined by this
connection to the utility grid [4]. PV capacity in the world, including stand-alone,
hybrid and grid connected, is rapidly increasing. Figure 9.1 shows the global PV
total capacity from 2004 to 2014.
It is clear from the Fig. 9.1 that, 40 GW capacity added in 2014 and also more
than 60% of all PV capacity in operation worldwide at the end of 2014 was added
over the past 3 years [5].
PV generation systems have two big problems; PV conversion efficiency is very
low and PV electricity generation is effected from changing of weather condition
[6]. PV output varies periodically in a year and in a day, and is not stable due to
environmental condition. Accordingly, in order to increase PV output and PV
efficiency, it is crucial to analyze PV output considering solar radiation, tempera-
ture, wind speed, shadow, etc.
Maximum power point trackers (MPPTs) are employed for extracting power
from photovoltaic (PV) panels. MPPTs enforce the solar modules to operate at MPP
under the fluctuations of ambient conditions. Therefore, they take a vital role for
9 Photovoltaic System: Case Studies 237
Fig. 9.1 PV total installed capacity from 2004 to 2014 in the world [5]
increasing of PV system efficiency. Different MPPT algorithms are used for the
determination of MPP. These algorithms are divided into two groups: direct and
indirect. In indirect algorithms, the operating point, where PV generator operates
with maximum power, is estimated either measuring current, voltage and radiation
values or with numerical approximations-mathematical expressions using experi-
mental data. In direct algorithms, the maximum power point is not obtained by
procedures on the contrary to indirect algorithms; the system is forced to operate at
MPP. Direct and indirect methods used for determination of maximum power point
are examined in the literature. A detailed review of these algorithms is done by
Salas et al. [2] and advantages and disadvantages of the algorithms are given. In a
study conducted by Berrera et al. [1], seven commonly used MPPT algorithms’
MPP tracking performance are compared for two different radiation profiles under
standard test conditions. Among the seven generally adopted algorithms,
Perturbation and Observe (P&O) algorithm shows the best performance for two
different radiation profiles. M. Berrera also states that, Incremental Conductance
(IC) algorithm can be a good alternative to P&O algorithm under rapid and con-
tinuous irradiance variations. Esram et al. [7], made a comparison of nineteen
different MPPT methods according to their cost and performance. The authors state
that different algorithms can be suitable for different practice areas. Hohm et al. [8]
focus on comparison of three MPPT methods i.e., P&O, IC and Constant Voltage
(CV) algorithms, using a PV array simulator. Their performance comparison results
show that P&O algorithm is very competitive against other MPP tracking algo-
rithms and can have a better performance in excess of 97%. The study carried out
by Hua et al. [9] shows the performance comparison of voltage feedback control,
power feedback control and widely used P&O and IC MPPT methods for two
different radiation condition. In result of their comparison, among three algorithms
IC method shows best performance under two radiation conditions. Reisi et al. [10]
compares different MPPT methods with simulation models under Matlab/Simulink.
Their study introduces a classification for MPPT methods based on three categories:
hybrid, online and offline methods. As a result of their study, they provide a
selection guide of appropriate MPPT methods. Subudhi et al. [11] makes a com-
prehensive comparison study based on features, like control variables, control
strategies, circuitry and approximate costs. Their comparison results offer a useful
tool not only for the MPPT users but also the designers and manufacturers of the
PV systems. Brito et al. [12] performs the comparison of usual MPPT methods
using solar array simulator. They made a comparison between twelve methods with
respect to the amount of energy obtained from PV. The authors state that perfor-
mance differences among the best MPPTs are very slight, and these algorithms must
be evaluated according to each situation.
In this part, the case studies of MPPT system, which includes stand-alone and
hybrid PV systems, will be briefly reviewed, followed by discussion of the MPPT
modeling, design, etc. Several stand-alone and hybrid MPPT application will be
presented. Latest developments in MPPT methods will be summarized. Finally some
of the present challenges facing the MPPT techniques will be explored. This chapter
structured as follows: Sect. 9.2 describes the MPPT techniques and MPPT algo-
rithms modeling; Sect. 9.3 presents the case studies of MPPT techniques, which
includes stand-alone and hybrid PV systems; in Sect. 9.4, performance comparison
of MPPTs are depicted and at last this part concluded with related discussion.
9.2 Maximum Power Point Tracker
Different MPPT algorithms are used for the determination of MPP. These algo-
rithms are divided into two groups: direct and indirect. The direct methods are;
sampling methods, methodology by modulation and other methods. The indirect
methods are; curve-fitting, look-up table, open-circuit voltage, short-circuit current
and open-circuit voltage of test modules.
9.2.1 Modeling of MPPT Algorithms
This part is presented as a summary of our published conference paper [13]. MPPT
algorithms control the PV output to take the PV power to the maximum power
point. This control is basically changing of duty-cycle of the dc-dc converter.
Duty-cycle changing decision depends on the: PV current in only current photo-
voltaic (OC) algorithm, short circuit current in short-circuit current (SC) algorithm,
PV current and voltage in P&O and IC algorithms. Basic grid-connected PV-battery
hybrid system is depicted in Fig. 9.2.
Control part of the dashed diagram (MPPT) in Fig. 9.2 is a MPPT algorithm.
Control and Duty part of the MPPT system is depicted as a Matlab/Simulink model
in Fig. 9.3.
Fig. 9.2 Basic grid-connected PV/battery hybrid system diagram
Fig. 9.3 Matlab/Simulink model of the MPPT algorithm
Algorithm blog of the Fig. 9.3 is an actual maximum power point control of the
MPPT system. In Fig. 9.3, PV output current, PV output voltage and C (the
step-size of the algorithm) input values of algorithm. Duty is an output of the
algorithm which is converted to PWM signal.
In this part, Matlab/Stateflow based modeling methodology of four commonly
used MPPT algorithms (SC, OC, P&O and IC) are presented. Stateflow is an envi-
ronment for modeling and simulating combinatorial and sequential decision logic
based on state machines and flow charts. Stateflow provides combine tabular and
graphical representations, including flow charts, state transition diagrams, state
transition tables and truth tables to model how your system reacts to events,
time-based conditions, and external input signals. With Stateflow you can design logic
for supervisory control, task scheduling, and fault management applications [14].
9.2.1.1 Perturbation and Observe Algorithm
Perturbation and Observe algorithm uses an iterative method to extract maximum

power from PV. P&O algorithm measures the power values of PV array and then
compares the measured power with prior power to perturb the operation point of
PV. If the PV power error is positive, then it changes the duty in same direction. On
the other hand, if the PV power error is negative, then it changes the duty in reverse
direction. This duty change continues until PV output power reaches the MPP
power level. Figure 9.4 illustrate a basic flowchart and Matlab/Stateflow model of
P&O algorithm.
Fig. 9.4 a Basic flowchart, b Matlab/Stateflow model of P&O algorithm [13]

Table 9.1 Matlab/Stateflow model description of the P&O algorithm

No. Comment No. Comment
1 Beginning 5 Decrease duty
2 Calculate PPV(t1), PPV(t1) and ΔP 6 Assign VPV(t1), IPV(t1)
3 Evaluate ΔP 7 Measure VPV(t2), IPV(t2)
4 Increase duty 8 Go to 2
Detailed description of the Matlab/Stateflow model in Fig. 9.4 is given in

Table 9.1 according to the numbers in the model.
9.2.1.2 Incremental Conductance Algorithm
IC algorithm forced the PV to the MPP based on the observation of conductivity by

taking the instantaneous output voltage and output current of PV. Conductivity
observation is based on the differentiation of PV power with respect to the PV
voltage and setting result to zero. Equations (9.1) and (9.2) give the conductivity
observation of the IC method [2].
dPPV dVPV dIPV dIPV

¼ IPV þ VPV ¼ IPV þ VPV ¼0 ð9:1Þ
dVPV dVPV dVPV dVPV
IPV dIPV
¼ ð9:2Þ
VPV dVPV
In (9.2), the right side of the equation is the incremental conductance and the left
side of the equations is the negative conductivity. Figure 9.5 shows the basic
flowchart and Matlab/Stateflow model of IC algorithm.
9.2.1.3 Only Current Photovoltaic Algorithm
In the OC algorithm, the PV is forced to operate at the maximum power point by

using only PV current. Equation (9.3) can be obtained from boost type dc-dc
converter equation.
PPV ¼ VPV IPV ¼ Vo ðIPV ð1 DÞÞ ¼ Pboost Vo ð9:3Þ
In the OC algorithm, PV current and PV voltage are measured and then PV

power is calculated. Then duty-cycle is changed. PV current and PV voltage are
measured again and then PV power calculated. PV power compares with prior one.
Fig. 9.5 a Basic flowchart, b Matlab/Stateflow model of IC algorithm [13]

Table 9.2 Matlab/Stateflow model description of the IC algorithm

1 Beginning 5 Evaluate conductance
2 Assign VPV(t1), IPV(t1) 6 Increase duty
3 Measure VPV(t2), IPV(t2) 7 Decrease duty
4 Calculate dV and dI 8 Go to 2
Table 9.3 Matlab/Stateflow model description of the OC algorithm

1 Beginning 5 Evaluate duty
2 Measure IPV(t1), calculate PPV(t1) 6 Increase/decrease duty
3 Measure IPV(t2), calculate PPV(t2) 7 Assign PPV(t1)
4 Evaluate PPV(t1) and PPV(t2) 8 Go to 3
As a result of the comparison, algorithm decides to increase/decrease the

duty-cycle. Figure 9.6 shows the basic flowchart and Matlab/Stateflow model of the
OC algorithm.
9.2.1.4 Short Circuit Current Algorithm
SC algorithm depends on the linear relation between maximum power point current
and PV short circuit current. The relation is actually a proportional constant ðkÞ.
The proportional constant mainly depends on the fill factor, solar cells fabrication
technology and the environmental conditions [2].
IMPP
k¼ ¼
~ Constant\1 ð9:4Þ
ISC
SC algorithm flowchart and Matlab/Stateflow model are depicted in Fig. 9.7.

9.3 Case Studies of the MPPT Algorithms
Many researches in the literature [7–13, 15–21] investigated different MPPT sys-
tems. Some of these studies are given below as stand-alone and hybrid application.
Fig. 9.6 a Basic flowchart, b Matlab/Stateflow model of OC algorithm [13]

Fig. 9.7 a Basic flowchart, b Matlab/Stateflow model of SC algorithm [13]

Table 9.4 Matlab/Stateflow No. Comment No. Comment

model description of the SC
algorithm 1 Beginning 5 Evaluate IMPP
2 Assign ISC(t1) 6 Decrease duty
3 Measure ISC(t2) 7 Increase duty
4 Calculate IMPP 8 Go to 2
9.3.1 Stand-Alone Applications
Stand-alone applications are given in this part as a summary of our previous

published paper [15, 16].
9.3.1.1 Widely-Used MPPT Algorithms Comparison
In this part, unlike the MPPT performance studies which are investigated before,
four commonly used MPPT algorithms performances are compared under real
ambient conditions. The MPP tracking systems are realized with an experimental
setup, which is capable of running four commonly used MPPT algorithms (P&O,
IC, OC and SC). As a result under real environmental condition, the performances
of the MPPT algorithms are measured and compared [16].
The realized experimental setup is constituted of five main elements: control
unit, the dc-dc converters, battery, two identical PV panels and a power analyzer for
measuring PV module output values. Figure 9.8 shows the block diagram of the
Fig. 9.8 Experimental test

system diagram for
performance evaluation of
MPPT methods [15]
Fig. 9.9 a Dc-dc converter circuit, b converter power boards [15]
experimental setup. Identical boost type dc-dc converters are connected identical
PV panels which are fixed at same position. Dc-dc converters are controlled by the
same controller (dSpace). 24 V battery bank is connected to the dc-dc converters’
outputs [15].
Step-up type dc-dc converters are employed for the comparison. All of four
above mentioned MPPT algorithms can be easily obtained by changing the control
algorithm in the control system. Detailed information about dc-dc converter is given
in [1]. Converters are operated at 35 kHz frequency. Dc-dc converters control is
based on the current and voltage measurements of PV panels which are obtained by
hall-effect sensors. Designed power boards and dc-dc converter circuit are depicted
in Fig. 9.9. Figure 9.9a also shows that, there is a S1 switch which is necessary for
measuring the short circuit current of PV for the SC method.
MPPT algorithms are designed in Matlab/Stateflow Toolbar. Dc-dc converters
are controlled by running the algorithms in the dSpace. Herein, algorithm codes are
generated in Matlab-Simulink. Basic control scheme of the test bench is shown in
Fig. 9.10.
Matlab/Simulink based dSpace control is used for performance comparison of
SC, IC, OC and P&O MPPT algorithms. MPP control diagrams for each of algo-
rithm modeled in Matlab/Simulink individually. The algorithm in Fig. 9.10 is a
Matlab/Stateflow based designed subsystem of the control unit. The input C (step
size of the algorithms) is selected as 0.01 to compare MPPT algorithms under the
same conditions.
9.3.1.2 Performance Comparison of MPPT Algorithms for Vehicle

Integrated PV Modules
This part presents performance comparison of IC, P&O and OC algorithms

examined under real environmental conditions. Performance of algorithms are
comparing as a double groups on a special test bench. Potential performance of
MPPT algorithms regulating output of solar modules on a vehicle is assessed by
controlled moving modules at the special test bench [16].
Fig. 9.10 Basic Matlab/Simulink based control scheme of MPPT system [15]
The test bench employed for performance evaluation of MPPT algorithms is

constituted from a measurement system (power analyzer) for measuring output
power of PV modules, two identical dc-dc converters, two identical PV modules
and control unit for employing three MPPT algorithms. The basic block scheme of
the test bench is illustrated in Fig. 9.11. PV modules are mounted on test platform
to simulate fluctuating irradiance of solar panels on a vehicle in motion. This
platform can change both azimuth angle and tilt angle of PV modules. In this study,
the platform is moved regarding to a defined motion loop to test realizations of three
MPPT algorithms with variation of solar irradiance.
Motion loop of the platform consist of four time intervals and motion ratio of
platform is increased at each time interval to simulate fluctuation of solar irradiance
with different ratios. This motion loop makes possible to verify responses of MPPT
algorithms to the variation of solar irradiance. Consequently, the test platform can
simulate instant solar irradiance changes that MPPTs will face for vehicle integrated
solar panels. Figure 9.12 shows an example of change in solar radiation of motion
loop.
Motion loop of the platform is composed of four time intervals to simulate
changing of solar irradiance with higher ratios at each time interval. This effect is
simulated by making solar panels move in four different ways with four different
motion loops of the platform. Between 0 and 50 s time intervals, only tilt angle
slightly changes by time. Between 50 and 100 s time intervals, tilt angle changes
are considerably much bigger than first time interval. 100–160 s time intervals are
Fig. 9.11 Basic block

scheme of the test bench [16]
Fig. 9.12 An example of

change in solar radiation in
motion loop [16]
for changing of both azimuth angle and tilt angle at the same time. Between 160
and 220 s time intervals, while tilt angle of platform is constant, only direction of
panels is altered rapidly. As it is understood from each interval, for getting maxi-
mum radiation variation, motion of platform is increased from first interval to last
interval. Motion loop is designed this way, to measure success of each MPPT
algorithms under high variation of solar irradiance which they should deal with for
vehicle integrated solar panels.
In the designed test bench, dSPACE is employed as a control unit. This controller
manages both control signals of dc-dc converters and defined motion loop simul-
taneously. Algorithm codes for defined motion loop and dc-dc converters are gen-
erated in Matlab/Simulink. Three different control algorithm codes for converters are
generated for each MPPT algorithm and implemented as double combinations of
three algorithms to compare each other. A sample control diagram, which is modeled
in Matlab/Simulink, for one of the comparison is shown in Fig. 9.13.
Fig. 9.13 Basic Matlab/Simulink control diagram of the test bench [16]
9.3.2 Hybrid System Application
Hybrid system applications are given in this part as a summary of our previous
published paper [17]. In this part, wind-solar-battery hybrid system constructed on
the roof of Electrical Engineering Department Building, Yildiz Technical
University, Istanbul, Turkey is studied. A wind turbine which is equipped with
permanent magnet synchronous generator (PMSG), two PV arrays and battery
group are employed as hybrid system equipment in the system. In addition, an
inverter, a MPPT controller and a hybrid charge controller are employed as power
conditioner unit. A measurement and a data logging system are also exist in the
hybrid system. It is planned to realize reliability analysis and wind/solar energy
potential of investigated area with measurements.
Hybrid system has eight PV modules. PV modules are divided into two groups.
One of the groups (PV1: four modules) is connected to hybrid charge controller
with wind turbines. The other PV group (PV2: four modules) is connected to the
MPPT charge controller. Hybrid charge controller and MPPT charge controller are
connected to 24 V DC bus. Battery group has two series (for reaching the DC bus
voltage level) and two parallel elements. Figure 9.14 illustrates the basic block
scheme of the wind-solar-battery hybrid system.
Fig. 9.14 Block diagram of the wind&PV hybrid system [17]
Hybrid charge controller has 1000 W nominal power which is designed for low
power wind-solar applications which controls the wind turbine and PV1. Hybrid
charge controller adjusts the sources depends on battery voltage level. Moreover,
controller also employed for some protection features such as solar wind turbine
automatic brake, battery over voltage, cells reverse charging, etc.
MPPT charge controller controls the PV2. MPPT charge controller adjust the
PV2 depends on the battery voltage level. Moreover, controller also employed for
some protection features such as battery over/deep voltage protection, PV reverse
current protection, etc.
The inverter is employed for supplying loads power from DC bus. Inverter has
1000 W nominal power and 3% THD value. Input DC voltage range is various
from 21 to 30 V to protect batteries from over charge and deep discharge.
As illustrated in Fig. 9.15, loads are halogen lamps (100 W). Load power is
supplied by hybrid system. Three 100 W halogen lamps are connected to system as
Fig. 9.15 Loads of the system

loads. Loads can be switched individual switches to analyze dynamic response of

the system.
Measurements, control and the monitoring elements of the hybrid system are
depicted in Fig. 9.16.
Fig. 9.16 Measurements, control and monitoring elements of the hybrid system
9.4 Performance Comparison of the MPPTs
9.4.1 Widely-Used MPPT Algorithms Comparison
In this section, the experimental system performances of four algorithms are pre-
sented. In this study two identical dc-dc converters are connected to two identical
PV panels. These dc-dc converters are controlled by two different MPPT algorithms
for a period of 240 s. This comparison process is carried out until all algorithms’
comparisons with each other are done. Because two algorithms are compared
together on same platform and environmental conditions are same for each PV
panel in all comparisons [15].
In order to analyze the performance of four MPP tracking algorithms, algorithms are
experimentally compared under medium-high (540–640 W/m2) radiation level. This
comparison process is carried out between 13:35 pm and 14:35 pm time intervals on
26 Dec. 2012. Figure 9.17 shows the wind speed and ambient temperature variation
between 13:35 pm and 14:35 pm (1 h) during the comparison process. As it can be
seen in Fig. 9.17, while ambient temperature is almost stable, wind speed change
between 1–5 m/s. Wind speed decreases the PV temperature which is one of the
affecting factor of PV performance. Wind speed variation is neglected in this study due
to the PV performance is mainly affected from temperature and radiation. The output
power values of PV modules employed by each algorithm are depicted in Fig. 9.18.
Numerical result of these comparisons are given in Table 9.5. The test results
show that real environmental conditions, IC algorithm is the most successful MPPT
algorithm. However P&O algorithm performance is very close to the IC algorithm.
When the P&O algorithm is optimized, the MPP tracking performance of IC and
P&O algorithms will be the same. The IC algorithm success based on [15];
• IC algorithm oscillates around the MPP less then P&O algorithm,
• IC algorithm does not diverge from MPP under rapidly changing radiation,
• IC algorithm uses PV current and voltage to track the MPP,
• IC algorithm does not cut the power flow for measuring the PV current/voltage.
Fig. 9.17 Wind speed and

temperature variation during
the comparison process [15]
Fig. 9.18 Comparison results of a IC and P&O, b OC&IC, c OC and P&O, d SC and OC, e IC
and SC, f SC and P&O algorithms [15]
9.4.2 Vehicle Integrated Solar System
This part presents performance comparison results related to the three MPPT
algorithms. Three different algorithms compared as couples between each other and
power values are measured. Power output values of each algorithm and the solar
radiation variation on PV modules are illustrated in Fig. 9.19.
In Fig. 9.19b, power output result of OC and IC algorithms are depicted.
Figure 9.19 confirms natural expectation, that power outputs of both algorithms are
Table 9.5 Numerical results Comparison Algorithm Energy Delta energy (%)
of the comparisons [15]
Delta %
P&O & IC P&O 6376 IC 1.835
IC 6493
OC & IC OC 6342 IC 2.806
IC 6520
OC & P&O OC 6291 P&O 4.991
P&O 6605
SC& IC SC 6097 IC 19.28
IC 7273
SC& P&O SC 6035 P&O 19.27
P&O 7198
SC&OC SC 5862 OC 14.35
OC 6703
Fig. 9.19 a Solar radiation and comparison results of b OC and IC, c OC and P&O, d IC and
P&O algorithms [16]
similar with solar radiation. On the other hand, power-time graph in Fig. 9.19b also
shows that IC algorithm is more successful at tracking MPP. Total energy data
acquired from the power analyzer that are given in Table 9.6 also proves it.
Percentage of energy differences in Table 9.7 indicates that efficiency of IC algo-
rithm is 5.536% higher than OC algorithm.
Table 9.6 Numerical results Comparison Algorithm Energy (J)

of the four section of motion
Sect. 1 Sect. 2 Sect. 3 Sect. 4
loop [16]
OC & IC OC 1234 1132 1233 826
IC 1309 1190 1306 865
OC & P&O OC 1290 1173 1273 881
P&O 1365 1247 1324 925
P&O & IC P&O 1348 1209 1305 969
IC 1430 1300 1385 1026
Table 9.7 Energy Comparison Delta Energy differences (Delta energy) (%)
differences results of the four
Sect. 1 Sect. 2 Sect. 3 Sect. 4 Total
section of motion loop [16]
OC & IC IC 6.077 5.123 5.920 4.721 5.536
OC & P&O P&O 5.813 6.308 4.006 4.994 5.284
P&O & IC IC 6.083 7.526 6.130 5.882 6.416
In Fig. 9.19c, power output result of OC and P&O algorithms are depicted. This
time P&O algorithm is more successful at tracking MPP. Total energy data in
Table 9.6 which is acquired from the power analyzer also verifies this inference.
Delta energy data in Table 9.7 reveals that P&O algorithm has 5.284% more total
efficiency as compared to OC algorithm.
In Fig. 9.19d, power output result of winners of first and second comparisons are
depicted. As depicted in Fig. 9.19d P&O algorithm is less successful than IC
algorithm at finding MPP. It is also stated in Tables 9.6 and 9.7 with total energy
data and energy differences data respectively. IC algorithm is more successful than
P&O algorithm.
The test results show that IC algorithm is the most successful maximum power
point tracker algorithm under fast altering solar irradiation. However this perfor-
mance is not based on success of IC algorithm at radiation fluctuation. This situ-
ation is detected from energy differences data in Table 9.7. Although radiation
variation values increased at every section, delta energy is not increasing regularly
with each section. With presence of these results, it is not possible to show that IC
algorithm is more adaptive to variation of irradiance than P&O and OC algorithms.
Besides it can be claimed that IC algorithm is the most successful one between three
algorithms at tracking MPP also in the case quick variations at irradiation.
9.4.3 Hybrid System
Hybrid system results are given in this part as a summary of our previous published
paper [18]. In this study, the effects of two different charge controllers on PV panel
performances are investigated as given in Fig. 9.20. The weather conditions and
electrical values of the system are recorded simultaneously with a weather station
Fig. 9.20 Block diagram of the investigated system [18]
and a data logger. The analyzed data are recorded between 10 Oct. 2013 and 30
Oct. 2013 for 20 days. The variation of solar radiation on the location is shown in
Fig. 9.21.
Current and voltage values of PV panels are recorded for duration of a minute.
Output voltage and current values of two panels are shown in Fig. 9.22. Vmppt and
Imppt define output voltage and current of PV2, Vhybrid and Ihybrid identify output
voltage and current of PV1, and Vbus describes DC bus voltage.
DC bus voltage variation is kept in desired limit between 21 and 30 V as clearly
seen in Fig. 9.22. The output voltages of both panels drop to 10 V when the solar
Fig. 9.21 Variation of the

solar radiation during the
comparison [18]
Fig. 9.22 a Output voltage, b output current of the PV system [18]
radiation is minimum. Output voltage of PV2 increases up to 45 V when the

radiation is maximum whereas maximum output voltage of PV1 stays under 30 V.
If the current variations are examined, although PV1 gives higher current than
PV2, similar characteristics can be seen in both systems. In the days between 9th
and 15th that solar radiation is maximum in, output current of PV2 is minimum
since Controller 2 monitors and controls the DC bus voltage to prevent the voltage
exceed 27.8 V.
Output power variations of panels are shown in Fig. 9.23. The output power
characteristic varies in proportion to solar radiation as seen in Fig. 9.23. Operation
characteristic of PV2 panel group controlled by Controller 2 depending on DC bus
voltage can be seen clearly in the Fig. 9.23.
Since DC bus voltage reaches to maximum value between 9th and 15th days,
Controller 2 restricts the related panel output power. This feature contributes to
system stability. In the same time interval, there is no limitation in other panel
output power. Although Fig. 9.23 creates perception that the generated power by
PV1 is higher than PV2, the total energy obtained from PV1 is 39,826 kWh
whereas PV2 is 45,366 kWh in the measurement interval. The energy generated
from PV2 is higher by 16.81% from PV1.
Fig. 9.23 a Output power of PVhybrid, b output power of the PVmppt [18]
The power difference between two panel groups is caused by Controller 2 that
always provides maximum power point operating. While output power of PV1
oscillates in high range, there is stable output power in PV2. It clearly seen from
Fig. 9.23a that DC bus voltage is higher than 27.8 V between the days 9th and 14th.
9.5 Conclusion
In this part, the case studies of MPPT system, which includes stand-alone and
hybrid PV systems, are briefly reviewed, followed by discussion of the MPPT
modeling, design, etc. Several stand-alone and hybrid MPPT application are pre-
sented then latest development in MPPT methods will be summarized and finally
some of the present challenges facing the MPPT techniques are explored. Case
studies and results are presented. Based on the chapter general results, main con-
clusions are as follows:
• MPPT system has a vital importance for PV applications.
• MPPT system improve the efficiency of PV systems.
• MPPT algorithms performance is affecting environmental conditions.
• The most successful MPPT algorithms among the four commonly used MPP
methods is IC.
• The success of IC algorithm is coming from its success in comparison with P&O
and OC algorithms at all ambient conditions.
• SC algorithm turned out to be the worst one.
• MPPT controllers have different algorithms and operating principle change the
performance of PV modules.
• Controller selection has much importance as panel type, application area and
battery size in photovoltaic system design.
References
1. Berrera M, Dolara A, Faranda R, Lova S (2009) Experimental test of seven widely-adopted

MPPT algorithms. In: IEEE Bucharest Power Tech Conference, 2009, pp 1–8
2. Salas V, Olias E, Barrado A, Lazaro A (2006) Review of the maximum power point tracking
algorithms for stand-alone photovoltaic systems. Sol Energy Mater Sol Cells 90:1555–1578
3. Kalogirou S (2009) Solar energy engineering: processes and systems. Academic Press, New
York
4. Markvart T, Castaner L (2003) Practical handbook of photovoltaics fundamentals and
applications. Elsevier, UK
5. Renewable Energy Policy Network for the 21st Century (2015) Renewables 2015 global
status report. http://www.ren21.net/ren21activities/globalstatusreport.aspx
6. Mutoh N, Ohno M, Inoue T (2006) A method for MPPT control while searching for
parameters corresponding to weather conditions for PV generation systems. IEEE Trans Ind
Electron 53(4):1055–1065

techniques. IEEE Trans Energy Convers 22(2):439–449
8. Hohm DP, Ropp ME (2003) Comparative study of maximum power point tracking
algorithms. Prog Photovolt Res Appl 11:47–62
9. Hua C, Shen C (1998) Comparative study of peak power tracking techniques for solar storage
system. In: IEEE applied power electronics conference and exposition, pp 679–685
10. Reisi AR, Moradi MH, Jamas S (2013) Classification and comparison of maximum power
point tracking techniques for photovoltaic system: a review. Renew Sustain Energy Rev
19:433–443
11. Subudhi B, Pradhan RA (2013) Comparative study on maximum power point tracking
techniques for photovoltaic power systems. IEEE Trans Sustain Energy 4:89–98
12. Brito MAG, Galotto L, Sampaio LP, Melo GA, Canesin CA (2013) Evaluation of the main
MPPT techniques for photovoltaic application. IEEE Trans Ind Electron 60:1157–1167
13. Durusu A, Nakir I, Tanrioven M (2014) Matlab/Stateflow based modeling of MPPT
algorithms. Int J Adv Electron Electr Eng 3:117–120
14. MatWorks (2015) Stateflow support documents (Support). http://www.mathworks.com/help/
stateflow/index.html
15. Durusu A, Nakir I, Ajder A, Ayaz R, Akca H, Tanrioven M (2014) Performance comparison
of widely-used maximum power power point tracker algorithms under real environmental
conditions. Adv Electr Comput Eng 14(3):89–94
16. Nakir I, Durusu A, Ugur E, Tanrioven M (2012) Performance assessment of MPPT
algorithms for vehicle integrated solar system. In: 2nd IEEE energy conference and
exhibition, pp 1034–1038
17. Arikan O, Isen E, Durusu A, Kekezoglu B, Bozkurt A, Erduman A (2013) Introduction to
hybrid system-Yildiz Technical University. IEEE Eurocon 2013:1145–1149
18. Arikan O, Kekezoglu B, Durusu A, Isen E, Erduman A, Bozkurt A (2014) Comparison of
charge controllers on PV performance: an experimental study. Int J Adv Electron Electr Eng
3:121–125
19. Nakir I, Durusu A, Akca H, Ajder A, Ayaz R, Ugur E, Tanrioven M (2016) A new MPPT
algorithm for vehicle integrated solar energy system. J Energy Resour Tech 138
(2):021601-1-9
20. Hsieh GC, Hsieh HI, Tsai CY, Wang CH (2013) Photovoltaic power-increment-aided
incremental-conductance MPPT with two-phased tracking. IEEE Trans Power Electron
28:2895–2911
21. Wang F, Wu X, Lee FC, Wang Z, Kong P, Zhuo F (2014) Analysis of unified output MPPT
control in subpanel PV converter system. IEEE Trans Power Electron 29(3):1275–1284
Chapter 10
Maximum Power Point Tracking
Algorithms for Partial Shaded PV Systems
Ibrahim Sefa, Necmi Altin and Saban Ozdemir
Abstract Photovoltaic modules have nonlinear current-voltage (I-V) characteristics.

Thus, output power of the photovoltaic module varies with module specification and
its operation point. It means that, the photovoltaic system generates maximum power
at a single operation point for an environmental condition such as the irradiation level
and angle, ambient temperature level etc. In addition, energy conversion efficiency
varies with load level and operation point of the photovoltaic system. Since these
parameters are variable, operation point of the photovoltaic system should be con-
trolled to get maximum output power and maximum energy conversion efficiency.
This action is called as maximum power point tracking. The maximum power point
tracking action is usually performed with a power electronics converter. A number of
maximum power point tracking methods have been introduced to obtain fast response,
especially in rapidly-changing atmospheric conditions, low oscillation and higher
energy conversion efficiency values. However, most of these methods are effective for
uniform solar irradiation conditions. If the solar irradiation is non-uniform, the
power-voltage (P-V) curve of the photovoltaic module or array has multiple peak
points: Several local maximum power points and one global maximum point. In this
case, traditional maximum power point tracking methods determine the nearest peak
power point, which may be a local maximum point. Thus, some improved maximum
power point tracking methods have been proposed to determine the global maximum
power point of the photovoltaic system even under partial shading conditions.
A discussion on different maximum power point tracking methods for the solution of
these problems will be given and the most powerful techniques in the literature will be
outlined.
I. Sefa N. Altin (&)

Department of Electrical and Electronics Engineering, Faculty of Technology, Gazi
e-mail: naltin@gazi.edu.tr
I. Sefa
e-mail: isefa@gazi.edu.tr
S. Ozdemir
Vocational School of Technical Sciences, Gazi University, Ostim, Ankara, Turkey
e-mail: sabanozdemir@gazi.edu.tr

262 I. Sefa et al.
Keywords Photovoltaic MPPT Partial shading condition Global maximum

power point
Abbreviation
ACA Ant colony algorithm
CC Constant current
CST Current sweep technique
CV Constant voltage
FLC Fuzzy logic control
GP Global peak
IC Incremental conductance
LP Local peak
MPP Maximum power point
MPPT Maximum power point tracking
PC Pilot cell
P&O Perturb and observe
RCC Ripple correlation control
PSC Partially shading condition
PSO Particle swarm optimization
PV Photovoltaic
10.1 Introduction
The increasing word power demand, limitation of the conventional energy sources
and increasing environmental concerns increase the number of the studies on
renewable energy sources. As a result of these studies, photovoltaic (PV) systems
which are safe, modular, clean, inexhaustible, maintenance free and nearly reach-
able in all over the word have gained importance. Their modular structure, long
lifetime, decreasing cost and recent advancements in static power conversion
technologies allows implementing PV systems from a few Watts to MW power
level. Because power level of the PV module is limited, numbers of modules are
connected in series and parallel to obtain the high power levels [1–5].
The power generated by PV modules is depend on the module specifications,
environmental issues such as pollution and temperature and solar irradiation level.
So, the output power varies with operation conditions and this may reduce the
module energy conversion efficiency. Therefore, maximum power point tracking
(MPPT) methods are used to change the operation condition according to these
effects to obtain maximum energy conversion efficiency. The perturb and observe
(P&O) method, the incremental conductance (IC) method, the constant voltage
(CV) method, the constant current (CC) method, the pilot cell (PC) method, the
10 Maximum Power Point Tracking Algorithms … 263
ripple correlation control (RCC) method, the current sweep technique (CST), the
parasitic capacitance method are the well-known MPPT methods. These methods
have different specifications in complexity, convergence speed, module depen-
dency, and track the maximum power point (MPP) of the PV system under uniform
insolation condition where only one peak point exists in the P-V curve [2–5].
Unfortunately, there are many factors that affect the operation of each module
and generate differences on the module characteristics. In this condition, multiple
peaks may appear in P-V curve of PV system. The conventional circuit models of
PV modules are changed because of passing clouds, manufacturing tolerances, dirt,
aging, soldering errors or micro cracks which occur during production, trans-
portation or installation periods, and this cause degradation of PV system perfor-
mance [6]. This effect results an extra heat on PV module and may cause localized
hot spot. A bypass diode is connected in parallel with each PV module to overcome
these negative effects [7]. Especially non-uniform insolation of PV system because
of passing clouds or another cause of shadow has great effect on system perfor-
mance. The preventive methods can be grouped into three categories as
(1) Modification of power converter topologies and their interconnections,
(2) Modifying of the PV module connection (or generating dynamic connection
patterns), (3) Modifying the MPPT algorithm in order to obtain maximum power
even in multiple peak power points [6].
Characteristics of the shaded module will be different from the other ones under
partially shading condition (PSC). So, using separate converter for each module
seems as an effective way to prevent the negative effects of this condition. Thus, the
power converter topologies are changed from centralized converters to string
converters or to module integrated converters [6]. Also different connection patterns
are investigated to mitigate the partial shading effects because performance
degradation rate is not depending on shading area but depends on shading pattern,
configuration of the array and location of the shaded module in the array []. In
Fig. 10.1, P-V curve examples of a PV array for uniform and non-uniform inso-
lation conditions are depicted [7]. As it is seen from the figure, the P-V curve can
differ significantly. In addition, reconfiguration methods those dynamically change
the array configuration with active switches have been proposed to adapt the PV
Fig. 10.1 P-V characteristics

of a PV array with bypass
diodes for different
irradiations
264 I. Sefa et al.
system different PSCs. But these methods increase the system cost and complexity
[8, 9].
Since the P-V curve of the PV system has multiple peaks with several local
peaks (LPs) and a global peak (GP), the MPPT action is more complicated under
PSCs. Conventional MPPT methods such as P&O and IC likely detect one of the
LPs and oscillate around the vicinity of this LP. So, these methods cannot track the
global MPP of partially shaded PV system. This may cause relevant energy loss.
These issue has investigated lately, because the initial PV applications are usually
rooftop and in rural areas [10]. Thus, enhanced MPPT methods have been inves-
tigated to improve the performance of the PV system under PSCs with the
increasing usage and power levels of PV systems. Initial studies on MPPT methods
under PSCs are usually based on modification of the conventional methods. But
with advancements of the microcontroller and their computational capability new
methods based on soft computing have also been investigated. Although main
objective of these methods and the conventional MPPT methods are same, they
have different specifications in terms of complexity, convergence speed, calculation
effort, flexibility, steady state oscillation and etc.
10.2 Characteristics of Partially Shaded PV Module
Output electrical parameters of photovoltaic modules exhibit nonlinear behavior.

Several PV cell equivalent circuit models are presented in the past literature.
Generally, the model called as double-diode model and one-diode model (which is
also called or five parameters model) are used [11, 12]. The one-diode model is
obtained from a reduction of double-diode model. Also, simpler equivalent circuits
are obtained by assuming that the equivalent series resistor of cell (Rs) is zero and
parallel equivalent resistor of cell (Rp) is infinity. The equivalent circuit of one
diode model is given in Figure “One-diode equivalent circuit of PV cell” (shown in
other chapter of this book).
Two-diode model which is one of the common PV equivalent circuit models is
obtained by adding second parallel diode to the model. The two-diode model
provides more accurate V-I characteristics than one-diode model. In one diode
model, the ideality factor A is assumed as constant. However, A is a function of
voltage and its value is closer to one at high voltage levels approaches two at lower
voltage. The junction recombination is modeled by adding a second diode in par-
allel with the first diode. This PV cell model is shown in Figure “Two-diode
equivalent circuit of PV cell” (shown in other chapter of this book). Equations for
the output current of the PV cell are also given in another chapter of this book for
one-diode and two-diode equivalent circuits.
The output current of a PV cell has nonlinear characteristics. This results a
nonlinear relation between the PV voltage and PV current and also between PV
voltage and PV power. In addition, this relation is affected from operation condi-
tions and environmental effects such as solar irradiation, ambient temperature,
Fig. 10.2 a The P-V and I-V curves of the PV module. b P-V curves of the PV module for
different solar irradiation conditions
pollution etc. The PV module current and power curve versus module voltage is
given Fig. 10.2a. The module voltage, current and power variations for different
solar radiation levels are given Fig. 10.2b. As it is seen from the figure, each solar
irradiation value defines new P-V and I-V curves. Therefore, the operation point of
the PV system has to be changed with the environmental effects to track the
maximum power points of each curve and to obtain maximum efficiency.
Otherwise, PV system may generate lower power than it can.
All curves in Fig. 10.2 are produced for ideal conditions. However, generally
this is not valid all the time. Practically, it observed that dirtiness, shadow, unequal
module specifications, aging of modules and usage of the by-pass diodes increase
the nonlinearity and affect these curves [13, 14].
10.2.1 Hot Spot and Partial Shading Effect
Hot spot effect can be defined as occurring high temperature in any place of
module. Shading, pollution, transportation damage, cracks during the installation
and wrong soldering are the most important reasons of hot spot. The hot spot effect
reduces the efficiency of the PV module and also if it goes on for a long time, it can
shorten the life of the module. Moreover, PV modules can be out of order because
of this effect [15].
Any size of PV installations from low power rooftop systems to high power PV
power plant can meet partially or fully shading effect. PV modules and connections
should be checked for maintenance purpose and some parameters should be
monitored for early intervention. Thus, early aging and hot spot effects can be
minimized and energy production can be maximized.
266 I. Sefa et al.
When the PV modules run into partially shading effects, they perform different
output functions depending on module connections and shading formation. The
shadowed cells in a serial, parallel or serial-parallel mixed connection stop the
energy generating source operation and operate like an energy consuming load.
This is the main reason of this variation on output function. The PV system contains
series or parallel connected modules to reach high power levels. Thus above
described process may be encountered any time in a day. The shadowed cell gets
reverse bias and starts to draw current. Namely, shadowed cell does not produce
energy but consume energy produced from the other cells. Then, total amount of
generated energy is getting down and hot spot occurs around the reverse biased
cells. This condition is depicted for serial connected PV modules in Fig. 10.3 [16].
Therefore, in a PV system which consists of several PV cell and modules
connected in serial, parallel or serial-parallel to increase voltage and/or current
levels, the P-V curve of entire PV system becomes different from ideal P-V curve
which is given in Fig. 10.2a. The resultant P-V curve can be changed according to
the module connection type and situation of the shaded module or modules. The
partial shading effect on P-V curve is depicted in Figs. 10.4, 10.5 and 10.6 for
serial, parallel and serial parallel mixed connection conditions, respectively [12].
Here, UR and PSC indicate the uniform insolation and the PSC, respectively.
The decrease in energy generation because of the partial shading is not only
related with the size of shaded area (or number of shaded modules) but also it is
related with the shaded pattern and situation of shaded PV modules and their
connection. Therefore, configuration of the PV system such as numbers of serial
and parallel connected modules has significant effect on performance of the PV
system under PSCs. So, some connection strategies are investigated to reduce
negative effects of the partial shading such as bridge link, total cross tied,
honey-comb structures, etc. [17].
Commercially available PV modules contain one or more parallel connected
by-pass diodes to protect the modules from the negative voltages caused by partial
Fig. 10.3 Shading effect of the serial connected PV modules

Fig. 10.4 The partial shading effect on P-V curve for serial connected PV modules
Fig. 10.5 The partial shading effect on P-V curve for parallel connected PV modules
Fig. 10.6 The partial shading effect on P-V curve for serial-parallel mixed connected PV modules
shading and so prevent hot spot effects. The effect of the connected by-pass diode
on PV module and serial connected cell string is depicted in Fig. 10.7 [18, 19].
As well as by-pass diodes, blocking diodes are also used between photovoltaic
module strings. These diodes protect the modules from reverse power flow between
strings. However, both bypass and blocking diodes cause power loses and increase the
system cost and nonlinearity of the string behavior. Although, using by-pass diodes is
a well-known protection method, some active structures like power converters are
also studied in the recent years instead of these by-pass diodes [13, 14].
While objects with high volume such as building, tree, chimney and etc. cause
partial shading, objects with low volume like pipe, wire, cable, rope can also cause
partial shading of the PV. However, there is no direct relation between shadowed
area of modules and degradation of efficiency. Therefore, degradation amount of
efficiency can be more than the amount of shadowed area. Namely, a wire or a pipe
based shadow can cause dramatic degradation on efficiency [20]. Cloudy or shaded
operating conditions cause degradation of produced energy up to 70% even in the
268 I. Sefa et al.
Fig. 10.7 The effect of the connected by-pass diode on a PV module, b serial connected string
power plants [21]. Moreover, if the reason of whole or partial shading is a moving
object then, degradation of energy production can be dynamic [22].
The conventional equations given for the PV cell can be easily extended for a
PV array. The PV modules are composed of serial, parallel and serial parallel mixed
connected cells. Although several connection possibilities exist to form the arrays,
serial, parallel and serial parallel mixed connected modules are usually used in
arrays. The superposition can be applied them to obtain the equations for a module
or an array [21]. If subscript M represents the module, NSM and NPM show the
number of serial connected modules and parallel connected modules, respectively.
The total number of modules can be written as NSM x NPM for an array. While the
PV array voltage is related with number of serial connected modules, the array
current is related with number of parallel connected module strings. Thus, current of
the array with NSM x NPM serial-parallel connected modules can be written as [21]:

NPM IscM qðVA þ RSA IA Þ
IA ¼ NPM IscM qVocM e AkB TNS NSM 1 : ð10:1Þ
e NS kT
where, IA is array current, IscM is short circuit current of module, A is the ideality
factor of diode, q is the electronic charge, kB is the Boltzmann’s constant, T is the
junction temperature, RSA is the equivalent series resistance of the array, VA is the
array voltage, VocA is open circuit voltage of the array, Ns is number of the serial
connected cells in modules. The Eq. 10.1 is valid for only uniform insolation
conditions and it should be revised for non-uniform insolation conditions such as
partial shading. The current equation can be written for array, which is composed of
NSM x NPM modules and has by-pass diodes, with some assumptions as given
below:
XNPM XNPM qðVA þ RSAx IAx VocAx Þ

IA ¼ I ¼
x¼1 Ax
I
x¼1 scAx
1 e kA kB TNS ðNSM NDx Þ
ð10:2Þ
Here, string represents the each parallel branch of the array which is composed
of serial connected modules and x is number of the string (x ¼ 1; 2; . . .; NPM ), the
voltage across on the by-pass diodes are assumed as zero, and NDx shows number of
the shadowed module in xth string [21].
10.3 Maximum Power Point Tracking Under Partially

Shading Conditions
Since the P-V curve of the PV array has multiple peak points, the conventional
MPPT methods which usually use hill climbing method or are usually based on
slope of the P-V curve may detect one of the local MPPs instead of the global
MPP. This causes considerable energy losses, which is very important for today’s
energy markets and also for investors. So, a requirement of MPPT algorithms which
can track the MPP of PV array even under PSCs is come out.
The PV inverters are used to export the energy generated by PV system to the
grid and can be grouped into two categories such as single-stage PV inverter and
two-stage PV inverter as shown in Fig. 10.8a, b, respectively. The main difference
between these inverter topologies are the number of the power converter stages.
Two-stage PV inverters exist a DC-DC converter and a DC-AC inverter. While the
DC-DC converter regulates the DC voltage and performs the MPPT, the inverter
converters DC to AC, regulates the AC voltage and/or current and exports the PV
supplied energy to the loads or to the grid. In single-stage inverters, only an inverter
exists and performs all required actions. The MPPT part is one of the key com-
ponents of the PV inverter, because it determines the inverter reference current
value and therefore inverter output power. Here, The MPPT method determines the
global MPP of the PV system by using some variables such as actual PV voltage
(V), actual PV current (I), short circuit current (Isc), open circuit voltage (Voc), solar
irradiation level (Ir), temperature (T) and etc. according to the applied method. As
explained in the previous sections, resultant P-V curve of the PV system is
dependent on both solar irradiation levels of the each modules and shading pattern.
Therefore, connection strategies of the PV modules have significant effect on the
PV system performance and different connection strategies are investigated.
Moreover, the MPPT methods examined in two categories as hardware based
MPPT methods and firmware based MPPT methods. The hardware based methods
include the module connection strategies, decentralization of the PV system. The
decentralization method recommends connecting smaller strings with less number
of modules to the grid through their own inverter or even connecting each module
to the grid through its own module integrated inverter, instead of connecting all
modules to the grid through the central inverter. The software based methods are the
algorithms proposed to determine the global MPP under any operation condition.
The most common software based methods are explained in the following sections.
270 I. Sefa et al.
Fig. 10.8 Overall control diagram of the PV inverter. a Two-stage PV inverter. b Single-stage PV
inverter
10.3.1 MPPT Methods for Partially Shaded PV Systems
In practical applications a number of PV modules are connected in series and

parallel to obtain the required power level. Therefore, non-uniform insolation
condition is very common. Because the energy conversion efficiency is very
important value for PV systems and many MPPT methods which can track the
global MPP even in PSCs are introduced [23–41].
10.3.1.1 Modified P&O Method
The P&O method is commonly used in MPPT systems because of its major
advantages such as high tracking speed and simple implementation specifications.
However, this method is based on hill climbing rule and cannot track the global
MPP when multiple peaks are occurring in P-V curve of PV system. Thus, some
modifications have done to extend the advantages of this method to PSCs. After the
experimental studies, some observations are reported about partially shaded PV
systems characteristics. It is seen that, the peak points of the P-V curve are still
about 0.8 Voc [17, 24]. In addition, while magnitudes of the peak points are in
tendency of increasing before the global MPP and they are in tendency of
decreasing after the global MPP. Therefore, P&O method is commonly used to
track the global MPP under PSCs.
Different modification schemes can be used for MPPT with P&O method under
PSCs. The PV system is started to operate at 0.85 Voc and the conventional P&O
method is used to track the MPP until a disturbance such as the PSC is detected or
an interrupt is generated. When the disturbance is detected or a timer interrupts
which is generated every predetermined time interval (for example 20–25 s) occurs,
the partial shading subroutine is activated. This situation is detected by monitoring
the power variation (DP) value. If the power variation value is higher than the
predefined critical power (DPcrit) value, this situation is labeled as PSC. The par-
tially shading subroutine searches the other peak points on the P-V curve. It is also
known that, voltage difference between the sequential peak points is almost equal to
80% of the open circuit voltage of the PV module (Voc_module). Therefore, the
operating voltage value is changed with a voltage variation step (DVx) which is
smaller than 0.8 Voc_module. The lower voltage variation value provides higher
possibility of tracking global MPP, but tracking time also increases. The large
values of DVx increase the risk of missing global MPP while reducing tracking time.
After this variation is applied, dP/dV slope of P-V curve for this operation point is
measured. If the slope is positive, the voltage variation action is repeated until
minimum operation voltage value Vmin is reached. If the slope is negative, this
means that there is a peak point in vicinity of this operation point and the P&O
method is activated to detect this peak point.
The PV power (Pmax_last) and voltage (Vmax_last) values are stored when the first
peak point is reached and these values are compared with the values obtained at
next peak point and higher power values is updated. This action is repeated towards
to left side of the P-V curve until Vmin or a peak point with lower power value is
reached. If the power value of the peak point is lower than the previous one,
operation point is slipped to last peak point with highest power value and voltage
variation action is applied towards to other direction. The dP/dV slope of the P-
V curve is measured for all voltage variation. The positive slope value represents
that there is peak point in vicinity of this operation point. This peak value is tracked
by activating the P&O method. The PV power and voltage values are read, and if
these values are higher than the Pmax_last and voltage Vmax_last, this operation is
repeated in same direction. If the obtained power value is lower than the Pmax_last or
Vmax which is determined as 0.85 Voc_module, operation point is transferred to
previous peak point [24]. This modified method operates like double Perturb &
Observe method. It generates relatively big perturbations to track the GMPP in case
272 I. Sefa et al.
of multiple peaks. However, the proposed modified P&O method may not track the
GMPP under rapidly changing solar irradiation conditions [25].
In another study, an alternative P&O method which compares instantaneous
measured power (Pm(t)) and instantaneous maximum power reference value is
proposed [26]. The instantaneous measured power is calculated by multiplying
instantaneous voltage and instantaneous current measurements. The P-V curve is
divided into two regions. The instantaneous maximum power is related temperature
and irradiance as Eq. 10.3:
PMPP ðtÞ ¼ PMPP ðTðtÞ; EðtÞÞ ¼ a½TðtÞbðEðtÞ ð10:3Þ
here, T(t) is temperature value, E(t) is irradiance value, a(t) is temperature factor and
E(t) is irradiance factor which determines the maximum power current. The tem-
perature factor is also called as voltage factor. In region I, maximum power point
current is obtained by reducing the PV array current as given below:
IðtÞ ¼ Imeas ðtÞ DI ð10:4Þ
The performance of this method is depending on the DI. In region II, current
reference is calculated by Eq. 10.5:
Pmeas ðtÞ
IðtÞ ¼ ð10:5Þ
aðTÞ
This algorithm is operated until the power error ratio becomes lower than the
predetermined level:
Pm ðtÞ Pref ðtÞ

\e ð10:6Þ
Pm ðt 1Þ
here, Pm(t) is the instantaneous measured power and Pref (t) is the instantaneous
maximum power reference. The P&O method is used to estimate a(T). This method
introduces various new coefficients which complicate the MPPT process to deter-
mine the global MPP in case of multiple local peak points.
Another P&O based MPPT method for DC-DC converter operating under PSCs
is voltage sweep method. In this method, the PV array is changed voltage from its
maximum value (near to VOC) to a minimum value (near to ISC) periodically [7]. At
each sampling cycle, the voltage reference and by the way output power of the
DC-DC converter is increased step by step and operating voltage and current are
measured and stored. This action is sustained until PV system voltage reaches the
predetermined minimum PV voltage. After determining the region of global MPP,
the conventional P&O method is run to achieve operation at the global MPP [10].
In addition, measuring the voltage of all PV modules is another method. This
method is based on the observation that the voltages of PV modules receiving
different irradiances are different [27]. The operation of this method is always start
with reference voltage of 0.85 Voc and calculates the number of PV modules. If
all voltages of PV modules are equal, this means that solar irradiation is uniform
and system operates with conventional P&O method. When the peak point is
determined, the system stores the information (voltage and current values) of this
point. Voltages of the PV modules are checked with a timer program. When voltage
of a module is different from another, this is defined as PSC and a global MPP
subroutine is called. This subroutine controls if any PV module voltage is less than
zero or not. If there is a PV module with voltage less than zero, this means that, last
peak point is left peak on the P-V curve. In this condition, reference voltage is set to
0.85 Voc and then P&O method is used to determine the global MPP. If all PV
modules’ voltages is higher than zero, this means that, last peak point is right peak
on the P-V curve. This means that some of PV modules are bypasses and do not
provide power to the output. In this condition, subroutine classified the PV modules
into two groups and determines the PV modules with low voltage. Then program
set the reference voltage to 0.85 Voc but, in this case number of modules with
lower voltages is not used calculating the array open circuit voltage (Voc). Again
P&O method is used to track the global MPP [27].
10.3.1.2 Modified IC Method
Since there are multiple peak points are observed in the P-V curve of the PV array,
conventional MPPT methods cannot track the global MPP of the PV system.
However, some of the conventional methods have been modified to track the
GMPP even in PSCs [25]. The conventional IC method is used with predetermined
linear function to track the MPP even in PSCs without any additional circuits [21].
In this method, PV voltage and current values are evaluated to determine the PSCs
by using Eqs. 10.7 and 10.8:
DVPV ¼ VPV ½n VPV ½n 1\DVSET ð10:7Þ

DIPV IPV ½n IPV ½n 1
¼ \DISET ð10:8Þ
IPV ½n 1 IPV ½n 1
here DVSET and DISET are predetermined voltage and current values and are defined
according to manufacturer data. After this situation is determined, voltage reference
of the system is set according to a linear function defined as given below:

Vo;rms
VPV ¼ IPV ½n: ð10:9Þ
Io;rms
The P-V and I-V curves of a PV system for PSC and operation of the modified IC
method are depicted in Fig. 10.9. As shown in the figure, the MPP of the PV system
under uniform irradiation is depicted as point A and this point is tracked with
conventional IC method. When a PSC occurs, operation point is changed to point
274 I. Sefa et al.
Fig. 10.9 Operation

principle of modified IC
method
B, while the real or global MPP is point D. In this condition, the IC method cannot
track the GMPP and operates in vicinity of point B. The modified IC method check
the Eqs. 10.7 and 10.8 and when both conditions are satisfied, the operation point is
changed to point C which is located at lower voltage area via the linear function
given in Eq. 10.9. When the operation point is reaches the point C, the conventional
IC method is used again to track the global MPP and to reach the point D. In
addition, using variable step size in conventional IC method improves the tracking
performance and efficiency of the system and reduces the loss [21]. Moreover, this
method is validated only two peaks condition and in case of multiple peak points at
the left side is not considered.
Furthermore, conventional IC method is modified with some observations to
track the MPPT even in PSCs. The critical observations which are also used in
modified P&O method are:
• The maximum power points occur at the multiples of 0.8 Voc which is also
used in modified P&O method is combined.
• The magnitudes of the peak points increase before the global MPP and decrease
after the global MPP.
According to these observations it is obtained that there are three possible options
for position of global MPP as shown in Fig. 10.10: (a) The global MPP is located
between the local peak points and local peaks are increasing until the global MPP
and decreasing after the global MPP, (b) The global MPP is located at left end of the
P-V curve and (c) The global MPP is located at right end of the P-V curve [25].
This method requires the Voc voltage value of the PV module and maximum
number of the serial connected modules (Nmax), and determines the duty cycle value
of the DC-DC converter. A reduced reference voltage value is determined and the
operation starts with the conventional IC method. The conventional IC method
tracks the first MPP that is depicted as peak point A in Fig. 10.10a. The duty cycle
of the converter and maximum power value for this first MPP is stored. Then the
Fig. 10.10 Possible locations of GMMP and other peaks
reference value is changed by adding 0.8 Voc to the voltage of the first MPP and
operation point of the system is moved to right side of P-V curve (around B in
Fig. 10.10a). The IC method is operated again and second MPP point is tracked.
The power and duty cycle value at this point is stored. If power value of the second
MPP is higher than the first MPP and Vmax which is equal to multiply of Voc and
Nmax is not reached, 0.8 Voc is added to the previous reference voltage value and
operation point is moved to right side of P-V curve according to observation 2. If
the next power value is lower than the previous one (for example peak point C in
Fig. 10.10a), the operation point is moved to the previous peak power point which
is stored, and then the conventional IC method is operated. If the next power value
is higher than the previous one (for example peak point H in Fig. 10.10c),
0.8 Voc is added to the previous reference voltage value and operation point is
moved to right side of P-V curve until the Vmax is reached [25].
If the power value at the second MPP is lower than power at the first MPP as
shown in Fig. 10.10b and minimum reference voltage is not attained yet, the first
MPP voltage is reduced by 0.8 Voc and operation point is moved to left side of
first MPP. The conventional IC method is used again to track the peak point. This
operation goes on until the minimum reference voltage is attained. A bias value for
the slope in conventional IC method can be used to prevent the duty cycle variation
in case of low power variations. Thus the total system efficiency is increased and
oscillation in steady state is reduced [25].
10.3.1.3 Dividing Rectangles (DIRECT) Algorithm
The DIRECT algorithm uses Lipschitz function to describe the power/voltage

relationship of PV cells and to track the global MPP of the partial shaded PV
system [28]. Under PSC, connection of PV cells can be re-organized and new
modules can be obtained with cells with similar irradiation condition. Then, these
modules can be considered in series and/or parallel connection according to shading
pattern to constitute a new PV array. Then, the output current (ia) and voltage
(va) of new configuration can be defined as given below:
276 I. Sefa et al.
X Np
ia ¼ i
b¼1 b
ð10:10Þ
X Ns
va ¼ v
j¼1 j
ð10:11Þ
where Np is number of parallel branches, Ns is number of serial branches, ib is the

bth branch current and vj is the voltage of jth module. The PV power can be
calculated as:
pa ¼ va ia : ð10:12Þ
The partial differentiation of power versus voltage is given below:
@pa @ia
¼ i a þ va 8va 2 ja; bj: ð10:13Þ
@va @va
Equation 10.14 can be written for the jth module of the bth branch:

n
K vbj þ ibj n s bj
Rs
@vj 1þ q
I RS e p bj
þ RS
¼
kTA
RSH
: ð10:14Þ
@ib n
K vbj þ ibj n s bj
Rs
qnp np
kTAns
bj
bj
IRS e p bj
þ ns
bj
bj RSH
It is seen from Eqs. 10.10–10.14 that, there is a valid @Pa =@Va value and this
value is bounded by a maximum value M. According to mean value theorem, there
is c 2 jv1 ; v2 j for every v1 ; v2 2 ja; bj such that:

jpðv1 Þ pðv2 Þj @p
¼ ¼ p0 ðcÞ; c 2 ðv1 ; v2 Þ: ð10:15Þ
jv1 v2 j @va c
Since M bounds the p0 ðcÞ, Eq. 10.16 can be obtained from Eq. 10.15:
jpðv1 Þ pðv2 Þj M jv1 v2 j 8v1 ; v2 2 ½a; b: ð10:16Þ
Equation 10.16 is known a Lipschitz condition and the function pðvÞ which
satisfies this condition with constant M is known as Lipschitz function.
If v1 is assumed as sampled point and v is a variable, Eq. 10.16 gives both lower
and upper bounds and values of function pðvÞ at other points can be obtained as:
pðv1 Þ M jv v1 j pðvÞ pðv1 Þ þ M jv v1 j: ð10:17Þ
If v1 is assumed at center of ½a; b, Eq. 10.17 can be rewritten as:

ba
pðvÞ maxv2½a;b pðvÞ pðv1 Þ þ M : ð10:18Þ
2
Thus, bound of the distance from global maximum is observed from sample
pðv1 Þ with just one sample. Here, both M and (b − a) might be big initially, but
(b − a) can be replaced by smaller length of subintervals by using more samples.

The DIRECT method is proposed to search the global extreme of this Lipschitz
function in a finite number of iteration [28].
When a DC-DC converter is used, boundaries of the duty cycle is [0,1]. However,
this bounded are can be narrowed by using PV parameters such as maximum and
minimum values of the PV short circuit current, open circuit voltage and equivalent
resistive load. Under PSC, while there are multiple peaks on P-V curve, I-V curve
shows multiple steps characteristics. It is assumed that, the PV system is operating at
its MPP depicted as Dm1 in Fig. 10.11a. If a PSC occurs at this time, operation point
of the PV system is move from 1 to 2 as shown in Fig. 10.11a.
This results decrease in output power and energy conversion efficiency of PV
system. Now, the DIRECT method starts the sampling interval and takes three
samples Di (i = 1,2,3) to determine operation condition if it is uniform irradiation or
PSC. Each step of the multistep I-V curve can be considered a part of the I-V curve
of a PV module under uniform irradiation. Therefore, the I-V can be considered in
three parts. If Idi and VDi are well separated, means Eqs. 10.19 and 10.20 are
satisfied, ID1 would be smaller than IMPP3 (MPP current of first step), and VD2
would also bigger than VOC of the middle step [28]:
DID3;1 ID3 ID1

¼ 0:1 ð10:19Þ
ID3 ID3
DVD1;2 VD2 VD1

¼ 0:2: ð10:20Þ
ID2 VD2
The 0.1 and 0.2 values are defined according the VMPP and IMPP and VOC and ISC
relations:
IMPP 0:9 ISC ð10:21Þ
VMPP 0:8 VOC : ð10:22Þ
Fig. 10.11 a Variation of the operation point. b Illustration of first three samples
278 I. Sefa et al.
If the I-V curve has more than three steps, two sequential steps always con-
sidered as one step. In addition, if the Eq. 10.19 or Eq. 10.20 is not satisfied, one of
the intervals should be divided into three subintervals and two new samples should
be taken. DIRECT evaluate the values obtained from these three samples and
determines the operation condition as uniform irradiation or partial shading.
If DIRECT determines the PSC, potentially optimal interval (for example jth
interval) is divided into three subintervals by taking two new samples (Dj2 and Dj3)
as second iteration. In third and fourth iterations, DIRECT divides other two
potentially optimal intervals into three subintervals by taking two new samples for
each potentially optimal interval as depicted in Fig. 10.12. At the end, there are nine
intervals and DIRECT evaluates the obtained data. In most application this oper-
ation is enough to determine some samples near the global MPP. If two samples
that have the highest power values are sequential, the global MPP is in this interval
Fig. 10.12 Iterations of DIRECT method

and DIRECT moves the operation point to this interval and tracks the global MPP
by using conventional P&O or IC methods. However, if these two samples are not
sequential, this means that system has two global MPPs or one local MPP closed to
global one. In this case DIRECT operates two more cycles and finally obtains 27
intervals [28].
10.3.1.4 Fuzzy Logic-Based MPPT Method
The fuzzy logic control (FLC) theory is vagueness and uncertainty based mathe-
matical discipline and allows one to use non-precise or ill-defined concepts.
The FLC removes the system’s complete mathematical model requirement, but it
requires prior knowledge about the system to provide proper design. The fuzzy
logic control has robust performance even under some parameters of the system
varies or and disturbances occur through its nonlinear and adaptive nature [3, 29,
30]. The FLC is applied to MPPT of PV system under uniform insolation condi-
tions and fast tracking speed with reduced oscillation is obtained. The inputs of the
FLC are usually defined as an error (E) and a change in error (DE), and these are
calculated as:
PðkÞ Pðk 1Þ PðkÞ Pðk 1Þ

EðkÞ ¼ or EðkÞ ¼ ð10:23Þ
VðkÞ V ðk 1Þ IðkÞ I ðk 1Þ
DE ¼ EðkÞ E ðk 1Þ ð10:24Þ
Thus, the P&O method is evaluated with fuzzy logic controller with the advance
of the variable size. Also the IC method can be evaluated by using change in PV
power and change in PV voltage values as inputs and faster tracking speed and
lower oscillation around the MPP are obtain with the help of variable step size [3].
Both two approaches combine the advantages of P&O or IC method such as fast
response, easy to implement with the adaptive nature of the FLC.
However, these FLC based MPPT algorithms unable to track the MPP of PV
system under PSCs because of the complex structure of P-V characteristics of the
PV system. So some modifications have been done to prevent the MPPT system to
trap the local MPP. The initial voltage tracking function is introduced for this aim.
If the PSC is determined, the operation point of the PV system is changed to the
computed initial voltage, and then the MPPT action is performed. The initial
voltage tracking function is defined as given below [30]:
VMPP
Vinitial ¼ I ð10:25Þ
IMPP
where VMPP and IMPP are MPP voltage and current values of PV array, respectively,
and I is the instantaneous PV current. The VMPP and IMPP values can be obtained
280 I. Sefa et al.
from I-V and P-V characteristics of the PV array operating at standard test condi-
tions [30].
In another method, the PV system is scanned and operating power of the PV
system is stored. Then, these values are used together with the FLC based MPPT
algorithm to track the global MPP even in PSCs. In this method, the operating area
of PV system is scanned and obtained maximum power values are stored during the
initial or changing weather conditions. A preset value representing admissible
difference between the determined maximum power and operating power is used in
control rules. If the actual power difference is greater than the pre-set power dif-
ference value, the duty cycle is increased; otherwise the FLC based MPPT algo-
rithm is applied. The flowchart of this modified FLC based MPPT method is given
in Fig. 10.13. This scanning and storing action and control rule prevent the system
to trap the local MPP.
Three scanning and storing methods can be performed in this method:
(i) Initializing system with maximum duty cycle, (ii) Increasing the duty cycle from
a minimum value to a maximum value with a fixed step, (iii) Applying a large
initial perturbation step to make a wide search. All three method guarantee to track
the global MPP under PSCs but their detection speeds are different.
Fig. 10.13 Flowchart of the

modified fuzzy logic based
MPPT method
10.3.1.5 Artificial Neural Network Based MPPT Method
Artificial Neural Networks (ANNs) are computational computer systems which are
designed to perform the characteristics of the human brain such as learning, pro-
ducing new knowledge via learning and discovering new knowledge and skills
automatically without any help. The ANN simulates the biological neural system by
using artificial neurons instead of biological ones. These neurons are connected to
each other with different weights to compose the artificial neural network which can
learn, store in memory and discover the relations between the data.
The ANN is used to track the MPP of the uniformly insolated PV systems with
its pattern recognition capability. The PV parameters such as PV voltage and PV
current or environmental parameters such as temperature and insolation or com-
bination of these two groups can be used as input parameters for the ANN. The
VMPP, the IMPP and the converter duty ratio are usually used as outputs. The
designers should decide parameters about the hidden layers. Design process of the
ANN is completed after the weights are determined with intricate training stage by
using the data obtained from simulation or experimental results. Although the ANN
based MPPT method has some advantages, its performance is depended on the
technical knowledge and skill of the designers, and PV module characteristics. The
differences between the commercially available PV module characteristics bring out
module dependency to performance of this method. In addition, retraining is
required because of variation of PV characteristics with time, with solar insolation
or with temperature. Therefore, number of studies on the ANN based MPPT
method under non-uniform insolation and PSCs is very less because of these
drawbacks. Instead, the ANN is used together with the other methods to provide
learning and remembering features.
The conventional IC method that cannot track the global MPP can be combined
with ANN to achieve improved performance in these conditions. The IC method is
usually used together with the CV method to increase the tracking speed. The CV
method uses the linear relation as given with Eq. 10.26 between the VMPP and VOC:
VMPP ¼ K VOC ð10:26Þ
Here K is usually 0.7–0.8, and this value is related with the module character-
istics. Thus, the operation point of the PV system transferred to the vicinity of MPP
with one step, and then the IC method tracks the real MPP with improved tracking
speed. However, these values are valid for only uniform insolation conditions.
Under non-uniform insolation conditions, the parameter K can be determined with
ANN to transfer the PV system to the vicinity of the global MPP and then IC
method can track the global MPP with fast speed. A feedforward ANN with two
inputs and one output can be used for this aim. Here, the PV voltage and PV current
can be used as input signals and the parameter K or directly VMPP can be used as the
output value. After determining the numbers of the hidden layer and neurons at the
hidden layer which have significant effect on ANN performance, the training
process is executed with the training and test values obtained from experimental
282 I. Sefa et al.
studies. After the successful training stage, the ANN that can perform nonlinear
input-output mapping for different operation conditions is obtained. Consequently,
a method that can perform fast MPPT performance under both uniform and
non-uniform insolation condition is obtained by combining the ANN and the IC
method [31].
Similarly, the ANN and the FLC can be combined to achieve a fast and efficient
MPPT method which can track the global MPPT even under PSCs. In this topol-
ogy, while the ANN is determining the voltage reference (that is the required
operation point), the FLC controls the power converter to transfer the actual
operation point to the required operation point. Unlike the previous method,
averaged solar insolation levels of modules and temperature value are used as
inputs. The modules of the PV system are divided into four groups, and average
insolation level value of each group is calculated by using Eqs. 10.27–10.30:
E1 ¼ 0:25 ðEM1 þ EM2 þ EM4 þ EM5 Þ ð10:27Þ
where EM1-EM9 are the solar insolation level of the PV modules. Thus, the resultant
P-V curve which is affected from the insolation level of each module and also the
shading pattern is aimed to model. The output of the ANN is again voltage refer-
ence. The difference between the voltage reference and the actual PV voltage and
the integral of this difference are used as input variables of the FLC. The output
variable of the FLC is duty ratio value of the converter. Thus, different insolation
levels of the PV modules, in other words shading pattern of the PV system, are
taken into account by the ANN used in voltage reference determination and the
operation point is transferred to the determined point via the FLC [32]. The ANN is
sufficiently accurate in mapping between a partially shaded condition and optimum
voltage and power of a PV array is obtained by this ANN-FLC combined method.
10.3.1.6 Particle Swarm Optimization Method
The particle swarm optimization (PSO) is population based and stochastic search
technique which is developed by inspiring the behaviors of bird flocks [33–35].
This technique is presented for the multi-variable and multi-parameter optimization
problems and successfully used in solution of nonlinear problems. This technique is
evolutionary computing method and starts with random solutions which are called
particles, and then optimum solution is obtained by updating the generations. Next
directions of the each particle are determined according to its own success and
success of the best particle in a neighborhood (Pbest) and success of the best particle
in the population (Gbest). The position of the particle xi is adjusted by using:
xi ðk þ 1Þ ¼ xi ðkÞ þ vi ðk þ 1Þ ð10:31Þ
Here vi is the velocity component representing the searching step size and can be
calculated by
vi ðk þ 1Þ ¼ wvi ðkÞ þ c1 r1 fPbesti xi ðkÞg þ c2 r2 fGbest xi ðkÞg ð10:32Þ
where c1 and c2 are the acceleration coefficients, w is the weight of the inertia, r1,
r2 2 U(0, 1), Pbesti is the individual best position of particle i, and Gbest is the best
position of the population [33–35].
While the PSO based MPPT method can be used to track the global MPP, this
method can also be used together with one of conventional method such as IC or
P&O. In this situation, the conventional MPPT method tracks the MPP under
uniform insolation conditions. When the PSC is detected, the PSO algorithm is
activated and global MPP is detected even multiple peak points are occurring on P-
V curve of the PV system.
Generally, solution vector with Np particles is defined as given below:
xki ¼ dg ¼ ½d1 ; d2 ; d3 ; d4 ; . . .::dn and n ¼ 1; 2; . . .:Np : ð10:33Þ
The objective function can be determined as:
f xki [ f ðPbesti Þ: ð10:34Þ
If the condition given in Eq. 10.34 is satisfied the individual best position Pbesti
is updated using Eq. 10.35:
Pbesti ¼ xki : ð10:35Þ
The algorithm generates three duty cycle values di (i ¼ 1; 2; 3) and the opti-
mization process is started and updates these three duty cycle values. After some
iteration, all three duty cycles reach the MPP of PV system. Owing to the low
velocity of the duty cycle variation, the MPPT action is obtained without any
oscillation. In addition, since this MPPT method is not based on the slope or sign
value of P-V curve of PV system, this method can successfully track the MPPT
even in PSCs [34].
However, some problems may appear especially if the small but fast change in
insolation occurs. In this case, small step sizes decrease the tracking speed of the
system. In addition, bigger step sizes may cause skipping the global MPP and
detecting local MPP instead of global MPP. Therefore, conventional PSO algorithm
should be modified. The modification has two phases. In first one previous duty
cycles are decreased and increased linearly according to the PV array power by
284 I. Sefa et al.
factor K1. In second phase, after the three duty cycles reach the MPP, two of the
three duty cycles are perturbed extremely in positive and negative directions with a
constant K2 [34].
The PSO algorithm can also be used together with conventional methods. In this
case, if the solar insolation is uniform, the conventional MPPT method activated
and tracks the MPP of the system. The operation condition of the system is checked
by using Eqs. 10.19 and 10.20. If these equations are satisfied, this means that
The PV system operates under PSCs and the PSO algorithm is activated to track the
global MPP of system [33, 35].
In this study, the PSO based MPPT method is designed to track the global MPP
for the PSCs. The number of the particles used in the PSO method is 11. The other
parameters of the PSO method are c1 = 1.5, c2 = 2.2 and w = 0.3. The commer-
cially available PV module, SunPower SPR305-WHT, is modeled with
MATLAB/Simulink and a sample PV system which consists of three strings with
five serial connected modules is built up. The solar insolation level of the modules
is changed to test the performance of the designed PSO based MPPT method. The
P-V curves of the PV system tested for the different insolation conditions and
determined operation points by the proposed method are depicted in Fig. 10.14. As
seen from the figure, the proposed scheme, determines the global MPP with great
performance both in accuracy and speed, and improves the efficiency.
10.3.1.7 Modified Fibonacci Search Based MPPT Method
The Fibonacci search method is recursive method that uses the Fibonacci numbers.
There is limitation about this method that the objective function is to be have only
single extremum point (single maximum or single minimum point). Therefore, this
method narrows down the search space and produces subintervals that includes
single maximum or minimum by using Fibonacci numbers. The searching action
can be performed in both directions. The values of the objective function at two
check points (e.g. PV voltage values V1 and V2 in MPPT action) are used to
determine the direction. The Fibonacci numbers that this method uses can be
determined with the Eq. 10.36 [36–38]:
Fn ¼ Fn1 þ Fn2 for n ¼ 2; 3; . . .: and F0 ¼ 0; F1 ¼ 1 ð10:36Þ
As a result, Fibonacci numbers are 0; 1; 1; 2; 3; 5; 8 . . .

In this method a variable x and a function f(x) are determined. The PV voltage,
PV current or duty cycle of the converter may be used as variable x and PV output
power is used as f(x). Since the main aim is tracking the MPP, a search space should
be determined as [xmin, xmax] for the variable. At initializing stage two operation
points (or voltage values), here variables x1 and x2 are determined as depicted in
Fig. 10.15. If f(x1) < f(x2), it means that the MPP should be in interval [x1, xmax] and
range is shifted to the right as given in Fig. 10.15. In next iteration, x2 becomes x1
Fig. 10.14 Operation points determined by the PSO based MPPT method for two different
shading patterns
and a new value for x2 is calculated with Fibonacci search formula given in
Eq. 10.37 [36]:

F ð n 1Þ
x2 ¼ xmin þ ðxmax xmin Þ ð10:37Þ
FðnÞ
where, F(n) is Fibonacci number. According to Eq. 10.37, the distance between the
samples is given in Eq. 10.38 and it is seen that the distance becomes narrower in
each iteration:
286 I. Sefa et al.
Fig. 10.15 Modified Fibonacci search algorithm based MPPT method
ai þ 1 þ bi þ 1 þ ai þ 1 ¼ ai þ bi ð10:38Þ
In the next iteration, if f(x1) < f(x2), again the previous process is performed. If f
(x2) < f(x1), at that time the search space is limited with x2, so xmax takes the value
of x2 and this means that the MPP should be in interval [x1, x2] and range is shifted
to the left. For the next iteration, x1 becomes x2 and a new value for x1 is calculated
with Eq. 10.39 [36]:

F ð n 1Þ
x1 ¼ xmax þ ðxmax xmin Þ ð10:39Þ
F ð nÞ
In next iteration again f(x1) and f(x2) values are compared. The shifting direction
is determined according to their values, and new variable values are calculated by
using Eq. 10.37 or Eq. 10.39 according to the direction. As declared above, the
distances between the samples are narrowed after each iteration. This process is
preceded until the MPP is reached [36–38].
10.3.1.8 Ant Colony Optimization Based MPPT
The ant colony algorithm (ACO) is a population based probabilistic algorithm for
finding optimal paths. This method uses the behavior of ants searching for food and
is used to solve difficult optimization problems [39]. In MPPT applications, the PV
array output power can be selected as the fitness function. After the initializing step,
the PV array power (here fitness value) is calculated for all the newly generated
particles in the current generation. Then, all fitness values of the newly generated
particles and the particles in the archive are ranked in ascending order. The archive
is updated with K best results. The particles in this generation are updated and
particles for the next generation are generated. This process is repeated until
reaching the maximum iteration number or termination condition. Here, the ter-
mination condition is reaching MPP of the PV system [39].
One of the major concerns of using ACO is algorithm speed. This is directly
related with the tracking speed. The conventional MPPT methods such as P&O or
IC method can be used to determine the first and the last local MPPs. Thus, these
values are used as boundary of the searching space and the solution space is
narrowed. Consequently, tracking speed is improved [39].
Other evolutionary algorithms such as the genetic algorithm, differential evo-
lution algorithm are also used for MPPT of PV systems under PSCs. These systems
can also be used together with the other methods to optimize the system parameters
[39–41].
10.4 Comparison of Global MPPT Methods
As it is declared that, the PSC has great effect on PV system performance. After
tests, it is obtained that power loss occurred at PV system as a results of partial
shading can be high as 70% [24]. Therefore, the MPPT algorithm for partially
shaded systems has gain importance and many algorithms have been investigated to
improve the PV system performance. Initial studies are based on the modification of
the commonly used MPPT methods like P&O and IC methods which are effective
for uniform insolation conditions with their advantages such as ease implementation
and fast tracking speed. So, a search algorithm is combined with conventional
MPPT methods to overcome the disadvantage of getting trapped at local peak of P-
V curve. This search algorithm may search the whole P-V curve periodically to
detect the global MPP. This requires long time and cause reduction on generated
power.
The DIRECT method is proposed to decrease the required search time and
prevent the energy deduction. However, in some partially shading cases, this
method cannot convergence to the global MPP in a few steps and therefore scan-
ning time may be as long as ordinary search algorithm. The open circuit voltage or
short circuit current of the PV module is also used to shorten the search time. Since
this method is required the knowledge of PV module, the module dependency is
come out. Besides, a declination on the PV module characteristics because of aging
effect reduces the efficiency of this method. Although the line search algorithm with
Fibonacci search has been used for tracking global MPP of partial shaded PV
systems, this method does not guarantee converging to the global MPP for any
PSCs. Similar to the conventional P&O or IC methods, the conventional
extremum-seeking based MPPT method also is trapped at a local peak and therefore
this cannot track the global MPP under PSCs. Thus, sequential extremum-seeking
based MPPT method is investigated. This method can track the global MPP and
eliminate the power ripple around MPP, but its convergence time is high because of
the global scan requirement [42]. However, this method has some drawbacks such
as high implementation difficulty, noise sensitivity [43]. In addition, global
288 I. Sefa et al.
perturbed-based extremum-seeking control method based on two band pass filters is

also proposed to track the global MPP of partial shaded PV system. This topology
reduces the noise sensitivities of band pass filter and improves the tracking per-
formance and speed [44].
In addition, artificial intelligence based methods such as FLC, PSO and ANN are
also used to track the global MPP of partially shaded PV system [45]. The FLC has
advantages of eliminating the mathematical model requirement, handling nonlin-
earities and uncertainties and having adaptive nature. However, the conventional
FLC method cannot guarantee to convergence to global MPP in PSCs. So modified
FLC based MPPT methods have been implemented to increase the tracking speed
and therefore efficiency of the PV system. The PSO is an optimization technique
that models the behavior of bird flocks. The PSO uses a collection of particles to
find the optimal solution for the problem. However, the PSO is used for
time-invariant problems. Since the environmental conditions such as solar insola-
tion and temperature vary with time, a detection and re-initialization procedures
should be included in the PSO to re-initialize the system when a variation on
environmental conditions is detected. The PSO with this specification has superior
advantages such as tracking global MPP effectively in all insolation conditions, fast
convergence speed and simple structure. However, initial particle positions affect
the performance of PSO based MPPT method, and improper values may cause
converging to a local MPP. The ANN methods are also used to track the global
MPP [7, 45]. Although different neural network structures have different advantages
and disadvantages, proper trained ANN based MPPT method can track the global
MPP under any PSC. However, this method requires significant computational
effort for training the ANN and also is model dependent. The main specifications of
these methods are summarized in Table 10.1.
The ACO is another soft computing method based on the swarm intelligence. It
has superior advantage as adapting its parameters according to changing environ-
mental conditions. However, it is difficult to implement this topology with low cost
microcontroller because of the great computational effort requirement. Therefore,
the convergence speed of this method is relatively low and this drawback limits its
applications. In addition, this method can be combined with another conventional
method such as P&O or IC to narrow the search space and to improve the con-
vergence speed especially for rapidly changing atmospheric conditions [39].
The chaos search method is stochastic search method which is based on chaos
theory. This technique generates knowledge and variables by using past experiences.
Therefore, it is superior to other techniques which use random variables. Although,
this method has features to track the global MPP even under PSCs, number of studies
are very limited and its performance has not been validated yet [43].
The Tabu search method is an iterative search algorithm and can be used to track
the global MPPT of the partial shaded PV applications. This algorithm uses the
previous data to prevent the staying at the local MPP. However, similar to the chaos
search method, number of the studies on global MPPT application of the Tabu
search method is also limited. The cuckoo search method is another search method
which becomes popular recently. This method uses the brood parasitism of the
Table 10.1 Comparison of MPPT methods under PSCs

Method Sensing Implementation Tracking Module Global MPP
variables complexity speed dependency tracking
capability
Modified P&O V, I Low Fast Yes Medium
Modified IC V, I Low Medium Yes Medium
DIRECT method V, I Medium Fast Yes High
FLC based V, I Medium Medium Yes High
method
ANN based V, I Medium Medium Yes High
method [31]
ANN based Solar Medium to high Slow to Yes Guaranteed
method [32] insolation, medium
temperature
PSO method V, I Medium Slow to No Guaranteed
medium
Modified V, I Medium Medium No Medium
Fibonacci search
based method
ACO method V, I Medium to high Slow No Guaranteed
cuckoo birds and presents superior performance on global optimization problems.

This method is has only two parameters to tune, therefore it is simpler than the
genetic algorithm and PSO method which have three parameters. The Levy flight is
used in this method to determine the next step. Thus, short steps, long steps or short
distance jumps can be determined as next step [43]. This method is used to track the
global MPP with these features and provides fast and accurate performance. The
differential evolution is another evolutionary algorithm and it can be used to track
the global MPP of the partial shaded PV system. This method is an independent
method but it has same disadvantages with the ACO based MPPT [39].
10.5 Conclusions
Obtaining maximum available power from PV system is very important. Therefore,

high efficient power electronic converters, solar tracking systems and MPPT
methods have been investigated to improve the efficiency of the PV system.
Although various MPPT methods have been implemented and considerable results
are obtained for uniform solar insolation conditions, these methods are not useful
for PSCs. When the PV system does not receive uniform solar insolation, the
resultant P-V characteristics of the entire PV system becomes more complex and
have multiple peak points on it. Therefore, tracking the global maximum point of
the P-V curve becomes more complicated and improved MPPT methods are
290 I. Sefa et al.
required. Following observations about the partial shaded PV system characteristics

should be taken account while investigating maximum power point search algo-
rithm to improve the partially shaded PV system efficiency [24, 27]:
• Under PSCs, the P-V curve of PV system has multiple peak points while I-
V curve of the system has multiple steps.
• The maximum PV power value and the PV voltage value that it occurs are not
only dependent on solar insolation and ambient temperature level but also on the
shading pattern and array configuration.
• The global MPP of P-V curve of PV system may lie on the left side of the load
line. Therefore, the two stage technique will not be able to track the global MPP
of PV system.
• The peak points on the P-V curve of the array occur nearly at multiples of 80%
of module open circuit voltage.
• Therefore, the minimum displacement (voltage difference) between any two
consecutive peak power points is nearly 80% of module open circuit voltage.
• After extensive study of P-V curves, it is reported that when the P-V curve is
traversed from either side, the magnitude of the peak points increases. After
reaching the global peak point, the magnitude of the consecutive peaks (if they
are present) decreases continuously.
References
1. Sefa I, Altin N (2009) Grid interactive photovoltaic inverters—a review. J Fac Eng Arch Gazi
Univ 24(3):409–424
2. Sefa I, Ozdemir S (2010) Multifunctional interleaved boost converter for PV systems. In:
IEEE international symposium on industrial electronics (ISIE 2010), pp 951–956
3. Altin N, Ozdemir S (2013) Three-phase three-level grid interactive inverter with fuzzy logic
based maximum power point tracking controller. Energy Convers Manag 69:17–26
tracking techniques. IEEE Trans Energy Convers 22(2):439–449
5. Subudhi B, Pradhan R (2013) A comparative study on maximum power point tracking
techniques for photovoltaic power systems. IEEE Trans Sustain Energy 4(1):89–98
6. Rodriguez JDB, Franco E, Petrone G, Paja CAR, Spagnuolo G (2014) Maximum power point
tracking architectures for photovoltaic systems in mismatching conditions: a review. IET
Power Electron 7(6):1396–1413
7. Koutroulis E, Blaabjerg F (2012) A new technique for tracking the global maximum power point
of PV arrays operating under partial-shading conditions. IEEE J Photovoltaics 2(2):184–190
8. Rani BI, Ilango GS, Nagamani C (2013) Enhanced power generation from PV array under
partial shading conditions by shade dispersion using Su Do Ku configuration. IEEE Trans
Sustain Energy 4(3):594–601
9. El-Dein MZS, Kazerani M, Salama MMA (2013) Optimal photovoltaic array reconfiguration
to reduce partial shading losses. IEEE Trans Sustain Energy 4(1):145–153
10. Ishaque K, Salam Z (2013) A review of maximum power point tracking techniques of PV
system for uniform insolation and partial shading condition. Renew Sustain Energy Rev
19:475–488
11. Fathy A (2015) Reliable and efficient approach for mitigating the shading effect on
photovoltaic module based on Modified Artificial Bee Colony algorithm. Renew Energy
81:78–88
12. Belhachat F, Larbes C (2015) Modeling, analysis and comparison of solar photovoltaic array
configurations under partial shading conditions. Sol Energy 120:399–418
13. Daraban S, Petreus D, Morel C (2014) A novel MPPT (maximum power point tracking)
algorithm based on a modified genetic algorithm specialized on tracking the global maximum
power point in photovoltaic systems affected by partial shading. Energy 74:374–388
14. Olalla C, Clement D, Rodriguez M, Maksimovic D (2013) Architectures and control of
submodule integrated DC–DC converters for photovoltaic applications. IEEE Trans Power
Electron 28(6):2980–2997
15. Moretón R, Lorenzo E, Leloux J, Carrillo JM (2014) Dealing in practice with hot-spots. In:
29th European photovoltaic solar energy conference and exhibition, pp 1–6
16. Kim KA, Seo GS, Cho BH, Krein PT (2016) Photovoltaic hot-spot detection for solar panel
substrings using AC parameter characterization. IEEE Trans Power Electron 31(2):1121–1130
17. Patel H, Agarwal V (2008) MATLAB-based modeling to study the effects of partial shading
on pv array characteristics. IEEE Trans Energy Convers 23(1):302–310
18. Batzelis EI, Georgilakis PS, Papathanassiou SA (2015) Energy models for photovoltaic
systems under partial shading conditions: a comprehensive review. IET Renew Power Gener
9(4):340–349
19. Ishaque K, Salam Z, Taheri H, Syafaruddin (2011) Modeling and simulation of photovoltaic
(PV) system during partial shading based on a two-diode model. Simul Model Pract Theor
19:1613–1626
20. Hidalgo-Gonzalez PL, Brooks AE, Kopp ES, Lonij VP, Cronin AD (2012) String-level
(kW-scale) IV curves from different module types under partial shade. In: 38th IEEE
photovoltaic specialists conference (PVSC), pp 1442–1447
21. Ji Y-H, Jung D-Y, Kim J-G, Kim J-H, Lee T-W, Won C-Y (2011) A real maximum power
point tracking method for mismatching compensation in PV array under partially shaded
conditions. IEEE Trans Power Electron 26(4):1001–1009
22. Silvestre S, Kichou S, Chouder A, Nofuentes G, Karatepe E (2015) Analysis of current and
voltage indicators in grid connected PV (photovoltaic) systems working in faulty and partial
shading conditions. Energy 86:42–50
23. Bidram A, Davoudi A, Balog RS (2012) Control and circuit techniques to mitigate partial
shading effects in photovoltaic arrays. IEEE J Photovoltaics 2(4):532–546
24. Patel H, Agarwal V (2008) Maximum power point tracking scheme for PV systems operating
under partially shaded conditions. IEEE Trans Ind Electron 55(4):1689–1698
25. Tey KS, Mekhilef S (2014) Modified incremental conductance algorithm for
photovoltaic system under partial shading conditions and load variation. IEEE Trans
Ind Electron 61(10):5384–5392
26. Carannante G, Fraddanno C, Pagano M, Piegari L (2009) Experimental performance of MPPT
algorithm for photovoltaic sources subject to inhomogeneous insolation. IEEE Trans Ind
Electron 56(11):4374–4380
27. Chen K, Tian S, Cheng Y, Bai L (2014) An improved MPPT controller for photovoltaic
system under partial shading condition. IEEE Trans Sustain Energy 5(3):978–985
28. Nguyen TL, Low K-S (2010) A global maximum power point tracking scheme employing
direct search algorithm for photovoltaic systems. IEEE Trans Ind Electron 57(10):3456–3467
29. Alajmi BN, Ahmed KH, Finney SJ, Williams BW (2013) A maximum power point tracking
technique for partially shaded photovoltaic systems in microgrids. IEEE Trans Ind Electron
60(4):1596–1606
292 I. Sefa et al.
30. Chin CS, Tan MK, Neelakantan P, Chua BL, Teo KTK (2011) Optimization of partially
shaded PV array using fuzzy MPPT. In: IEEE colloquium on humanities, science and
engineering (CHUSER), pp 481–486
31. Punitha K, Devaraj D, Sakthivel S (2013) Artificial neural network based modified
incremental conductance algorithm for maximum power point tracking in photovoltaic system
under partial shading conditions. Energy 62:330–340
32. Syafaruddin, Karatepe E, Hiyama T (2009) Artificial neural network-polar coordinated fuzzy
controller based maximum power point tracking control under partially shaded conditions.
IET Renew Power Gener 3(2):239–253
33. Ishaque K, Salam Z, Shamsudin A, Amjad M (2012) A direct control based maximum power
point tracking method for photovoltaic system under partial shading conditions using particle
swarm optimization algorithm. Appl Energy 99:414–422
34. Ishaque K, Salam Z, Amjad M, Mekhilef S (2012) An improved particle swarm optimization
(PSO)–based MPPT for PV with reduced steady-state oscillation. IEEE Trans Power Electron
27(8):3627–3638
35. Ishaque K, Salam Z (2013) A deterministic particle swarm optimization maximum power
point tracker for photovoltaic system under partial shading condition. IEEE Trans Ind
Electron 60(8):3195–3206
36. Ramaprabha R, Mathur B, Ravi A, Aventhika S (2010) Modified Fibonacci search based
MPPT scheme for SPVA under partial shaded conditions. In: IEEE third international
conference on emerging trends in engineering and technology, pp 379–384
37. Tumbelaka HH, Miyatake M (2010) Simple integration of three-phase shunt active power
filter and photovoltaic generation system with Fibonacci-search-based MPPT. In: IEEE
symposium on industrial electronics and applications (ISIEA 2010), pp 94–99
38. Miyatake M, Inada T, Hiratsukai I, Zhao H, Otsuka H, Nakano M (2004) Control
characteristics of a Fibonacci-search-based maximum power point tracker when a
photovoltaic array is [1] partially shaded. In: IEEE the 4th international power electronics
and motion control conference, (IPEMC 2004), pp 816–821
39. Jiang LL, Maskell DL (2014) A uniform implementation scheme for evolutionary
optimization algorithms and the experimental implementation of an ACO based MPPT for
PV systems under partial shading. In: IEEE symposium on computational intelligence
applications in smart grid (CIASG), pp 1–8
40. Tajuddin MFN, Ayob SM, Salam Z (2012) Tracking of maximum power point in partial
shading condition using differential evolution (DE). In: IEEE international conference on
power and energy (PECON), pp 384–389
41. Ramli MAM, Ishaque K, Jawaid F, Al-Turki YA, Salam Z (2015) A modified differential
evolution based maximum power point tracker for photovoltaic system under partial shading
condition. Energy Build 103:175–184
42. Elnosh A, Khadkikar V, Xiao W, Kirtely Jr. JL (2014) An improved extremum seeking based
MPPT for grid connected PV systems with partial shading. In: IEEE international symposium
on industrial electronics, pp 2548–2553
43. Liu Z-H, Chen J-H, Huang J-W (2015) A review of maximum power point tracking
techniques for use in partially shaded conditions. Renew Sustain Energy Rev 41:436–453
44. Bizon N (2015) Global maximum power point tracking based on new extremum seeking
control scheme. Prog Photovoltaics Res Appl. doi:10.1002/pip.2700
45. Syafaruddin, Hiyama T, Karatepe E (2010) Investigation of ANN performance for tracking
the optimum points of PV module under partially shaded conditions. In: Proceedings of
international power energy conference, pp 1186–1191
Chapter 11
Solar Energy Harvesting in Electro
Mobility
Aytaç Gören
Abstract Based on the experiences of five solar cars designed and manufactured in
11 years, participations in establishments of solar charging stations and local solar
power plant projects, this chapter involves modeling energy harvesting and storing
parts of solar cars, differences between maximum power point tracker topologies in
implementations, the structures of brushless direct current motors (BLDCMs) and
batteries as loads and the similarities of brushless direct current motors; briefly,
solar energy harvesting for electro mobility. Light weight is one of the keys for
efficiency in electro mobility. This enforces implementations of new technologies in
manufacturing light weight electric vehicles. The end of the first section of this
chapter is about using polymer composites in manufacturing process of solar cars.
On the other hand, if energy harvesting should be separated from the vehicle,
modular on or off-grid solar charging stations might be an efficient solution and an
implementation of this type of energy harvesting is presented in the second section
of this chapter. The last section in this chapter is about hybrid off-grid systems
which also includes smart solutions. Implementations of this chapter are manu-
facturing process chassis and body of a solar car using polymer composites, a
model of an off-grid PV charging station for electric vehicles (EVs) in a campus
area, electrical units of a solar car for World Solar Challenge.
Keywords Solar car

Solar system modeling
Stand-Alone Photovoltaic

(SAPV) Brushless Direct Current Motors (BLDCMs) Vacuum Assisted Resin
Transfer Molding (VARTM)

AM Air Mass
BLDCM Brushless Direct Current Motor
CAD Computer Aided Design
A. Gören (&)
Automatic Control and Robotics Laboratories, Department of Mechanical Engineering,
Dokuz Eylul University, Izmir, Turkey
e-mail: aytac.goren@deu.edu.tr

294 A. Gören
EMF Electromotive Force

EV Electric Vehicle
HAWT Horizontal Axis Wind Turbines
ICE Internal Combustion Engine
NOCT Normal Operating Cell Temperature
PEM Proton Exchange Membrane
P&O Perturb & Observe
PV Photovoltaic
SAPV Stand-Alone Photovoltaic
SC Solar Car
SCRIMP Seemann Composites Resin Infusion Moulding Process
SOC State of Charge
STC Standard Test Conditions
VARTM/VARIM Vacuum Assisted Resin Infusion Moulding
VAWT Vertical Axis Wind Turbines
VBRTM Vacuum Bag Resin Transfer Moulding
WT Wind Turbine
11.1 Modeling of Solar Power Systems
11.1.1 Modeling of Semi Flex Silicon Solar Panel
Equivalent circuit of a PV cell can be simply modeled as a current source in parallel

with a resistor and a diode those are connected in series with another resistor
(Fig. 11.1). The output of the current source is directly proportional with the light
on the cell. The open circuit voltage of the cell is quite different from the cell that is
connected with the load which is shaped with Rs resistor in the same figure.
Besides, the temperature effects the output of the cell as well. On the other hand, the
mathematical model of the cell is defined using the Shockley Diode Equation in
most of the researches about PVs (Eqs. 11.1 and 11.2) [1–3]. The parameters in
Fig. 11.1 PV cell models. a One diode model. b Two diode model
11 Solar Energy Harvesting in Electro Mobility 295
Table 11.1 Parameters of cell model

Parameters for cell current calculations
Iph Light generated current (A) V Voltage (V)
IS Cell saturation of dark current (A) k Boltzmann cons.
(1.38 10−23 (j/K))
TC Cell temperature (K) A Ideal factor ( )
q Electron charge (1.6 10−19 (V)) I Current (A)
IS1 First diode saturation current (A) IS2 Second diode saturation current (A)
N1 Quality factor of D1 ( ) N2 Quality factor of D2 ( )
Vt Thermal voltage (V), (k*Tc)/q Rs Internal series resistance (Ω)
Rp Internal parallel resistance (Ω)
Parameters given by manufacturers
VOC Open circuit voltage @25 °C (V) ISC Short circuit current @25 °C (A)
Vm Voltage @MPP@25 °C (V) Im Current @MPP@25 °C (A)
Pm Maximum power @25 °C (W)
Eqs. 11.1 and 11.2 and the parameters which are given by the manufacturers might
be seen in Table 11.1.

qV
I ¼ Iph Is exp 1 ð11:1Þ
kTc A

V þ I * Rs V þ I * Rs ðV þ I * Rs Þ
I ¼ Ip Is1 exp 1 Is2 exp 1
N1 * Vt N2 * Vt Rp
ð11:2Þ
In Eqs. 11.1 and 11.2, the first one is the simple form of the solar cell which is
represented by one diode or saturation current whereas the second equation is the
cell representation with two diodes. Both models can be seen in Fig. 11.1a, b
respectively.
Semi flex PV panels are the panels which are flexible in one direction. This
advantage makes it possible to cover a surface of the solar car whose shape is
formed using fluid dynamics analysis. On the other hand, the final shape of a solar
car might be also similar with a tube for some designs. In that case, the solar
irradiation received by the surface might differ from a flat surface that is perpen-
dicular to the sun rays and for rough calculations, the average energy generation
values might be used. Remembering that only one region receives the sun rays
perpendicularly for this case, it is a big disadvantage for energy generation. So, for
solar cars, there are some tube-like designs with transparent vehicle bodies and flat
PV modules inside.
296 A. Gören
Fig. 11.2 PV curves. a Voltage-current. b Voltage-power @diff. temp. c Voltage-power @ diff.

irradiance
In Fig. 11.2, current-voltage and power-voltage graphs of a PV cell at different

cell temperatures might be seen. As it is known, the Silicon photovoltaic structure,
in fact accidently discovered during photo diode experiments in 1950s. This was a
discovery of a structure that does not control the current due to light, a structure that
generates electrical energy due to light. So, the similarities in the graphics below
come from the semi-conductor characteristics with electricity. It should be con-
sidered that the values in these graphs are taken under standard illumination
1000 W/m2 at 1.5 AM that is generally called as 1 sun. However, PVs can get more
Fig. 11.3 A simple PV cell model in MATLAB Simulink
sun levels using optics. This systems are called as concentrators. Since the thickness
of the body shell and mass are important parameters for solar cars, Fresnel Lenses
are mostly used for this purpose.
For modeling the photovoltaic cell in MATLAB, it is common to form a model
of a current source, that is effected by the parameters of light intensity (radiation)
and temperature. Considering the one or two diode equations, using the parameters
in Table 11.1, a more convenient model might be used. Chapter 2 in this book
gives a detail photovoltaic cell model. Besides, a simple MATLAB model might be
seen in Fig. 11.3.
Power generation optimization or finding the maximum point for the power
generation is based on finding the maximum rectangular area in Fig. 11.2a or
finding the peak point of the curve depending on the cell temperature and irradiance
for Fig. 11.2c. The way of doing this is increasing or decreasing the output voltage
level of a DC/DC converter (please see Fig. 11.7). One of the common techniques
is to change the output voltage with a little difference and observe the output power
with the previous output power value (perturb and observe). On the other hand, if
the source is not just PV, but there are wind turbine(s), water turbine(s) or diesel
generator in the system, total system output should be optimized due to energy
efficiency or cost.
298 A. Gören
11.1.2 Types of Energy Loads for Stationary Solar Powered

Systems and on Vehicles
11.1.2.1 Energy Loads for Stationary Solar Powered Systems
Stand alone photo voltaic (SAPV) systems are the system which have only PV to
generate electricity and are more implemented systems than hybrid ones because of
their simplicity in design and implementation. Energy consumption of a house
differs with the area of the house, the location, the number of the people use the
house, the characteristic of usage and some other parameters, the energy con-
sumption of the house should not only be analyzed in details but the analysis should
also cover the whole year usage. In some countries, like Turkey, the generated
power can be also sold using the grid. So, it is the initial point to design the system
if the system will be off-grid (not connected to the grid) or on-grid (connected to the
grid). If the system is on-grid, it is more like that a huge capacitance is connected to
the system; but, this time charging the capacitor is selling the energy generated to
the grid and discharging the capacitor is buying energy from the grid. So, it is
mostly not reasonable to connect a battery pack to the system. On the other hand, if
it is an off-grid system, commonly a battery pack that its 80% can be charged by the
PVs in 5 h mostly convenient in Aegean Region of Turkey. Table 11.2 shows
energy consumption of some loads. Battery, as a load, modeled in next part of this
section with different types of batteries.
If the system is an off-grid SAPV system, to be more efficient, loads are rec-
ommended to use direct current, since the inverter (the unit converts DC to AC)
also draws energy. Loads connected to the system are mostly considered as resistive
loads.
11.1.2.2 Energy Loads for Solar Powered Mobile Systems
Brushless Direct Current Motor (BLDCM) Model

The mathematical model of a linear BLDCM is generally related with the
Lorentz force as in electric motors. Of course, there is a nominal force this time, but
not a nominal torque.
~ v ~
F ¼ q ~ B ð11:3Þ
If the variations of the stator self inductance with rotor position and the mutual
inductance between the stator windings considered as negligible; the electrical
dynamics of a BLDCM may be modeled in an electrically balanced system [4–7].
In that case,
Table 11.2 Power of some loads in houses

Device Power (W) Device Power (W) Device Power (W)
Coffee machine (Turkish cf.) 200 Washing m. 1200–1500 Tumble dryer 4000
Coffee machine (filter) 800 Hair dryer 1000–2000 Heater (res.) 1500–2500
Toaster 800–1500 Vac. cleaner Air Condition
Big 200–700 Room 1000
Compact 100 Central 2000–5000
Pop corn m. 250 Sewing m. 100 Fan 10–50
Blender 300 Iron 1000 Table fan 10–25
Microwave Oven 600–1500 Shaver 15 Electric blanket 200
Washing machine 500 Electric pan 1200 Computer
Laptop 20–50
Desktop 80–150
TV CD/VCR/DVD Player 35–50 Satellite receiver 30
(25″ LCD) 150
(19″ LCD) 70
9″ Angle grinding 1200 Wireless Phone sender 40–150 El. clock or radio cl. 1–3
Receiver 5
Cold saw 900–1400 Radio/CD/MP3 Player 10–30 Ham radio 5
12″ Saw 1100 Radio/MP3 player for car 8 Printer 100
1″ Drill 1000 Illumination 25 W Fluorsc. 28 Refrigerator/freezer
50 WDC inc. lamp 50 Old 475–540
40 WDC hlg. lamp 40 New 60–112
20 W Fluorsc. 22
disi
usi V0 = Ls þ Rs isi þ ei ; i ¼ 1; 2; 3 ð11:4Þ
dt
X
isi ¼ 0 ð11:5Þ
In Eq. 11.4, Rs and Ls are the stator resistance and inductance, usi is the motor
terminal voltage, isi is the phase current and ei is the back-EMF associated with the
ith phase. The potential of the motor neutral terminal in wye-connected windings is
denoted as V0. The back-EMF induced in each phase is;
dwir @wir dh @w
ei ¼ ¼ ¼ x ir ð11:6Þ
dt @h dt @h
If wir is the mutual magnetic flux between the permanent magnet and the stator
windings in the ith phase, h is the rotor position, and x is the rotor speed. If the
model is assumed as linear;
dwir ¼ Lir ir ð11:7Þ

300 A. Gören
The mutual inductance Lir is expressed, using the terms of the trigonometric
Fourier series, as:
X
K
2p 2p
Lir ¼ ðLirak coskðph ði 1ÞÞ + Lirbk sinðph ði 1ÞÞÞ ð11:8Þ
k¼1
3 3
The back EMF can be derived from (11.6) and (11.7) as;
@Lir
ei ¼ i r x ð11:9Þ
@h
The electromagnetic motor torque can be derived as;
X
3 X
3
@Lir 1 @Lrr
T¼ Ti þ Ticogg ¼ ir ð isi Þ þ i2r ð11:10Þ
i¼1 i¼1
@h 2 @h
The first three terms in Eq. 11.10 are mutual torques caused by interaction
between the permanent magnet field and the phase currents. Ticogg is the
cogging-torque, due to the attraction of the permanent to the salient portions of the
stator iron. So, even in the absence of the phase currents, the cogging-torque is
present.
If Tl is the load torque, J is the rotor inertia, B is the viscous friction coefficient
C is the Coloumb friction coefficient, mechanical dynamics model of the motor is
defined as;
1
h_ ¼ x; x_ ¼ ðT Tl Bx CsignðxÞÞ ð11:11Þ
J
Since efficiency of each part has importance for a solar car, hub motor use is very
common for solar cars. Hub motor is the motor which the rim itself is also the rotor
of the motor. Hence, there is no transmission system and the efficiency of the motor
can reach 99% [4].
Model of the Batteries
Next step of generation energy comes with a harder engineering problem that is to
store it. If it is a system that can be connected to grid, it is not needed to consider
chemical reactions or heat up a mass to store energy; but if it is an off-grid one, the
energy that is generated needs to be transformed in different forms or capacities.
Capacity might be a water tank for a PV system that is for irrigation to reduce the
costs and to increase the system life (see Fig. 11.17a). But, mostly it is the chemical
reactions help us. Lead acid, NiCd, NiMH, LiIon, LiIonPo (or LiPo shortly) are
mostly known rechargeable types of batteries for that purpose [8]. It is very
common to use LiPo and LiIon batteries in electric vehicles (EVs) because of their
advantage of high energy density (kWh/kg) values. Lead Acid types of batteries are
used in off-grid SAPV or hybrid systems mostly because of their costs [9, 10].
Fig. 11.4 Quasi-static model

of equivalent electrical circuit
of a battery
The quasi-static model of equivalent electrical circuit of the LiPo batteries which is
based on Thevenin model (Fig. 11.4) is commonly used, but to understand the
limits of the LiPo batteries, dynamic behavior should be also take into account
especially for solar vehicles [11].
In Fig. 11.4, Eth is the open circuit voltage, Rbat is the resistance of the battery,
Ubat is the voltage of the battery and Ibat is the current of the battery.
The lead-acid battery voltage during charging can be expressed with Eq. 11.12
using the internal resistance of the battery (rbat), the electrode potential (Eo), the
number of cells in the battery (N), the charging current of the battery (Ibat), state of
charge (SOC), η is the overpotential and a1 is a parametric constant depends on the
phase of charge [9, 12]:
1
h_ ¼ x; x_ ¼ ðT Tl Bx CsignðxÞÞ ð11:12Þ
J
For discharging process, on the other hand, it can be defined with the following
Eqs. 11.13 and 11.14.
Ubat = U0 g10 ð11:13Þ

1 4 0:27
Ubat ¼ ½2:85 0:12ð1 SOCÞ þ þ 0:02
C10 1 þ I1:3 SOC1:5 ð11:14Þ
þ ð1 0:007DTÞ
11.1.3 Energy Needs of Solar Cars
Efficiencies of PVs for the known researches are between 6 and 46% for NREL [13]
in year 2015. As for the same laboratory report, the most efficient ones are four
junction or more GaAs ones whereas the maximum efficiency for Si panels is
27.6% which are the concentrated ones. As known, the efficiency of solar panel is
calculated using Eq. 11.15. In this equation, ηPV is the efficiency of the panel, Pm is
the output power in Wp at standard test conditions (STC) for one square meter of
area under 1000 (W) of global irradiation at 25 °C and air mass (AM) of 1.5.
302 A. Gören
Pm
gPV = ð11:15Þ
E Ac
On the other hand, in real conditions the cell temperature is not at laboratory
conditions first of all. And for a vehicle that is travelling with different velocities at
different weather conditions to different directions, the energy taken from the solar
array varies. For the first step of being realistic, power can be taken from the solar
array at a known temperature and global irradiation can be calculated using the
temperature difference Eq. 11.16. NOCT is the normal operating cell temperature
whereas the / is the solar irradiance. The difference between the temperature of the
ambient air temperature (Ta) and the cell (Tc) is [14]:
NOCT 20
TC TA ¼ / ð11:16Þ
800
Energy need of a vehicle is calculated with the assumptions of all mechanisms

run in ideal tolerances and in design phase it is mostly a prediction using these
calculations. Unlike internal combustion engine cars, the electric car (EV) has
highest torque at start and nearly a constant torque through running region.
Considering the efficiencies of PVs which have the maximum efficiency of 46%
even for the GaAs panels, for a vehicle which generates its energy from sun, the
efficiencies and the robustness of the units have great importance. This also means
nearly zero losses and very efficient systems. Electric motors of solar cars, for
instance, have the efficiency of over 93% (most efficient 99.7%), the motor drivers
mostly 99% and there is generally no transmission system.
Considering the weights and technologies of steel body with internal combustion
engine vehicles, travelling from place to another place feels like asking a traveler if
he travels to the place you travel and planning to go to your target city without
forgetting the cities he will visit. Minimizing energy needs and optimizing energy
usage starts with minimizing travelling mass and analyzing way of doing. Unlike
classical concept of mass production cars, fibers whether they are carbon, aramid or
glass make the polymer composite structure more and more strength, light and
durable. Mass of a car we use in daily life is approximately 1200 kg whereas a solar
car with the dimensions of 4.7 m 1.8 m 1.1 m 200 kg. This case not only
reduces the energy needed to travel, but also gives a chance to manufacture the
vehicle form more aerodynamics [15–18].
In designing phase of a solar car, the basic equation to calculate the energy needs
is the Eqs. 11.18 and 11.17 might be used for the energy predicted for a solar car
race [18, 19]. The parameters are defined in Table 11.3. In Fig. 11.5, the resistances
affect the energy need of a solar car can be seen.
Table 11.3 Challenge strategy calculations for a solar car

Parameters for energy need calculations
WT Total resistive forces (N) x Distance (m)
WR1 Rolling resistance force (1) (N) W Weight of the vehicle (N)
WR2 Rolling resistance force (2) (N) Crr1 Rolling resistance const. (1) ( )
WB Acceleration resistance Crr2 Rolling resistance const. (2) (Ns/m)
force (N)
WST Gradient resistance force (N) N Number of wheels ( )
m Total mass of the vehicle (kg) h Total height vehicle climbs (m)
η Motor, controller and drive train Na Number of times the vehicle will
efficiency ( ) accelerate in a race day ( )
ηb Watt-battery eff. ( ) g Acc. due to gravity constant (m/s2)
ηPV Eff. of PV ( ) v Av. velocity through the route (m/s)
Eb Energy available in the va Average velocity of the wind towards the
batteries (joule) vehicle (m/s)
Pm Average power estimated from k Factor for rolling mass’ ( )
solar arrays (W)
Fig. 11.5 Energy need of a car

Px 1 Na Wv2
g gb E þ ¼ WCrr1 þ NCrr2 v þ qCd Av x + Wh þ
2
ð11:17Þ
v 2 2g
WT ¼ WL þ WR1 þ WR2 þ WB + WST ð11:18Þ

Pm :x
g gb Eb + ¼ WT ð11:19Þ
v
11.1.4 MPPT’s on Solar Cars
The electric vehicles (EVs) were built earlier than internal combustion engine
(ICE) vehicles. Between late 1827 and 1839, different electric vehicles are built in
Hungary, Netherlands, Scotland and US [20–23]; however, lack of usable
rechargeable, high energy density batteries and the high popularity of ICE with easy
304 A. Gören
reloading of fuel. Since 1840s, EVs got popular for some time intervals like in
1910s with mass production concepts, late 1940s with invention of semiconductor
structures, in 1970s with energy and petrol crisis. Besides, solar cars are not novel
research areas, since it is known that the first was announced in 1955. The first solar
car (SC) invented has 12 Selenium PV cells and a small electric motor rotating a the
rear wheel shaft (Sunmobile, W.G. Cobb, 31.08.1955). Although the first solar car
race was in 1985 (Tour Del Sol), it became popular with the Australian World Solar
Challenge (WSC) in year 1987. The concept of WSC was devised by Hans
Tholstrup, who is a Danish-born adventurer and traveled from Perth to Sydney
(4130 km) in 20 days with a solar car called Quiet Achiever. Different concept solar
car races are being organized in different continents nowadays. WSC (Australia,
road challenge), European Solar Challenge (ESC, Europe, circuit race), American
Solar Challenge (NASC, North America, combination of road and circuit races),
Moroccan Solar Car Race (MSCR, road challenge), Alternative Energies Cup
(Japan, circuit race), TUBİTAK Formula G (Turkey, circuit race), South African
Solar Challenge (SASC, South Africa, road challenge) and Atacama Solar
Challenge (CSA, Chile, road challenge) are some popular solar car races. In
addition to these, some races are organized for one or two times in different
countries. Today, SCs in races are classified with international motorsport federa-
tions and International Solar Car Federation which was formed in 1991 to give
support to organizers.
A solar car is powered by PVs (mostly semi flex) which cover the surface of the
car. The vehicle is designed with the concept of ‘least energy need to travel’ (see
Figs. 11.6 and 11.7). After achieving the DC formed electrical energy from the
Fig. 11.6 Solaris S7 solar car in world solar challenge

Fig. 11.7 DesTech Solaris (2015) solar car PV cell layout
Fig. 11.8 a Solar team design MPPT’s in Solaris solar car projects. b Developed telemetry
interface of MPPTs
mono/poly Si or GaAs PVs, MPPTs provide high efficiencies of charging the high
energy density batteries, mostly LiPo or LiIon (Fig. 11.8). With their polymer
composite bodies and chassis, it is possible to achieve to manufacture a car that is
150 kg with the dimensions of 5.00 m 1.8 m 1.0 m (l/w/h) as will be told in
following sections of this chapter. Some cars use solar concentrators, mostly
306 A. Gören
Fig. 11.9 a Maximum power point tracking. b Common MPPT topologies for SCs
Fresnel Lenses to concentrate the light intensity on the PV cell. This case, however
means an additional weight to the car which makes the optimal energy usage
problem a little bit more complex.
Power for a resistor in direct current (DC) is the product of applied voltage and
the current. So, the most efficient implementation of a solar cell is maximizing the
area in Fig. 11.7. PV cells are assumed as DC formed constant current generators
which might be differ in real implementations, but for engineering calculations
which have tolerances inside, said to be true. With this assumption, the VI curve of
a cell is a horizontal line starting from ISC then a line like curve falls down to VOC
with a slope which is connected to the first line with a radius simply (Fig. 11.9a).
However, the output current is affected by the radiance directly and the temperature
affects the point that the line like curve intersection point with the horizontal line, so
the open circuit voltage. On the other hand, solar radiation and temperature change
instantaneously in daytime as might be seen in Fig. 11.10. Maximum power point
trackers are devices those find the point where the system gets maximum power
from the photovoltaic.
In Fig. 11.10, irradiation levels taken by the pyranometer which is mounted on
the roof of the main solar charging station in Tinaztepe Campus Location of Dokuz
Eylul University can be seen. These data are used to make the results of calculations
more real for solar car energy harvesting. This station is one of the stations gen-
erates the energy need of electric vehicles which are used in Dokuz Eylul
University Tinaztepe Campus Location (please see Fig. 11.24b).
Consider two points on V-I curve of a PV cell. Output power is the product of
output voltage and current. The greatest output power can be achieved with the
maximum area of the rectangle. Knowing that the MPPTs are devices those are
connected between PVs and loads, MPPT is a DC/DC converter which increase or
Fig. 11.10 Irradiation levels @ DEU Tinaztepe Campus Location on 24.12.2014
decrease the output voltage, to have the maximum area for the PV used (please see
Fig. 11.9a). If the output voltage level of the MPPT is higher than input voltage
level, it is boost type. If the output voltage level is lower than the input voltage
level, it is buck type (please see Fig. 11.9b). Energy efficiency improvements using
different MPPTs and different MPPT algorithms might be found in relevant
chapters of this book.
Team Solaris of Dokuz Eylül University (Izmir) is a community of academi-
cians, young investors and researchers from undergraduate to Ph.D. level, who
gathered around the idea of designing/creating a car powered by solar energy in
December 2003. As per date, over two hundred and fifty researchers took part, five
solar cars and four EVs were designed and manufactured in Solaris Projects. Solaris
Cars took part in WSC (Australia), ESC (Belgium), MSC (Morocco), Formula G
(Turkey) and European Electric Vehicle Event (Austria). Just between May 2013
and November 2016, Team Solaris participated in eight international and five
national solar car/EV races. Most of the implementations and information told
about solar cars in this chapter are the technical experiences gained during Solaris
projects design, manufacturing and challenging race phases.
Although some cars are still using stationary system type MPPTs for some solar
car teams, it is common to use special designed boost type DC/DC converters with
adaptive algorithms and wireless telemetry embedded circuits for this purpose.
Designing the PV array and battery capacity of a stationary solar power plant is
mainly a cost optimization problem in engineering with more constant parameters.
However, designing a solar array and energy storage system for a solar car is more
complex which includes flow analysis, battery and PV combination calculations,
energy need and optimization, strength analysis, efficiency analysis and more. The
308 A. Gören
voltage level itself for example changes the speed of the vehicle whereas torque
generated by the motor is a result of battery type, parallel branches of the battery
pack, C rating, racing conditions and regulations of the challenges. Every details
have to be considered. For instance, if the by-pass diode that is needed for
bypassing the module that has a shaded cell even has to be a fast, low voltage drop
type diode in order not to cause a big loss. Or sometimes teams spray water to
surface of cells in order to decrease the cell temperature. If we consider an area of
6 m2 is used to be covered with PVs for solar cars, for use of Si cells which have
approximately 22% efficient, the solar array can generate more than 1300 W
@1000 W/m2 global radiation @25 °C. However, the cell temperature almost
never 25 °C, the global radiation level is limited just with laboratory conditions, the
solar cells are just for a little time interval in daytime perpendicular to sun rays and
1000 W/m2 is a rare value with limited time interval and affected with weather
conditions. For an optimist prediction of average 1000 W output for a time interval
from the solar array, the car itself should be very efficient in means of air flow,
rolling resistances, the motor, driving unit and the batteries.
Photovoltaics are sources which the current generated by them changes due to
incident light intensity. As mentioned before, every unit or every system model
block on the solar car should be used in maximum efficient regions. On the other
hand, the load connected to the system changes instantaneous, so the unit which
simulate the load as maximum for PVs is called as maximum power point tracker
[24–26]. This is explained in Fig. 11.9 in previous sections. Commercial product
MPPTs which are used on stationary systems are also used on some solar cars. This
case causes inefficiency, because of not only the commercial products are designed
in run as boost, buck and voltage regulation regions, but also are heavy products
with unneeded cases. A more common and effective way is to design the MPPT just
as buck (input voltage is greater than output voltage) or boost type (output voltage is
greater than input voltage). With the assumption of using the optimal number of
series and parallel groups of PVs on a solar car, the most efficient type of MPPT
topology for a solar car mainly the boost type and a common algorithm is
hill-climbing/P&O (perturbation and observe) [1, 24–27]. The main disadvantage
of this technique is the local maximum points that can cause not tracking the real
maximum point. In Fig. 11.11, simplified electrical diagram of a solar car which
designed by Team Solaris, might be seen. In this implementation, the cells are
grouped in three groups. Each group has five MPPTs those are connected in series
to charge a series of 32 LiPo cells connected also in series. The combination of the
battery pack, however can be changed to different capacity of LiIon or LiPo cells
and also the number of series-connected cells. If the values changes, developed
bi-directional telemetry system and its software is used to change the charging
voltage. The output voltage level is limited with the efficient test regions for MPPT.
MPPTs in series can increase the efficiency, especially when some cells are shaded.
A method of using solar energy more efficient is to rotate the PVs of the solar car
towards the sun to be sure that the arrival of the sun’s rays are as close to per-
pendicular as possible (see Fig. 11.12). This is very common when the solar car is
not travelling, but even they are rare, there are also some implementations of using
Fig. 11.11 DesTech Solaris solar car electrical connections
Fig. 11.12 DesTech Solaris upper car body orientation towards the sun
310 A. Gören
during the race like a group of sun trackers. In that case, not only the route direction
and car direction should be good analyzed, but also the energy used to control the
angle of the panels and the energy generated by the panels and the change in center
of gravity should be well analyzed.
Before DesTech Solaris Solar Car, two common combinations of MPPTs are
tested. First one was to divide the PV surface into two groups and to use two
MPPTs. This selection, naturally was generally related with the voltages of battery
and PV combination. As might be expected, this caused important power losses
during the races because of damaged cells or connections of PV modules. Revised
solution was to change the battery voltage level and PV groups. The groups
increased to three with three boost type MPPTs whose algorithm also improved.
This revision increased the efficiency level significantly. DesTech Solaris, on the
other hand, uses fifteen MPPTs (five series, three parallels) which are optimized for
shades and connection damages (please see Fig. 11.11).
11.1.5 Using Polymer Composites for Increasing Efficiency

of Energy Harvesting
Wind turbine (WT) electric generation systems have less subsystems than in
thermal, hydroelectric or nuclear power plants. And one of the great advantages of a
wind turbine is, like PVs, it is easier to scale the structure. WT has three main parts,
the electric generator, blades and the wind turbine tower. It has control systems to
control the angles of the blades or the tail and for energy transfer. The electric is
generally a three phase form which has 120° between each other. Based on the
orientation of the axis of rotation, wind turbines have two main types: Horizontal
axis wind turbines (HAWT) and vertical axis wind turbines (VAWT). Figure 11.13
shows (a) VAWT and (b) HAWT. HAWTs can be scalable from 500 W to MWs,
but VAWTs are generally chosen for their motion of less vibration and esthetics,
however they are less efficient and less scalable (mostly 10 kW). The output form
of both types is alternate current form, as seen in previous section that is on Hybrid
System. Blades, mechanical connection of blades and if the dimensions of the
turbine has a small value in scale, sometimes the WT tower can be also polymer
composites and they are manufactured using vacuum assisted resin transfer method
which will be told with solar car body production in this section.
There has been a growing interest to use composite materials in structural
applications ranging from aircraft and space structures to automotive and marine
applications instead of conventional materials. This is because advanced compos-
ites exhibit desirable physical and chemical properties that include high specific
stiffness and strength, dimensional stability, temperature and chemical resistance,
and relatively easy processing. A variety of manufacturing methods can be used
according to the end-item design requirements. Most commercially produced
composites use a polymer matrix with textile reinforcements such as glass, aramid
Fig. 11.13 Blades of wind turbines. a Vertical axis type. b Horizontal axis type
and carbon. Besides, the structures and parts which need light weight but also
greater strength values need to be optimized in more cycles. Analyzing, recon-
struction using CAD model, redesigning and remanufacturing phases include more
details and loops. Solar Cars, ballistic parts, fifty meter long wind turbine wings are
some examples of these light weight technological parts. This section is based on
solar car manufacturing steps and techniques which is very similar with wind
turbine blade manufacturing.
Vacuum assisted resin infusion techniques have become popular in manufac-
turing of these composites. In the literature, vacuum infusion is known under
different acronyms. The most popular terms to describe vacuum infusion processes
are: VARTM-Vacuum Assisted Resin Transfer Moulding, VARIM-Vacuum
Assisted Resin Infusion Moulding, SCRIMP™-Seemann Composites Resin
Infusion Moulding Process, VBRTM-Vacuum Bag Resin Transfer Moulding,
VARI-Vacuum Assisted Resin Infusion process and so on. All involve basically the
same technology, and describe methods based on the impregnation of a dry rein-
forcement by liquid thermoset resin driven under vacuum [28]. Stages can be
simplified as:
1. Identifying the important parameters for design,
2. Preliminary design with constraints,
3. Identifying the types of fibers and analysis,
4. Improving design with the conclusions of analysis,
5. Model production (1:1),
6. Mould production,
7. Moulding.
312 A. Gören
Fig. 11.14 Fluent analysis of Solaris S7 (2013)
In manufacturing stages of polymer composite products, the analysis’ of a solar

car are based on strength, vibration, fluent and efficiency [18, 29]. Figure 11.14
shows the fluent analysis of Solar Car Solaris S7 (2013) whereas Fig. 11.15 shows
the fluid mechanics analysis of Solar Car DesTech Solaris (2015). DesTech Solaris
is 4500 mm in length and 1800 mm in width. Foam mold is formed from modular
small foam parts. You may find the static structural analysis of Solar Car DesTech
Solaris in Fig. 11.16.
In flow analysis of the body, the regions which has turbulences or the regions
against the flow are found. If vacuum is formed during flow, this causes the vehicle
pulled backwards by the flow which is called as parachute effect. And if the analysis
shows high regional forces against flow, this shows the need of smoothing the
edges or angles of the form. Considering these two main cases, car body is opti-
mized once more to be efficient. On the other hand, static and dynamic mechanical
analysis determines the textile fiber angles and density per area (Fig. 11.16). These
analysis determine the types (carbon, aramid or different fiber glass types) and
layers of textiles. Analyzing person should not consider just material characteristics,
but also some special characteristics of the selected types. For example, carbon,
Fig. 11.15 Fluent analysis of DesTech Solaris S8 (2015)
even it is in fiber formed textile, a good conductor which is not so needed when
transmitting wireless signals or if there are non-isolated poles in a box of carbon
polymer composites.
Vacuum Assisted Resin Infusion Moulding (VARIM or VARTM) is a polymer
composite manufacturing process to produce high-quality large-scale components.
In this process, dry preform fabrics are placed in an open mold and a plastic vacuum
bag is placed on the top of the mold. The one-sided mold is connected with a resin
source and a vacuum pump. The liquid resin infuses into the reinforcing fibers
thanks to the vacuum drawn through the mold. Curing and de-molding steps follow
the impregnation process to finish the product. Curing process is very important to
produce composite parts with optimum mechanical properties [30]. All regions in a
composite part are supposed to be produced almost at the same curing conditions.
So, a uniform temperature distribution through the mould surface is of high
importance.
The main steps of the process are:
a. A dry fabric or preform and accompanying materials such as release films, peel
plies are laid on tool surface.
314 A. Gören
Fig. 11.16 Static structural analysis of DesTech Solaris (2015)
b. The preform is sealed with a vacuum bag and the air is evacuated by a vacuum
pump.
c. Liquid resin with hardener from an external reservoir is drawn into the com-
ponent by vacuum.
d. The liquid resin with hardener is infused into the preform until complete
impregnation.
e. Curing and de-molding steps follow the impregnation to finish the product.
The components of the infusion process utilized in this work are illustrated in
Fig. 11.17. The function of the each component, given in Fig. 11.17, during
manufacturing can be summarized as:
Vacuum bagging films are sealed to the edge of the mould with vacuum bag sealant
tape to create a closed system.
Fig. 11.17 Vacuum assisted resin transfer molding (VARTM) process control diagram
Fig. 11.18 Model production of Solaris Erke solar car for moulding
Double side bag sealant tapes are used to provide a vacuum-tight seal between the
bag and the tool surface.
Release films are typically placed directly in contact with the laminate. They sep-
arate the laminate from the distribution medium. Release films are often perforated
to ensure that any trapped air or volatiles, which may compromise the quality of the
laminate, are removed.
Release fabrics and peel plies are placed against the surface of the laminate. They
are woven products which are strong and have good heat resistance. Release films
316 A. Gören
Fig. 11.19 VARTM process for solar car Molding
impart a gloss finish on the cured laminate, whereas peel plies and release fabrics
leave an impression of the weave pattern. Peel plies provide a clean, uncontami-
nated surface for subsequent bonding or painting (Fig. 11.18).
Tool release materials are used to release the product from tools easily and obtain a
smooth surface finish. For this purpose, either self adhesive Teflon films or liquid
release agents are utilized. In certain situations Teflon films can also temporarily
solve tool porosity problems.
A highly permeable layer called “resin distribution medium” placed on the top of
the preform spreads the resin quickly over the lateral extent of the part.
Bleeder/breather fabrics are non-woven fabrics allow air and volatiles to be
removed from within the vacuum bag throughout the cure cycle. They also absorb
excess resin present in some composite lay ups.
For moulding (VARTM or VARTM) process, model of the product which has the
scale of one is manufactured mostly (Fig. 11.18). However, for some cases which the
time is very limited for manufacturing the product or if the product is a prototype that
will be improved or the products will be manufactured in small numbers; the mold
itself might be also manufactured without manufacturing the model (Fig. 11.19).
The next step for manufacturing polymer composites is to tool the real model of
the CAD model. This is needed for manufacturing the mould. However, having the
exothermic reaction precautions of the curing process of polymer composite
moulding, it is reasonable to tool the mould itself without manufacturing the real
model if the product is a solar car body.
Manufacturing the chassis of the solar car, is another challenging process which
includes mechanical stress analysis and structural optimization. After optimization,
one of the most efficient way of forming the chassis is to manufacture flat polymer
composite sandwich parts and cut them using water jet and then form the chassis
from cut modular parts. In Fig. 11.20, this process and fixing the chassis with the
shell of the vehicle can be seen. Fixing the chassis with body needs reference points
in order to have the exact solid model designed. This makes a even 150 kg total
weight achievable as a solar car (pilot weight excluded).
Figure 11.21 shows the latest solar car of Solaris Projects (Solaris 8
Project/2013–2015). This solar car is designed and manufactured for WSC 2015, so
most of the technical specifications are the results of regulations of this event. It is a
four wheel vehicle with one motor on the left rear side. The total weight of the
Fig. 11.20 Combining chassis and body of the DesTech Solaris Solar Car (2015)
318 A. Gören
Fig. 11.21 DesTech Solaris Solar Car (2015)
vehicle is 182 kg for the first version, but increased to 199 kg because of the extra
weights to improve the stabilization of rear suspensions, electrical isolation and
upper body orientation apparatus. Carbon polymer composite monocoque body has
different textiles in different locations of the car. The car has also a telemetry system
to analyze the energy used, generated and also to sense the operating conditions.
Telemetry system send the data of voltage and temperature of each thirty series
battery group, battery current, MPPTs (so, the current and voltage of PV groups),
motor and vehicle speed and motor driver current. Voltage and temperature of each
Table 11.4 Technical specifications of DesTech Solaris

Name DesTech Solaris (S8)
Production year 2015
Dimensions 4.5 (m) 1.80 (m) 1.05 (m)
Weight 182 (kg) (without driver)
Body and chassis Carbon fiber sandwich—monocoque body and chassis
Wheels 2 front, 2 rear
Power of PV array 1341 (Wp)
Number of PV cells 392
Efficiency of PV cells 22.8%
Nominal voltage of the system 111 (V)
Max. voltage of the system 126 (V)
Battery capacity 4.7 (kWh) 30 series and 8 parallel groups—5.3 (Ah) battery
cells
Battery type Lithium polymer
Motor 3 (kW) BLDC hub motor, max eff. 95%
MPPTs Boost type, 98% max eff., connected as 5S3P
eight parallel—thirty series LiPo battery cell and the current from battery pack are
sensed and also the cells are protected by battery management system (please see
Table 11.4).
11.2 Modular Off-Grid Solar Energy Charging Stations

for EVs
Project EYLEM is a project supported by Izmir Development Agency (IZKA) and

Dokuz Eylul University which is aimed to be a model for e-mobility in campus
areas [22, 31]. Not only three EVs are designed and manufactured by Solaris Team,
Fig. 11.22 A 1 kWp off grid charging station connection

320
Fig. 11.23 A 4 kWp off grid charging station connection

A. Gören
Fig. 11.24 A Charging stations (DECharge 1, 2 and 3) in campus area and charging EVs
(DEMobils) at the station
but also three solar powered charging stations are established in campus area (see
Fig. 11.15). Project is proposed with the analysis of daily usage of ICE vehicles by
campus security which is resulted as approximately 50 km. This limited the
capacity of the batteries on EVs, since the batteries are gel lead acid type of
batteries for the EVs in this project. A main solar charging station of 54 m2 that has
a PV group of 5 kWp (4 kWp + 1 kWp) and a battery group of 28.8 kWh
(21.6 kWh + 7.2 kWh) is located at the center of the campus area. The main station
is also aimed to be used as a workstation for EVs. The second and third solar
charging stations are located on 21 m2 and have 1 kWp PVs and 7.2 kWh battery
groups (see Figs. 11.22 and 11.23). All stations are off-grid designed. So, the
reasons of having battery packs at charging stations are to schedule the charging at
night, to use batteries for illumination and hand-held electric motor operated tool
purposes.
An enhanced version of these EVs is also designed for EV car races in Turkey
which is more light and efficient than other three (see Fig. 11.24).
The electrical connection diagram of the main solar charging station (DECharge
1) and two other stations might be seen in Fig. 11.25a, b. System output is 48 VDC
for all. An additional DC/DC converter of 48/72 is used when the nominal voltage
of the EV which is connected for charging is 72 V.
322 A. Gören
Fig. 11.25 a Hybrid PV and wind mill system without batteries for irrigation. b Hybrid electric
generation system
11.3 Hybrid Combination of Power Generation Systems
A system is called off-grid, if it is not connected the interconnected electrical grid

whereas on-grid if it is connected. Storing energy is a more challenging engineering
problem than generating mainly. If the energy generation system has its own simple
natural solutions inside, it is more efficient and cost effective. A simple example of
this is a solar energy generation system for irrigation. It is more efficient and cost
effective to replace batteries with a water tank which is also a capacity for the
system model (please see Fig. 11.25a). In this system, electricity generated by PVs
runs the pump motor and water tank is a capacity that is charging for a potential
energy level and discharging it through the system which is similar with the battery
pack replaced. It is sure that if the system is on-grid, the grid itself behaves like an
infinite capacity that can be used with the unit currency/kWh.
Figure 11.25b shows another way of forming a hybrid system which is more
common than Fig. 11.25a. A hybrid system which generates electricity from solar,
wind and kinematic energy from water flow might be seen in Fig. 11.7b. Hybrid
charge controller is the point where different sources are combined. Mobile phone
base stations in outback are good implementations of hybrid systems since hybrid
systems fills the lack of each other.
It is well known that the source of a system should be well analyzed and
designed due to the energy generation efficiencies, cost and dimensions. An
important point is that the only DC generation is the PV cell part. Mostly, three or
mono phase AC forms are the outputs of a turbine even it is a wind or another type
of energy conversion implementation. If it is in mono phase form which is the
voltage is generally represented with the Eq. 11.20 whereas if it is in three phase
form that the voltages are represented with the Eqs. (11.21)–(11.23) (remember that
x ¼ 2pf). In that case, the balanced system calculations are used which is based on
the assumption that all three phases are following each other with 120° angles and
all loads are compensated.
vm = Vmax sinð2pftÞ ð11:20Þ

pffiffiffi
vA ¼ 2V sinxt ð11:21Þ
pffiffiffi
vB ¼ 2V sinðxt 120 Þ ð11:22Þ
pffiffiffi
vC ¼ 2V sinðxt 240 Þ ð11:23Þ
Figure 11.17b shows a generalized schematic of a hybrid system which is

composed of a PV group, a wind turbine, a generator (it can be either a water
turbine generator, diesel generator, etc.), a battery system (generally for backup in
case of no electricity generation), system load(s) and a hybrid charge controller that
optimizes the source usages due to power generation of each source. Eliminating
the peaks of switching on or off, calculation of power of a direct current system is

Energy Harvesting and Energy Efficiency: Nicu Bizon Naser Mahdavi Tabatabaei Frede Blaabjerg Erol Kurt Editors

Uploaded by

Copyright:

Available Formats

Energy Harvesting and Energy Efficiency: Nicu Bizon Naser Mahdavi Tabatabaei Frede Blaabjerg Erol Kurt Editors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Energy Harvesting and Energy Efficiency: Nicu Bizon Naser Mahdavi Tabatabaei Frede Blaabjerg Erol Kurt Editors

Uploaded by

Copyright:

Available Formats

Lecture Notes in Energy 37

More information about this series at http://www.springer.com/series/8874

Frede Blaabjerg Erol Kurt

ISSN 2195-1284 ISSN 2195-1292 (electronic)

© Springer International Publishing AG 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

October 2016 Lucian Dascalescu

experimental systems are described in detail in order to compare their output

Piteşti, Romania Nicu Bizon

1 Energy Saving and Efﬁcient Energy Use By Power Electronic

Part I Energy Harvesting

Part II Energy Efﬁciency of the Photovoltaic Systems

8 Maximum Power Point Tracking (MPPT) Algorithms

Part III Energy Efﬁciency of the Hybrid Power System

Part IV Technical Strategies, Efﬁcient Methods and Applications

19 Determining the Optimal Battery Model for a Speciﬁc

Necmi Altin Department of Electrical and Electronics Engineering, Faculty of

Aytaç Gören Automatic Control and Robotics Laboratories, Department of

Mircea Raceanu National Research and Development Institute for Cryogenics

AAHP Air-to-Air Heat Pump

BDR Battery Discharge Regulator

FFT Fast Fourier Transform

O-QPSK Offset Quadrature Phase Shift Keying

TT&C Telemetry, Tracking, and Telecommand

Frede Blaabjerg, Huai Wang, Pooya Davari, Xiaohui Qu

F. Blaabjerg H. Wang P. Davari (&)

© Springer International Publishing AG 2017 1

Abbreviation and Acronyms

1.2 Electrical Energy Generation and Consumption

1.3 Power Electronics for Adjustable Speed Drives

1.3.1 Motor Drive Applications

Fig. 1.5 Relation between

Figure 1.4a exempliﬁes a direct connected AC motor to a main where a mechanical

1.3.2 System Architecture

Fig. 1.7 Block diagram of a multi-drive conﬁguration of ASD units

1.3.3 System Components

Every component contributes to system losses. In power electronics systems

1.4 Power Electronics for Lighting

1.4.1 Evolution of Lighting Technology

1.4.2 Energy Efﬁcient and Intelligent Lighting Power

Fig. 1.10 Typical LED

1.5 Emerging Opportunities and Challenges

Table 1.1 Typical lifetime target in different power electronics applications

1. Blaabjerg F, Ionel DM (2015) Renewable energy devices and systems—state-of-the-art

Nazenin Gure, Abdulkerim Kar, Erturul Tacgin, Alper Sisman

Abstract In this millennium, the methodologies to harvest existing dissipated

N. Gure (&) A. Kar E. Tacgin

© Springer International Publishing AG 2017 17

Keywords Hybrid energy harvesting

Abbreviation and Acronyms

2.2 Comparison of Piezoelectric and Electromagnetic

EM energy harvesters (EMHs) generate power based on faraday’s law of induction,

2.3 Classic HEH Systems

While tremendous amounts of multimode energy harvesters (EHs) are possible,

2.3.1 Fixed-Frequency Classic HEHs

Fig. 2.2 Classic HEH

Fig. 2.3 a Two-pole and

2.3.2 Broadband Classic HEHs