Hosting Capacity For Smart Power Grids

Ahmed
F. Zobaa
Shady H.E. Abdel Aleem
Sherif M. Ismael
Paulo F. Ribeiro Editors
Hosting
Capacity
for Smart
Power Grids
Hosting Capacity for Smart Power Grids
Ahmed F. Zobaa • Shady H. E. Abdel Aleem
Sherif M. Ismael • Paulo F. Ribeiro
Editors
Hosting Capacity for Smart

Power Grids
Editors
Ahmed F. Zobaa Shady H. E. Abdel Aleem
College of Engineering Mathematical, Physical and Engineering
Design and Physical Sciences Sciences Department
Brunel University London 15th of May Higher Institute of Engineering
Uxbridge, UK Cairo, Egypt
Sherif M. Ismael Paulo F. Ribeiro

Electrical Engineering Division Advanced Power Technologies
Engineering for the Petroleum and Innovations in Systems
and Process Industries (Enppi) and Smart Grids Group
Cairo, Egypt Federal University of Itajuba
Itajuba, MG, Brazil
ISBN 978-3-030-40028-6 ISBN 978-3-030-40029-3 (eBook)

https://doi.org/10.1007/978-3-030-40029-3
© Springer Nature Switzerland AG 2020

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, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Today’s power systems are becoming much more complex, expandable, dynamic,
and non-predictable; therefore, accurate and up-to-date hosting capacity (HC) anal-
ysis studies are necessary to face the significant challenges associated with the
massive integration of renewable energy resources into modern electrical power
systems. This book is designed to be essentially an introduction to the concepts and
historical developments of the hosting capacity assessment. It will be beneficial to
distribution system operators, network planners, distribution generation investors,
and researchers in this field as it presents past, present, and future possibilities
regarding HC research. It is expected that the HC will play a significant role in
future power systems and smart grids. Using the HC approach to drive network
requirements could steer distributed generation (DG) toward areas of the network
where it could have the greatest positive impact on network reliability and win-win
benefits with DG owners. Finally, smarter DG integration into future electrical
systems can be met if utilities have a clear forecast of their potential network
HC. As can be seen, the chapters in each section maintain their own thematic
continuity and at the same time have significant overlaps with chapters in other
sections. Therefore, one may read the book in its entirety or focus on individual
chapters. Due to its broad scope, this book will be an ideal resource for students in
advanced graduate-level courses and special topics in the field of smart grids, hosting
capacity assessment, and enhancement in modern electrical power systems. HC
research is a key enabler for affordable, reliable, and renewable energy sources, so
it is possible to transition away from traditional high-carbon energy sources. There-
fore, it is imperative that novel solutions be sought to enable networks to cope with
future developments to realize resilient distribution networks that can host the pushy
DG penetration while enhancing the system reliability of power supplies and
controlling the over-hosted areas. Based on the editors’ experiences, we believe
that this book is an important guide to the researchers, distribution system operators,
and network planners who are interested in the integration of renewable energy
resources in the electrical power systems. Comprehensive presentation of the hosting
v
vi Preface
capacity concepts, calculation methods, performance limits, enhancement tech-

niques, and case studies will be introduced in this book.
The book is mainly focused on the concept of hosting capacity applied to modern
power grids, but each of the book’s chapters begins with the fundamental structure of
the problem required for a basic understanding of the methods described. It is sorted
out and organized in 10 chapters.
Chapter 1: This chapter presents an overview of the hosting capacity concept and
its developments and defines the main HC performance limits and its up-to-date
enhancement techniques. It also presents the concept of dynamic hosting capacity as
well as its application to energy storage systems.
Chapter 2: This chapter presents a new methodology in quantifying the system
risk encountered by distribution networks during high PV penetration. The estab-
lishment of a number of operational risk indices is achieved. The network’s hosting
capacity is calculated using sparse grid technique which showed superior perfor-
mance compared with the Monte Carlo technique from both computational time and
calculation accuracy perspectives.
Chapter 3: In this chapter, the concept of local hosting capacity is applied through
the evaluation of allowed voltage rises due to harmonic distortion. Moreover, a brief
discussion on the definition of background harmonic distortion of electrical systems
resulting from additional connections, such as consumers, distributed degeneration,
and nonlinear loads, are presented. It is clearly concluded that the high levels of
background harmonic distortion can result in equipment malfunction, decreasing the
likelihood of meeting international standards, for example, IEEE 519 or IEEE 1547,
as well as the impact on the hosting capacity value of electrical systems.
Chapter 4: In this chapter, the effect of harmonics caused by different distributed
energy resources on the hosting capacity is examined. The analysis also considers
voltage violation and ampacity constraints. The entire modelling and simulation is
performed in MATLAB and validated through Typhoon HIL to prove its efficacy for
real-time implementation. Various distribution test systems are simulated with
different loading and uncertainty in source conditions to validate the effect of
different distributed energy resources in the hosting capacity assessment.
Chapter 5: This chapter presents a comprehensive method to identify the mini-
mum required battery energy storage systems (BESS) for increasing the hosting
capacity of a system, considering the uncertainties associated with the distributed
energy resources and loads. Then, an economic model has been developed to assess
the minimum required BESS from an economic perspective. Finally, the perfor-
mance of the developed models is assessed on a real agricultural feeder in Australia.
Chapter 6: In this chapter, two categories of market power indices were presented
to assess the capacity withholding of generation companies (GenCos) in the presence
of DERs. These indices are effective and can be used by independent system
operators (ISO) besides the technical assessment of hosting capacity. In addition,
the mitigation of capacity withholding of GenCos and the impacts of distributed
generation units on capacity withholding are described.
Preface vii
Chapter 7: This chapter elaborates a generalized deterministic approach to eval-

uate the PV hosting capacity for LV distribution networks under different operating
conditions. The proposed approach establishes safe limits for solar PV hosting
capacities (cumulative values) for a given distribution feeder based on the locational
and operational aspects of the solar PV units employed. Safe limit of hosting
capacity has been developed employing a number of sensitivity analyses considering
the influencing factors. From a distribution system planning perspective, the use of
such a deterministic approach is convenient and more practical compared to the use
of extensive simulations. Furthermore, the proposed method can be used as an
approximate guide or a rule of thumb to evaluate solar PV hosting capacity at a
given location of an LV distribution feeder without using complex stochastic
techniques.
Chapter 8: In this chapter, a strategy is proposed to maximize DG integration in
the distribution network without violating system operational limits. The strategy is
based on the available active and reactive powers provided by a substation. The
proposed strategy aims to prepare the distribution network using distribution net-
work reconfiguration (DNR) and soft open points (SOPs) installations to accommo-
date more distributed generation (DG) installations in the distribution system. In
order to solve the MINLP problem under investigation, a new metaheuristic optimi-
zation technique, called expanded Invasive Weed Optimization (exIWO), is used. A
case study was conducted to maximize the HC of the 83-bus distribution system. The
proposed strategy has proven its ability to increase the hosting capacity of the
distribution network at three loading levels.
Chapter 9: In this chapter, the mathematical model for incorporating hosting
capacity using power flow equations in large signal and transfer function model in
the small signal analysis is examined. This chapter presents the Internet of Things
(IoT) applications in the study. Moreover, it discusses the communication-based
control strategy implemented in Gasa Island, South Korea, and the Taiwanese
microgrid under normal and disturbance conditions that have been implemented
with a MAS platform in Taiwan using agent-oriented programming.
Chapter 10: This chapter introduces the concept of market hosting capacity,
expressing the maximum amount of renewables that can be connected to a given
power network while preserving their profitability in a deregulated environment. It
has also developed a suitable mathematical model to quantify this market hosting
capacity, founded on bi-level optimization principles. This is a new measure for
hosting capacity from a market point of view, which is useful to understand how
various factors, such as network investments, subsidies, and generation flexibility,
can effectively support higher investments in renewable generation.
In conclusion, this book aims to introduce good practice with new research
outcomes, programs, and ideas that connect the past, current, and future roles of
hosting capacity for modern power grids. It introduces innovative research out-
comes, programs, algorithms, and approaches that consolidate the present and future
and opportunities and challenges of the hosting capacity approach from various
viii Preface
perspectives. It is a useful tool for the network planners, designers, operators, and
practicing engineers of modern power systems who are concerned with renewable
energy resources and their challenges, reliability, and security. Likewise, it is a key
resource for advanced students, postgraduates, academics, and researchers who had
some background in electrical power systems.
In an experiment in criticism [1], the author suggests that books allow us to see
the world through other people’s eyes and view it as someone else sees it—and
experience their view “from the inside.” This is what gives books a profound and
mysterious power. At the same time, each of us sees the world through our own
perspective. Thus, the editors wish you may find this book useful to stimulate your
imagination and go even further by applying the concept of HC, but doing it with
your own sensitive and critical eyes.
Reference
1. Lewis, C. S. (1961). An experiment in criticism. Cambridge: Cambridge Univer-

sity Press.
Contents
1 An Overview of Hosting Capacity for Modern Power Grids . . . . . . 1

Tiago E. Castelo de Oliveira, Paulo F. Ribeiro, Ahmed F. Zobaa,
Shady H. E. Abdel Aleem, and Sherif M. Ismael
2 Hosting Capacity Determination via Risk Analysis Approach
Involving Likelihood and Severity of Operational Constraint
Violation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Hassan Al-Saadi
3 A Methodology Proposal for Calculation of Hosting Capacity
Including Different Power Quality Phenomena . . . . . . . . . . . . . . . . 35
Tiago E. Castelo de Oliveira and Paulo F. Ribeiro
4 Effect of Harmonics due to Distributed Energy Resources
on Hosting Capacity of Microgrid: A Hardware in Loop-Based
Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Sourav Kumar Sahu, Debomita Ghosh, and Dusmanta Kumar
Mohanta
5 A Scenario-Based Approach for Storage Capacity Determination
to Improve the Hosting Capacity of Distribution Systems . . . . . . . . 87
Mohammad Seydali Seyf Abad, Jin Ma, and Jing Qiu
6 Impact of Capacity Withholding on Hosting Capacity
Analyzing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Saeed Salarkheili and Mehrdad SetayeshNazar
7 A Generalised Deterministic Approach to Evaluate PV
Hosting Capacity of LV Distribution Networks
Under Different Operating Conditions . . . . . . . . . . . . . . . . . . . . . . 149
D. Chathurangi, U. Jayatunga, S. Perera, and A. Agalgaonkar
ix
x Contents
8 Hosting Capacity Maximization Based on Optimal

Reconfiguration of Distribution Networks with Optimized
Soft Open Point Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Ibrahim Mohamed Diaaeldin, Shady H. E. Abdel Aleem,
Ahmed El-Rafei, Almoataz Y. Abdelaziz, and Ahmed F. Zobaa
9 Generation Regulation Control Systems . . . . . . . . . . . . . . . . . . . . . 195
Krishnan Manickavasagam, Ilango Karuppasamy, and Vineetha
Puttaraj
10 Exploring the Concept of Hosting Capacity from an Electricity
Market Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Elias Valenzuela, Rodrigo Moreno, Dimitrios Papadaskalopoulos,
Francisco D. Muñoz, and Yujian Ye
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Chapter 1
An Overview of Hosting Capacity
for Modern Power Grids
Tiago E. Castelo de Oliveira, Paulo F. Ribeiro, Ahmed F. Zobaa,

Shady H. E. Abdel Aleem, and Sherif M. Ismael
1.1 Introduction
Conventional fossil fuel-based resources are affected by several problems, such as

diminishing fuel resources, global fuel price instabilities, and greenhouse gas emis-
sions, particularly gases such as carbon dioxide and carbon monoxide which
adversely affect the environment. Currently, renewable energy-based power gener-
ation is rapidly emerging across the world in response to technical, economic, and
environmental developments, as well as political and social initiatives. Now, renew-
able energy-based power generation is being rapidly developed, worldwide, thanks
T. E. Castelo de Oliveira (*)

Eindhoven University of Technology, Eindhoven, the Netherlands
e-mail: t.e.castelo.de.oliveira@tue.nl
P. F. Ribeiro
Advanced Power Technologies and Innovations in Systems and Smart Grids Group,
Federal University of Itajuba, Itajuba, MG, Brazil
e-mail: pfribeiro@ieee.org
A. F. Zobaa
College of Engineering, Design and Physical Sciences, Brunel University London,
Uxbridge, UK
e-mail: azobaa@ieee.org
S. H. E. Abdel Aleem
Mathematical, Physical and Engineering Sciences Department, 15th of May Higher Institute
of Engineering, Cairo, Egypt
e-mail: engyshady@ieee.org
S. M. Ismael
Electrical Engineering Division, Engineering for the Petroleum and Process Industries (Enppi),
Cairo, Egypt
e-mail: shriefmohsen@enppi.com
© Springer Nature Switzerland AG 2020 1

A. F. Zobaa et al. (eds.), Hosting Capacity for Smart Power Grids,
https://doi.org/10.1007/978-3-030-40029-3_1
2 T. E. Castelo de Oliveira et al.
to various technical, economic, and environmental factors and political and social
initiatives [1–3].
However, the booming integration of distributed generation (DG) units, if not
properly assessed, may lead to various problems such as over- and under-voltages,
excessive line losses, overloading of transformers and feeders, protection failure, and
high harmonic distortion levels exceeding the limits of international standards.
These problems occur when the number of DG units exceeds the maximum permis-
sible penetration level, in short when the system exceeds its hosting capacity
(HC) limit. The HC concept is essential to evaluate the system capacity to host
these DG newcomers without exceeding the allowable operational performance.
Practically speaking, a conflict of interest has been found among the DG owners/
investors and distribution system operators (DSOs) in deregulated energy markets. It
is because the DG investors are looking forward to more and more DG integration
into electrical networks, while DSOs are concerned about excessive DG penetration
problems. Therefore, the HC concept was proposed to resolve this conflict, as it
provides a fair and transparent solution that clearly indicates when to accept or to
reject new DG integration requests [4–6].
Today’s power systems are becoming much more complex, expandable,
dynamic, and non-predictable; therefore, accurate to-date HC analysis studies are
necessary to face the significant challenges associated with the high penetration of
renewable energy resources into modern electrical power systems.
HC is a key enabler for affordable and reliable renewable energy sources; hence,
it is possible to transition away from traditional high-carbon energy sources. It is
expected that the HC will play a significant role in future power systems and smart
grids. Using an HC approach to drive network requirements could steer DG toward
areas of the network where it could have the greatest positive impact on network
reliability and win-win benefits with DG owners. Thus, it is imperative that novel
solutions be sought to enable networks to cope with future developments to realize
resilient distribution networks that can host the pushy DG penetration while enhanc-
ing the system reliability of power supplies and controlling the over-hosted areas,
that is, the areas that have problems because of violation of technical performance
metrics.
Not long ago, quick but conservative methods were used to limit integrated DG
capacities. Nowadays, the HC concept has been developed to define the maximum
amount of DG that can be integrated into the electrical network accurately, without
exceeding any operational performance limits. In addition, the HC is a specific,
measurable, practical, and fair concept that facilitates the ease of use of clear
performance limits as evaluation criteria for DG penetration.
Currently, enhancing the system HC is considered one of the important goals for
DSOs, worldwide [1]. In the literature, common HC enhancement techniques were
investigated, namely, reactive power control, automatic voltage control techniques
such as on-load tap changer (OLTC) transformers, active power curtailment, net-
work reconfiguration and reinforcement, and harmonic mitigation techniques.
The authors believe that along with these HC enhancement techniques, energy
storage (ES) technologies will play a significant role in the near future as a promising
1 An Overview of Hosting Capacity for Modern Power Grids 3
HC enhancement tool. ES systems help in overcoming the overvoltage resulting

from high DG penetration, thus allowing the increase of the system’s HC. ES
systems allow the demand and generation of electricity to be mutually decoupled.
Even though ES is still expensive, it offers unique benefits that cannot be achieved
using other means. Proper sizing and allocation of the ES technologies may postpone
DSO’s plans for network reinforcements. Accordingly, customer-owned and
DSO-owned ES systems should be carefully investigated to help increase the
allowed DG penetration.
In the next sections, the concept of dynamic hosting capacity (DHC) will be
presented. Further, an application of ES in a system is addressed as a powerful tool to
improve a distribution system performance.
1.2 Dynamic Hosting Capacity Approach
In [7–14], HC issues related to voltage regulation and reverse power flow are
presented. However, it is necessary to consider also the time-varying impact of
harmonics on the voltage regulation in the presence of renewables. This poses a
problem as HC conditions vary in time. The main limitation of determining the HC is
its reliability, particularly when the system is under constant harmonic injection due
to the aggregated electronic devices connected to the electrical system, thus making
it a dynamic harmonic injection issue. One question that needs to be raised is how a
dynamic harmonic injection could affect the global HC value due to its impact on the
voltage profile.
A related hypothesis maintains that a dynamic harmonic injection, originating
from external conditions, such as the background harmonic injection, as well as
distortion from the inverters (e.g., in photovoltaic (PV) systems), leads to a DHC
profile which follows the tendency of their harmonic injection. This suggests that it
presents a constructive curve in time, showing that harmonic distortion problems
should be considered when planning improvement strategies to be applied in
electrical distribution systems.
The fundamental characteristic of DHC can be described as the allowance to
increase and expand the usual approach of HC by applying a time variance, which
can designate its behavior against both external and internal variations regarding the
system, such as load conditions, background harmonic distortion, and irradiation
index, among others [15, 16]. Furthermore, in accordance with [15], the HC at the
point of common coupling (PCC) can be defined as the sum of maximum power
possible to be injected for each frequency. Furthermore, time variation is set as a
vital factor in (1.1) within this analysis.
Fig. 1.1 Dynamic hosting Unacceptable

capacity (DHC) approach region
Dynamic HC
[9] Maximum HC
Acceptable region
Minimum HC
Time

X
h Xh V max ,h V max ,h V h ðt Þ
1 g g o
g ðt Þ ¼
Pmax max ,h
PPV ðt Þ ¼ ð1:1Þ
h¼1
Rf h¼1 1 þ tan ðφÞ hX f
Rf
where Pmaxg is the system’s HC; Pmax

PV is the maximum active power generated by the
DG unit; Rf and Xf are the Thevenin’s resistance and reactance at the PCC, respec-
tively; and V hg and V ho are the hth voltage magnitude at the load and the source,
respectively.
It is important to highlight the maximum quantity and the minimum quantity of
power that will be assumed regarding the harmonic voltage values which will be
defined for each instant of time [15]. A descriptive graph of the DHC approach is
shown in Fig. 1.1.
According to Fig. 1.1, it is shown that the maximum and the minimum HC values
have analytical consequences to the electrical grid. Firstly, the minimum value is
deterministic to describe the efficiency of the grid, and it is usually determined
according to the regulations and technical performance metrics of a grid. In other
words, the harmonic backgrounds at the PCC, as well as the voltage values, are close
to cross their respective limits imposed by local standards [15, 16]. Therefore, the
maximum value will regulate the region where the best power quality indices can be
found. Clearly, to achieve these values, it is necessary to perform conditioning
improvements. The determination of these ranges suggests an acceptable region
that can be defined between the minimum and maximum HC values [15]. Conclu-
sively, the larger the acceptable region is, the better the system performance will
be. This concept can determine that the DHC analysis is an important tool to
determine a further analysis regarding better conditions on the grid, considering
the insertion of the distributed energy resources (DER).
Fig. 1.2 Conventional

power flow
P
Voltage
Load
Distance
'V
Fig. 1.3 Bidirectional P Pg DERs

power flow due to high DER
penetration
P = Pg – Pc Pc
Voltage
Load
'V
Distance
1.3 Dynamic Hosting Capacity Considering Storage

Systems
Conventional electrical networks consider the total load flow directly from the
substation to the load. In regard to this fact, there is a voltage drop throughout the
transmission line, as shown in Fig. 1.2. This unidirectional load flow allows for easy
planning of the distribution transformers and the conductor section of the
feeders [17].
When there is a high penetration of DER in the electrical network, there may be a
reversal and/or bidirectionality of the power flow. As an example, the DER absorbs
the consumption of the load to which it is connected, and the additional power will
be injected into the distribution network, feeding other loads nearby. The phenom-
enon is shown in Fig. 1.3. Due to the bidirectional power flow, some negative
aspects can be registered, for example, a voltage rise at the connection point where
the DERs are installed, which can exceed standard indices, leading to possible
losses [17].
Alternatively, storage systems could help the system at the connection point by
absorbing excessive energy generation of renewables. Several researchers have
called this into question, claiming that storage systems can lead to improvements
into the distribution system, such as avoiding the voltage rises due to the bidirec-
tional power flow, as well as improving the HC of the system. Figure 1.4 shows the
storage system’s impact at the connection point. It is possible to understand that the
voltage would not rise in the system because there is no bidirectional load flow into
the system.
Fig. 1.4 The storage P’ Pg DERs

system and its operational
impacts Ps Pc
Voltage
Storage Load
'V |0 Pg –(Pc + Ps) = 0
Distance
Fig. 1.5 The DHC with and

Maximum HC with
without the ES system storage
Maximum HC
without storage
HC
Time
This problem can be outlined in terms of the power of the system. In this case, the
power’s sum will be zero due to the relation Pg (Ps + Ps), where Pg is the power
generated by DERs, Pc is the power consumed by loads, and Ps is the power which
will be stored by the storage system. It is important to highlight that the sum will be
zero, as long as there is a generation from the DERs, for example, in regard to PVs,
only at the moment when there is solar irradiation. This allows for a formal solution
to be found, where ΔV 0. In theory, the lesser the voltage rise, the better the HC
will be.
It is easily verifiable that there are two possible options in order to bring
improvement at the connection point: either regulate the voltage through regulators
such as OLTC or connect some storage systems into the connection point. It is
necessary to analyze which option is better economically speaking. In this analysis,
the storage systems will be under investigation with the intention of bringing
improvement regarding the DHC profile for the system, as shown in Fig. 1.5.
This proposed hypothesis examines how to solve the voltage rise problem in
electrical grids using energy storage systems in order to improve the DHC approach.
The ES system is examined with the intention of finding the best contribution to
decrease the voltage rises, as well as keeping at zero the reverse power flow, thus
improving the hosting capacity. This is a feasible but expensive solution for distri-
bution systems in relation to voltage rise indices due to DERs connected at the
electrical grid.
1.4 Formulation of the DHC with Energy Storage Systems

(ESS)
In the following sections, the importance of considering the maximum amount of

energy and its dependence on the harmonic voltage values are highlighted, which are
defined for each instant of time.
The DHC can be used to calculate the hosting capacity considering time-varying
energy storage systems, which can be defined as the dynamic-storage-hosting
capacity (DSTHC). Consequently, we need to formulate a hypothesis for it. The
DSTHC concept is applied to the allowed voltage magnitude for the system, in this
case, determined by local standards. In theory, the concept of hosting capacity, either
static or dynamic, makes it clear that the maximum voltage magnitude allowed is the
determining factor to find its global value.
The DSTHC was prepared using the same procedure as for dynamic HC, but
taking into account one specific consideration, which has already been mentioned
before. In this way, this assessment considers the voltage rise caused by the
bidirectional load flow into the system, thus determining the outline conditions.
Instead of basing the maximum voltage magnitude allowed for the system on local
standards, the allowed voltage magnitude can be assumed to be equal to the voltage
rise for systems without an ES system. Moreover, the voltage reference will be
nominal, which means equal to 1.0 (p.u). Those considerations are shown below.
(
V gmax ,1 ¼ V o ðt Þ
V 1o ðt Þ ¼ 1:0 ðp:uÞ
This method represents an alternative to analyze the storage system behavior

because, in this case, the maximum voltage rises already existent at the system,
considering the DERs, are the determinant factor to find a hosting capacity value. In
addition, the maximum storage power is calculated only for the fundamental fre-
quency, where the voltage rise is an important impact factor. Thus, Eq. (1.1) can be
given in (1.2).
1 V o ðt ÞðV o ðt Þ 1, 0Þ
ΔPmax
storage ðt Þ ¼
ð1:2Þ
Rf 1 þ tan ðφÞ XRff
Taking advantage of DSTHC, we can hypothetically propose that the maximum

power stored found in (1.2) is a complement of the maximum power generated given
in (1.1). Combining (1.1) and (1.2), we have a total maximum power generation for
the system considering an energy storage system, given in (1.3).
Pgmax 0 ðt Þ ¼ Pmax
g ðt Þ þ ΔPstorage ðt Þ
max
ð1:3Þ
It is easily verifiable that the maximum dynamic HC can be extended by (1.3) in

(1.4), where it can be defined by linear regression.
max 0
DSTHCtotal ðt Þ ¼ HCmax
total ðt Þ þ ΔHCstorage ðt Þ
max
ð1:4Þ
The new voltage profile for the system after the energy storage system can be
verified. There will not be a voltage rise. The voltage drop can be determined by
(1.2). Thus, the new voltage value is given in (1.5).
V 0o ðt Þ ¼ V o ðt Þ ΔV storage ðt Þ ð1:5Þ
It is important to highlight that the local DER system, considering an ES system,

will be dimensioned by (1.4) and (1.5), in order to ensure future improvement in the
current installation.
1.5 Further Remarks
HC research is a key enabler for affordable, reliable, and renewable energy sources,
so it is possible to transition away from traditional high-carbon energy sources.
Therefore, it is imperative that novel solutions be sought to enable networks to cope
with future developments to realize resilient distribution networks that can host the
pushy DG penetration while enhancing the system reliability of power supplies and
controlling the over-hosted areas.
This chapter presents an overview of the HC concept and its developments,
defines the main HC performance limits, and its up-to-date enhancement techniques.
Then, the chapter presents the concept of dynamic HC as well as its application to
energy storage systems. The authors presented a new concept for calculating the
local HC as a contribution to the context of planning and improving DER installa-
tions into distribution systems. The definition of DHC has been discussed, as well as
its application to energy storage systems.
References
1. Ismael, S. M., Aleem, S. H. E., Abdelaziz, A. Y., & Zobaa, A. F. (2019). State-of-the-art of
hosting capacity in modern power systems with distributed generation. Renewable Energy, 130,
1002–1020. https://doi.org/10.1016/j.renene.2018.07.008.
2. Ismael, S. M., Aleem, S. H. E., Abdelaziz, A. Y., & Zobaa, A. F. (2018). Practical consider-
ations for optimal conductor reinforcement and hosting capacity enhancement in radial
distribution systems. IEEE Access, 6, 27268–27277. https://doi.org/10.1109/ACCESS.2018.

2835165.
3. Ismael, S. M., Aleem, S. H. E., Abdelaziz, A. Y., & Zobaa, A. F. (2019). Probabilistic hosting
capacity enhancement in non-sinusoidal power distribution systems using a hybrid PSOGSA
optimization algorithm. Energies, 12(6), 1018. https://doi.org/10.3390/en12061018.
4. Sakar, S., Balci, M. E., Aleem, S. H. E. A., & Zobaa, A. F. (2016). Hosting capacity assessment
and improvement for photovoltaic-based distributed generation in distorted distribution net-
works. In 2016 IEEE 16th International Conference on Environment and Electrical Engineer-
ing (EEEIC) (pp. 1–6). Florence. https://doi.org/10.1109/EEEIC.2016.7555515.
5. Sakar, S., Balci, M. E., Aleem, S. H. E. A., & Zobaa, A. F. (2017). Increasing PV hosting
capacity in distorted distribution systems using passive harmonic filtering. Electric Power
Systems Research, 148, 74–86. https://doi.org/10.1016/j.epsr.2017.03.020.
6. Sakar, S., Balci, M. E., Abdel Aleem, S. H. E., & Zobaa, A. F. (2018). Integration of large-scale
PV plants in non-sinusoidal environments: Considerations on hosting capacity and harmonic
distortion limits. Renewable and Sustainable Energy Reviews, 82(part 1), 1–1684. https://doi.
org/10.1016/j.rser.2017.09.028.
7. Pesaran H.A, M., Huy, P. D., & Ramachandaramurthy, V. K. (2017). A review of the optimal
allocation of distributed generation: Objectives, constraints, methods, and algorithms. Renew-
able and Sustainable Energy Reviews, 75(2015), 293–312.
8. Tan, W. S., Hassan, M. Y., Majid, M. S., & Abdul Rahman, H. (2013). Optimal distributed
renewable generation planning: A review of different approaches. Renewable and Sustainable
Energy Reviews, 18, 626–645.
9. Bollen, M., & Hassan, F. (2011). Voltage magnitude variations. In Integration of distributed
generation in the power system (pp. 141–222). Hoboken: Wiley.
10. Carvalho, P. M. S., Correia, P. F., & Ferreira, L. a. F. (2008). Distributed reactive power
generation control for voltage rise mitigation in distribution networks. IEEE Transactions on
Power Apparatus and Systems, 23(2), 766–772.
11. De Oliveira, T. E. C., Bonatto, B. D., Filho, J. M. C., Ribeiro, P. F., & Santos, I. N. (2015).
Análise Econômica da Hospedagem de Fontes de Geração Distribuída no Sistema Elétrico de
um Campus Universitário. In An. do CBQEE 2015 – XI Conferência Bras. sobre Qual. da
Energ. Elétrica.
12. Bollen, M. H. J., & Rönnberg, S. K. (2017). Hosting capacity of the power grid for renewable
electricity production and new large consumption equipment. Energies, 10(9), 1325.
13. De Oliveira, T. E. C., Ribeiro, P. F., & Santos, I. N. (2016). Determining the harmonic hosting
capacity of PV sources for a university campus. In Proceedings International Conference on
Harmonics and Quality of Power (ICHQP) (Vol. 2016–Decem, pp. 836–841).
14. Baccino, F., De Nigris, M., Gianinoni, I., Grillo, S., Massucco, S., & Tironi, E. (2012). A
methodology for evaluating PEVs hosting capacity margins in distribution grids. In IEEE
Power Energy Society General Meeting (pp. 1–6).
15. de Oliveira, T. E. C., Bollen, M., Ribeiro, P. F., de Carvalho, P. M. S., Zambroni, A. C., &
Bonatto, B. D. (2019). The concept of dynamic hosting capacity for distributed energy
resources: Analytics and practical considerations. Energies, 12(13), 1–18.
16. De Oliveira, T. E. C., Carvalho, P. M. S., Ribeiro, P. F., & Bonatto, B. D. (2018). PV hosting
capacity dependence on harmonic voltage distortion in low-voltage grids: Model validation
with experimental data. Energies, 11(2), 465.
17. Paludo, J. A. (2014). Avaliação dos Impactos de Elevados Níveis de Penetração da Geração
Fotovoltaica no Desempenho de Sistemas de Distribuição de Energia Elétrica em Regime
Permanente.
Chapter 2
Hosting Capacity Determination via Risk
Analysis Approach Involving Likelihood
and Severity of Operational Constraint
Violation
Hassan Al-Saadi
2.1 Introduction
The high-power penetration into a distribution network exported from photovoltaics

(PVs) during noon times has been cited to have negative impacts on normal
operational condition. While a low penetration could be associated with positive
impacts like voltage improvement or transformer relief, high penetration can have
adverse impacts on voltage regulators, LV feeder congestion, direction of the power
flow, frequency, etc. [1]. One problem is that the power injected into the network
from PVs can intermittently influence the voltage profile of the distribution lines,
causing them to be either over or under specified limits. For another example,
Automatic Q Control (AQC) that compensates reactive power will be significantly
overloaded especially when most distributed generators (DGs) such as PVs export
only active power. When the power penetration is high, the active power directed
from a substation toward the low-stream buses will be low, unbalancing active-
reactive power adversely. The power factor (PF) correction will be triggered inter-
mittently in which the capacitor banks and tap-changing transformers might be
exposed into a situation that is not designed to operate. On-line voltage regulator
(OLVR) may disfunction its working principle conditions as well. During normal
operation (without high power penetration from PVs), OLVR reacts in accordance
with measurement units located usually at the end of the feeder to maintain the
voltage within utility standards. During high power penetration, PVs connected at
the end of the distribution feeder may allow the measurements to send deceptive
voltage readings. As a result, voltage dip may dominate feeder voltage profile, which
H. Al-Saadi (*)
School of Electrical and Electronic Engineering, University of Adelaide, Adelaide, SA,
Australia
e-mail: hassan.al-saadi@adelaide.edu.au

https://doi.org/10.1007/978-3-030-40029-3_2
12 H. Al-Saadi
specifically could occur in the mid of the feeder [2]. The problems mentioned above,
and others not mentioned here, give rise to a new expression in the taxonomy of
power system called “hosting capacity” (HC). While the general definition may
simply be referred to a maximum limit of a power network in hosting PVs without
jeopardizing connected appliances, the determination of HC is still an attractive topic
that has not reached the consensus yet. It has been found that the definitions of HC
seem to converge in their implication in the published literature. Examples of these
definitions are as follows:
The hosting capacity of the LV grid for dispersed generation is restricted by the maximum
permissible voltage rise within the grid and the maximum short-term loading of the
transformer and the cables, due to the diurnal cycle of the PV [3]
The maximum amount of new production or consumption that can be connected without
endangering the reliability or quality for other customers” and “the acceptable degree of
DER1 penetration under given circumstances [4]
The upper limit of DG before network congestion occurs [5]
The largest PV generation that can be accommodated without violating the feeder’s oper-
ational limits [6]
In this chapter, the adopted definition of hosting capacity (HC) is the maximum
capacity of DGs connected in a distribution network without disrupting the normal
operational conditions of that network.
2.2 Recent Methods on Hosting Capacity Determination
This section provides a quick summary of different techniques used to determine HC

in the recent web available sources. In the existing literature for HC determination,
DGs’ types, sizes, locations, and network’s specifications factorize the value of
HC. The HC is network oriented and based on DGs’ types and number with other
details explained in Table 2.1 for existing methods for assessing high DG impacts
and HC. In addition, a summary for a number of studies conducted with different
approaches tackling HC is provided. In the table, the studies are classified into
steady-state planning [7] or dynamic planning [8–12]. The attempt for steady-state
planning is to determine the HC of network regardless of the daily time variation,
whereas the dynamic related studies are meant to determine the HC according to a
specific date and daytime. Thus, the HC has been evaluated with different criteria,
methods, and power flow directions. The direction of power flow is assessed to be
either unidirectional, from distribution transformer to the consumers, or bidirec-
tional, from local distribution transformers to consumers or from consumers to
nearby transformers. The bidirectional power flow occurs when the total amount
of generation from DGs suppresses the local demands in which the network starts
1
DER is the abbreviation of distributed energy resource.
Table 2.1 The summary on the existing different methods for HC studies
Case study Criteria HC and penetration method Grid type
[7] Two buses with PV connected into Voltage and line current limits Analyze the steady-state bus voltage and line current Bidirectional
one bus and the another is substa- behavior when PV penetration increases with different sce-
tion bus with constant voltage in narios such as different PFs, percentages of load penetration,
addition to the assumption of loads and substation voltages
summed up and connected into the
PV bus
[8] Use an actual and large distribu- For HC purposes, ANSI C84.1 Create different scenarios wherein the PV penetration, at Unidirectional
tion feeder in California with standards as overvoltage criteria each scenario, is increased by the insertion of the additional
commercial and residential urban were taken. Other criteria taken PVs on the randomly selected locations
and rural feeders, modelled using for other analyses such as fre- The PVs inserted are selected from a set of PV sizes (43,502
OpenDSS quency distortion residential preferred PVs, 2625 commercial preferred PVs
which are based on SCADA and AMI, provided from Cal-
ifornia Solar Initiative Survey) in which that set is probabi-
listically distributed according to customer services of PV
sellers
[9] 302 low-voltage rural and subur- 3% voltage variation, no Classification approach is about classifying the HC capabil- Unidirectional
ban grids overloading electrical ity of the grids into five categories (very weak up to very
components strong) using Weibull fitting model
HC capability of a grid is probabilistically treated using
Weibull distribution as well after evaluating the single grid
with a number of DG random sizes with different configu-
rations
DG penetration is performed in piecewise increase with
discrete interval 0.1 kW
(continued)
2 Hosting Capacity Determination via Risk Analysis Approach Involving Likelihood. . .
13
Table 2.1 (continued)
14
Case study Criteria HC and penetration method Grid type

[10] 232 residential feeders in the North Line to neutral voltage limit Then at each penetration, 1-day analysis (5 min) is Unidirectional
West of England with line ampacity performed using real PV and load profiles from Elexon
PV penetration is done by the increase of the percentages of
customers with PV panels (e.g., 10% of customers with PVs
or 20% of customers with PV), 3 kW for each PV. HC is
calculated once one of the criteria is met
The actual feeders are classified according to their parame-
ters from HC perspective in an extrapolated way
[11] 400 kV national grid in Sweden. Bus voltages and transformer The 100 stepwise increases of two virtual pre-allocated wind Bidirectional
and cable ampacity turbines are performed with the use of weather-recorded data
of 2 years (2009 and 2010) resulting in 2 million LF calcu-
lations
At each increase, the criteria are checked to draw a system
performance line in relation with wind power penetration
[12] Medium- and low-voltage network Coincident hours with voltage Piecewise ranges of load and WT percentages are created to Bidirectional
(16 buses, 61 buses) and line thermal current find the joint probability of occurrence among these ranges
constraints to create different scenarios (e.g., 10% demand with 20%
WT generation or 80% demand with 50% WT generation).
The annual profiles are used for finding the periodic joint
probabilities
The summarized steps are as follows:
1. Specify a range of load and DG (e.g., 10–20% load and
70–80% DG connected) that needs to be evaluated
Perform hourly LF for 1 year using 2003 recorded profiles
of one or two pre-allocated WT outputs and one profile for
all loads. When the range in step 1 occurs, record the dura-
tion of the number of hours that coincide at this range which
is called “coincidence hours”
Go to step 1 and specify different range and then perform
steps 2 and 3 till all ranges are reached
H. Al-Saadi
2 Hosting Capacity Determination via Risk Analysis Approach Involving Likelihood. . . 15
exporting power back to the bulk grid through grid supply point. Each referenced
method is intensively reviewed and analyzed within this table.
The method that grabbed attention most and is adopted in the current work is
introduced by Bollen and Hassan [4]. They explain their concept for HC determi-
nation and introduced the general fundamentals to seize the impacts of DGs. It is
stated that “The impact of distributed generation can be quantified only by using a set
of performance indicators.” The further explanation of this concept is provided in
[13]. The summarized steps are as follows:
First stage Establish a number of performance indices.
Second stage Specify a number of standardized permissible limits for the
established indices in the 1st stage such as a limit can be defined
by EN50160 [14] or any other standards defined by the local utility.
Third stage Find the mathematical relationship between the indices and the
increases in the number of DG connections.
Fourth stage Identify network HC by streamlining the different states (acceptable
deterioration, unacceptable deterioration or critical deterioration)
according to the exceedance of the limits in second stage.
The last mentioned method is depicted in Fig. 2.1. The x-axis displays DG
penetration increases starting from a prespecified amount of power penetration.
The y-axis displays the degree of deterioration accompanied with these increases.
The region shaded in light green indicates that the system is under tolerable
deterioration. The region shaded in red is where the system is considered under
intolerable deterioration. The network HC limit is identified once the deterioration
index crosses from one region into another. There is a possibility of having critical
region that could deliver two different values of HC limits. This region is shown in
the figure with partially yellowish color. In this chapter, this method is promoted,
thereby establishing new risk indices for operational performance which is intro-
duced in the next section.
Fig. 2.1 HC determination method. Adapted from [4]

16 H. Al-Saadi
2.3 System Formulation and Evaluation
2.3.1 System Evaluation
One of the deterring obstacles in the topics of risk assessment in power system is the
analytical expression of relative frequency of occurrence. The number and different
sources of uncertainties in the network make the task of finding analytical represen-
tation non-trivial, especially in linking the input variables to the response variable.
Instead of the analytical solution, the power of numerical computing techniques is
proven to provide feasible answers for many engineering problems. Therefore, in
this chapter, the nonintrusive method is employed for a set of scenarios. For each
scenario, the numerical simulation is adopted for the evaluation of the system
through probabilistic distributions and statistical analysis.
To formulate the problem mathematically, two quantities are considered: uncer-
tain quantities and fixed ones, noted by x and d, respectively. While the uncertain
quantities, x, stand for the intermittent output power of the connected PV or loads,
the fixed quantities, d, represent the level of PV increase connection and the time of
the day. In Fig. 2.2, two stages are essential in using nonintrusive method: formu-
lating the uncertainty, x, and propagating in the form of quantity of interest.
The procedures of system evaluation can be explained in three steps.
2.3.1.1 Step A: Uncertainty Characterization
This step characterizes the involved uncertainties by scenario-based method or

probability density distribution or fuzzy numbers. This characterization is usually
achieved through intensive work in a multidisciplinary environment.
The common practice in power system is to consider the grid-connected PV as a
power injector with linear relation to the solar radiation incident on the surface.
However, if the uncertain quantity known as clearness index, kt, is considered
instead of solar radiation, then the output power of PV is treated with nonlinear
relation to this index. The clearness index represents the degree of the transparency
in the atmosphere during a specified time in the day, t. It is bounded with minimum
Fig. 2.2 System evaluation using nonintrusive method

and maximum values, noted kt min and kt max, respectively. The more tendency of this
quantity to the maximum values means having more cloudless atmosphere. With the
assumption that each PV is equipped with maximum power point tracker, the output
power of PV as a function of clearness index, PPV(kt), is provided as follows:
PPV ðkt Þ ¼ A ∙ η ∙ I βt ðkt Þ ð2:1Þ
where η is the efficiency of the entire PV unit including the aging degradation; A is
the surface area; and I βt ðkt Þ is the radiation incident on the surface with the incline
angle, β. The latest is a function of clearness index which can be expressed as
follows [15]:
8
>
> D ∙ k t þ D0
kt
D > 0&D0 0
>
> 0
< 1 þ eBðB kt Þ
β
I t ðk t Þ ¼ kt ð2:2Þ
>
> D ∙ k t D0 0 D > 0&D0 < 0
>
> 1 þ e ðB kt
B Þ
:
0 D0
where D and D0 are composed of different fixed parameters, which can be tracked
back following the same reference. B and B0 are the fixed parameters of the logistic
function that shapes the relational curve between the diffuse fraction and kt. The
probabilistic representations of this quantity have been addressed with different
probability density functions (PDFs) such as Weibull, Gaussian, double beta,
Boltzmann, single gamma, beta, bi-exponential, double Gaussian, triple Gaussian,
Weibull-logistic, etc. Single gamma [16] is adopted for this work.
The performance of the presented model is depicted in Fig. 2.3. In this figure, the
solar radiation received on the surface of the connected PV is represented in z-axis
(black striped area edged with red line) and varies according to y-axis that represents
different times during a single day. For demonstration purposes, three different times
(8:00 AM, 12:00 PM, and 1:00 PM) are considered to represent the probability
densities, which are aligned according to the axis of day hour. The x-axis, named
“power flow net,” represents the actual power import/export from or to the utility
grid. In each single hour, the probability density of solar irradiation is depicted
against the power flow net, showing the most likely occurrence of export/import
power from a PV during single hour in a day. In this step, only single hour is
considered for the uncertainty characterization.
2.3.1.2 Step B: System Evaluation via Nonintrusive Method
The large number of factors considered in the evaluation of the system increases the
complexity of the evaluation task. The nonintrusive method is applicable for numer-
ical based evaluation taking into the account the recent advances in computing
power. It has been utilized in the recent studies that concern the problems of
18 H. Al-Saadi
Fig. 2.3 The probabilistic representation of output power produced by grid-connected PV in

relation to the probability density of solar irradiation. Three PDFs across different daytimes
(8:00 AM, 12:00 PM, and 1:00 PM) are shown
power system such as the computation of the probabilistic power flows [17, 18]. The
nonintrusive methods govern the response variable to the uncertain input quantities
without the need to deal with the internal mathematical relation. It views the system
as a black box and allows us to analyze the stochastic performance cheaply through
the statistics such expectations, medians, quantiles, etc.
Let us consider the system to have uncertain functions represented by
n dimensional random vector, X, and parametrized by the fixed quantities
represented by d; then the function of the black box can be considered as g(X, d ),
and the response vector of variables of interest, represented by Y, can be expressed as
follows:
Y ¼ gðX, dÞ ð2:3Þ
Herein, the number of variables, n, in each vector is shared equally likely for PVs
and loads; the function g(X, d ) comprises internal computational relation that is not
checked by the author of this chapter. The internal relations are based on the used
software for computing power load flows.
The nonintrusive method simplifies the use of statistics such as sampling, quad-
rature treating, and regressions, by employing m stochastic runs of deterministic
evaluations.
2.3.1.3 Step C: Hypothesis Testing
Hypothesis testing is when the variables of interest are analyzed, and statistics-based
decisions are made. It should be noted that it is not possible to 100% accept or reject
the decision made out of a hypothesis. Instead, these decisions should be assessed
carefully in accordance with common traditional practices such as confidence
intervals, standard deviations, and variances. In the case of this work, the statistical
results of the sparse grid technique to the Monte Carlo technique is tested by relative
error and used for driving the statistical analysis such as the use of expectation in the
histogram and bar-stem diagram.
2.3.2 Sparse Grid Technique
In multidimensional environment, the univariate quadrature rule can be employed to

solve the numerical integration if the random variables are mutually independent.
Commonly, tensorization is used when applying such rule. With the high number of
dimensions, however, the classical tensorization would become arduous to be solved
due to the curse of dimensionality. The solution is to use sparse grid technique that
requires a smaller number of points within the dimensions.
Let us consider Λlr denotes one-dimensional integration involving a sequence of
univariate quadrature rules for a function, g, given its weight function, f, at i-th
dimension, as follows:
XM l
Λli ¼ ki ¼1
i
gðxli ,ki Þf ðxli ,ki Þwli ,ki , ð2:4Þ
where the index li denotes the accuracy’s level for a dimension i, i ¼ 1, 2, . . ., n,

which depends on the nodes’ number mli ; wli ,ki represents precomputed quadrature
weights; and xli ,ki 2 ½0, 1 represents nodes (abscissas). The exactness of Λli improves
with higher li, where li 2 ℕ .
In the classical tensor, the curse of dimensionality emerges as a big problem when
high level of accuracy is required with large dimensional problem. The classical
tensor leads to computation degree to be O(mn) which requires the increase of
computing power exponentially if the number of dimensions increases. The sparse
grid technique, in the meanwhile, which is introduced first by Smolyak [19], deals
20 H. Al-Saadi
with the finest number of nodes, letting the degree of computation to be O

(m ∙ (Log m)n 1) maintaining similar accuracy.
Therefore, employing the sparse grid technique on tensor product results in a
weighted sum of product rules, indexed by the level of accuracy, as follows [20]:
Xℓ X
Ξn ðℓ Þ ¼ Uℓ
z¼ℓnþ1 z l2Θnz
ðΛl1 ⨂ ⨂Λln Þ, ð2:5Þ
where Θnz denotes the index set of multi-index vectors, fℓi gni¼1 , such that Θnz ¼
n P o
ℓ 2 fℕgn : jℓ j ¼ di¼1 ℓ i ¼ z þ n &f∅ : z < 0g , and U ℓz ¼ ð1Þℓz n1 ℓz .
The extended formula of the tensor product is as follows:
Xm1 Xm2 XM n
1
Λl1 ⨂ ⨂Λln ¼ i1 ¼1 i2 ¼1
rn ¼1
w i1 ⨂ ⨂wn
in

g x1i1 , , xnin ∙ f x1i1 , , xnin : ð2:6Þ
Due to the nesting feature, the selection of Kronrod-Patterson rule is adopted for
univariate quadrature. This rule adds more nodes to the preceding level of accuracy,
li 1, thereby improving the exactness in the dimension evaluation.
2.4 Proposed Risk Analysis Approach for Hosting Capacity

Determination
2.4.1 Risk Definition
The involvement of the risk assessment require clarification and definition for the
sake of approaching clear and directive framework. In this regard, the term “risk” has
been introduced by ISO/IEC [21] to be “The combination of probability of an event
and its consequence.” ISO/IEC also introduces “risk analysis” to be “systematic use
of information to identify and to estimate the risk.” Later, ISO replaced the risk’s
definition into “Effect of uncertainty on objectives” [22]. In Australia/New Zealand,
AS/NZS ISO 31000:2009 adopted the last definition [23]. In addition, the definition
of the term “risk” has been addressed, by Kaplan and Garrick [24], using triplet
questions: what are the possible events, what are the consequences of these events,
and what are the associated probabilities of an event? Afterward, these triplet
questions become the main contributor for defining the so-called quantitative risk
assessment (QRA), especially in engineering applications, such as in [25]. After
identification and description of possible hazardous phenomena, the relative fre-
quency of occurrence (occurrence probability) can be used to quantify the identified
risk [26].
2.4.2 Deterioration Indices and Hosting Capacity
2.4.2.1 Risk Indices
The violations of the prespecified rated current in the line and standardized voltage
ranges are two problems identified previously in [27] and discussed in [28, 29] to
have a negative impact on the operational performance of distribution network when
large number of PVs is connected. These steady-state operational constraints can be
used to assess the degree of risk exposure to the network. The degree might be eased
or hardened depending on the technical standards adopted by the distribution power
utility.
During an hourly assessment of the operational performance (as discussed in
Fig. 2.3), the violations concerning bus voltages and ampacity of lines are consid-
ered for formulating the deterioration indices, signified as D.Index 1 and D.Index
2, respectively. Each risk can be quantified based on the number of violation and on
the extent of consequence according to [26]. Both can characterize the likelihood and
severity of risk components. Therefore, in each risk index, the number of violation
(s) occurring in the entire network, denoted by N , and the extended consequence of
the violation(s), denoted by D , are considered. The extended consequence, D, is
formulated here as an accumulative sum of the percentage of the overvoltage
subtracted from the maximum acceptable voltage bound, symbolized by Vmax. The
following steps counted for overvoltage violation, D.Index 1, are used to determine
the values of N and D during each stochastic pocess run:
First step Set N 0 ¼ 0 and D 0 ¼ 0.
Second step If |Vbus(i)| > Vmax, then N i ¼ N i1 þ 1 and
Third step D i ¼ D i1 þ jV bus ðViÞmax
jV max ðiÞ
ðiÞ , where i refers to the node number.
Fourth step Repeat step 2 until all buses are covered.
Fifth step Store the data so that N ) D:Index 1N and D ) D:Index 1D .
After performing m-times of evaluations, each risk index can be quantified by
computing the expectation of its likelihood of violation’s occurrence and severity of
its consequence. Following the black box in Fig. 2.2, functional relation between
joint PDFs denoted by fjoint(X, d ) and internal functions of the system denoted by g
(X, d) can be used to estimate the expectation of likelihood as follows:
Z
½D:Index 1N ¼ D:Index 1N ðgðX, dÞÞ ∙ f joint ðX, d Þ dX, ð2:7Þ
X
and the expectation of the severity is estimated as follows:

22 H. Al-Saadi
Z
½D:Index 1D ¼ D:Index 1D ðgðX, dÞÞ ∙ f joint ðX, dÞ dX, ð2:8Þ
X
Similarly, the expectations of the likelihood and severity of D. Index 2 are

estimated in the same way.
2.4.2.2 Hosting Capacity Determination
The probability-consequence diagrams are common for risk visualization. Others are
risk bubble representations, curve approaches, uncertainty boxes, and strength-of-
evidence assessment [30]. However, a precise diagram has not been recommended
yet for the era of the current power grids. This leaves the debate open as the
informative graphical diagram for risk display may rely on the nature of the problem,
which can be ultimately decided by the risk analysts/reporters.
The expectations of the system performance when increasing multiple PV con-
nections are displayed in terms of two directions of risk index, i.e., likelihood and
severity. The likelihood takes the vertical axis, whereas the severity takes the
horizontal axis. The stepwise increases of PV connections are depicted as a third
axis using scatter points and distinguished by the size/color according to
jet-colored bar.
Using expected values, streamlining different regions, to extinguish the state of
risk from being negligible into unacceptable, is edged by a stripped red line. To
clarify further, the striped red line partitions diagram into two areas: not shaded area
where the PV connections expose no thread to system’s performance. The red
shaded area indicates that PV connection is expected to expose the system into a
thread of technical violation. In Fig. 2.4, three risk indices are drawn, and the values
of these indices when crossing among the specified areas are shown in the jet-colored
bar. Depending on the significance of each index, the limit of HC is decided
considering these three indications. Mathematically speaking, the limit of HC can
be formulated if the maximum bound of risk index is preassigned such as maximum
bound of likelihood, Lmax, and maximum bound of severity, Smax, for D.Index 1, as
follows:

ðD:Index 1N , D:Index 1D Þ : D:Index 1N Lmax
HCðD:Index 1Þ ¼ , ð2:9Þ
ε _ D:Index 1D Smax ε, ε 2 ℕ
where ε means the degree of allowance error that depends on the resolution of the
stepwise PV increase.
Similarly, other HC limits for other D.Index can be driven and concluded for final
HC determination.
Fig. 2.4 Risk-based analysis method
2.5 Case Studies
Two case studies are presented in this chapter to show the effectiveness of this
approach. The work is implemented using MATLAB R2018b on a desktop com-
puter (Intel® i7-2600 4-core with 3.4GHz processor). In this section, the simulation
of superior of spare grid technique is tested against Monte Carlo technique. Then, the
approach of HC is implemented on two distribution feeders and networks named
11 bus distribution feeder and large distribution network. The operational voltage at
the main substation transformer for both cases is kept constant via On-Load Tap
Changer (OLTC) to be 1 pu all the time (the effect of OLTC on HC is out of the
scope of this chapter). The PDF of the clearness index, kt, is set to follow the time
characteristics of Adelaide city (35oS), Australia. The values of parameters of the PV
model introduced in (2.2) for three different times are shown in Table 2.2. The loads
are characterized by Gaussian PDFs using the 3-σ principle around the load maxi-
mum values. The nominal voltage is expected to be 1 pu, and voltage exceeding
1.1 pu is considered as an overvoltage violation, as specified by voltage standard
EN50160 [14]. An overcurrent is counted when the current flow at any line within
the targeted zone exceeds 150% of that line’s rated-current capacity.
2.5.1 Simulation Test
In the comparison between Monte Carlo technique and sparse grid technique, two
quantities are tested: relative error and computational time. The reference result is
computed by the number of 40,000 samples using Monte Carlo technique. The aim
24 H. Al-Saadi
Table 2.2 PV data for the PV model parameters PV model specifications

uncertainty model at different
Time D D0 β 15o
daytimes in a day in
December 12:00 1.327 0.0465 η 0.7
10:00 1.174 0.0246 θ 15o
8:00 1.094 0.0122 ρ 0.3
Fig. 2.5 Relative errors against the number of samples for fair comparison between Monte Carlo
technique and sparse grid technique
of using the relative error is to compare the modelling accuracy. The results are
shown in Fig. 2.5 displaying the values of up to 1000 samples. The relative errors
computed using sparse grid technique shows a steady tendency toward the optimal
values, i.e., without deviating arbitrarily. This is because the coverage of Monte
Carlo technique is completely random and not expected to be steadily improving,
unless a large number of samples are used, which is computationally inefficient.
Following the aforementioned analysis, the sample number used for the computation
of Eqs. (2.7) and (2.8) is 1000 that induces the relative error equal or less than the
value of 0.003.
Another feature of sparse grid technique is the time speed performance. In
Fig. 2.6, Monte Carlo technique seems to perform faster with the number of samples
less than 200. The last result is not beneficial due to the lack of stable relative errors.
Instead, sparse grid technique performs in the sample number of 1000 as five times
faster as the performance of Monte Carlo technique. Therefore, sparse grid technique
is employed only in the following simulation results. It is obvious that sparse grid
proves its superiority over Monte Carlo technique, explained by selecting the finest
number of samples through a unique quadrature rule mentioned earlier. It should be
noted that this feature is considerably important in the purpose of providing accurate-
fast hourly risk assessment that can be incorporated easily with other time-based
automation approaches in the modern power grids.
Fig. 2.6 The computational

time in seconds consumed
for both techniques during
the assessment against the
number of samples
Fig. 2.7 Radial distribution network with nine points of customer connections feeding
36 PV-installed households
Table 2.3 Details of the maximum contractual loads (kW) in each house, point of customer
connection, and feeder
Level Middle feeder Sided feeder
Each house 3.3 3.3
Each point 9.9 19.8
Entire feeder 59.7 29.7
2.5.2 11 Bus Actual Feeder
An actual distribution network with a 11/0.4 kV substation (see Fig. 2.7) is used to
implement the proposed risk-based approach. The technical details are shown in
Tables 2.3 and 2.4 (refer to [31] as the original source of this feeder). The rated
power of the substation transformer is 125 kVA. With an average of maximum
contractual load of 3.3 kW in each house and typical load profile shown in Fig. 2.8,
the performance of the network is estimated through the aforementioned risk indices.
The assessment is conducted through a 10-kW stepwise increase of PV connec-
tions at the point of common coupling which is, in this case, the same as the point of
customer connection. At each step, stochastic run is performed with the use of
26 H. Al-Saadi
Table 2.4 Technical details Specification Middle feeder Sided feeders

of the distribution feeder
Cable type 3 50 + 25Cu 3 50 + 25Cu
Resistance (ohm/km) 0.391 0.078
Reactance (ohm/km) 0.391 0.078
Total length (m) 240 300
Max ampacity (A) 166 166
Fig. 2.8 Daily load profile

for typical residential
electric consumptions
two-dimension sparse grid technique of 13 levels. In Fig. 2.9, three different

daytimes (8:00 AM, 10:00 AM, and 12:00 PM) are considered. The risk indices,
D.Index 1 and D.Index 2, are estimated for each time accordingly; see the subfigures
(a–f). The threshold for severity for D.Index 1 is set to be 150% of the specified
ampacity. The threshold for D.Index 2 is set to be 0.1 pu over the nominal voltage of
the point of customer connection.
For D.Index 1, it is obvious that introducing PVs improves the system perfor-
mance during the PV increase from 0 kW up to 10 kW. This is especially clear when
the time changes from 8:00 AM, subfigure (a), to 12:00 PM, subfigure (e). At
12:00 PM, the PVs are expected to contribute the most, thereby relieving the
network stress. However, at more than 30 kW of PV connection, the network
exacerbates rapidly, suggesting actions/regulations to be taken/drawn.
It is important to mention that the household load has been set to follow the
percentages out of its maximum contractual values, as indicated in Fig. 2.8. This
timely averaged load varies significantly according to statistics of the dweller’s
behavior in energy consumption, which is commonly modelled considering the
diversity factor. The severity of this index is recorded as a violation when connecting
30 kW at 10:00 AM and 20 kW at 12:00 PM. No severe threat is indicated when
increasing PV connections from 0 to 60 kW during the hour around 8:00 AM.
In terms of D.Index 2, the network is expected to experience overvoltage issues if
60 kW or more of PV connection is placed during the hour around 8:00 AM; see
subfigure (b). At 10:00 AM, “1.5” bus in any feeder is expected to experience an
overvoltage problem if 20 kW of PVs is installed in each point of customer
connection; see subfigure (d). At 12:00 PM, 20 kW is expected to increase the
likelihood of number of buses with an overvoltage issue; see subfigure (f) with
likelihood at “2.8.” However, from subfigures (d) and (f), it is clear that the value,
20 kW, of PV connections is not expected to cause a severe overvoltage in the points
Likelihood Severity Severity’s threshold
2 4
8:00AM D.Index 1 D.Index 2
Likelihood & Severity
1.5 3
1 2
0.5 1
0 0
a) b)
6 6
10:00AM
D.Index 1 D.Index 2
5 5
4 4
3 3
2 2
1 1
c) 0 d) 0
8 8
12:00AM
D.Index 1 D.Index 2
6 6
4 4
2 2
0 0
e) f)
0 10 20 30 40 50 60 0 10 20 30 40 50 60
PV increase (kW) PV increase (kW)
Fig. 2.9 Three risk assessments for D.Index 1 and D.Index 2 at three different daytimes (8:00 AM,
10:00 AM, and 12:00 PM)
of customer. The network is expected to experience a severe overvoltage problem

during 10:00 AM and 12:00 PM when 30 kW or more is connected; see the stem (red
pointed) lines at subfigure (d) and subfigure (f).
28 H. Al-Saadi
Table 2.5 Hosting capacity limits for 11 bus feeder according to likelihood only
Time Load percent (%) HC limit1 (kW) HC limit2 (kW)
8:00 AM 35 <50 <60
10:00 AM 60 <20 <20
12:00 PM 80 <20 <20
Table 2.6 Hosting capacity limits for 11 bus feeder according to likelihood and severity
Time Load percent (%) HC limit1 (kW) HC limit2 (kW)
8:00 AM 35 >60 >60
10:00 AM 60 <30 <30
12:00 PM 80 <20 <30
These two important indices can be used to identify two values of HC values: HC
limit1 and HC limit2. If only likelihood is considered which is the number of
technical violation(s), HC limits can be indicated as shown in Table 2.5. It is prudent
to first start by stating that the values of HC limits change according to the day hours
and loading percentages. In addition, more precise HC limit can be indicated if high
resolution of PV increasing steps is considered (10 kW is considered). In this table,
during the maximum PV production, which is at 12:00 PM, the HC limit1 and limit2
are 20 kW for both. These HC limits can be eased when considering the severity of
the violation as another factor to indicate the risk as shown in Table 2.6. It is
noticeable that some of HC limits are higher in Table 2.6 compared in Table 2.5,
indicating that not counting the severity may lead to overestimate the risk of PV
connections at some point. For instance, the violation of overvoltage could be
spotted in “1.5” buses in the entire feeder at 10:00 AM during 20 kW of PV
connection; see subfigure (d). But this violation does not pose a severe thread
which can be tolerated, subjected to the operators or policy makers, allowing more
PVs to be connected. According to Fig. 2.7, each house should be allowed to export
PV output power no more than 6.66 kW in the sided lines or 5 kW in the mid-line
during the hour centered at 12:00 PM to avoid overcurrent and overvoltage issues or
no more than 30/3 kW for the sided line and 30/4 kW for the mid-line if overcurrent
issue is tolerated but no voltage violation.
2.5.3 11-kV Grid with 91 (11/0.4 kV) Distribution

Transformers
The approach is carried out on a real existing distribution network situated in South
Australia (see Fig. 2.10). The technical details of this network can be found in
[32]. The total number of 11/0.4 kV distribution transformers is 91. The network is
segmented into two zones: zone A and zone B. Zone A is assessed entirely as well as
two specified feeders: feeder 1 and feeder 2. The characteristics of these zones and
feeders are listed in Table 2.7. The cable codes mentioned in this table are 1 ¼ 61/
2.50-Flat (3 300 mm2) AL XLPE, 2 ¼ 6/1/3.75 ACSR, 3 ¼ 6/1/2.75 ACSR, and
4 ¼ 30/7/.102 ACSR. This network is modelled in OpenDSS, an advanced distri-
bution network modeller introduced by an American institute named EPRI
[33]. OpenDSS is co-simulated with MATLAB through Component Object Model
(COM) interface. The PV connections are increased in a firm stepwise and divided
by the maximum load of the transformer. The assessment for each zone or feeder is
done separately.
The use of bar and stem figures to display the risk seems to be directing toward
the HC indication, but not determination. It is because of the severity of each risk
index that cannot be judged due to the overlap with likelihood’s scale. For that
reason, the theoretical diagram shown in Fig. 2.3 is used to show the scale of both
likelihood and severity as appropriate as it can be to match the severity’s requirement
of determining HC. In this diagram, scaled axes of likelihood and severity are
displayed individually, allowing a fairer judgment for consideration. For example,
Fig. 2.10 Two zones and feeders under assessments within a large distribution network located in
South Australia
Table 2.7 Data of the zones and feeders under assessments

Dist. from Number Number Number of Total Cable
Segment substation (km) of buses of lines transformers length (km) type
Zone A 1.4 56 29 19 6.3 1, 2
Feeder 1 4.2 15 7 6 1.1 3, 2
Feeder 2 3.9 8 4 4 1.3 1, 2
Zone B 8.1 30 14 12 5.5 1, 2, 4
30 H. Al-Saadi
Fig. 2.11 HC
determination according to
the likelihood of violation
for two risk indices when
performing 0.5–4.5 PV
increases/load and estimated
during an hour duration
around 12:00 PM
Fig. 2.11 shows the expectations of D.Index 1 and D.Index 2, computed at 12:00 PM
for the entire zone A. For D.Index 1, the reddish shaded area in these figures
(as discussed previously in Sect. 2.4.2.1) is cropped for 1.5 severity, i.e., 150%
over specified rated ampacity, and 1 likelihood, i.e., one line expected to have
ampacity violations. For D.Index 2, the reddish shaded area in these figures
(as discussed previously in Sect. 2.4.2.1) is cropped for 0.1 severity, i.e., 1.1 pu
over nominal bus voltage, and 1 likelihood, i.e., one bus expected to have overvolt-
age violation. According to the jet-colored bar, HC limits can be determined as HC
limit1 to be 1.7 pu and HC limit2 to be 2 pu.
The selection of HC limits is meant to be conservative in Fig. 2.11. The reason for
conservativeness is that these violations are not expected to cause a severe overvolt-
age or severe ampacity exceedance. With these HC limits, the severity is just 5% of
rated ampacity (Fig. 2.11a) and 1.007 pu of over nominal voltage (Fig. 2.11b).
Following these low values of severities, the likelihood can be compromised with
higher values up to the thresholds which are 1.5 (150% over ampacity) and 0.1 (1.1
over nominal voltage). Therefore, HC limits can be further determined according to
severity; see Fig. 2.12. In this figure, the HC limit1 and HC limit2 are determined to
be 2.1 pu and 2.9 pu, respectively. In other words, the ampacity violation occurs first
and then the overvoltage violation in the scale of PV increase connections. That
would lead to a critical area that is shaded with yellow and requires more attention
from the side of policy makers, analysts, and researchers for finding ways for optimal
utilization; see Fig. 2.13.
Fig. 2.12 HC
determination according to
the severity of violation for
two risk indices when
performing 0.5–4.5 PV
increases/load and estimated
during an hour duration
around 12:00 PM
Fig. 2.13 HC HC limit2

determination using
proposed risk-based method HC limit1
Critical
0.5 2.5 4.5
2.1 2.9
We assess other feeders and zones: feeder 1, feeder 2, and zone B. The outcomes
are presented in Table 2.8. These assessments reveal that the hosting capacity is
larger when the size of network or feeder is smaller in number of buses and lines.
However, these need a thorough investigation as well as other important key factors
such as the distance from the substation, the total length of the feeder, the adopted
standards for streamlining indices’ threshold, type of cables, the nominal voltage at
the substation, the percentage of loading, etc. In Table 2.8, the results that are shown
for only two HC limits indicate that the ampacity violation always occurs before the
overvoltage occurrence. The results of both risk indices according to Fig. 2.4 are
shown in Fig. 2.14 for zone B. We used the severity’s threshold to determine the
values of HC limits. It is advised that this particular early violation is to be
investigated further with more number of zones and feeders to find out whether
this is a coincidence or network related.
32 H. Al-Saadi
Table 2.8 HC determination Segment HC limit1 HC limit2

for the zones and feeders
Zone A 2.1 2.9
Feeder 1 4 >13
Feeder 2 3.5 10.5
Zone B 5 8
Fig. 2.14 HC determination for zone B when performing 1–13 PV increases/load and estimated
during an hour duration around 12:00 PM
2.6 Conclusion
This chapter presents a methodology in quantifying the system risk encountered by

distribution networks during high PV penetration. The establishment of a number of
operational risk indices is achieved. Two risk components, likelihood and severity,
are considered in the formulation of these risk indices and estimated using the
relative frequency (occurrence probability) of violations. The determination of HC
is achieved and implications are highlighted. Sparse grid technique is introduced as a
superior means compared with Monte Carlo technique in both computational time
and calculation accuracy.
A realistic 11/0.4 kV distribution network and large distribution network located
in South Australia are employed to show the implementation and the effectiveness of
the approach. The method is location specific and time varying in which HC is
estimated for different daytimes. The mathematical expressions are provided. The
nonintrusive method is utilized to handle the computation of multiple sources of
uncertainties. The risk indices are estimated through expectations while stepwise
increases are performed. Regions for different type of impacts that PVs could impose
on the performance are described and analyzed. Finally, HC limits are identified
according to the described regions.
The discussed findings are significant and can play an important role legislating
new regulations in power distribution sectors. It is especially appealing for
residential networks being transferred to be active and feeding the nearby commer-
cial or industrial buildings or complexes around noon time. In addition, determina-
tion of HC can benefit a wide range of interested parties such technical analysts,
policy makers, distribution network operators, etc.
Future work is recommended for further development of this area. The fact that
the sparse grid technique proves to be superior over Monte Carlo technique is
explained by selecting the finest number of samples through a unique quadrature
rule mentioned earlier. The technique can be incorporated easily with other time-
based automation approaches in the modern power grids. The effect of OLTC on the
value of HC is worthy of investigation. Therefore, it is recommended to further
improve the PDFs of loads, rather than adopting Gaussian ones. Also, the critical
area is shaded with yellow and requires more attention from the side of policy
makers, analysts, and researchers for finding ways for optimal utilization.
References
1. Ismael, S., Aleem, S., Abdelaziz, A., & Zobaa, A. (2019). State-of-the-art of hosting capacity in
modern power systems with distributed generation. Renewable Energy, 130, 1002–1020.
2. Sarkar, M., Meegahapola, L., & Datta, M. (2018). Reactive power management in renewable
rich power grids: A review of grid-codes, renewable generators, support devices, control
strategies and optimization algorithms. IEEE Access, 6, 41458–41489.
3. Stetz, T., Marten, F., & Braun, M. (2013). Improved low voltage grid-integration of photovol-
taic systems in Germany. IEEE Transactions on Sustainable Energy, 4(2), 534–542.
4. Bollen, M. H., & Hassan, F. (2011). Integration of distributed generation in the power system.
Hoboken: Wiley.
5. McNab, S., Crownshaw, T., Lemon, S., Strahan, R., & Pandey, S. (2016). Determining network
congestion. Auckland: Electric Power Engineering Centre.
6. Dubey, A., & Santoso, S. (2017). On estimation and sensitivity analysis of distribution circuit’s
photovoltaic hosting capacity. IEEE Transactions on Power Systems, 324, 2779–2789.
7. Shayani, R. A., & de Oliveira, M. A. G. (2011). Photovoltaic generation penetration limits in
radial distribution systems. IEEE Transactions on Power Systems, 26(3), 1625–1631.
8. Rylander, M., & Smith, J. (2012). Stochastic analysis to determine feeder hosting capacity for
distributed solar PV. Palo Alto: EPRI technical report 1026640.
9. Breker, S., Claudi, A., & Sick, B. (2015). Capacity of low-voltage grids for distributed
generation: Classification by means of stochastic simulations. IEEE Transactions on Power
Systems, 30, 689–700.
10. Rigoni, V., Ochoa, L. F., Chicco, G., Navarro-Espinosa, A., & Gozel, T. (2016). Representative
residential LV feeders: A case study for the North West of England. EEE Transactions on
Power Systems, 31(1), 348–360.
11. Etherden, N., & Bollen, M. H. (2013). Dimensioning of energy storage for increased integration
of wind power. IEEE Transactions on Sustainable Energy, 4(3), 546–553.
12. Ochoa, L. F., Dent, C. J., & Harrison, G. (2010). Distribution network capacity assessment:
Variable DG and active networks. IEEE Transactions on Power Systems, 25(1), 87–95.
13. Etherden, N., Bollen, M., Ackeby, S., & Lennerhag, O. (2015, June). The transparent hosting-
capacity approach–overview, applications and developments. In International Conference and
Exhibition on Electricity Distribution.
14. CENELEC EN 50160. (2010). Voltage characteristics of electricity supplied by public distri-
bution systems.
34 H. Al-Saadi
15. Al-Saadi, H., Zivanovic, R., & Al-Sarawi, S. F. (2017). Probabilistic hosting capacity for active
distribution networks. IEEE Transactions on Industrial Informatics, 13(5), 1–1.
16. Hollands, K. G. T., & Huget, R. G. (1983). A probability density function for the clearness
index, with applications. Solar Energy, 30(3), 195–209.
17. Xiao, Q. (2015). Dimension reduction method for probabilistic power flow calculation. IET
Generation, Transmission & Distribution, 9(6), 540–549.
18. Ni, F., Nguyen, P. H., & Cobben, J. F. (2017). Basis-adaptive sparse polynomial chaos
expansion for probabilistic power flow. IEEE Transactions on Power Systems, 32(1), 694–704.
19. Smolyak, S. A. (1963). Quadrature and interpolation formulas for tensor products of certain
classes of functions. Soviet Mathematics Doklady, 4, 240–243.
20. Heiss, F., & Winschel, V. (2008). Likelihood approximation by numerical integration on sparse
grids. Journal of Econometrics, 144(1), 62–80.
21. British Standards Institution. (2002). Risk management - vocabulary - guidelines for use in
standards: ISO/IEC guide 73(2002). ISBN 580401782.
22. International Organization for Standardization. (2009). Risk management vocabulary: ISO
guide 73(2009).
23. Standards Australia/Standards New Zealand Standards Committee. (2009, November) AS/NZS
ISO 31000:2009: Risk management-principles and guidelines.
24. Kaplan, S. (1997). The words of risk analysis. Risk Analysis, 17(4), 407–417.
25. Nowak, J., Sarkani, S., & Mazzuchi, T. (2015). Risk assessment for a national renewable energy
target part I: Developing the model. IEEE Systems Journal, 9(3), 1045–1056.
26. Aven, T., Baraldi, P., Flage, R., & Zio, E. (2013). Uncertainty in risk assessment: The
representation and treatment of uncertainties by probabilistic and non-probabilistic methods
(pp. 1–186). Hoboken: Wiley.
27. Schwaegerl, C., Bollen, M. H. J., Karoui, K., & Yagmur, A. (2005, June). Voltage control in
distribution systems as a limitation of the hosting capacity for distributed energy resources. In
IET CIRED Conference. Turin.
28. Navarro-Espinosa, A., & Ochoa, L. F. (2016). Probabilistic impact assessment of low carbon
technologies in LV distribution systems. IEEE Transactions on Power Systems, 31(3),
2192–2203.
29. Mulenga, E., Bollen, M., & Etherden, N. (2020). A review of hosting capacity quantification
methods for photovoltaics in low-voltage distribution grids. International Journal of Electrical
Power & Energy Systems, 115, 105445.
30. Goerlandt, F., & Reniers, G. (2016). On the assessment of uncertainty in risk diagrams. Safety
Science, 84, 67–77.
31. Conti, S., & Raiti, S. (2007). Probabilistic load flow using Monte Carlo techniques for
distribution networks with photovoltaic generators. Solar Energy, 81(12), 1473–1481.
32. Al-Saadi, H., Zivanovic, R., & Al-Sarawi, S. (2017). Probabilistic hosting capacity for active
33. Dugan, R. C., & McDermott, T. E. (2011, July). An open source platform for collaborating on
smart grid research. In 2011 PESGM IEEE. Detroit.
Chapter 3
A Methodology Proposal for Calculation
of Hosting Capacity Including Different
Power Quality Phenomena
Tiago E. Castelo de Oliveira and Paulo F. Ribeiro
3.1 Background Harmonic Distortion
One of the key questions in the summarization of background distortion is related

with the type of sources that can be aggregated, which will either cause positive or
negative impacts. Their effect will depend on the nature of these loads and sources.
There is a considerable number of studies in the background distortion field. Mostly,
the knowledge of harmonic background distortion is based on harmonic injection in
the majority of the cases. In this section, an analytical analysis that has been
conducted on background distortion will be put forward, aiming at clarifying
concepts that will later be used to define the concept of hosting capacity. In general
terms, harmonic distortion, of either current or voltage, consists in a way to describe
the principle of background distortion.
Background harmonic distortion (BD or BHD) can be defined as the aggregated
level of harmonic distortion present in the supply voltage at a point of common
coupling (PCC) excluding the connection of a new linear or nonlinear load or
generator to the PCC. This distortion may vary throughout the day in shape and/or
in magnitude as well as phase of individual components. Thus, it may consequently
increase or reduce the resulting distortion after the connection of a new nonlinear
load or generator at the PCC.
T. E. Castelo de Oliveira (*)

Electrical Energy Systems, Technological University of Eindhoven, Eindhoven,
the Netherlands
e-mail: t.e.castelo.de.oliveira@tue.nl
P. F. Ribeiro
Advanced Power Technologies and Innovations in Systems and Smart Grids Group,
Federal University of Itajuba, Itajuba, MG, Brazil
e-mail: pfribeiro@ieee.org

https://doi.org/10.1007/978-3-030-40029-3_3
36 T. E. Castelo de Oliveira and P. F. Ribeiro
3.1.1 Modelling and Methods of Background Distortion

Source
As mentioned before, in order to define the contribution level of nonlinear loads, it is

assumed that the polluting loads can be modelled as an equivalent Norton, which
means they can be modelled as a simple current source as per [1]. Thus, according to
[2], the maiming power system can be assumed to be linear under the analysis.
Moreover, the terms “primary emission” and “secondary emission” are introduced to
explain the source of emission. In this case, the primary emission is the emission
originating from devices connected at the PCC, whereas the secondary emission is
the emission originating from outside the device where a nonfundamental compo-
nent in the background voltage distortion will result in a distorted current. For
example, the distortion from high frequency fluorescent lamps is defined as a
primary emission. On the other hand, the current distortion caused by inverters at
terminals of a wind turbine at the PCC is classified as a secondary emission [2].
In order to represent the voltage distortion caused by the background harmonic
distortion, the equivalent Norton of the load at the PCC can be transformed into an
equivalent Thevenin as well, where a power system voltage for the harmonic order
h is represented by a voltage source Em(h), an internal impedance defined as Zm(h),
and the equivalent current defined as Im(h), according to [3]. The voltage at the PCC
point is given by Vm(h). Moreover, the harmonic impedance of the linear loads
connected at the PCC is represented by Zp(h). Thus, the current source of the polluting
loads connected at the PCC is characterized by Ip(h). Finally, the shunt harmonic
impedance is represented as Zs(h). The model described is shown in Fig. 3.1.
The method given in [4] is accurate considering an equivalent of the circuit
regarding two harmonic equivalent sources: utility voltage Em(h) and costumer
voltage Ec(h), as well as the utility impedance Zm(h) and customer impedance
Zc(h), defined at a harmonic order h. In this case, the Norton equivalent can be
turned into the Thevenin equivalent on the costumer side. The system is shown in
Fig. 3.2.
These contributions can be calculated using Eqs. (3.1) and (3.2).
Z m ðhÞ
E m ðhÞ ¼ E ð hÞ ð3:1Þ
Z m ð hÞ þ Z c ð hÞ c
Zm(h)
Im(h)
Em(h). Vm(h). Zp(h). Ip(h). Zs(h).
Fig. 3.1 Power system representing the background distortion model [1]
3 A Methodology Proposal for Calculation of Hosting Capacity Including Different. . . 37
Fig. 3.2 Power system Zm(h) PCC Zc(h)

representing the background
distortion model as a two
Thevenin equivalent [4]
Em(h) Vm(h) Ec(h)
Fig. 3.3 Voltage phasor

diagram based on IEC [4]
Z c ð hÞ
E c ðhÞ ¼ E ð hÞ ð3:2Þ
Z m ð hÞ þ Z c ð hÞ m
It is important to highlight that for these models, both sides must be analyzed as
harmonic contribution, which means that the contribution of utility and the contri-
bution from the prosumer will be implemented by using a superposition method in
order to determine the background distortion at the PCC [4].
In order to understand the equivalent voltage at the PCC, the IEC voltage phasor
method cited in [4] allows us to find the total voltage harmonic distortion at the PCC
considering the background voltage and the customer installation. In this case, the
voltage harmonic distortion, before any nonlinear load on the costumer’s side is
connected, will be defined as V background. Moreover, the voltage phasor V emission is the
voltage vector that estimates the voltage drop on the costumer’s side. Thus, the
voltage vector at the PCC V PCC , which consists of the voltage background and the
customer’s harmonic voltage emission, is shown in Fig. 3.3.
3.1.2 Example of the Impacts of Background Harmonic

Distortion
The presence of harmonic voltage distortion in electrical networks can lead to

several disturbances, such as increasing losses, heating of components, loss of
equipment life, power factor, and others [5]. As defined before, nonlinear loads are
the main cause for pollution of the waveform in the electrical system and must be
considered.
Moreover, televisions can cause considerable disturbance depending on the time

of the year around the world, for instance, during the World Cup [5–8]. At this time
of the year, television viewers will switch on their equipment at the same time
considering different time zones. The impact of television viewing and other elec-
tronic loads on the utility grid harmonic distortion may be of great relevance as
discussed in [9]. It is commonly considered that in industrial countries, the 5th
harmonic will be significant and coincidently characterized by the peak of television
viewing. Given these points, television receivers have power supplies which create
current harmonics into the system. Thus, based on these phenomena, it is safe to
assume that the hosting capacity for the system will be decreased as a function of the
increase of harmonic background level [10]. Furthermore, an important relation
between the harmonic background level, voltage rise, and hosting capacity will be
defined in this chapter, where the bigger the background is, the bigger the voltage
rise will be, and, as a consequence, the lower the hosting capacity acceptable region
will be, according to [11].
Although many papers have discussed the effects of background voltage distortion
through different devices and throughout the years, few studies have addressed the
problem of background distortion during televised events, particularly the FIFA
Soccer World Cup, when a large number of TVs are connected into the distribution
grids, causing significant levels of background distortion during the matches [6–8, 12].
As mentioned before, harmonic distortions on the electrical grid can become
much more serious on specific occasions, when massive events are occurring, such
as the FIFA World Cup or other worldwide televised event. Moreover, the impact of
this harmonic distortion will depend on the topology of the grid, electrical param-
eters, and even cultural subjects. In Fig. 3.4, an example of a measurement during the
FIFA World Cup 2010 of Brazil matches can be seen [6].
It is easy to notice that during the matches, the harmonic distortion increased
substantially, which will directly impact the background distortion of the system.
Thus, the distortion reduction at half-time is related to a temporary load demand
causing a damping and lower impedance at all frequencies [6].
Fig. 3.4 THD profile at different days of Brazil matches [6]

Fig. 3.5 THD (Total Harmonic Distortion) of voltage at 230 kV during a Brazilian match on the
World Cup 2010 [6]
Still in accordance to [6], Fig. 3.5 in the 5th harmonic voltage at 230 kV in the
Northeast of Brazil during the Brazil versus Portugal match was measured.
According to [6], a typical 5th harmonic voltage on a typical day is under 0.5%.
The same source concludes that the impact of TV viewing must be considered on
normal days, as well as during highly popular events, as a way to analyze the impact
of the background distortion in the system. It is clear from these results that a
significant increase of harmonic pollution can take place, especially in the 5th
harmonic order. Those results show that the harmonic background can be severe
due to nonlinear loads connected at the same time.
In this section, the definition and requirements for the nonlinear contributions and
their background distortion were determined. It is well-known that harmonic distor-
tion at the PCC is caused by the harmonic sources on both sides of the system: the
utility, as well as the customer’s facilities. The theoretical aspects of the background
distortion have been reviewed in order to select the correct model. It is important to
highlight that the background distortion model has an important contribution to the
hosting capacity value regarding distribution systems.
3.2 Analysis of Hosting Capacity Through Different Power

Quality Phenomena
Both concepts and applications of hosting capacity have been studied by researchers
in order to elaborate a new significant strategy to plan and improve network systems,
especially for new generation sources. For distributed energy resources, the hosting
capacity approach has been introduced as a communication tool between stake-
holders concerning the connection of DG to the electric grid [13]. In this context,
hosting capacity is understood as the maximum generation that can be integrated

into the distribution grid without causing excessive disturbances in the power
quality, where the term excessive means beyond the limits imposed by the standards
maximum voltage rise, harmonics, flicker, etc. [1–7, 13]. Still according to the
literature [11, 13], it is important to take into consideration uncertainties when
hosting capacity analysis is being conducted on electrical systems. The uncertainties
can be classified as customers’ type of installations (three-phase or single-phase) and
customers’ connection phase if single-phase or in case of photovoltaic installations,
the panel orientation, the type of the inverter, and implemented reactive power
control if any, among others.
As shown in Fig. 3.6, the curve family below will represent different range of
hosting capacity analytical analysis. Moreover, the hosting capacity can be defined
as the worst hosting capacity level, the ideal hosting capacity level, and, finally, the
best hosting capacity level. Besides, the acceptable deterioration can be defined as
the region between the best hosting capacity range (the line designating the best
hosting capacity and the best background level) and the worst capacity range (the
line designating the worst hosting capacity and the worst background level), which
defines the region where it is possible to keep the system operating, without
uncertainties, while respecting its limit index defined by local standards. It is
important to take into account that the system will usually be operating in its ideal
curve and will also be defined for a certain instant of time. Therefore, the system can
vary from the worst level to the best level depending on the background distortion
level at a given instant of time. In other words, the hosting capacity of the system will
be variable throughout a period of time, and it sometimes is necessary to choose one
specific instant of time to analyze the effects on the system [13].
Fig. 3.6 Range curves of a generic performance index versus the amount of distributed generation
[13]
According to the previous analysis, the impact of a certain amount of DG can be

quantified using a set of indicators that accentuate its performance on an electrical
system, where these indicators can be any power quality indexes or even a mix of
these indexes [10–13]. When the power quality phenomena are set up at a certain
hosting capacity analysis, the hosting capacity investigation is expected to be as
close to what is estimated in a real case study as possible. In other words, power
quality phenomena are not severable, and they must be taken into account as a set of
indicators. In some cases, the system’s performance will improve after the DG
connection, or there will be negative impacts on the system (these performance
classifications depend on the background analysis at a certain instant of time).
In the next section, the concept of integrated local hosting capacity will be applied
through the evaluation of allowed voltage rises due to the harmonic distortion, as
well as through its mathematical development based on [13].
3.3 Integrated Hosting Capacity: Voltage Rise Due

to Harmonic Injection
Nowadays, more and more power electronics are being integrated into our electrical
system, especially at the distribution system level [9–13]. As is well-known, power
electronics are likely to inject harmonic currents and, consequently, cause a certain
level of disturbance through their background distortion. In general, the larger the
device is, the greater the impact will be, either positively or negatively [10]. In other
words, any class or type of power electronic device connected to the network will be
injecting harmonic currents, and consequently, voltage disturbance will occur as
well as voltage rise as a result of the root mean square (RMS) magnitude dependence
on voltage waveform. For example, a power electronic device connected at the PCC
in the system can be modelled as a set of controlled current sources, in parallel,
integrated with its respective impedance for each harmonic order, where each
individual source will be injecting a certain amount of harmonic current, in magni-
tude and phase, and as a consequence, a voltage RMS will appear. It is important to
highlight that the power electronic device will be inserting a harmonic current at the
PCC until order n, according to Fig. 3.7.
Fig. 3.7 Integration of PCC

harmonic current source of
power electronic devices
through controlled current
source R1 R2 Rn-1 Rn
h=1 h=2 h=n-1 h=n
...
GND1 GND1 GND1 GND1
Fig. 3.8 Feeder model Pg+jQg

integrated with a DG
Source Xf
Rf
DG
Vo PL+jQL
Vg
Load
The sum of these injected currents via the described model will impact the RMS
voltage at the PCC. These values are the product of the harmonic current magnitude
and the impedance linked to each controlled harmonic current source. Since the
RMS voltage is the square root of the sum of squares of each harmonic voltage, it can
be expressed using the following equation (3.3):
vffiffiffiffiffiffiffiffiffiffiffiffiffi
u 51
uX
V RMS t V 2h ð3:3Þ
h¼1
To consider harmonic injection by the source and distortion at the PCC, our
model needs to be rewritten based on typical parameters of an LV feeder model [10–
13]. Details of the feeder model used and corresponding voltage rise formulae are
illustrated in Fig. 3.8.
Based on the model shown in Fig. 3.8 for which the Thevenin equivalent
impedance at the PCC is given by Rf + jXf. The complex power injected by a general
DG model system and the complex power absorbed by the local load are defined as
Sg ¼ Pg + jQg and SL ¼ PL + jQL, respectively. The grid injected active and reactive
powers at the PCC are defined as P ¼ Pg PL and Q ¼ Qg QL, respectively
[10]. The fundamental frequency voltage difference between the PCC and the main
supply can be expressed by (3.4).

P þ jQ
V_ G V_ o ¼ ðRf þ jX f Þ ð3:4Þ
V_ G ejθ
The voltage at the PCC is denoted by V_ G :ejθ , and the voltage at the source is V0.
j0
e .
In order to simplify our assumptions, let us assume that cosθ 1 and sinθ 0,
and (3.4) can be approximated by (3.5) for the fundamental frequency h ¼ 1.
PRf þ QX f
ΔV 1 ð3:5Þ
Vg
Then, Eq. (3.5) can be simplified for other harmonic orders h 6¼ 1 as in the
following:
Ph Rf þ Qh hX f
ΔV h ð3:6Þ
V hg
To assume the maximum voltage rise at the PCC, the load consumption (L) has to
be as minimum as possible, and the generation (G) has to be at its maximum value.
Thus, one can use (3.6) to compute an upper required for voltage rise assuming that
the power factor (PF) of net load can be controlled by the DG at the PCC system,
which means tan(φ) ¼ Q1/P1. Based on these assumptions, Eq. (3.6) can be rewritten
through (3.7) and (3.8).
ΔV 1 :V 1g P1 Rf þ Q1 X f ð3:7Þ
ΔV 1 :V 1g P1 ðRf þ tan ðφÞX f Þ ð3:8Þ
In these assumptions, the maximum power injected by a DG can now be

calculated using (3.9) assuming that the maximum of ΔV1 is V gmax ,1 V 10 , where
V gmax ,1 is the voltage limit for the fundamental frequency defined by local or
international standards. Moreover, assuming that loads have negligible harmonic
content:

V gmax ,h V gmax ,h V ho
Pgmax ,h ¼ ð3:9Þ
R 1 þ tan ðφÞ hX Rf
f
Going back to the assumption presented in (3.3), the active power generated in
relation to the RMS voltage can be rewritten as (3.10), using (3.3):
qffiffiffiffiffiffiffiffiffiffi
P 2ffi2
Vh P
V 2 V 2h
Pg ¼ RMS ¼ ¼ ð3:10Þ
Rf Rf Rf
In (3.9), the maximum power that can be injected by a DG was calculated, thus
defining the hosting capacity at the PCC as the sum of the maximum power possible,
based on (3.10), to be injected for each frequency as per (3.11) [10]:

X
h X51 V max ,h V max ,h V h
1 g g o
Pmax
g ¼ max ,h
PPV ¼ ð3:11Þ
h¼1
R h¼1 1 þ tan ðφÞ hX f
Rf
where:
V gmax ,h is the allowed voltage magnitude for harmonic order h
V ho is the actual voltage magnitude of harmonic order h at the PCC
φ is the arcos(PF), where the PF is the power factor at the PCC
If we agree to use the IEC and notice that, for V 1 kV, standards impose that
individual voltage distortion be below 2%, for instance, for h 2 and below 110%
for the fundamental frequency, h ¼ 1, then, the hosting capacity is given in (3.12):

1:10 1:10 V 1o Xh
0:02 0:02 V ho
Pmax
g ¼ þ ð3:12Þ
Rf 1 þ tan ðφÞ XRff h¼2 Rf 1 þ tan ðφÞ R
hX f
f
3.3.1 Sensitivity Analysis of an Integrated Hosting Capacity

Approach
Several factors may affect the performance of calculating the hosting capacity in an
electrical system, such as the harmonic voltage, power factor, minimum load of the
distribution feeder, and voltage regulation equipment, among others [10]. We will
discuss the harmonic distortion impact on the presented Eq. (3.12).
Using (3.12), it is possible to find the value of the hosting capacity considering the
voltage harmonics from the fundamental order up to the 50th harmonic order, as well
as performance index limits, which depend on local standards [10], for instance,
IEEE 519 or IEEE 1547 or even IEC. Moreover, the injection of harmonic current
into the system depends on the power electronic equipment connected at the PCC. It
is important to highlight that the power electronics can be controlled by a prosumer,
for instance, due to the fact that the equipment is manufactured according to
international standards. According to [10], the background voltage at the PCC
cannot be controlled as it is almost impossible to know what is causing the harmonic
voltage distortion. Therefore, one extreme scenario may be imposed to analyze how
the harmonic injection can impact the hosting capacity.
Consider a scenario where the voltage for the fundamental order is set at 1.00 p.u.
and the other harmonic voltages for h 2, up to 15, are set at 0.02 p.u. For this
scenario, a high distortion is considered as it is dependent on the background and all
the electronic devices connected at the PCC. The other scenarios consider a non-high
voltage distortion for h 2 up to 15, and the voltage distortion will set in at 0.01 p.u.
In reality, a decrease in harmonic voltage would increase the hosting capacity as
shown in the equation below:
P
Pgmax 0 1:1ð1:1 1Þ þ ð0:02 0:01Þ0:02 0:113
¼ P ¼ ¼ 1:027
Pmax
g 1:1ð1:1 1Þ þ ð0:02 0:02Þ0:02 0:11
It is possible to observe the decrease in harmonic voltage increases the hosting

capacity by almost 2.7%. It is important to highlight that the improvement of the
voltage distortion signals at the PCC might be dependent on electronic elements such
as harmonic filters, for instance. On the other hand, regarding voltage regulation
equipment, keeping the harmonic background constant, and considering a reduction

of V 1o to 0.98 p.u., for instance, could result in 1.5 times better hosting capacity [10],
which can be controlled by FACTS elements. More details concerning the integrated
hosting capacity approach and its sensitivity analysis can be found in [10], which
provides other examples in relation to a real case scenario considering the installa-
tion of PVs (Photovoltaic Generators).
References
1. Ghijselen, J. A., Ryckaert, W. R., & Melkebeek, J. A. (2002). Required load behaviour for
power quality improvement. In IEEE Power Engineering Society summer meeting (Vol. 3, pp.
998–1003).
2. Ronnberg, S. K., Members Zavoda, F., Convenor CA Corresponding Members, et al. (2018,
March). Power quality and EMC issues with future electricity networks.
3. Ryckaert, W. R. A., Ghijselen, J. A. L., & Melkebeek, J. A. A. (2003). Harmonic mitigation
potential of shunt harmonic impedances and the influence of background distortion. In IEEE
Bologna Power Tech conference proceedings (Vol. 2, pp. 663–670).
4. Papic, I., et al. (2018). A benchmark test system to evaluate methods of harmonic contribution
determination. IEEE Transactions on Power Delivery, 8977(1), 1–8.
5. Browne, N., Perera, S., & Ribeiro, P. F. (2007). Harmonie levels and television events. In 2007
IEEE Power Engineering Society general meeting, PES (pp. 7–12).
6. Ribeiro, P. F., et al. (2011). Harmonic distortion during the 2010 FIFA World Cup. In 2011
IEEE Power and Energy Society general meeting (pp. 1–8).
7. Macedo, J. R., Martins, A. G., Carneiro, J. R. V., & Siqueira, M. J. V. (2006). The impact of
FIFA World Cup 2006 on power quality in the electric distribution systems. In 2007 9th
international conference on Electrical Power Quality and Utilisation.
8. Leitão, J. J. A. L., Fonseca, L. C. A., Lira, M. M. S., Soares, L. R., & Ribeiro, P. F. (2007).
Harmonic distortion on a transmission system during games of the Brazilian National Team in
the 2006 world cup. In 2007 IEEE Power Engineering Society general meeting, PES (pp. 1–2).
9. Testa, A., & Langella, R. (2007). Harmonic pollution in Italian distribution networks in
coincidence with important sport events. In 2007 IEEE Power Engineering Society general
meeting, PES (pp. 1–7).
10. de Oliveira, T. E. C., Carvalho, P. M. S., Ribeiro, P. F., & Bonatto, B. D. (2018). PV hosting
capacity dependence on harmonic voltage distortion in low-voltage grids: Model validation
with experimental data. Energies, 11(2), 1–13.
11. F. E. Engineering. 2015, February. PQ-SmartGrid laboratory (pp. 1–16).
13. Dubey, A., & Santoso, S. (2017). On estimation and sensitivity analysis of distribution circuit’s
photovoltaic hosting capacity. IEEE Transactions on Power Apparatus and Systems, 32(4),
2779–2789.
Chapter 4
Effect of Harmonics due to Distributed
Energy Resources on Hosting Capacity
of Microgrid: A Hardware in Loop-Based
Assessment
Sourav Kumar Sahu, Debomita Ghosh, and Dusmanta Kumar Mohanta
Abbreviations
CSP Concentrated solar power
DERs Distributed energy resources
DNI Direct normal irradiance (W/m2)
DNIc Clear-sky direct normal irradiance (W/m2)
EHC Enhanced hosting capacity
FPGA Field-programmable gate array
HC Hosting capacity
HIL Hardware in loop
HPR Harmonics penetration ratio
PCC Point of common coupling
SCADA Supervisory control and data acquisition
SPV Solar photovoltaic
THD Total harmonics distortion
WECS Wind energy conversion system
Nomenclature
Ai Area of the wind turbine blades
A Global irradiance on 2-axis tracking plane (W/m2)
Ac Global clear-sky irradiance on 2-axis tracking plane (W/m2)
Ad Diffuse irradiance on 2-axis tracking plane (W/m2)
S. K. Sahu · D. Ghosh · D. K. Mohanta (*)

Department of Electrical & Electronics Engineering, Birla Institute of Technology,
Mesra, Ranchi, India
e-mail: dkmohanta@bitmesra.ac.in

https://doi.org/10.1007/978-3-030-40029-3_4
48 S. K. Sahu et al.
An Real N junction characteristics

Cq Short circuit current temperature coefficient
CI Distribution function
Egap Energy band gap of the material used in the cell
G Global irradiance on a fixed plane (W/m2)
Gc Global clear-sky irradiance on a fixed plane (W/m2)
Gd Diffuse irradiance on a fixed plane (W/m2)
GIS Global information system
IC,SC Cell short circuit current at reference radiation and temperature
IG Global irradiation
IExt Extraterrestrial solar radiation
IPVC Photovoltaic current
IN Nth harmonics component of current
IPV PV array output current
IRES,I Reverse saturation current at TI
k Constant
K Boltzmann’s constant
k1,k2 Constants used for shape and scale
m Mass of the wind
NSe Number of series modules connected
NS Number of modules connected in parallel
PRated Rated power of the wind turbine
ρISN Density of air
q Charge of an electron
R Solar irradiation (mW/cm2)
RSer Series resistance
RShunt Shunt resistor
T Cell temperature
TI Reference cell temperature
vinf Velocity of wind obstruction at infinite distance
VCutout Cut-out wind speed
Vo PV array output voltage
Vrated Rated wind speed
Vcut-in Cut-in wind speed
4.1 Introduction
The transition of conventional power system toward smart grid has led to the
penetration of nonconventional energy sources in the form of distributed energy
resources (DERs) [1]. This is due to changes in type of energy production, that is,
shift from large-scale generation unit to small units integrated in distribution system;
type of customers, that is, the movement toward sustainable energy system is driving
a change to efficient energy system; decrement of fossil fuel; and environmental
4 Effect of Harmonics due to Distributed Energy Resources on Hosting Capacity. . . 49
aspects [2]. However, increased integration of DERs may hinder the grid perfor-
mance, leading to violation of voltage, thermal limits, added injection of harmonics,
and mal-operation of protection [3]. Thus, integration of these DERs should be done
as per the network hosting capacity (HC). The HC is defined as the quantity of added
generation or utilization that can be integrated to the grid without endangering the
operating limits or network reliability [4].
To maintain the operational parameters within the predefined standards and to
avoid mal-operation of the equipment, node voltage violation, ampacity of the
conductors, and uncertainty should be taken as constraints while calculating the
HC [4]. Parameter such as node voltage play a significant role in the hosting capacity
assessment due to the fact that voltage violation limit is reached far earlier than the
current limit. Upon violating the voltage limits, the distribution side may face the
uneven voltage profile which may cause local circulating currents. These circulating
currents may contribute to ohmic losses and may lead to false tripping of protecting
equipment. Ampacity should be the subsequent deciding factor, as extreme deviation
in the ampacity may result in thermal breakdown of the lines, equipment, and
conductor insulation [5]. As most of the DERs constitute solar and wind, a high
uncertainty is associated with both these sources. Any impulsive change in these
sources may result in sustainable disturbance in the grid. So to utilize these energy
resources to their optimal level, power electronic converters are used. Semiconduc-
tor switches such as IGBTs and MOSFETs which are used in these converters are
dependent on the switching frequency requirement and the uncertainty in the input
sources. The switching of these converters introduces harmonics in the system
[6, 7]. In addition to these, different nonlinear loads such as arc furnaces, saturated
magnetic core of various machines, and induction motors also contribute to the
harmonics of the utility [8].
The presence of harmonics in the distribution system causes various problems
such as degradation in voltage profile, heating of the wires and machines, and low
injection of DERs. These issues in the utility collectively hampers the HC of the
network. It also creates a problem in optimizing the size of DERs [9–12]. Although
these energy sources have negligible harmful environmental effect, they create
harmonics to a considerable extent [13, 14]. With the advancement of the maximum
power point tracking, for solar photovoltaic as well as wind power plants, the system
can get maximum power irrespective of varying sources [15, 16]. Due to the
arrangement of maximum power, the firing in the converter changes dynamically,
which creates additional distortion in the grid current and in voltage. The renewable
sources are location and time dependent. If the plant is in a geographical location
where the uncertainties of the resources are frequent, then the utility will be exposed
to more such disturbances. Due to these disturbances, the operational difficulty such
as variation in HC is a must, which is undesirable [17]. To address these operational
problems, accurate analysis of these resources is necessary to achieve a reliable and
effective operation of the utility without any voltage or ampacity limit violation.
In this chapter, effect of harmonics on HC of microgrid is tested for different
DERs. The DERs, particularly solar and wind, are uncertain in nature. Based on the
uncertainty of the different DERs, total harmonic distortion (THD) with voltage and
ampacity limits as the major constraint is measured. The effect of THD on the HC of
the network is monitored, which is a significant aspect for planning the size and
number of DERs in the microgrid. The entire analysis is done in MATLAB and
validated in Typhoon HIL (a real-time simulator). Typhoon HIL simulation is a
platform that includes the complexities of the plant and tests them in real time with a
controlled environment. All power system dynamics can be added in the system, and
by simulating them, the actual microgrid response can be predicted. This technique
is a realistic technique for testing, system improvement, and predicting its dynamic
behavior [13, 14]. The analysis is tested on 5-bus and IEEE 33-bus distribution
network with different DERs at a specific geographical location of India obtained
through global information system (GIS), to efficate the results. The results are also
verified at variable load conditions. Further on analyzing all the constraints, the
possible HC is proposed.
The arrangement of this chapter is as follows. The concept of HC is highlighted in
Sect. 4.2. Section 4.3 details the microgrid and its components. Section 4.4 illustrates
the effect of DER on hosting capacity. Section 4.5 details the tools used. Section 4.6
depicts the case study and results. Section 4.7 concludes the findings and highlights
the possible future work.
4.2 Concept of Hosting Capacity (HC)
In the twenty-first century, energy consumption per capita is considered to be the one
of the most important indicators to know how much an individual is coping up with
the current time. In order to align with the demand of the situation, the energy
consumption pattern, quantity, and quality are drastically changing. As the demand
is increasing in never before rate, the generation should also match the increasing
demand. Due to continuous evolution in the field of science and technology,
renewable sources like solar photovoltaic, small hydro, and wind energy conversion
system are no more exorbitant. In this context, solar photovoltaic and wind energy
conversion system are most commonly used. Use of these energy sources is contin-
uously changing the dynamics of the power system, from unidirectional power flow
to bidirectional power flow. Thus, to regulate and to guide this integration process,
hosting capacity term is coined in power system. Hosting capacity can be defined as
the production or consumption of new energy considering the indicator/indicators to
lie within its allowable limits. Some of the indicators for hosting capacity are
voltage, current, power, and frequency. As frequency and power remain same as
compared to voltage and current, the indicators, that is, voltage and current are
widely used in the distribution system HC analysis. This hosting capacity can be
divided into two types:
(i) Dynamic hosting capacity
(ii) Static hosting capacity
Utility
Thevenin
Impedance
Various Distributed Point of Common

Energy Sources Coupling
Gen
Loads
Fig. 4.1 DER integration for HC analysis
In dynamic hosting capacity, the parameters that contribute to the calculations

change with the variation in the load and system conditions. Sudden change in
network configuration due to the closing of the circuit breaker will also be included
under dynamic hosting capacity investigation.
In static hosting capacity, the steady-state network parameters such as voltage and
ampacity are used, and the value remains fixed when loads are fixed over a period of
time. Fewer complexities are involved in this calculation as the dynamic parameters
are not considered.
The hosting capacity for a simple circuit as shown in Fig. 4.1 can be calculated by
applying Kirchhoff’s voltage law (KVL) considering the network impedance at the
bus under consideration.
4.3 Microgrid and Its Components
Microgrids are the modern local grids comprising of renewable and non-renewable
DERs that can operate in two different modes, namely, grid-connected mode and
standalone mode. Figure 4.2 depicts a typical microgrid structure with different
types of DERs and load integration. Some of the DERs are discussed in the
following subsection.
Utility
Gen
Fig. 4.2 Typical microgrid
4.3.1 Wind Energy
The wind energy conversion system (WECS) consists of various parts, such as rotor,
generator, gear transmission, tower, boost converter, inverter, and wind direction
sensor [18]. These are some of the most essential parts which are prerequisites for
further analysis. In addition to it, some of the operational parameters such as cut-in
wind speed, cut-out wind speed, cut brake wind speed, and Betz limit are also very
essential for the smooth operation of the WECS. Operational parameters such as
cut-in wind speed and cut-out wind speed play an important role in wind power
generation. To convert the wind current into useful electric power, there is a limit;
that is, all the energy content in the air particles cannot be converted into useful
power, which is detailed by Betz limit. The cut-in speed and cut-out speed is
normally 12.6 kmph and 90 kmph, respectively [19]. Figure 4.3 shows the cut-in
and cut-out speed graphically. The velocity of wind in between these values is
responsible for energy generation.
The wind turbine generates power by rotating the shaft with the help of the flow
of the wind through its blades. The wind which passes through the wind turbine
slows down as it gives up its energy. The kinetic energy that can be converted by the
wind turbine rotor is 59.3%, which is obtained when the residual wind, that is, the
outgoing wind after hitting the turbine blades from the wind turbine, is one-third of
the free wind. This is known as the Betz limit [20]. If the blades have ideal efficiency,
then the residual wind will result in zero value, that is, no wind condition leading to
the stop of operation of wind farms. The power content in the wind (PWind) is
depicted in Eq. (4.1).
1
PWind ¼ ρ Ai v3inf ð4:1Þ
2
Fig. 4.3 Typical wind

turbine power
characteristics
Power(kW)
Rated Power
Cut-in Rated Cut-out

Speed Speed Speed
Wind Speed
The wind power plant produces energy when the wind speed is in between cut-in
speed and cut-out speed. Equation (4.2) depicts the amount of power generated (PG, WT)
by the wind turbine.
8 9
>
<0 80 < v < vcut‐in >
=
PG,WT ¼ Jvk þ LPRated, 8vCut‐in < v < vrated ð4:2Þ
>
: >
;
PRated 8vCut‐in < v < vCutout
where
!
PRated
J¼ 3
V rated V 3 Cut‐in
!
PRated
L¼ 3
V Cut‐out V 3 Cut‐in
4.3.2 Solar Energy
Solar energy is one of the widely harvested renewable sources of energy due to easy
application and installation. Solar energy can be used in two methods, namely,
concentrated solar power (CSP) and solar photovoltaic (SPV) power. In CSP, solar
thermal energy is used to make steam and run the turbine, but in SPV, plants photo
panels are used for conversion of solar energy to electric energy. Because of its
compactness, easy availability in the market, and low maintenance cost, SPV is
widely used in the modern power system. The block diagram of a typical solar
photovoltaic plant is shown in Fig. 4.4.
Fig. 4.4 Block diagram of a

Solar PV
solar photovoltaic system
Array
Converter/ Energy Storage

Inverter
Arrangement Device
Utility Grid
IPV
RSe
RS RRSC V
IPhoto
IRSC
Fig. 4.5 Simplified equivalent model for photovoltaic cell
In the SPV plants, photovoltaic cells are the fundamental components of the
photovoltaic array. These are semiconductor devices which convert solar energy to
electric energy. This phenomenon is called the photovoltaic effect. The electric
power generated at a particular plant is a function of the geographical location,
irradiation, and optimal tilt angle. Each of these values has significant impact on
power generation. For a particular power rating of the solar photovoltaic plant, PV
arrays are connected in series to match the voltage rating and in parallel to provide
required load current. Photovoltaic cell can be modelled by a current source named
photovoltaic current (IPVC). The equivalent circuit diagram is as in Fig. 4.5, where
RSer is the series resistance and RShunt is the shunt resistor. Generally, the value of
these resistances are very less, so they are neglected to have a less complex analysis
of the circuit [21].
By solving the circuit shown in Fig. 4.5 with the above stated assumptions, the
PV array output current can be written as in Eq. (4.3) below. This current is the
resultant of NS number of parallel modules.

qV o
I PV ¼ N S I SN N S I RES exp 1 ð4:3Þ
kTAn N Se
The reverse saturation current (IRCS) in the cell is dependent on the temperature,
which is as in Eq. (4.4).
3
T Egap 1 1
I RSC ¼ I RES,I exp q ð4:4Þ
TI kAn TI T
The photocurrent (IPhoto) is a function of solar irradiance and temperature at a

particular geographical location. This is given by Eq. (4.5):
R
I Photo ¼ Cq ðT T I Þ þ I C,SC ð4:5Þ
100
Thus, PV array power can be calculated by Eq. (4.6):
P ¼ V I PV

qV o
P ¼ N S I SN V N S I RES V exp 1 ð4:6Þ
kTAn N Se
Storage of the surplus power during the off-peak hour is one of the major aspects
of renewable integration. In this context, battery storage is very vital. Presently, the
invention of high-density Li-ion batteries is cost-effective, and these batteries have
very high energy density and can store an appreciable amount of energy. They are
used for avoiding overloading and peak saving. These batteries can be connected to
the microgrid to improve HC as well [22, 23]. Supercapacitors are also an alternative
to energy storage mainly due to their high energy density and quick charging-
discharging capacity [22].
4.3.3 Biofuel-Based Power Plant
Any fuel that is derived from the biomass is called biofuel. It is considered to be
renewable energy unlike fossil fuel as the source of biofuel is readily available like
animal waste and crop waste. Other sources such as animal fat, rapeseed oil, and
soybean oil can be used as a biofuel for generators. The net heating value can vary
between 42 MJ/kg for liquid biofuel and 39.7 MJ/kg for rapeseed oil. An example
for such system is Wartsila 9L20, which gives 1539 kW of electrical power output at
11 kV [24].
4.4 Effect of DER on Power Flow
Rapid advancement in the distributed energy technology is leading to the integration

of more and more renewable with the utility [25]. But if these energy resources are
integrated into the system without any planning/power calculation, then the network
performance will deteriorate. So it is most important to estimate the capacity that can
be hosted by or connected to the network. Above the hosting capacity limits, the
utility should not be allowed to integrate with the grid to avoid under/over-voltage
problem, thermal limit of the conductor, puncture of the insulation, and increased
harmonic distortion.
Various factors such as randomness in the sources, harmonics, ampacity of the
conductor, and voltage limit can affect the HC of the network to a greater extent. The
randomness in solar irradiation value is contributed by two parameters, deterministic
and stochastic variations. Instochastic parameter variation, various environmental
parameters like air quality and the shading of cloud. In deterministic parameter
variation, the sun’s position in the sky is considered. The combination of these two
indices is called clearness index (CI) which changes with time, that is, throughout the
day and seasons of the year, as shown in Eq. (4.7). CI is given as
IG
CI ¼ ð4:7Þ
I Ext
On the other hand, the randomness for wind energy conversion system can be
determined by considering the positional and time-dependent wind flow. As the
wind pattern is totally random, various distribution functions can be implemented by
considering the best fitting of the curve. From [26], the Weibull distribution function
is taken for representing the wind speed variation as in Eq. (4.8).
k 1 k1
k 1 v 1 kv2
F Wind ðvÞjk1 ,k2 ¼ e ð4:8Þ
k2 k2
The nonlinearity due to the loads and converters is a prominent source of

harmonics which affects HC considerably. Harmonics have various frequencies
that are integral multiple of the fundamental. Due to the various frequency compo-
nents, the current/voltage waveform deviates from the pure sinusoidal nature. The
estimation of THD can be done by comparing voltage/current waveforms against the
fundamental; the THD can be estimated as in Eq. (4.9):
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
I 22 þ I 32 þ þ I N 2
THD ¼ ð4:9Þ
I1
The vast use of the nonlinear types of equipment and renewable energy sources
distorts the sinusoidal current and voltage signals, which forms different types of
harmonics in the power system networks. These distorted voltage and current signal
causes various problems such as heating of equipment, false tripping of the protec-
tion equipment, and power quality issues. The decrement in the power quality and
heat formation should be kept minimal by restricting the harmonics to a safer limit as
per the standard, but the decrement in the power quality may lead to the restriction in
addition of renewables to the utility. Other than harmonics, voltage profile and
ampacity of the conductors are the key factors in the hosting capacity evaluation.
Dueto the addition of DERs, the voltage profile at point of common coupling (PCC)
and the ampacity will change. The recommended allowable deviation for different
voltage level is detailed in IEEE 519 [27, 28]. This change in the voltage or change
in the load current will cause various deformations in the network parameter profile
which may lead to decrease in the hosting capacity [29]. Keeping all the above
discussed factors in mind, the optimal sizing of the DERs can be decided for the
improvement of the HC. The optimal size of the DERs can be decided by taking
application and location into account. To optimize the sources, various methods,
parameters, and constraints can be considered [30, 31]. Various optimizing methods
such as particle swarm optimization (PSO), ant colony optimization (ACO), and
genetic algorithm (GA) can be implemented by considering the different application
of the network [32]. Mainly, the objective function and the constraints are required to
decide which optimization is applicable to the problem, and accordingly, size of
DER and types of DER along with its location can be decided. Network parameters
such as X/R ratio play an important role in optimization as the convergence of the
solution is greatly affected by this ratio [33]. Network reinforcement is one of the
methods to enhance the HC in the distribution system. The injection of the DER is
limited by the network constraints like voltage rise, thermal limit, network ampacity,
and the protection of the equipment. Along with the technical constraints, utility
rules may also restrict the DER investors to add more renewable to the utility grid.
4.5 Tools Used to Analyze the Effect of Harmonics on HC

by DERs
4.5.1 MATLAB
MATLAB is a high-performance numeric computing platform which is extensively

used in all aspects of engineering and mathematics. The MATLAB-based real-time
emulators are very expensive, so in this chapter, a Typhoon-based HIL is used.
4.5.2 Typhoon Hardware in Loop (HIL)
Typhoon HIL is a scientific research software and hardware package for the real-
time testing of the power system. With its advanced features, it can simulate the
accurate condition with industry standard. Typhoon HIL simulators can be config-
ured in various modes where the user can define time according to the dynamics of
the plant. If the plant dynamics is very fast, Typhoon HIL ultra-high-fidelity
simulators can simulate the plant as low as a 1 μs time step. If the plant requires
high time steps for a specific application, then the time step can even be scaled to
10 μs. For a very large plant with multiple controllers and machines, the emulator
can simulate the plant with different time steps. These time steps are independent of
each other. Typhoon HIL real-time emulators also allow the user to optimize the
simulation so that the cores of the processor can be optimally used. Typhoon HIL is
an adaptive HIL manufacturer, which is contributing to the real-time emulator by
consistently improving the designing, testing, and verification environment. The
fast-growing microgrid library for the HIL application is revolutionary as it will help
the designers to design the microgrid with minimal approximation. Minimizing the
approximation will result in obtaining better real-time response of the system. The
FPGA-based HIL platform ensures quality performance of the emulator, and this
emulator can be used in various fields, such as:
(i) Power system and its application
(ii) Microgrid and its applications
(iii) Various FACTS devices and their application to the power network
(iv) Rapid control prototyping
To perform Typhoon HIL simulation, various series of emulators are present such
as 402, 600, 602, 602+, 603, and 604. The selection of emulator depends on various
parameters, such as number of machines, number of converters, and number of pulse
width modulation (PWM) channel requirements. Typhoon HIL emulator gives
precise control over power flow, voltage, and current at different magnitude and
phase. The various realistic grid phenomena such as sag and swell in voltage, spikes,
Fig. 4.6 Front I/O ports of HIL402

Fig. 4.7 Rear USB port and Ethernet port
Fig. 4.8 Typhoon HIL402 connected to the system
and rate of change of frequency can be simulated easily. Typhoon HIL can be easily
connected to the system by USB cable. It can also be connected by an ethernet port.
The interface of Typhoon HIL402 emulator is shown in Figs. 4.6, 4.7, and 4.8.
Typhoon HIL emulator works on FPGA platform with core count starting from
four in numbers. In case of higher number of core requirement, the researcher may
opt for a higher-end model, which can operate in standalone mode as well as parallel.
Each switching element in the library is assigned a weight, and each core can run in a
predefined value of weight. While simulating a model in HIL device, all the weights
and core counts are considered. The following steps are elaborated with figures for
understanding the operation through Typhoon HIL. The following steps may be
executed for connecting the HIL with the plant.
Fig. 4.9 Selection of emulator
Fig. 4.10 Selection of configuration
4.5.2.1 Step 1
Consider the core count and weight that will be used in the model. Choose a suitable
emulator, followed by a suitable configuration from the device table constituting of a
number of machine solvers as in Figs. 4.9 and 4.10, respectively.
Fig. 4.11 Compilation and model execution
Fig. 4.12 Selection of device (V-HIL or actual HIL) for model execution
4.5.2.2 Step 2
Create the model which is to be tested in HIL schematic editor. If multiple weights
are present in the model, then use core separator from Typhoon HIL library to assign
different weights to different physical cores of the emulator according to the
configuration table as shown in Fig. 4.10 and discussed in Step 1.
4.5.2.3 Step 3
Compile the model and run the model by clicking the button as shown in Fig. 4.11. If
there will be an error, it will appear at the bottom section, and required action needs
to be taken to fix it. After successful compilation, a command titled “compilation
finished successfully” will appear on monitor.
4.5.2.4 Step 4
Load the model into the real FPGA-based Typhoon HIL device connected to the
system by clicking on it as in Fig. 4.12.
Fig. 4.13 Representation of SCADA panel
4.5.2.5 Step 5
Design the supervisory control and data acquisition (SCADA) interface according to
the output requirements, that is, voltage waveform, current waveform, frequency
response, power measurement, THD calculation, etc. Data recorder to record data for
further analysis in various formats such as “.csv and .mat” is also available. For
changing the input during the execution time, there are action widgets such as knob
and slider in the SCADA panel as shown in Fig. 4.13.
4.5.2.6 Step 6
As shown in Fig. 4.13, on the right side, there are model settings. In model setting,
input all the data in the SCADA input section.
4.5.2.7 Step 7
Make sure the required contactors in the model settings are in CLOSE mode.
4.6 Case Studies and Results
For analysis purposes, two test systems are considered: 5-bus distribution system
and IEEE 33-bus distribution system with variable solar irradiation and variable
wind speed, and biofuel-based power are used. The networks are also simulated with
variable load conditions. For testing, two networks, MATLAB and Typhoon
HIL402, are used.
Table 4.1 Solar irradiation data
Location: Arunachal Pradesh February
Inclination of plane: 43 (28 , 97 )
Time 06:52 07:07 07:22 07:37 07:52 08:07 08:22 08:37 08:52 09:07 09:22 09:37
Irradiation 0 0 0 0 0 0 414 429 442 452 462 469
Time 09:52 10:07 10:22 10:37 10:52 11:07 11:22 11:37 11:52 12:07 12:22 12:37
Irradiation 476 481 486 490 493 495 497 498 499 499 498 497
Time 12:52 13:07 13:22 13:37 13:52 14:07 14:22 14:37 14:52 15:07 15:22 15:37
Irradiation 495 493 490 486 481 476 469 462 452 442 429 414
Time 15:52 16:07 16:22 16:37 16:52 17:07 17:22
Irradiation 396 374 347 313 268 0 0
4 Effect of Harmonics due to Distributed Energy Resources on Hosting Capacity. . .
63
Table 4.2 Wind speed data

Wind speed 6 (m/s) 12 (m/s) 18 (m/s)
Table 4.3 Line and load data for 5-bus distribution system
Line data Load data Hosting capacity
Sl. From To R X P Q 70% load Full load
no. bus bus (Ohm) (Ohm) (kW) (kVA) (kVA) (kVA)
1 1 2 0.092 0.0477 0 0 605.0039 479.144
2 2 3 0.493 0.2511 100 60
3 3 4 0.366 0.1864 90 40
4 4 5 0.3811 0.1941 120 80
The solar irradiation input is taken from a specific geographical location [34]
(Table 4.1).
The irradiation from different geographical location can be extracted from geo-
graphical information system (GIS) data. These data are used as an input for
simulation to predict the exact behavior of the DER in different irradiation
conditions.
To analyze the realistic scenario of THD due to wind power plant, a realistic wind
data is taken from the east coast of India and depicted in Table 4.2.
Based on these inputs, both networks are simulated to analyze the effect of these
DERs on HC.
4.6.1 5-Bus Distribution System

Assumptions:
1. All the sources are connected one by one to study the different real-time
conditions.
2. The grid harmonics are neglected to study the harmonic effect of the DERs.
3. The test system is uncompensated.
4. The sources are placed in all nodes of the network.
The line and load data for the 5-bus distribution system with calculated HC is given
in Table 4.3, and the corresponding figure is as in Fig. 4.14.
In the 5-bus system, considering voltage limit violation and ampacity as the
measuring constraint, the HC is calculated by using MATLAB as shown in
Fig. 4.15. Different loads are considered and power injections are done at each
node according to the HC of the nodes. From Fig. 4.15, it can be observed that the
voltage is well within the limits according to IEEE standard.
Pv Plant
Gen Bio-Fuel
Plant
Gen
Wind Power
Plant
Bus 5
Load
Gen Gen
Gen
Bus 1 Bus 2 Bus 3 Bus 4

Load Load Load
Fig. 4.14 5-bus distribution system with all possible DG connections at all buses
Fig. 4.15 MATLAB-based 5-bus voltage indicator while injecting the calculated HC
Fig. 4.16 Solar irradiation 268 W/m2 for 5-bus system
Fig. 4.17 Short time harmonics measurement at 268 W/m2



The results from MATLAB are then imported to the real-time environment where
the THD is analyzed as shown in Figs. 4.16, 4.17, 4.18, 4.19, 4.20, 4.21, 4.22, and
4.23. Here in the plots, THD is plotted against time in seconds for different
irradiation conditions.
Table 4.4 THD values at different solar irradiation and load

Very short time harmonics
measurements(IEEE 519)
Irradiation Bus Number of Voltage Ampacity
(W/m2) Load no. deviation Value deviation deviation
268 Full 2 No 13.78 Within limits Within limits
load
70% 2 1 18.49
load
374 Full 2 1 20.22
load
70% 2 1 16.18
load
414 Full 2 No 12.88
load
70% 2 No 13.30
load
499 Full 2 No 12.03
load
70% 2 No 14.44
load
Fig. 4.24 THD at different buses at a wind speed of 6 m/s for 5- bus system
4.6.1.1 Case I: Solar Photovoltaic at all Buses
In this 5-bus system, solar photovoltaic sources are connected to all buses according
to the HC of the network. At this injection condition, THD is maximum at bus
2, which can be observed from Figs. 4.16, 4.18, 4.20, and 4.22. As bus 2 is
connected to maximum capacity line, its THD is slightly higher, but under the
Fig. 4.25 Short time harmonics measurement at 6 m/s
standard in most of the time, which can be inferred from Figs. 4.17, 4.19, 4.21, and
4.23.
The results are also tabulated in Table 4.4 at different loading conditions. It can be
inferred that the THD violation is very minimal as per the standard.
Table 4.5 THD at different buses due to variable wind speed

Very short time harmonics measurements
(IEEE 519)
Wind Bus Number of Voltage Ampacity
speed Load no. deviation Value deviation deviation
6 (m/s) Full 5 1 32.9 Within limits Within limits
load
70% 2 4 30.7, 32.3,
load 26.8, 15.7
5 1 15.6
12 (m/s) Full 2 3 17.4, 19.3, 19.7
load 5 1 43.4
70% 2 1 24.1
load 3 1 29.1
18 (m/s) Full 2 1 19.8
load 5 1 16.1
70% 2 1 36.5
load 5 2 21.4, 16.4
4.6.1.2 Case II: Wind Power Plant at all Buses
Similar at all the nodes, wind power plant is connected in the considered 5-bus
network. As the uncertainty in the wind is very high, it can be observed from
Figs. 4.24, 4.25, 4.26, 4.27, 4.28, and 4.29 that the variation in the value of THD
for wind power plant is maximum among all other considered DERs.
Fig. 4.30 THD at different buses for 5-bus system
Table 4.6 THD at different buses due to biofuel-powered generator

Generating 5-bus system maximum Voltage Ampacity
capacity Load THD deviation deviation
As per HC limit Full 0.0015 Within limits Within limits
load
70% 8.5081e-04
load
The change in wind speed causes change in the THD quantity which can be
verified from Figs. 4.24, 4.25, 4.26, 4.27, 4.28, and 4.29 as well as from Table 4.5.
As wind profile is very dynamic in nature, thus, to have a stable output from wind
power plant, multiple converters are used. Due to this extensive use of converters,
the current and voltage waveforms get distorted and create THD in the utility.
Table 4.5 depicts the change in value of THD at various buses due to change in
wind profile and also shows the maximum number of violation of IEEE standard.
4.6.1.3 Case III: Biofuel-Based Power Plant at all Buses
The biofuel-powered generators have the lowest THD contribution due to the
absence of the switching devices.
Biofuel-powered generators are the best substitute for the diesel generator due to
the low environmental impact. In these generators, normally, no power electronic
converters are used, so the distortion in the current and voltage waveform is minimal.
This can be seen from Fig. 4.30, where current and voltage THDs are shown in the Y-
axis. The values in the Y-axis are very small which resembles very low THD. As the
Table 4.7 Line and load data for IEEE 33-bus distribution system [35]
Line data Load data
Sl. no. From bus To bus R (Ohm) X (Ohm) P (kW) Q (kVA)
1 1 2 0.092 0.0477 0 0
2 2 3 0.493 0.2511 100 60
3 3 4 0.366 0.1864 90 40
4 4 5 0.3811 0.1941 120 80
5 5 6 0.819 0.7070 60 30
6 6 7 0.1872 0.6188 60 20
7 7 8 1.7114 1.2351 200 100
8 8 9 1.030 0.7400 200 100
9 9 10 1.04 0.7400 60 20
10 10 11 0.1966 0.0650 60 20
11 11 12 0.3744 0.1238 45 30
12 12 13 1.468 1.1550 60 35
13 13 14 0.5416 0.7129 60 35
14 14 15 0.591 0.5260 120 80
15 15 16 0.7463 0.5450 60 10
16 16 17 1.289 1.7210 60 20
17 17 18 0.732 0.5740 60 20
18 18 19 0.164 0.1565 90 40
19 19 20 1.5042 1.3554 90 40
20 20 21 0.4095 0.4784 90 40
21 21 22 0.7089 0.9373 90 40
22 22 23 0.4512 0.3083 90 40
23 23 24 0.898 0.7091 90 50
24 24 25 0.896 0.7011 420 200
25 25 26 0.203 0.1034 420 200
26 26 27 0.2842 0.1447 60 25
27 27 28 1.059 0.9337 60 25
28 28 29 0.8042 0.7006 60 20
29 29 30 0.5075 0.2585 120 70
30 30 31 0.9744 0.9630 200 600
31 31 32 0.3105 0.3619 150 70
32 32 33 0.341 0.5302 210 100
Hosting capacity Full load 4.8 MVA
70% load 3.5098 MVA
THD content is very low, maximum biofuel-powered generators can be connected at

each node. It is also observed that due to the absence of uncertainty, the THD content
is very low, which is reflected in Table 4.6.
Gen
26 27 28 29 30 31 32 33
23 24 25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
S/S
19 20 21 22
Fig. 4.31 IEEE 33-bus distribution system
Fig. 4.32 MATLAB-based 33-bus voltage indicator while injecting the calculated HC
4.6.2 IEEE 33-Bus Distribution System
Assumptions:
1. All the sources are connected one by one to study the different real-time
conditions.
2. The grid harmonics are neglected to study the harmonic effect of the DERs.
3. The sources are placed only in one location, bus no. 2 since this bus is connected
to the maximum capacity line.
4. The test system is compensated.
The line data, load data, and base voltage with calculated HC are tabulated in
Table 4.7, and the corresponding single line diagram is shown in Fig. 4.31.
Similarly, to analyze the effect of THD, in case of -IEEE 33 bus distribution
system, firstly, keeping voltage limit and ampacity as constraints, the analysis is
done for HC. This can be verified from Fig. 4.32. It can also be verified that with the
increase in load, the voltage decreases.
All the variation of the voltages and its effect on THD for 33-bus distribution
system can be validated through the HIL results as in Figs. 4.33, 4.34, 4.35, 4.36,
4.37, 4.38, 4.39, and 4.40.

4.6.2.1 Case I: Solar Photovoltaic at Bus 2 in IEEE 33-Bus Distribution

System
In this considered IEEE 33-bus distribution system, current THD at each node is
measured. This THD is mainly due to the SPV injection at bus 2 only. Here, bus 2 is
considered for the power injection as this bus is connected to the maximum power
line. The changing THD values can be observed in Figs. 4.33, 4.34, 4.35, 4.36. The
change in THD value in Fig. 4.33 for current at bus 2 can be observed; however, due
Table 4.8 THD values for different solar irradiation values

Voltage deviation
Irradiation 33-bus system Value beyond Ampacity
(W/m2) Load maximum THD Bus no. limit deviation
268 Full 0.4251 2, 3, 4, 5 +5%, +6%, +2%, Within limit
load +1%
70% 0.4250 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
374 Full 0.4245 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
70% 0.4250 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
414 Full 0.4245 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
70% 0.4250 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
499 Full 0.4245 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
70% 0.4249 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
to scaling of the axes, bus 2 current plot appears to be smooth. Minute variation in
the THD values can be observed in all other buses.
Results in Figs. 4.33, 4.35, 4.35, and 4.36, is obtained for different irradation
conditions. In the foresaid Figs. 4.33, 4.35, 4.35, and 4.36, “X” axis shows the time
in seconds, and “Y” axis shows the THD magnitude. Table 4.8 summarizes all the
values of THD which are showcased in Figs. 4.33, 4.34, 4.35, and 4.36. From
Table 4.8, it can be observed that very small change in THD is taking place due to
the change in irradiation at different time of the day.
4.6.2.2 Case II: Wind Power Plant at Bus 2 in IEEE 33-Bus Distribution
System
Variation of wind speed throughout the day is a normal phenomenon that is faced by
the wind power plants. The effects of change in wind speed on THD values are
shown in Figs. 4.37, 4.38, and 4.39.
From the summary that is presented in Table 4.9, the values of THD are constant
at the lower wind speed. When the wind speed is higher, the value of THD changes
with variation in the load.
Fig. 4.37 THD at different buses at a wind speed of 6 m/s for 33 bus
Table 4.9 THD at different buses due to variable wind speed

Voltage deviation
Wind 33-bus system Value beyond
speed Load maximum THD Bus no. limit Ampacity deviation
6 (m/s) Full 0.4237 2, 3, 4, 5 +5%, +6%, Within limit
load +2%, +1%
70% 0.4237 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
12 (m/s) Full 0.4253 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
70% 0.4237 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
18 (m/s) Full 0.4347 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
70% 0.4237 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
Fig. 4.40 THD at different buses for 33-bus system with bio-fuel powered generator
Table 4.10 THD at different buses due to biofuel-powered generator

Voltage deviated
Generating 33-bus system Value beyond Ampacity
capacity Load maximum THD Bus no. limit deviated
As per HC Full 0.4540 2, 3, 4, 5 +5%, +6%, Within limit
limit load +2%, +1%
70% 0.4556 2, 3, 4, 5 +5%, +6%,
load +2%, +1%
4.6.2.3 Case III: Biofuel-Based Power Plant at Bus 2 in IEEE 33-Bus

Distribution System
The THD content due to the biofuel-powered generator is as shown in Fig. 4.40,
where “X” axis is the time axis and “Y” axis is the THD axis.
The summarized values in Table 4.10 show that the maximum value of THD is
0.4556. The diesel generator does not have variation in its power source, that is, fuel.
In Fig. 4.40, the THD values are stable, and at bus 2 which is connected to maximum
capacity line, the THD is slightly increasing at the initial stage due to starting of the
generator.
Although the THD values for the biofuel-powered generator are comparable to
other injections (SPV, WECS), it is of stable nature.
Thus, the overall HC as per THD analysis can be summarized as in Eq. (4.10):
(
HCCAL > HCBF PG > HCSPV > HCWECS , without compensation
¼ ð4:10Þ
HCCAL > HCBF PG > HCSPV =HCWECS , with compensation
Where,
HCCAL ¼ Calculated hosting capcity.
HCBF_PG ¼ Hosting capacity due to bio-fuel powered generator.
HCSPV ¼ Hosting capacity due to solar powered generator.
HCWECS ¼ Hosting capacity due to wind power generation system.
8
>
< HCMax , When all the sources are bio‐fuel powered genertors
HCTHD ¼ HCMin , When all the sources are wind powered generators
>
:
HCMax > HC > HCMin , When combination of sources are connected
ð4:11Þ
Based on the conclusion drawn in Eqs. (4.10) and (4.11), the HC of the network
can be planned to serve quality power without hampering the THD limits as well as
over-voltage and ampacity of the conductor so as to inject power optimally.
4.7 Conclusion and Future Work
The sprawl of power system network has led to the injection of DERs in the form of
solar, wind, biofuel, etc. The planning of the size and quantity of these DERs in the
network is essential to obtain reliable and quality power. The DERs, especially solar
and wind, are totally nature dependent and are highly uncertain. Thus, the network
HC due to these DERs needs careful assessment. Due to the uncertain nature of these
sources, the THD, distorting the fundamental signal and violating the voltage and
ampacity limits, plays a significant role in HC assessment. The effect of THD on HC
at different irradiation conditions for SPV and different wind speed is analyzed both
in MATLAB and in real-time emulator using Typhoon HIL. The analysis is also
compared with biofuel-powered generators to monitor the varying effect of THD
on HC.
As the energy consumption per capita is exponentially growing, meeting the
energy demand is the biggest challenge. To optimally handle the emerging power
crunch problem, each bus should be injected with the maximum possible power, that
is, enhanced hosting capacity (EHC), which will not only solve the power crunch but
also encourage the individual to install DERs.
References
1. Harvey, R., Xu, Y., Qu, Z., & Namerikawa, T. (2017). Dissipativity-based design of local and
wide-area DER controls for large-scale power systems with high penetration of renewables. In
2017 IEEE conference on control technology and applications (CCTA), pp. 2180–2187.
2. Tsikalakis, A. G., & Hatziargyriou, N. D. (2007). Environmental benefits of distributed
generation with and without emissions trading. Energy Policy, 35(6), 3395–3409.
3. Munkhchuluun, E., & Meegahapola, L. (2017). Impact of the solar photovoltaic
(PV) generation on long-term voltage stability of a power network. In 2017 IEEE innovative
smart grid technologies – Asia (ISGT-Asia), pp. 1–6.
5. Zhou, Y., & Li, H. (2014). Analysis and suppression of leakage current in cascaded-multilevel-
inverter-based PV systems. IEEE Transactions on Power Electronics, 29(10), 5265–5277.
6. Stratford, R. P. (1980). Rectifier harmonics in power systems. IEEE Transactions on Industry
Applications, IA-16(2), 271–276.
7. Wang, J., Guo, Q., Yang, Z., Wang, Y., Bai, Z., & Ma, H. (2015). Modeling and research of
three-phase inverter powering nonlinear load., IEEE 24th International Symposium on Indus-
trial Electronics (ISIE), Buzios, 2015 pp, 331–336.
8. Sebaay, A. E. L., Ramadan, M., & Adma, M. A. A. (2017). Studying the effect of non-linear
loads harmonics on electric generator power rating selection. European Scientific Journal, ESJ,
13(18), 548.
works. In EEEIC 2016 – International conference on environmental and electrical engineering,
pp. 2–7.
10. Cundeva, S., Bollen, M., & Schwanz, D. (2017). Hosting capacity of the grid for wind
generators set by voltage magnitude and distortion levels. In Mediterranean conference on
power generation, transmission, distribution and energy conversion (MedPower 2016),
pp. 73–79.
11. Gupta, S. C., Kumar, Y., & Agnihotri, G. (2008). Optimal sizing of solar-wind hybrid system.
In ICTES, pp. 282–287.
12. Alturki, M., Khodaei, A., Paaso, A., & Bahramirad, S. (2018). Optimization-based distribution
grid hosting capacity calculations. Applied Energy, 219(March), 350–360.
13. Patsalides, M., et al. (2017). The effect of solar irradiance on the power quality behaviour of grid
connected photovoltaic systems. Renewable Energy and Power Quality Journal, 1(05),
323–330.
14. Al-Saadi, H., Zivanovic, R., & Al-Sarawi, S. F. (2017). Probabilistic hosting capacity for active
15. Selvakumar, S., Madhusmita, M., Koodalsamy, C., Simon, S. P., & Sood, Y. R. (2019). High-
speed maximum power point tracking module for PV systems. IEEE Transactions on Industrial
Electronics, 66(2), 1119–1129.
16. Fathabadi, H. (2017). Novel maximum electrical and mechanical power tracking controllers for
wind energy conversion systems. IEEE Journal of Emerging and Selected Topics in Power
Electronics, 5(4), 1739–1745.
17. Sun, W., Harrison, G. P., & Djokic, S. Z. (2012). Distribution network capacity assessment:
Incorporating harmonic distortion limits. In IEEE power and energy society general meeting,
pp. 1–7.
18. Cantero, A. S. (2018). Modeling and characterization of a wind turbine emulator. In Advances
in renewable energies and power technologies (pp. 491–508). Elsevier https://www.elsevier.
com/books/advances-in-renewable-energies-and-power-technologies/yahyaoui/978-0-12-
812959-3.
19. Neill, S. P., Hashemi, M. R., Neill, S. P., & Hashemi, M. R. (2018). Offshore wind. In
Fundamentals of ocean renewable energy (pp. 83–106). Academic Press https://www.
elsevier.com/books/fundamentals-of-ocean-renewable-energy/neill/978-0-12-810448-4.
20. Huleihil, M., & Mazor, G. (2012). Wind turbine power: The Betz limit and beyond. In Advances
in wind power. IntechOpen.
21. Singh, G. K. (2013). Solar power generation by PV (photovoltaic) technology: A review.
Energy, 53, 1–13.
22. Poulios, V., Vrettos, E., Kienzle, F., Kaffe, E., Luternauer, H., & Andersson, G. (2014).
Optimal placement and sizing of battery storage to increase the PV hosting capacity of low
voltage grids. Master thesis. In International ETG Congress 2015; Die Energiewende – Blue-
prints for the new energy age, pp. 85–92.
23. Etherden, N., & Bollen, M. (2014). The use of battery storage for increasing the hosting
capacity of the grid for renewable electricity production. In Conference Innovations Security.
24. Kapusta, Ł. J., Sundell, J. P., & Teodorczyk, A. (2011). Liquid biofuels-promising energy
source for a small scale power plants. Silniki Spalinowe/Combustion Engines, 3(146), 1–6.
25. Shayani, R. A., Member, S., Aurélio, M., De Oliveira, G., & Member, S. (2011). Photovoltaic
Generation Penetration Limits in Radial Distribution Systems," in IEEE Transactions on Power
Systems, 3(26), 1625–1631.
26. Sharma, S. K., Chandra, A., Saad, M., Lefebvre, S., Asber, D., & Lenoir, L. (2017). Voltage
flicker mitigation employing smart loads with high penetration of renewable energy in distri-
bution systems. IEEE Transactions on Sustainable Energy, 8(1), 414–424.
27. Carnovale, D. J. (2014). Application of IEEE Std 519-2014 harmonic limits. In IEEE IAS
Atlanta Section, pp. 1–19.
28. C. Studies. (2014). IEEE_STD_519_1992vs2014.
29. Alalamat, F. (2015). Increasing the hosting capacity of radial distribution grids in Jordan.
30. Akram, U., Khalid, M., & Shafiq, S. (2018). Optimal sizing of a wind/solar/battery hybrid grid-
connected microgrid system. IET Renewable Power Generation, 12(1), 72–80.
31. Xu, L., Ruan, X., Mao, C., Zhang, B., & Luo, Y. (2013). An improved optimal sizing method
for wind-solar-battery hybrid power system. IEEE Transactions on Sustainable Energy, 4(3),
774–785.
32. Laksmi Kumari, R. V. S., Nagesh Kumar, G. V., Siva Nagaraju, S., & Babita Jain, M. (2018,
January). Optimal sizing of distributed generation using particle swarm optimization. In 2017
international conference on intelligent computing, instrumentation and control technologies
ICICICT 2017, pp. 499–505.
33. Alizadeh, S. M., Ozansoy, C., & Alpcan, T. (2016). The impact of X/R ratio on voltage stability
in a distribution network penetrated by wind farms. In Proceedings 2016 Australasian Univer-
sities Power Engineering Conference AUPEC 2016, pp. 1–6.
34. JRC photovoltaic geographical information system (PVGIS) – European Commission.
[Online]. Available: https://re.jrc.ec.europa.eu/pvg_tools/en/tools.html. Accessed 23 September
2019.
35. Rajaram, R., Sathish Kumar, K., & Rajasekar, N. (2015). Power system reconfiguration in a
radial distribution network for reducing losses and to improve voltage profile using modified
plant growth simulation algorithm with distributed generation (DG). Energy Reports, 1,
116–122.
Chapter 5
A Scenario-Based Approach for Storage
Capacity Determination to Improve
the Hosting Capacity of Distribution
Systems
Mohammad Seydali Seyf Abad, Jin Ma, and Jing Qiu
5.1 Introduction
The penetration of distributed energy resources (DERs) in distribution systems has

been increasing in recent years. However, the capacity of distribution systems to host
DERs is bounded by different technical constraints. Different methods such as
network augmentation [1], reactive power control [2, 3], and active power curtail-
ment [3–5] have been proposed to increase the hosting capacity (HC) by addressing
the technical challenges incurred due to high penetration of DERs. However, these
methods have some technical and/or economic drawbacks. For instance, reactive
power control of DERs can effectively increase the HC only when the X/R ratio is
high. But X/R ratio is usually low in distribution systems. Further, although network
augmentation methods can effectively increase the HC, they are usually costly for
the utility [1]. Similarly, active power curtailment methods can effectively increase
the HC. However, power curtailment-based methods are usually not popular among
the DER owners as the curtailed energy is not totally compensated.
Using battery energy storage systems (BESSs) to address the issues associated
with DERs is another popular method. The application of BESSs includes
preventing over-voltage [6], smoothing the output of renewable resources [7],
peak shaving [8], and energy arbitrage [9]. Generally, studies that focused on the
application of BESSs in resolving the technical issues associated with high penetra-
tion of DERs can be divided into two categories: (i) those that resolved the technical
issues by controlling the BESSs [10, 11] and (ii) and those that resolved the technical
issues by optimally allocating the BESSs [12–14]. In the first category, the size and
M. S. S. Abad (*) · J. Ma · J. Qiu

University of Sydney, Sydney, NSW, Australia
e-mail: mohammad.seidaliseifabad@sydney.edu.au; j.ma@sydney.edu.au; jeremy.
qiu@sydney.edu.au

https://doi.org/10.1007/978-3-030-40029-3_5
88 M. S. S. Abad et al.
location of DERs and BESSs are given, and the unknown variables are the active and
reactive power of the BESSs. However, if BESSs are to be utilized for increasing the
HC of a system, their sizes and locations are unknown. Therefore, the methods in the
first category cannot be used directly for increasing the HC of a system. Neverthe-
less, the second category, which is based on the optimal allocation of BESSs could
have a better performance as the optimal size, location, and control action can be
identified to increase the HC.
Studies related to the allocation of BESSs have mainly focused on the following
aspects: authors in [15] allocated distributed energy storage systems to increase the
efficiency of distribution and transmission systems by using bilevel optimization. A
robust model to minimize the procurement cost by optimally allocating energy
storage systems and renewable energy sources was presented in [16]. Authors in
[17] allocated energy storage systems to maximize the benefit of the system operator.
The study exploited an accelerated Benders decomposition to solve the problem.
Energy storage systems were optimally allocated in [18] for maintaining the power
balance in active distribution networks considering the efficiency and life cycle of
the storage units. In [12], a methodology for allocating BESSs in a distribution
network is developed to minimize the wind energy curtailment as well as the annual
supply cost of the electricity. Authors in [19] defined a multi-objective optimization
to allocate the BESS by identifying a trade-off between the technical and economic
goals by minimizing the voltage deviation, network losses, feeder congestions, and
costs of supplying loads. In [13], a method for optimal placement, sizing, and control
of a community BESS to increase the HC of rooftop PVs in low-voltage
(LV) networks is proposed. Authors in [20] developed a stochastic optimization-
based method for BESS allocation and operation to ensure a desired voltage stability
margin in wind-intensive distribution systems. In [21], a convex optimization
approach has been used for BESS planning and operation in systems with high PV
penetration level. In that study, the minimum BESS required for mitigating the
technical impacts of high uptake of PVs has been estimated by using OpenDSS
simulation tool. Then, an optimization model has been used to identify the location
of BESSs. Authors in [22] proposed a heuristic strategy based on voltage sensitivity
in combination with an optimal power flow (OPF) framework to identify the optimal
number, size, and location of BESSs that prevent over-voltage and under-voltage in
an LV distribution network. To properly model the uncertainties associated with the
output of DERs and loads, authors in [6] proposed a scenario-based method to
identify the minimum required BESSs that prevents over-voltage in distribution
system with high PV penetration level. The proposed method in that study is
based on the voltage sensitivity of the scenarios associated with the customers’
injected power.
Nevertheless, to the best of the authors’ knowledge, the common drawback of the
aforementioned methods that utilize BESSs for HC improvement is that they did not
consider the uncertainty associated with the location and size of DERs. Hence, they
have only identified the required BESS capacity for a given scenario. Therefore, in
this chapter, a comprehensive method has been proposed to identify the minimum
required BESSs for increasing the HC of a system considering the uncertainties
5 A Scenario-Based Approach for Storage Capacity Determination to Improve. . . 89
associated with the DERs and loads. Then, an economic model has been developed
to assess the minimum required BESS from an economic perspective. Finally, the
performance of the developed models is assessed on an agricultural feeder in
Australia.
5.2 Energy Storage System Classification and Application
As shown in Fig. 5.1, energy storage technologies can be divided into five catego-
ries, as follows:
1. Electrochemical: These energy storage devices have been becoming more popu-
lar due to the decreasing trend of their costs as well as their high performances.
The main idea in these storages is storing electrical energy in the form of chemical
energy in batteries. The most popular electrochemical storages include lithium-
ion, lead-acid, nickel-cadmium, sodium-sulfur (NaS), and vanadium redox
(VRB) batteries.
2. Mechanical: These storages can store energy in different mechanical forms.
Mechanical storages include flywheel, pumped hydro, and compressed air energy
storages. Flywheels are electromechanical systems that can store mechanical
energy in the angular momentum of a rotating mass. Pumped hydro storages
are based on storing and releasing the potential energy of water. The charging
process consists of converting the electrical energy to potential energy by
pumping water from a lower to an upper reservoir. The discharging process
Energy storage
technologies
Electrochemical Thermal Mechanical Chemical Electrical
Lithium-ion Thermo- Compressed Super-

Hydrogen
battery chemical air capacitor
Sensible Pumped
NaS battery Methane
heat hydro
Lead-acid
Latent heat Flywheels
battery
Redox flow
battery
Nickel-
cadmium
Fig. 5.1 Classification of energy storage technologies

consists of rotating a turbine by releasing water from the upper to lower reservoir
to generate electricity. Compressed air energy storage’s working principle is
accumulating energy by compressing and storing air in sealed tanks or under-
ground vessels. However, during the discharge process, the compressed air is
heated and released through a turbine to generate electricity.
3. Chemical: The basic idea in these storages is storing hydrogen as a compressed
gas or in the liquid form and synthesizing it to methane and finally generating
electricity by using fuel cells, combustion engines, or gas turbines.
4. Thermal: The working principle of these storages is storing energy in the form of
heat in different types of materials and generating electricity using steam turbines.
Generally, there are three types of thermal storages, namely, sensible heat, latent
heat, and thermo-mechanical storages. Please refer to [23] for more details.
5. Electrical: These storages can store electricity in the form of an electric field.
Electrical storages include supercapacitors, which have high efficiency and long
lifetime.
The electrochemical storages have been significantly progressing in recent years
[24]. The leading countries on electrochemical storages in terms of accumulated
rating power and number of installations are presented in Fig. 5.2, which is based on
all storage statuses, including operational, under construction, under repair,
announced, off-line, contracted, and decommissioned. Figure 5.3 shows the accu-
mulated rating power and the number of installed electrochemical BESSs that are
operational in 2015 [24]. As shown, the USA with a total capacity of 354 MW
1200 400
1000
350 NUMBER OF STORAGE PROJECTS
TOTAL CAPACITY (MW)
300
800
250
600 200
150
400
100
200
50
0 0
Estimated capacity (MW) Number of storage projects
Fig. 5.2 Estimated battery storage capacity including operational, under repair, off-line, under
construction, contracted, and decommissioned by country in 2015
400 250
NUMBER OF STORAGE PROJECTS

350
200
TOTAL CAPACITY (MW)
300
250 150
200
150 100
100
50
50
0 0
Estimated capacity (MW) Number of storage projects
Fig. 5.3 Estimated operational battery storage by country in 2015
Fig. 5.4 Percentage of 612 projects 2444 MWh 1350 MW

100
different electrochemical
10.8
battery project in the world
80 11.8 30.6
43
12.6
5.7
60
Share in (%)
12.3
11.1
40 6
62.6
48.7
20 39.2
0
Projects MWh Capacity MW Capacity
Lithium ion Lead acid Flow battery Sodium salt Nickel based Others
(192 installations) is in the first place, Japan with a total capacity of 97 MW

(35 installations) is in the second place, and China with 48 MW (53 installations)
is in the third place.
Analysis of the existing electrochemical storages shows that lithium-ion storages
make up for more than 50% of the total projects. Figure 5.4 shows the share of
different electrochemical battery technologies. This figure is based on analyzing
612 battery projects across 42 countries with a total installed capacity of 1350 MW/
2444 MWh. As can be seen, 99.1% of all the considered projects belongs to lithium-
ion, lead-acid, flow battery, sodium-salt, and nickel-based categories.
5.3 Hosting Capacity Estimation
In this section, a framework that estimates the HC of a radial distribution system is

presented, which is based on a comprehensive probabilistic approach, proposed in
[25]. This modular framework is based on an optimization model that aims to
maximize the injected power to the system at each time step. Figure 5.5 summarizes
the main components of the framework. The first module is a scenario generator,
which is designed to address the uncertainties associated with the type and location
of DERs. The second module comprises a time series analysis to address the
uncertainties associated with loads and the output power of DERs using an HC
optimization model. The third module is the probabilistic analysis of the results
obtained from the second module.
The core of the framework is the HC optimization model, as follows:
X
maximize
|fflfflfflfflfflffl{zfflfflfflfflfflffl} pgi,t , ð5:1Þ
N gi,t i2DG
subjected to
X
Pij,t ¼ Pjk,t þ pdj,t pgj,t 8j 2 DG, ð5:2Þ
k:j!k
X
Pij,t ¼ Pjk,t þ pdj,t 8j 2 fN ∖DG g, ð5:3Þ
k:j!k
X
Qij,t ¼ Qjk,t þ qdj,t qgj,t 8j 2 DG, ð5:4Þ
k:j!k
Module 1 Module 2 Module 3
Scenario Time series Probabilistic

Generator analysis based on analysis
the optimization
model
Fig. 5.5 General framework for HC estimation

X
Qij,t ¼ Qjk,t þ qdj,t 8j 2 fN ∖DG g, ð5:5Þ
k:j!k
pgj,t ¼ ηgj,t Capgj N gj,t 8j 2 DG, ð5:6Þ

qgj,t ¼ pgj,t tan ϕgj 8j 2 DG, ð5:7Þ

υ j,t ¼ υi,t 2 r ij Pij,t þ xij Qij,t 8ði, jÞ 2 B, ð5:8Þ
υi,t υ 8i 2 N , ð5:9Þ
N gj,t 1 8j 2 DG, ð5:10Þ
where Pij, t and Qij, t are the active and reactive power flow from bus (i) to bus ( j) at
the time step (t); pdi,t and qdi,t are the active and reactive load at bus (i) and time step
(t); pgi,t and qgi,t are the active and reactive generation at bus (i) and time step (t); ϕgi is
the power factor angle at bus (i) and time step (t); υi, t ¼ |Vi, t|2, where |Vi, t| is the
voltage magnitude at bus (i) and time step (t); Capgi and ηgi,t are the capacity and
capacity factor of the DER at the time (t) and bus (i); N gi,t represents the number of
DER at the time (t) and bus (i); N and DG are the set of all buses and the set of DG
locations, respectively; B is the set of all lines and υ represents the maximum
allowable voltage; rij + jxij is the impedance from bus (i) to ( j); (i, j) and j ! k
represents the line from bus (i) to bus ( j).
According to the third module, the HC of a distribution system could be divided
into three regions, as shown in Fig. 5.6. Region (A) shows the DER penetration
levels that do not cause any violation in technical constraints, regardless of DER
locations. Region (B) represents the DER penetration levels that do not cause any
technical violation when DERs are located at certain locations. It was demonstrated
Fig. 5.6 Possible regions as Cumulative probability curve

well as the distribution curve Probability distribution curve
of HC
A B C
Increasing DER penetration

in [25] that the probability distribution curve of the HC in the region (B) could be
approximated by a Gaussian-shape distribution. The start point of the probability
distribution curve is the border between regions (A) and (B). This start point is
defined as the minimum HC of the system. Region (C) represents all DER pene-
tration levels that would cause a violation of technical constraints, regardless of the
location of DERs. The border between regions (B) and (C) is defined as the
maximum HC. BESS is one of the options that can be used to increase both the
minimum and the maximum HC. Section 5.4 presents a methodology to identify the
minimum required BESS to increase the minimum and maximum HC up to a
certain level.
5.4 Storage Sizing Framework
In this section, a comprehensive method and an approximated method for identifying

the minimum BESS capacity required to increase the minimum HC are proposed.
The proposed methods are based on the estimation of the required active power
curtailment.
5.4.1 Comprehensive Method
As shown in Fig. 5.6, the HC depends on the location of DERs in the region (B).
Therefore, any HC in the region (B) would cause a technical constraint violation for
some DER location scenarios. However, it is possible to resolve the constraint
violations by curtailing DERs. The optimization model for calculating the minimum
active power curtailment required to avoid technical constraint violation is as
follows:
X
maximize
|fflfflfflfflfflffl{zfflfflfflfflfflffl} pgi,t APCgi,t , ð5:11Þ
APCgi,t i2DG
subjected to
X
Pij,t ¼ Pjk,t þ pdj,t pgj,t þ APCgj,t 8j 2 DG, ð5:12Þ
k:j!k
X
Pij,t ¼ Pjk,t þ pdj,t 8j 2 fN ∖DG g, ð5:13Þ
k:j!k
X
Qij,t ¼ Qjk,t þ qdj,t qgj,t 8j 2 DG, ð5:14Þ
k:j!k
X
Qij,t ¼ Qjk,t þ qdj,t 8j 2 fN ∖DG g, ð5:15Þ
k:j!k

υ j,t ¼ υi,t 2 r ij Pij,t þ xij Qij,t 8ði, jÞ 2 B, ð5:16Þ
υi,t υ 8i 2 N , ð5:17Þ
APCgj,t pgj,t 8j 2 DG, ð5:18Þ
pgj,t ¼ ηgj,t Capgj 8j 2 DG, ð5:19Þ

qgj,t ¼ pgj,t APCgj,t tan ϕgj 8j 2 DG: ð5:20Þ
Solving the optimization models (5.11), (5.12), (5.13), (5.14), (5.15), (5.16), (5.17),
(5.18), (5.19), and (5.20) results in the active power curtailment required at each time
step. To identify the minimum required BESS for a scenario, the models (5.11),
(5.12), (5.13), (5.14), (5.15), (5.16), (5.17), (5.18), (5.19), and (5.20) should be
solved over the studied period. The maximum curtailed power during an hour over
the studied period is the minimum rating power required for the BESS for that
scenario. Similarly, the maximum curtailed energy during a day over the studied
period is the minimum energy rating of the required BESS. The rating power and
energy capacity of the required BESS should be identified for different scenarios that
would result in the same HC. Figure 5.7 represents the framework for identifying the
minimum BESS required to increase the minimum HC to a certain level (HCref) in
the region (B). The main idea of the framework is generating expansion scenarios
that would result in the total DER capacity of HCref. Then, the required rating power
and energy capacity of BESS should be identified. Finally, the required BESS is
estimated based on the designed BESS for the generated scenarios. In the following,
the steps of the framework are explained in detail.
• Step A: The focus of this step is addressing the uncertainty associated with the
size of DERs. Depending on the DER technology, there is a distribution curve for
the size of the units based on the historical statistics. For instance, Fig. 5.8
represents the distribution curve of PV systems in California. Generate a scenario
by sampling from the distribution curve of DERs. The sum of the samples should
be equal to the minimum HC level (HCref).
• Step B: This step is designed to address the uncertainty associated with the
location of DERs. To do so, the location of generated DER sample in Step B is
selected randomly, using a uniform distribution from the pool of potential
locations.
• Step C: After carrying out steps A and B, the optimization models (5.11), (5.12),
(5.13), (5.14), (5.15), (5.16), (5.17), (5.18), (5.19), and (5.20) should be used to
identify the required active power curtailment over the study period. Then, based
on the obtained curtailed power, determine the required BESS for the scenario. If
the curtailed power over the studied period is zero, the scenario is not a valid
scenario as it does not cause any technical constraint violation over the study
period. Thus, the following indicator for each scenario is defined to identify the
validity of each scenario.
• Step A: Generate a scenario by sampling from the distribution curve of DERs. The
sum of the samples should be equal to the minimum HC level (HCref).
• Step B: The location of generated DER samples in Step B is selected randomly

using a uniform distribution from the pool of potential locations.
• Step C: Solve the optimization model (11)-(20) to identify the required active
power curtailment over the study period for the defined scenarion in steps A and B.
• Step D: Repeat steps A, B and C until repetition condition is not violated.
Fig. 5.7 The proposed framework for minimum BESS estimation

Frequency
Capacity (kW)
Fig. 5.8 The distribution curve of PV systems based on California solar statistics [26]

1 Total APC > 0
ΘBESS ð jÞ ¼ ð5:21Þ
0 Otherwise
• Step D: Repeat steps A, B, and C until repetition condition is not violated. As it

was mentioned, the HC in region (B) has a distribution function. Therefore, it is
possible to calculate the probability of constraint violations for the new minimum
HC (HCref) as follows:
Z HCref
ProbCV ¼ pdf HC , ð5:22Þ
0
where pdfHC is the probability distribution function of the HC. It was also mentioned
that the active power curtailment over the study period is zero for some of the
generated scenarios in steps A and B. The probability of valid generated scenarios is
as follows:
PN BESS
j¼1 ΘBESS ð jÞ
PrAPC ¼ , ð5:23Þ
N BESS
where NBESS is the total generated scenarios. Repeat steps A, B, and C while the
following condition is held:
N BESS 100 &&, ð5:24Þ

0 < PrAPC ProbCV , ð5:25Þ
5.4.2 Approximated Method
Estimating the required BESS using the proposed comprehensive method is based
on solving an optimization model over the study period for different expansion
scenarios, which means that the comprehensive method could have a considerable
computational burden. However, an important question is how to estimate the
required BESS without extensive computational burden? An approximated method
to estimate the BESS size is presented in this section. The proposed method only
uses the probability distribution function of the HC as well as the load and generation
profiles in the system. The approximation method is based on the maximum back-
feed in the system. To have a better understanding of the method, consider a
simplified distribution system as shown in Fig. 5.9. The back-feed can be calculated
as follows:
pbf ¼ pgen pLd , ð5:26Þ

pgen ¼ sgen PFgen , ð5:27Þ
where sgen and PFgen are the capacity and power factor of DER and pbf represents the
active power that is injected to the upstream grid via distribution system transformer.
In this system, back-feed is the main limiting factor of the HC as it would result in an
over-voltage issue in bus (2). Thus, the HC of the system could be used to calculate
the maximum allowable back-feed as follows:
1 2
pbf, qbf pgen, qgen
Grid Transformer pLd, qLd

V1 V2
Fig. 5.9 A simple representation of a distribution system
bf ¼ sgen PFgen pLd ,

psafe ð5:28Þ
max
where smax
gen is the maximum DER capacity that could be accommodated in the
system, i.e., HC of the system. The maximum allowable back-feed can be used as
an index to estimate the minimum BESS required to increase the HC. Please refer to
[27] for more details on limitations of back-feed in real distribution systems. Any
DER with a higher capacity than smax safe
gen would result in a higher back-feed than pbf ,
hence the over-voltage problem. The excess back-feed power that would cause the
constraint violation can be calculated as follows:
pexcess ¼ sgen PFgen psafe

bf , ð5:29Þ
where sgen smax

gen . To resolve the constraint violation, a BESS can be designed to
store the excess back-feed power. The minimum rating power required for the BESS
is the maximum excess back-feed power during an hour over the studied period.
Similarly, the minimum energy rating of the required BESS is the maximum excess
back-feed power during a day over the studied period. The same concept can be
applied to an actual distribution system. Therefore, maximum back-feed power can
be defined for any distribution system based on the minimum HC of the system. The
steps for estimating the minimum required BESS based on the safe back-feed
concept are as follows:
• Step 1: Identify the maximum back-feed for the system by simplifying the
distribution system to the model presented in Fig. 5.9. To do so, the loads should
be aggregated at the secondary side of the transformer, and the network config-
uration should be neglected. Then, the minimum HC should be allocated next to
the aggregated load. Next, the back-feed power should be calculated using the
aggregated load, minimum HC, as well as the load and DER profiles. Finally, the
maximum of the back-feed power should be identified over the studied period.
• Step 2: Identify the excess back-feed power. To do so, the new minimum HC
should be connected next to the location of the aggregated load. Then, the back-
feed power should be calculated using the aggregated load, new minimum HC, as
well as the load and DER profiles. As shown in Fig. 5.10, the excess back-feed
power can be identified by comparing the back-feed with the maximum back-feed
over the study period.
Back-feed power Excess back-feed

Maximum back-feed power
Substation power
time
Maximum excess
back-feed power
Fig. 5.10 Determination of excess back-feed power (red area) by comparison of maximum back-
feed power (dashed blue line) and back-feed power (solid green line)
• Step 3: Estimate the minimum required BESS based on the excess back-feed over
the study period. The maximum excess back-feed power during an hour over the
studied period is the minimum rating power required for the BESS. Similarly, the
maximum excess back-feed power during a day over the studied period is the
minimum energy rating of the required BESS.
5.5 Economic Assessment
This section explains the proposed method for performing the economic assessment
of the designed BESSs. It should be mentioned that the economic assessment in this
section is based on the annual costs and benefits of the BESSs.
5.5.1 The Annual Costs of the Estimated BESSs
The annual cost of BESS comprises three main terms as follows:

• Total annual capital cost: The total capital cost of an energy storage system
includes the capital costs required for different parts of a BESS. Figure 5.11
shows the structure of a BESS. Three terms should be considered as the capital
costs of the BESS: (1) energy cost for BESS, which is the cost of the devices that
store the energy; (2) cost of power, which is the cost of power electronic devices
(inverter) in the BESS; and (3) the total cost for the balance of plant, which is
Battery Inverter
Grid
Supporting structure
Fig. 5.11 The structure of a BESS
defined as the cost of all the auxiliary systems of a BESS including the trans-
former and supporting structures that are required to exchange the energy.
• Operation and maintenance cost: This term is defined to identify the annual cost
of maintenance and operation of the BESSs.
• Replacement cost: This term is designed to cover the cost of battery replacement
at the end of the life cycle of the battery. If the planning horizon is longer than the
life cycle of batteries, then this term should be considered.
The calculation of the total annual capital cost is as follows [28]:
EESS
BEC ¼ UCSC Total
, ð5:30Þ
Eff
where BEC and UCSC represent the total cost ($) and the unit cost ($/kWh) of
storage device, respectively; E ESS
Total is the designed energy capacity of the storage
device (kWh); and Eff represents the BESS efficiency.
PEC ¼ UCPE SESS

Total , ð5:31Þ
where PEC and UCPE represent the total cost ($) and the unit cost ($/kW) of power
electronic device, respectively, and SESS
Total is the rated power of the BESS (kW).
E ESS
BOP ¼ UBOP Total
, ð5:32Þ
Eff
where BOP and UBOP represent the total cost ($) and the unit cost ($/kWh) of the
balance of plant, respectively. Thus, the total capital cost can be calculated as
follows:
TCC ¼ BEC þ PEC þ BOP, ð5:33Þ

where TCC denotes the total capital cost. Then, the total annual capital cost is as
follows:
ACC ¼ TCC CAF, ð5:34Þ
where
IRð1 þ IRÞhzn
CAF ¼ , ð5:35Þ
ð1 þ IRÞhzn 1
where IR is the annual interest rate and hzn is the planning horizon. Further, the
operation and maintenance cost can be calculated as follows:
AOMC ¼ FOM SESS

Total , ð5:36Þ
where AOMC and FOM denote the total annual cost ($/year) and the fixed unit cost
($/(kW.year)) of the maintenance and operation of BESS, respectively. Finally, the
replacement cost can be calculated as follows:
EESS
ARC ¼ AFactor Total
, ð5:37Þ
Eff
where
h i
AFactor ¼ F factor ð1 þ IRÞr þ ð1 þ IRÞ2r þ CAF, ð5:38Þ
CycleTotal
r¼ , ð5:39Þ
CycleDay N year
where Ffactor represents the future value of the battery replacement cost ($/kWh);
CycleTotal and CycleDay denote the number of charge/discharge cycles of the storage
during its life cycle and during a day, respectively; and Nyear is the number of
operating days in a year for the storage system. Therefore, the total annual cost
(TAC) can be calculated as follows:
TAC ¼ ACC þ AOMC þ ARC: ð5:40Þ
5.5.2 The Annual Costs of the Active Power Curtailment
Compensation of the curtailed energy highly depends on the contract between the
off-taker and the generator owner. Usually, the compensation of active power
curtailment of renewable generators is based on the type of curtailment, the tech-

nology of renewable generation, and the amount of curtailed energy. Most com-
monly, generators are compensated based on the market value for the curtailed
energy. However, usually, this compensation does not include revenue lost from
support mechanisms such as green energy credits. Some countries such as Ireland
and Romania have such a compensation policy. In some countries, the curtailment
compensation only covers a fraction of the curtailed energy. This fraction could vary
from 15% to 50% or more. For instance, Greece only compensates 30% of the
curtailed energy of the wind facilities. However, there is no curtailment compensa-
tion for other technologies in Greece. Moreover, the classification and reason behind
the curtailment are important in the compensation policies. For example, the con-
gestion curtailments are usually compensated, while the curtailments related to the
security of the system are not. This policy has been used in Belgium and Germany.
Other countries might use a different dichotomy. For instance, real-time curtailments
are compensated, while the scheduled ones are not [29]. If it is supposed that there is
no compensation for the curtailed energy, then the curtailed energy is an opportunity
cost for the DER owner. In such a case, the annual cost of the active power
curtailment for all DER owners can be calculated as follows:
XX
8760
CostAPC ¼ T grid
t APCgi,t , ð5:41Þ
i2DG t¼1
where T grid
t is the network electricity price at time (t).
5.5.3 The Annual Benefit of BESSs for the Utility
The focus of this section is identifying the benefits of the integration of BESSs in the
system for the utility. Two obvious options could benefit the utility, which are the
reduction in the network losses as well as the energy arbitrage between peak and
off-peak periods. Both of these objectives can be achieved by minimizing the costs
of supplying energy from a utility perspective. Therefore, the objective function can
be defined as follows:
X X
8760 XX
8760
Minimize costs ¼ T grid
t r ij ℓij,t i,t T t :
pESS grid
ð5:42Þ
ði, jÞ2B t¼1 i t¼1
The decision variables are the optimal size and location of BESSs. The constraints of
the optimization problem include the BESS installation and operation constraints,
network constraints, and DER constraints, which are given as follows:
1. BESS installation constraints
X
BESS
i BESS
total , ð5:43Þ
i
BESS
i SESS,
i
min
SESS
i BESS
i SESS,
i
max
, ð5:44Þ
BESS
i E ESS,
i
min
EESS
i BESS
i E ESS,
i
max
, ð5:45Þ
X
SESS
i SESS
Total , ð5:46Þ
i
X
E ESS
i E ESS
Total , ð5:47Þ
i
BESS
i 2 f01g, ð5:48Þ
where BESStotal is the total number of locations in which BESSs could be installed;
Bi represents the binary variable to allocate BESS at bus (i); SESS
ESS
i and EESS
i
denote the power rating and energy capacity of a given BESS at bus (i),
respectively; SESS,i
max
and EESS,
i
max
represent the maximum power rating and
energy capacity of a given BESS at bus (i); and SESS, i
min
and EESS,
i
min
denote the
minimum power rating and energy capacity of a given BESS at bus (i). Con-
straints (5.46) and (5.47) limit the total power rating and energy capacity of the
BESS to the minimum required BESS, which was identified in Sect. 5.4.
2. BESS operation constraints [19]
X
t
i,t 0 Δt þ Lossi,t 0
pESS E ESS 8t 2 H , ð5:49Þ
ESS
i,t 0 ¼0
t 0 ¼0
X
t
pESS
i,t 0 Δt Lossi,t 0
ESS
EESS
i E ESS
i,t 0 ¼0 8t 2 H , ð5:50Þ
t 0 ¼0
i,t ¼ E i,t1 pi,t Δt Lossi,t Δt,

EESS ð5:51Þ
ESS ESS ESS
SESS
i pESS
i,t Si
ESS
8t 2 H , ð5:52Þ
εmin
i E ESS
i E ESS
i,t εi
max
E ESS
i 8t 2 H , ð5:53Þ

i,t ¼ 1 ηi
LossESS ESS
pESS
i,t
8t 2 H , ð5:54Þ
i,t pi,t1 þ pi
pESS 8t 2 H , ð5:55Þ
ESS RURD
i,t pi,t1 pi
pESS 8t 2 H , ð5:56Þ
ESS RURD
where H and Δt are the set of time steps and the length of each time step,
i,t 0 ¼0 represents the initial stored energy in BESS at bus (i); εi ,
respectively; EESS min
εmax
i , and ηi
ESS
denote the minimum and maximum allowable stored energy level
as well as the efficiency of the BESS at bus (i), respectively; and pRURD i is the
charging/discharging rate of BESS at bus (i). Constraint (5.49) is defined to make
sure that the sum of all stored and taken energy over all the steps of the study
period is less than the initial stored energy. Similarly, constraint (5.50) is defined
to keep the sum of all stored and taken energy over all the steps of the study
period below the initial available capacity of BESS. Constraint (5.51) models the
relationship between the state of charge and the output power of BESS at the time
(t) and bus (i). Constraint (5.52) keeps BESS power below their rating power;
constraint (5.53) is defined to set the minimum and maximum state of charge of
BESS at bus (i); constraint (5.54) is defined to model the energy losses in each
BESS at each time step; constraints (5.55) and (5.56) define the maximum
charging and discharging rate of BESSs. Finally, constraint (5.54) can be written
in a linear form by defining new variables as follows:
i,t ¼ XX i,t YY i,t ,

pESS ð5:57Þ
0 XX i,t SESS,
i
max
, ð5:58Þ
0 YY i,t SESS,
i
max
, ð5:59Þ

0 LossESS
i,t 1 ηi
ESS
ðXX i,t YY i,t Þ 2 SESS,
i
max
DD2,i,t , ð5:60Þ

i,t 1 ηi
0 LossESS ESS
ðYY i,t XX i,t Þ 2 SESS,
i
max
DD1,i,t , ð5:61Þ
DD1,i,t þ DD2,i,t ¼ 1, ð5:62Þ
DD1,i,t , DD2,i,t 2 f0, 1g, ð5:63Þ
where DD1,i,t and DD2,i,t are binary variables that identify the charging/
discharging status of the BESS. DD1,i,t ¼ 1 when the BESS is discharging, and
DD2,i,t ¼ 1 when the BESS is charging.
3. Network constraints
These constraints include the power flow as well as over-voltage constraints. Any
other technical constraints such as thermal capacity could be also defined as a
network constraint. The network constraints considered in this study are as
follows:
X
Pij,t ¼ Pjk,t þ pdj,t pgj,t pESS
j,t þ r ij ℓ ij,t 8j 2 DG, ð5:64Þ
k:j!k
X
Pij,t ¼ Pjk,t þ pdj,t pESS
j,t þ r ij ℓ ij,t 8j 2 fN ∖DG g, ð5:65Þ
k:j!k
X
Qij,t ¼ Qjk,t þ qdj,t qgj,t þ xij ℓij,t 8j 2 DG, ð5:66Þ
k:j!k
X
Qij,t ¼ Qjk,t þ qdj,t þ xij ℓ ij,t 8j 2 fN ∖DG g, ð5:67Þ
k:j!k

υ j,t ¼ υi,t 2 r ij Pij,t þ xij Qij,t þ r 2ij þ x2ij ℓ ij,t 8ði, jÞ 2 B, ð5:68Þ
P2ij,t þ Q2ij,t
ℓ ij,t ¼ 8ði, jÞ 2 B, ð5:69Þ
υi,t
υ υi,t υ 8i 2 N , ð5:70Þ
where υ represents the minimum allowable voltage in the system. Further,

constraints (5.64), (5.65), (5.66), and (5.67) represent the active and reactive
power balance in the network; constraints (5.68) and (5.69) represent the power
flow equations in the line (i, j); constraint (5.70) would make sure that the voltage
of the buses would stay in the operation range. Considering the objective function
and the constraints, the defined optimization is an optimal power flow (OPF)
problem. The relaxed OPF problem, which is formed by eliminating voltage and
current angles from power flow equations, is still non-convex due to the quadratic
constraint (5.69). This quadratic term can be approximated using a piecewise
linear function. Generally, a nonlinear function f(t) can be approximated over an
interval [a, b] by a piecewise linear function bf ðt Þ by using the break points
a ¼ t0 < t1 < t2 < . . . < tk ¼ b as follows:
X
k X
k
t¼ λv t v , λv ¼ 1, λv 0 8v 2 f0, 1, 2, . . . , k g, ð5:71Þ
v¼0 v¼0
X
k
bf ðt Þ ¼ λvbf ðt v Þ, ð5:72Þ
v¼0
where t0 < t1 < t2 < . . . < tk represent the ending points of the pieces. Further,
λv 8 v 2 {0, 1, . . ., k} are special order set of type 2 variables. This means that only
two adjacent λv are non-zero. This restriction can be removed by using binary
variables as follows:
0 λ0 w 0 , ð5:73Þ
0 λv wv1 þ wv , 8v 2 f1, 2, . . . , k 1g, ð5:74Þ
0 λk wk1 , ð5:75Þ
X
k1
wv ¼ 1, ð5:76Þ
v¼0
wv 2 f0, 1g: ð5:77Þ
This linearization technique is only valid for single variable functions. However,
ℓ ij, t in constraint (5.69) is a function of (P, Q, v). Nevertheless, if it is supposed
that v is fixed, then constraint (5.69) turns into a separable function. A function
such as f(Var1, Var2, . . ., Varn) is called separable if it could be expressed as a
summation of n single variable function such as f1(Var1), f2(Var2), . . ., fn(Varn).
Thus, each single variable function can be approximated with a piecewise linear
function. Please refer to [27] for more information regarding approximating the
nonlinear term (5.69) with a piecewise linear function.
4. DER constraints
pgi,t ¼ ηgi,t Capgi 8i 2 DG, ð5:78Þ

qgi,t ¼ pgi,t tan ðϕgi Þ 8i 2 DG, ð5:79Þ
where ηgi,t is the capacity factor of DER at time (t) and Capgi and ϕgi are the
capacity and power factor angle of DER at bus (i).The optimization model for
allocating the BESSs to minimize the costs of supplying loads is a mixed-integer
linear programming (MILP).
5.5.4 Cost/Benefit Assessment
The total benefits for DER owner as well as the utility are compared to the annual
cost of installing the required BESS. If the annual benefit is higher than the annual
costs, using BESSs to improve the HC is justifiable.
5.6 Different Energy Storage Technologies
There are different electrical energy storages, including lead-acid (LA), sodium-
sulfur (NaS), vanadium redox (VRB), and zinc/bromine (ZnBr). The comparison of
different BESSs is presented in Table 5.1 based on the development status, energy,
and power capital costs [30]. As can be seen, there is a considerable difference
between the capital cost of these BESS technologies, which can affect the feasibility
of the required BESS capacity from an economic perspective.
Table 5.1 Comparison of the BESS technologies based on cost and development status indexes
Energy capital cost Power capital cost
Technology Development status (US $/kWh) (US $/kW)
Lead-acid (LA) Mature 200–400 200–600
Sodium-sulfur (NaS) Commercialized 275–500 1000–3000
Vanadium redox (VRB) Early commercialization 600–1500 150–1000
stage
Zinc/bromine (ZnBr) Developed 200–400 150–1000
Energy storage installation cost ($/kWh)
1000
900
800
700
600
500
400
300
200
100
0
2016 2030
Lead acid (Flooded) Lead acid (VRLA)
Sodium-sulfur Sodium-Nickel chloride
Flow (VRFB) Flow (ZBFB)
Lithium-ion (NMC/LMC) Lithium-ion (LFP)
Fig. 5.12 Installation cost estimation of battery storages in 2016 and 2030
It should be mentioned that the capital cost of BESSs has been decreasing, which
means that although a BESS technology might not be economically feasible now, it
may become feasible in the future. Figure 5.12 shows the energy capital cost
estimation for different battery technologies in 2016 and 2030 [31]. As can be
seen, it is expected to have up to 66% decrease in capital costs by 2030. Similarly,
it is expected to experience an increase in the efficiency of battery storages. Fig-
ure 5.13 presents the expected change in the energy capital cost as well as the
efficiency of battery storage technologies in 2016 and 2030 [31].
Energy storage installation cost ($/kWh)

1000
900 2016
800
700
600 2016
500
400 2016 2016
2016 2016
2030
300 2016
200 2030
2030 2030 2030 2030
2016 2030
100
2030
0
68 73 78 83 88 93 98
Efficiency (%)
Lead acid (Flooded) Lead acid (VRLA)
Sodium-sulfur Sodium-Nickel chloride
Flow (VRFB) Flow (ZBFB)
Fig. 5.13 Estimated changes in the installation costs as well as the efficiency of battery storages
from 2016 to 2030
5.7 Numerical Results
In this section, simulations are carried out to assess the performance of the proposed
methodology. Initially, the effectiveness of the proposed method for identifying the
minimum BESS capacity is evaluated. Then, the economic feasibility of the mini-
mum required BESS for different technologies is discussed. Finally, the effect of
DER technologies on the minimum required BESS is assessed.
5.7.1 Test System
The proposed methodology is examined on a balanced distribution system. The

system is a 33-kV feeder in Australia, as shown in Fig. 5.14 [32]. This feeder is the
representative of agricultural feeders in Australia, which supplies loads such as
irrigation pumps and dairies. The green circles are the load points, and the total
load is 4.48 MW in the test system. The maximum allowable voltage rise is 5%.
5.7.2 Input Data
Figure 5.15 represents the normalized load profile, which is used to model the load
variation in the test network. The load profile is derived from the data made available
1 Substation 80
166 102 91
164 165 113
163 134 33
160 161
162 167 44
14 15
159 11 55
12 13 155 153
158 10 22 77
8
157 154 152 81
147 148 149 150 151 63 62 61
146 82
16 17 71 70 69
144 83
65 64
142 143
141 84
73 72 60 59
18 19 20
140 57 85
138 139
137 58 86
21 23
136 87
24 25
135 92 90
133 130 132 37 40 93

95 94
129 131
128 29 36 39 96
26 27
2 28 35 38 109 108
127 99
100
123 122 114 112
101
124 111 110 107
47 49 115 51 103
48 50 116 52 104
117 53
Fig. 5.14 Topology of an agricultural feeder in Australia
by the Australian Energy Market Operator (AEMO) [33]. The normalized PV and
wind profiles are presented in Figs. 5.16 and 5.17, which are derived from [34, 35],
respectively. The distribution of PV and wind sizes are derived from [25].
0.9
0.8
Load profile
0.7
0.6
0.5
0.4
0.3
300
200 20
15
100 10
5 Hour
Day
Fig. 5.15 Normalized hourly load profile for New South Wales from 1/1/2014 to 30/12/2014
0.8
0.6
PV profile
0.4
0.2
300
200 20
15
100 10
5
Day Hour
Fig. 5.16 Normalized hourly PV profiles for New South Wales from 1/1/2014 to 30/12/2014
Fig. 5.17 Wind production profile for New South Wales from 1/1/2014 to 30/12/2014
Table 5.2 Parameters for BESS economical assessment

Parameter LA VRLA NaS ZnBr VRB
Efficiency 0.75 0.75 0.77 0.7 0.7
Unit cost for power electronic 175 175 1000 175 Included in the unit
inverter (US $/kW) cost for battery
storage
Unit cost for battery storage 305 360 500 225 740
(US $/kWh)
Unit cost for balance of plant 50 50 0 0 30
(US $/kWh)
Fixed operation and maintenance cost 15 5 20 20 20
(US $/kW)
Future replacement cost (US $/kWh) 305 360 500 225 222
Number of charge/discharge life cycle 3200 1000 2500 10,000 10,000
Table 5.3 Electricity price data

Anytime energy Off-peak energy Shoulder energy Peak energy
Electricity price scheme (c/kWh) (c/kWh) (c/kWh) (c/kWh)
Flat 29.8 – – –
Time of use (ToU) – 19.6 36.8 45
Table 5.2 represents all the parameters required for calculating the annual instal-
lation cost of BESS for different technologies [28]. Table 5.3 shows the electricity
price rate for two different schemes from Origin Energy [36]. Peak period mentioned
in Table 5.3 is from 1 to 8 pm; shoulder period is from 7 am to 1 pm and from 8 to
10 pm; and off-peak period is from 10 to 7 am.
Fig. 5.18 PV HC
probability curve for the
agricultural feeder
CDF
5.7.3 Hosting Capacity Discussion
PV and wind HC of the test system are estimated based on the HC calculation
method presented in Sect. 5.3. Figure 5.18 presents PV HC curve for the test system,
which can be approximated using the Gaussian-shape distribution as (5.80) with
α ¼ 0.56, μ ¼ 19.97 MW, and σ ¼ 6.42 MW. The minimum PV HC for the test
system is 3.198 MW.
α ðxμÞ2
HC α N μ, σ 2 ¼ pffiffiffiffiffiffiffiffiffiffi e 2σ2 : ð5:80Þ
2πσ 2
Figure 5.19 presents wind HC curve for the test system, which can be approxi-
mated using the Gaussian-shape distribution as (5.80) with α ¼ 0.956,
μ ¼ 37.21 MW, and σ ¼ 0.7092 MW. The minimum wind HC for the test system
is 2.92 MW. As it can be seen, the HC distribution for wind is significantly different
from PV HC. This is mainly due to the difference between the wind and PV
generation profiles as well as the distribution of wind and PV sizes.
5.7.4 Minimum Required BESS
The performance of the proposed methods for estimating the minimum required
BESS is examined in this section. The minimum HC for PV and wind generation is
3.198 MW and 2.92 MW, respectively. Table 5.4 presents the estimated required
BESS to increase the minimum HC to 4 MW.
0.6
Histogram
Gaussian PDF
0.5
0.4
Probability
0.3
0.2
0.1
0
5 10 15 20 25 30 35
Hosting Capacity (MW)
Fig. 5.19 Wind HC probability curve for the agricultural feeder
Table 5.4 Minimum required BESS to increase the HC to 4 MW for PV and wind technologies
using the comprehensive and approximated methods
Comprehensive method Approximated method
Energy capacity Rating power Energy capacity Rating power
DER technology (MWh) (MW) (MWh) (MW)
PV 5.275 0.6697 5.515 0.75
Wind 9.879 0.5212 10.279 1.07
As can be seen, the minimum required BESS obtained by the approximated

method is higher than that of the comprehensive method. However, the advantage
of the approximated method over the comprehensive method is that it does not
require the network model. Therefore, it can be calculated without any computa-
tional burden. The drawback of the approximated method is that it would
overestimate the required BESS capacity, which could affect the feasibility of the
required BESS from an economic perspective. Moreover, the minimum required
BESS energy capacity for increasing the wind HC to 4 MW is 87.28% higher than
that of PV HC. Further, the minimum power rating of the required BESS for wind
technology is 22.17% lower than of that for PV technology. This shows that the
minimum required BESS capacity in a system depends on the type of distributed
energy resources that would be accommodated in that system.
5.7.5 Economic Feasibility Assessment of the Required BESS
This section discusses the economic feasibility of the estimated BESS that is
required for different technologies. To do so, the total annual cost of the required
BESS for different technologies is compared with the annual benefits of the
corresponding BESS. Table 5.5 shows the total annual cost for the required BESS
estimated using the proposed comprehensive method for PV and wind technologies,
respectively. As can be seen, using the ZnBr battery would result in the lowest total
annual costs for both PV and wind technologies, while the valve-regulated lead-acid
(VRLA) battery has the highest total annual costs.
If the total annual profit of the designed battery is higher than the minimum
annual costs, the designed BESS is economically feasible. Tables 5.6 and 5.7 present
the optimal allocation of the required BESS in the system for both PV and wind
technologies when the electricity price follows a flat rate. As can be seen, the
required BESS capacity is not distributed equally. Further, the optimal BESS
location for PV technology is different from that of wind technology.
Tables 5.8 and 5.9 present the optimal allocation of the required BESS in the
system for both PV and wind technologies when the electricity price follows a ToU
rate. Observe that the required BESS capacity is not distributed equally. Further, the
optimal BESS location for PV technology is different from that of wind technology.
Another important factor that has a considerable impact on the allocation of the
Table 5.5 Total annual costs of the required BESS for different technologies in US $ (2014)
BESS technology
DER technology LA VRLA NaS ZnBr VRB
PV 902,323 2,124,994 1,696,292 370,760 1,056,548
Wind 1,656,803 3,953,947 3,036,043 657,629 1,964,038
Table 5.6 Optimal allocation of the required BESS in the test system for PV generation when the
electricity price follows the flat rate
BESS1 BESS2
Location 149 159
Energy capacity (MWh) 3.6 1.675
Rating power (MW) 0.3 0.3697
Table 5.7 Optimal allocation of the required BESS in the test system for wind generation when the
electricity price follows the flat rate
BESS1 BESS2
Location 151 162
Table 5.8 Optimal allocation of the required BESS in the test system for PV generation when the
electricity price follows the ToU rate
BESS1 BESS2
Location 149 159
Table 5.9 Optimal allocation of the required BESS in the test system for wind generation when the
electricity price follows the ToU rate
BESS1 BESS2
Location 151 162
200
Profit in US $ (Thousands)
180
160
140
120
100
80
60
40
20
0
PV Wind
Flat rate ToU rate
Fig. 5.20 Utility annual profit from installing the required BESS for PV and wind generation for
both flat and ToU electricity price rates
required BESS for both PV and wind technologies is the electricity price. Consid-
ering Tables 5.6 and 5.8, it can be noted that the optimal location of BESSs for both
flat and ToU electricity prices is the same. However, the energy capacity and the
rating power of the optimal BESSs are different.
Figure 5.20 demonstrates the annual benefit that utility would gain from installing
the BESSs for both PV and wind technologies. As can be seen, the annual profit of
the utility with ToU tariff is higher than that of the flat tariff. Further, the BESS
benefit for PV generation is much higher than that of wind generation in the test
system. Moreover, the total annual profit of the utility is below the total annual costs
of the minimum required BESS for different storage technologies, which might
convey that BESS is not an economically feasible solution for increasing the
4500
4000
3500
3000
US $ (Thousands)
2500
2000
1500
1000
500
0
Utility DERs Utility DERs total Total LA cost VRLA NaS cost ZnBr VRB
profit profit profit profit profit Profit cost cost cost
with flat with flat with with with flat with
rate rate ToU rateToU rate rate ToU rate
PV Wind
Total annual profit of BESS for PV Total annual profit of BESS for wind
Fig. 5.21 Total annual profit compared with the total annual cost for different BESS technologies
HC. Nevertheless, the utility benefit is only a part of the BESS benefit. DER owners
also gain some benefit from avoiding active power curtailment. Figure 5.21 shows
the total annual profit of both utility and DER owners compared with the total annual
costs of different BESS technologies. As it can be seen, the total annual profit of the
BESS with both flat and ToU electricity price rates is still lower than the total annual
costs of all BESS technologies, which means that although BESSs can increase the
HC of the system, they are not a feasible option from an economic perspective.
However, it was also noted that the total annual profit of BESS for PV technology,
when the electricity price followed ToU rate, was 325,353.00$, which is consider-
ably close to the total annual cost of ZnBr battery technology, i.e., 370,760.00$.
Another considerable point in Fig. 5.21 is that the total annual profit of BESS for
wind generation is considerably lower than the total annual cost of all BESS
technologies. This implies that there are a few days that the spilled wind energy is
very high, but the spilled wind energy for the rest of the days is at a low level.
Finally, it should be mentioned that although some of BESS technologies are
economically infeasible now, it does not mean that they will continue to be infeasible
in the future. As it was mentioned in Sect. 5.6, it is predicted that the price of BESSs
would decrease and their efficiency would increase. Figure 5.22 presents the total
annual profit of both utility and DER owners compared with the expected annual
costs of different BESS technologies in 2030. As can be seen, ZnBr battery
3500
3000
2500
US $ (Thousands)
2000
1500
1000
500
0
Utility DERs Utility DERs total Total LA cost VRLA NaS cost ZnBr VRB
profit profit profit profit profit Profit cost cost cost
with flat with flat with with with flat with
rate rate ToU rateToU rate rate ToU rate
PV
Wind
Total annual profit of BESS for PV with ToU rate
Total annual profit of BESS for wind with ToU rate
Total annual profit of BESS for PV with flat rate
Fig. 5.22 Expected total annual profit in 2030 compared with the total annual cost for different
BESS technologies
technology is expected to be economically feasible in 2030 with both flat and ToU
electricity rates for PV technology. However, even the future decrease in the price of
energy storages would not make BESSs a feasible option for increasing the HC for
wind technology.
5.8 Summary
Technical issues such as over-voltage and overloading of lines and transformers

bound DERs that could be accommodated in distribution systems. Therefore, it is of
great importance to know how distribution systems can accommodate a higher level
of DERs. This chapter proposed a comprehensive method and an approximated
approach to identify the minimum required battery energy storage to increase the HC
of a system to a certain level. The comprehensive method aims to maximize the
injected power in the system while minimizing the curtailed energy. The outcomes
of this methodology are the minimum required power and energy capacity rating of
BESSs to increase the HC of the system. The second step of the methodology is the
economic assessment of the determined BESSs. The performance of the
comprehensive and approximated methods is assessed on an agricultural distribution

system in Australia. It was shown that the approximated method result is the solution
of the comprehensive method with an error of 5%. Then, the cost/benefit analysis is
carried out to assess the feasibility of the obtained BESSs. It was shown that the
BESS technology highly affects the feasibility of BESSs for HC improvement.
Further, it was demonstrated that none of the considered BESS technologies is
economically feasible for increasing the HC of the test system. Next, the impact of
DER technology (i.e., photovoltaics and wind) on the size as well as the economic
feasibility of the required BESS was assessed. It was shown that the required BESS
for increasing the HC of the test system for PV technology is 46.6% less than what is
required for wind technology. Moreover, it was demonstrated that the use of BESS
for increasing the HC of the system for PV technology would be economically
feasible in 2030. However, the use of BESSs for increasing the HC would not be a
feasible option for wind technology in the test system in 2030. Another factor that
has been assessed in this chapter was the effect of electricity price on the allocation
of the required BESS. It was shown that the ToU electricity price rate would yield in
a higher profit level for the utility than the flat electricity price rate, i.e., 12.78% for
PV and 126.8% for wind technology. Further, it was noted that the electricity price
rate did not affect the optimal location of the required BESSs. It, however, would
affect the energy capacity and rating power of the allocated BESSs in the system.
5.9 Future Works
As it was shown, community BESS could be an economically feasible solution to

increase the HC. However, there are other viable options such as using the on-load
tap changer (OLTC) of transformers, reactive power control of DERs, and reactive
power resources such as reactors. Although the aforementioned options could
increase the power losses in the system, they would decrease the APC by increasing
the HC. Therefore, these options could also be feasible from an economic point of
view. Thus, to choose the best option to increase the HC, a techno-economical
assessment should be performed to compare the profits of community BESS, trans-
former’s OLTC, DER-based reactive power control, and reactive power sources
such as reactors.
The second research question that could be addressed in the future is whether
residential BESSs could substitute community BESSs. The penetration level of
residential BESSs has been increasing in the recent years. Although residential
BESSs could be used to store the excess energy, they are usually controlled in a
way to maximize the profit of the prosumers. However, the best control scheme from
the prosumers’ point of view could be different from the best control scheme from
the network viewpoint. Thus, there is no guarantee that the aggregated residential
BESSs would increase the HC in the same way as community BESSs. Therefore, if it
is supposed that the required community storage to increase the HC to a certain level
(HCref) is (BESSCOM), we would like to estimate the required residential BESSs that
would increase the HC to the same HC level (HCref).
References
1. Shayani, R. A., & de Oliveira, M. A. G. (2011). Photovoltaic generation penetration limits in

radial distribution systems. IEEE Transactions on Power Systems, 26(3), 1625–1631.
2. Stetz, T., Marten, F., & Braun, M. (2013). Improved low voltage grid-integration of photovol-
taic systems in Germany. IEEE Transactions on Sustainable Energy, 4(2), 534–542.
3. Bletterie, B., et al. (2017). Voltage control with PV inverters in low voltage networks-in depth
analysis of different concepts and parameterization criteria. IEEE Transactions on Power
Systems, 32(1), 177–185.
4. Ghosh, S., Rahman, S., & Pipattanasomporn, M. (2017). Distribution voltage regulation
through active power curtailment with PV inverters and solar generation forecasts. IEEE
Transactions on Sustainable Energy, 8(1), 13–22.
5. Chalise, S., et al. (2016). Impact of active power curtailment of wind turbines connected to
residential feeders for overvoltage prevention. IEEE Transactions on Sustainable Energy, 7(2),
471–479.
6. Hashemi, S., Ostergaard, J., & Yang, G. Y. (2014). A scenario-based approach for energy
storage capacity determination in LV grids with high PV penetration. IEEE Transactions on
Smart Grid, 5(3), 1514–1522.
7. Omran, W. A., Kazerani, M., & Salama, M. M. A. (2011). Investigation of methods for
reduction of power fluctuations generated from large grid-connected photovoltaic systems.
IEEE Transactions on Energy Conversion, 26(1), 318–327.
8. Yang, Y., et al. (2014). Sizing strategy of distributed battery storage system with high penetra-
tion of photovoltaic for voltage regulation and peak load shaving. IEEE Transactions on Smart
Grid, 5(2), 982–991.
9. Damavandi, M. G., Marti, J. R., & Krishnamurthy, V. (2018). A methodology for optimal
distributed storage planning in smart distribution grids. IEEE Transactions on Sustainable
Energy, 9(2), 729–740.
10. Hashemi, S., & Ostergaard, J. (2018). Efficient control of energy storage for increasing the PV
hosting capacity of LV grids. IEEE Transactions on Smart Grid, 9(3), 2295–2303.
11. Zafar, R., et al. (2018). Multi-timescale model predictive control of battery energy storage
system using conic relaxation in smart distribution grids. IEEE Transactions on Power Systems,
33(6), 7152–7161.
12. Atwa, Y. M., & El-Saadany, E. F. (2010). Optimal allocation of ESS in distribution systems
with a high penetration of wind energy. IEEE Transactions on Power Systems, 25(4),
1815–1822.
13. Divshali, P. H., & Soder, L. (2019). Improving hosting capacity of rooftop PVs by quadratic
control of an LV-central BSS. IEEE Transactions on Smart Grid, 10(1), 919–927.
14. Bocker, B., et al. (2018). Storage valuation in congested grids. IEEE Transactions on Smart
Grid, 9(6), 6742–6751.
15. Hassan, A., & Dvorkin, Y. (2018). Energy storage siting and sizing in coordinated distribution
and transmission systems. IEEE Transactions on Sustainable Energy, 9(4), 1692–1701.
16. Hooshmand, E., & Rabiee, A. (2018). Robust model for optimal allocation of renewable energy
sources, energy storage systems and demand response in distribution systems via information
gap decision theory. IET Generation, Transmission, 13(4), 511–520.
17. Lin, Z., et al. (2019). Optimal ESS allocation in distribution network using accelerated
generalised benders decomposition. IET Generation, Transmission, 13, 2738–2746.
18. Lei, J., et al. (2018). Optimal allocation of a VRB energy storage system for wind power
applications considering the dynamic efficiency and life of VRB in active distribution networks.
IET Renewable Power Generation, 13(4), 563–571.
19. Nick, M., Cherkaoui, R., & Paolone, M. (2014). Optimal allocation of dispersed energy storage
systems in active distribution networks for energy balance and grid support. IEEE Transactions
on Power Systems, 29(5), 2300–2310.
20. Jalali, A., & Aldeen, M. (2019). Risk-based stochastic allocation of ESS to ensure voltage
stability margin for distribution systems. IEEE Transactions on Power Systems, 34(2),
1264–1277.
21. Li, Q., Ayyanar, R., & Vittal, V. (2016). Convex optimization for DES planning and operation
in radial distribution systems with high penetration of photovoltaic resources. IEEE Trans-
actions on Sustainable Energy, 7(3), 985–995.
22. Giannitrapani, A., et al. (2017). Optimal allocation of energy storage systems for voltage control
in LV distribution networks. IEEE Transactions on Smart Grid, 8(6), 2859–2870.
23. Dusonchet, L., et al. (2019). Technological and legislative status point of stationary energy
storages in the EU. Renewable & Sustainable Energy Reviews, 101, 158–167.
24. Telaretti, E., & Dusonchet, L. (2017). Stationary battery technologies in the US: Development
trends and prospects. Renewable & Sustainable Energy Reviews, 75, 380–392.
25. Seydali Seyf Abad, M., et al. (2018). Probabilistic assessment of hosting capacity in radial
distribution systems. IEEE Transactions on Sustainable Energy, 9(4), 1935–1947.
26. Dubey, A., Santoso, S., & Maitra, A. (2015). Understanding photovoltaic hosting capacity of
distribution circuits. In 2015 IEEE Power & Energy Society General Meeting, Denver, CO,
pp 1–5.
27. Seydali Seyf Abad, M., Ma, J., & Han, X. (2019). Distribution systems hosting capacity
assessment: Relaxation and linearization. In Smart power distribution systems (pp 555–586).
Cambridge, Massachusetts, USA: Elsevier.
28. Poonpun, P., & Jewell, W. T. (2008). Analysis of the cost per kilowatt hour to store electricity.
IEEE Transactions on Energy Conversion, 23(2), 529–534.
29. Lori Bird, J. C., & Wang, X. (2014). Wind and solar energy curtailment: Experience and
practices in the United States. Golden: National Renewable Energy Laboratory.
30. Akinyele, D., Belikov, J., & Levron, Y. (2017). Battery storage technologies for electrical
applications: Impact in stand-alone photovoltaic systems. Energies, 10(11), 1760–1799.
31. Pablo Ralon, M. T., Ilas, A., Diaz-Bone, H., & Kairies, K.-P. (2017). Electricity storage and
renewables: Costs and markets to 2030. Australia: IRENA.
32. Berry, A., et al. (2013). National feeder taxonomy; describing a representative feeder set for
Australian networks. Newcastle: CSIRO.
33. AEMO, National transmission network development plan 2015. 2015.
34. Pfenninger, S., & Staffell, I. (2016). Long-term patterns of European PV output using 30 years
of validated hourly reanalysis and satellite data. Energy, 114, 1251–1265.
35. Staffell, I., & Pfenninger, S. (2016). Using bias-corrected reanalysis to simulate current and
future wind power output. Energy, 114, 1224–1239.
36. Energy, O. (2018). NSW residential energy price fact sheet. NSW, Australia: Origin Energy.
Chapter 6
Impact of Capacity Withholding
on Hosting Capacity Analyzing
Saeed Salarkheili and Mehrdad SetayeshNazar
6.1 Introduction
Many vertically integrated power companies have been restructured worldwide

during the last three decades for improving the competitiveness and efficiency of
energy industry [1]. However, a perfectly competitive power market does not exist in
real world, and most of the present electricity markets can be considered as oligopoly
electricity markets [2]. Some generation companies (GenCos), in these imperfect
competition situations, possess market power to obtain more profits at the cost of
other market players. Strategies of market power can be categorized as capacity
withholding and financial withholding. When either strategy is successful, higher
profits of GenCos and lower market performance are expected [3–7].
Various indices and models should be used by market monitoring to analyze
GenCos’ market power in electricity markets [3–7]. In addition, regulators should
use market power mitigation problem. A proper program for the mitigation of market
power can force the extra producer surpluses acquired by the GenCos’ market power
back to the customer side. The 2000–2001 California electricity crisis demonstrated
that the programs for market power mitigation play a key role for electricity markets.
In fact, the capacity withholding of GenCos which was aggravated by the tight
conditions of market during the summer of 2000 combined with the absence of
proper programs for market power mitigation to create energy market crisis. It means
that the GenCos’ capacity withholding can lead to high market prices and scarcity of
generation capacity. Implementing capacity withholding analyzing in the process of
S. Salarkheili (*)
Faculty of Electrical Engineering, Shahid Beheshti University, A.C, Tehran, Iran
Tehran Regional Electric Company (TREC), Tehran, Iran
e-mail: s_salarkheil@sbu.ac.ir
M. SetayeshNazar
Faculty of Electrical Engineering, Shahid Beheshti University, A.C, Tehran, Iran

https://doi.org/10.1007/978-3-030-40029-3_6
122 S. Salarkheili and M. SetayeshNazar
hosting capacity (HC) analyzing can cause to select proper capacities and locations
of DGs to mitigate capacity withholding in power markets. In this chapter, capacity
withholding analyzing of GenCos is described and different indices are proposed. In
Sect. 6.2, the effect of generation limits on the capacity withholding of GenCos is
analyzed. In the next section, the capacity withholding of GenCos is evaluated in
transmission-constrained power markets. In Sect. 6.4, the mitigation of capacity of
withholding of GenCos and the impact of DGs on capacity withholding are
described.
6.2 Capacity Withholding Analyzing by Considering

Capacity Constraints
In an oligopoly market GenCos is able to exercise market power through different

bidding strategies. As mentioned above, capacity withholding and financial (eco-
nomic) withholding are the two parts of market power. Financial withholding means
that GenCos bid their full capacity but increase their bidding prices extremely above
their marginal cost to acquire higher profits. Since the market operator almost knows
GenCos’ marginal costs, financial withholding is usually easily detected and, then,
hard to implement in electricity markets with working market monitoring mecha-
nisms in place. On the other hand, GenCos’ capacity withholding can be justified by
pointing to operational requirements, which are not easy to verify and challenge by
authorities. In addition, capacity withholding can cause high-cost power units to
operate and increase the electricity market price. The ability of a GenCo to exercise
capacity withholding depends on its capacity limits, parameters of cost function,
transmission network limits, and the elasticity of electricity demand. In this section,
the effects of capacity constraints and demand elasticity are assessed.
6.2.1 Power Market Modeling
Noncollusive game models such as Cournot and supply function equilibrium (SFE)
models are broadly used for analyzing of market power [3–30]. The Cournot model
is used in this chapter to represent the strategic GenCos’ behavior in the presence of
elastic demands. Therefore, the theory of Cournot-NE (CNE) is used to find the
market equilibrium considering the GenCos’ capacity withholding.
Suppose in the market there are N GenCos and each GenCo has the following cost
function:
ci ¼ 0:5ai q2i þ bi qi i ¼ 1, . . . , N ð6:1Þ

6 Impact of Capacity Withholding on Hosting Capacity Analyzing 123
where qi is the quantity generated by GenCo i and ai and bi are the coefficients of the
GenCo’s cost function with ai 0 and bi 0.
The aggregate demand function is
λ ¼ αY þ β
ð6:2Þ
α>0
P
N
where Y ¼ yi and the transmission loss is ignorable. α and β are parameters of the
i¼1
aggregate demand function and λ is the electricity market price.
Each GenCo seeks in the market to maximize its profit (Ωi), and the optimization
problem faced by GenCo i is determined as follows:
max Ωi ¼ qi λ c i
ð6:3Þ
s:t: qmin
i qi qmax
i
where qmax
i and qmin
i are upper and lower production constraints of GenCo i,
respectively. The Lagrange function of the problem (6.3) is

Li ¼ qi λ ci þ θmax
i qi qmax
i θmin
i qi qmin
i ð6:4Þ
where θmax
i and θmin
i are Lagrange multipliers associated with lower and upper
production limits of GenCo i, respectively.
Differentiating the (6.4) with respect to qi and setting it as zero obtains
∂Li ∂λ
¼ qi þ λ ai qi bi þ θmax θmin ¼0 ð6:5Þ
∂qi ∂qi i i
In a perfectly competitive market, all GenCos are not price maker and the term
∂π/∂yi in (6.5) is equal to zero. But in an oligopoly market, GenCos are not price
taker and they are price taker and ∂π/∂yi 6¼ 0 in (6.5) can be yields. So qi of GenCo
i under perfect competition can be obtained by
λ bi θmax þ θmin
qi ¼ i i
ð6:6Þ
ai
According to (6.5) and by using (6.2), qi each GenCo under Cournot model can
be obtained by
λ bi θmax þ θmin
qi ¼ i i
ð6:7Þ
ð α þ ai Þ
6.2.2 Capacity Withholding Indices
Many indices have been proposed for market power assessment. There are different
categorizations for indices of market power. One of them categorizes indices of
market power into comparison and concentration indices. In comparison indices, the
oligopoly market’s results are compared with those of perfectly competitive market,
and concentration indices are based on the GenCos’ market shares. As another
categorization, the indices of market power can be classified into ex ante indices,
based on the simulation tools and market models, and ex post indices, in which the
results of real market are used. As a third categorization, there are global and nodal
indices. The nodal indices are used for analyzing of market power in each bus of
power system, and global indices assess the market power in the whole system.
Nodal indices are appropriate for transmission-constrained power markets [26–29].
A set of comparison indices is used in this chapter to analyze the capacity withhold-
ing of GenCos. In the proposed indices, the results of oligopoly market are compared
with those of two types of markets: perfectly competitive market with the nodal
prices of oligopoly market and perfectly competitive market. The superscripts e, p,
and p(e) are used for the values associated with the equilibriums of oligopoly market,
perfectly competitive market, and perfectly competitive market with the nodal prices
of oligopoly market, respectively. It should be noted that in this section, the
transmission constraints are ignored and nodal prices are equal. Based on (6.6) and
(6.7), the generation of GenCo i under perfectly competitive market (qpi ) can be
determined by
λp bi θpi max þ θpi min

qpi ¼ ð6:8Þ
ai
where θpi max and θpi min are Lagrange multipliers associated with upper and lower
production constrains of GenCo i under perfectly completive market, respectively.
The generation of GenCo under oligopoly market (qei ) can be written as
λe bi θei max þ θei min

qei ¼ ð6:9Þ
ð ai þ α Þ
where θei min and θei max are Lagrange multipliers associated with lower and upper
production limits of GenCo i under oligopoly market, respectively. The generation
of GenCo under perfectly competitive market with the nodal prices of oligopoly
market can be also written as
pðeÞ max pðeÞ min

pðeÞ λ e bi θ i þ θi
qi ¼ ð6:10Þ
ai
Fig. 6.1 Capacity p

withholding indices
Strategic offer Marginal cost

curve
le
Dldistortion
lp Demand curve
Dqidistortion
e
qei q
p qp(p )i q
i

where θi and θi are Lagrange multipliers associated with upper and
lower production constraints of GenCo i under perfectly competitive market with the
nodal prices of oligopoly market, respectively.
Aforementioned results were used to formulate the capacity withholding index
(Δqwithheld
i ), capacity distortion index (Δqdistortion
i ), and price distortion index
distortion
(Δλ ). These indices were defined as the following:
pðeÞ
Δqwithheld
i ¼ qi qei ð6:11Þ
Δqdistortion
i ¼ qpi qei ð6:12Þ
Δλdistortion ¼ λe λp ð6:13Þ
ΔQdistortion
DWI ¼ ð6:14Þ
ΔQwithheld
As shown in Fig. 6.1, λ p is the competitive market clearing price and qpi is the
GenCo i’s competitive generation. The output of GenCo i decreases to qei , and the
pðeÞ
market clearing price increases to λe due to the capacity withholding. qi is the
competitive output which would be generated at λ by GenCo i. In (6.14),
e
P
N P
N
ΔQwithheld ¼ Δqwithheld
i and ΔQdistortion ¼ Δqdistortion
i . The capacity withholding
i¼1 i¼1
indices are reported in Fig. 6.1. DWI ranges from 0 to 1 as it analyzes the market
power of GenCos from the perspective of their potential ability for capacity with-
holding. DWI indicates a decrement in the ability of GenCos for capacity withhold-
ing when it approaches 1. In contrast, when this index is lower, it is more likely that
producers have a high potential ability for capacity withholding. It can be seen that
when market power is low, the GenCos will have less incentive to raise the market
pðeÞ
price. It means that λe, qi , and Δqwithheld
i approach λ p, qpi, and Δqdistortion
i ,
respectively. It means that DWI approaches 1, as a result of the changes in both
Δqwithheld
i and Δqdistortion
i .
Substituting (6.10) and (6.9) into (6.11) yields
α e Δθwithheld
Δqwithheld ¼ q þ i ð6:15aÞ
i αi i ai
Δθwithheld
i ¼ θi þ θei min θi θei max ð6:15bÞ
It can be seen that Δqwithheld

i is influenced by the slope of aggregate demand
function and slope of marginal cost functions of GenCos (α/ai ratio). Therefore, a
GenCo with small slope of marginal cost function has a high potential ability to
withhold the capacity. Substituting (6.8) and (6.9) into (6.12) produces
α e Δθdistortion
Δqdistortion
i ¼ qi ΔQdistortion þ i ð6:16aÞ
ai ai
X
N
Δqdistortion
i ¼ ΔQdistortion ð6:16bÞ
i¼1
Δθdistortion
i ¼ θpi max þ θei min θpi min θei max ð6:16cÞ
It can be seen that the value of capacity distortion index may be negative or
positive. It depends on slope of the marginal cost functions of GenCos and the slope
of aggregate demand function. It is to be mentioned that α/ai ratio is significant for
analyzing the ability of GenCos for capacity withholding. For GenCos with large
ratio of α/ai, capacity distortion index may be positive. This shows that the gener-
ation of these GenCos has been decreased with respect to the perfectly competitive
market. It means that these GenCos indirectly can control the market price and make
the GenCos with small α/ai ratio become the marginal producers by capacity
withholding. It may cause the producers with small α/ai to produce more than their
perfectly competitive generations (qpi), and for them qei > qpi (Δqdistortion
i < 0) can be
achieved.
Δθwithheld Δθdistortion
The terms i ai and i ai represent the effect of the capacity limits of GenCos
on capacity withholding. Then, we would have
8
pðeÞ
>
> when qmin qi qmax
<0 i i
pðeÞ
Δθi
withheld
¼ < 0 when qi ¼ qi max ð6:17Þ
>
>
: pðeÞ
> 0 when qi ¼ qmin i
8
>
<0 when qi qpi qmax
min
i
Δθdistortion ¼ < 0 when qpi ¼ qmax ð6:18Þ
i
>
:
i
> 0 when qpi ¼ qmin
i
pðeÞ
It should be noted that according to (6.8), (6.9), and (6.10), qei qpi qi can be
pðeÞ
obtained. It means that if qi is between qmin i and qmax
i , qei and qpi are also between
qmin
i and qmax
i .
By using (6.2), Δπ distortion can be obtained by
Δλdistortion ¼ αY e þ β ðαY p þ βÞ ¼ αΔQdistortion ð6:19aÞ

X
N
Ye ¼ qei ¼ Qe ð6:19bÞ
i¼1
X
N
Yp ¼ qpi ¼ Qp ð6:19cÞ
i¼1
Ideally, with the increase of the total capacity distortion (ΔQdistortion), the
expected outcome would be the increment in Δqdistortion
i . (6.19) represents that
Δqdistortion
i is affected by the aggregate demand function’s slope. It indicates that in
a market with a large slope of demand function, producers will have more potential
ability to increase the market price.
By summing capacity distortions of all GenCos in (6.16), ΔQdistortion can be
obtained by
N
P
α Δθdistortion
ai qei þ i
ai
i¼1
ΔQdistortion ¼ ð6:20Þ
P
N
α
1þ ai
i¼1
Based on (6.20), DWI can be determined by

0 1
0 1 PN
Δθdistortion
α e
B a qi þ C
i
1
ΔQdistortion B CBi¼1 i
ai
C
DWI ¼ ¼@ PN AB N
α @P C ð6:21Þ
ΔQwithheld 1þ withheld A
α e Δθi
i¼1
ai q
ai i þ ai
i¼1
DWI can be obtained by
DWI ¼ DWIuc CF ð6:22aÞ

1
DWIuc ¼ ð6:22bÞ
PN
α
1þ ai
i¼1
N
P
α Δθdistortion
a i qi þ
e i
ai
i¼1
CF ¼ N ð6:22cÞ
P α Δθwithheld
ai qei þ i
ai
i¼1
The superscript uc shows the unconstrained production. It can be seen that in the
case of limited production, this index is composed of two parts. The first part is
unconstrained DWI (DWIuc) which is independent of the market model. The second
part is constraint factor of DWI (CF) which indicates the decrease of the potential
ability of GenCos for capacity withholding due to the capacity limits. If no GenCo is
pðπ e Þ
binding at its generation constraints (i.e., qmin i < qi < qmax
i ), that means
Δθiwithheld
¼ Δθidistortion
¼ 0 and CF ¼ 1, and then DWI ¼ DWIuc can be obtained.
6.2.3 Case Study
We present the case study of GenCos’ capacity withholding by using a three GenCo
power system. The GenCos’ cost parameters are shown in Table 6.1. The aggregate
demand parameters are α ¼ 2 $/(MW2) and β ¼ 90 $/MW. We have used GAMS
software and CONOPT solver to solve the optimization problems of the case study.
The following simulation cases are carried out:
Case A Perfect competition
Case B Cournot-type competition
Case C Cournot-type competition with decrease in demand elasticity
Case D Cournot-type competition with decrease in demand elasticity and capacity
of GenCos
The simulation results of cases A and B are listed in Table 6.2. It is assumed that
pðeÞ
the generation upper limits of GenCos are greater than qi . Therefore, the produc-
tion constraints can be disregarded in the analyzing of capacity withholding indices.
These indices are listed in Table 6.3. It can be observed that the capacity withholding
of GenCo 1 is higher than the others due to its small slope of the marginal cost.
To analyze the impacts of capacity limits of GenCos and demand elasticity on
capacity withholding, two cases C and D are performed. In these cases, GenCos use
Cournot-type competition and the demand function’s slope is decreased from α ¼ 2
Table 6.1 Cost parameters

GenCo no. Cost parameter ai ($/MW2) Cost parameter bi ($/MW) qmin
i (GW) qmax
i (GW)
1 1 12 0.3 30
2 1.5 10 0.8 25
3 2 8 1 20
Table 6.2 Simulation results for cases A, B, C1, C2, C3, and C4
pðeÞ
Case GenCo no. qei (GW) qpi (GW) qi (GW) λ ($/MW)
A 1 – 13.375 – 25.375
2 – 10.250 –
3 – 8.687 –
B 1 9.11 – 27.330 39.34
2 8.38 – 19.553
3 7.83 – 15.660
Table 6.3 Capacity withholding indices for case B

case GenCo no. Δqwithheld
i (GW) Δqdistortion
i (GW) Δλdistortion ($/MW) DWI
B 1 18.220 4.262 13.96 0.187
2 11.173 1.868
3 7.830 0.851
5
4
3
DWI
2
1
0
α=1. α=1. α=1. α=0. α=0. α=0. α=0.
α=2 α=1
7 5 3 8 5 4 3
DWI(Case C) 0.187 0.214 0.235 0.262 0.315 0.365 0.488 0.532 0.606
DWI(Case D) 1 0.816 0.672 0.536 0.339 0.365 0.488 0.532 0.606
CF(Case D) 5.34 3.82 2.85 2.04 1.07 1 1 1 1
Fig. 6.2 Illustration of the simulation results for cases C and D
$/(MW2) to α ¼ 0.3 $/(MW2) while keeping the demand function’s intercept

constant. It is also assumed that the generation upper constraints of GenCos 1, 2,
and 3 in case D are decreased to 13 GW, 10 GW, and 9 GW, respectively. It can be
observed in Fig. 6.2 that in case C and in the considered range of incremental
demand elasticity, DWI is increased from 0.187 to 0.606. In case D, DWI first
decreases from 1 to 0.339 with the decrease of the slope of demand function from
α ¼ 2 $/(MW2) to α ¼ 1 $/(MW2). Then, with further increase of demand elasticity,
it increases from 0.339 to 0.606. The coefficient accounts for the marginal impacts of
capacity limits on DWI. It can be seen that DWI will be decreased if CF is greater
than 1. The opposite changing trends of DWI in cases C and D indicate that the
potential ability of GenCos for capacity withholding is increased up to the point
where α is equal to 0.8 $/(MW2). It means that, with the increase of demand
elasticity, the expected result in terms of DWI would be a decreasing pattern along
with the releasing capacities of generation until the minimum value of CF (CF ¼ 1)
is reached. The value 1 of CF indicates that the values of DWI in cases C and D are
pðeÞ
equal and for all GenCos (qi < qmax i ).
6.3 Capacity Withholding Analyzing

in Transmission-Constrained Market
In the electricity markets, the potential ability of GenCos for capacity withholding
increases during the transmission congestion [3, 4, 15, 18, 25, 28, 29]. In Sect. 6.2, a
method for the capacity withholding analyzing of GenCos in electricity markets
based on some comparison indices has been presented. These indices can be used for
the capacity withholding assessment of GenCos. However, the impacts of transmis-
sion network congestion have been not considered in them, and they cannot be used
for nodal capacity withholding analyzing during the transmission congestion. This
section presents the development of the aforementioned method and offers some
contributions in comparison with the previous section.
1. Market model is extended to evaluate the transmission limits and other parame-
ters that affect GenCos’ capacity withholding.
2. Nodal capacity distortion index, nodal price distortion index, and nodal capacity
withholding index are formulated to assess the effect of transmission network
congestion on GenCos’ capacity withholding. In addition, DWI is reformulated to
include the transmission network limits.
3. In order to evaluate the potential ability of GenCos for exerting capacity with-
holding by considering the transmission congestion, a new index, called nodal
withholding-supply ratio (NWSR), is defined in this section.
6.3.1 Power Market Modeling
The main goal of this subsection is to model the capacity withholding of GenCos
during transmission congestion. This will be done by modeling a game theory-based
formulation. The Cournot model is consider to model the strategic behavior of
generators in the presence of demand elasticity. Then, the model of Cournot-Nash
equilibrium (CNE) is applied to calculate the market equilibrium considering the
GenCos’ capacity withholding. For the sake of simplicity, the market equilibrium
model is first formulated with any constraints; then it is extended for the market with
transmission congestion.
Suppose that there are I GenCos in the power market, and each of them has a
generation unit. The cost function of GenCo at bus i is determined as follows:
ci ¼ 1=2ai q2i þ bi qi : ð6:23Þ
The consumption at bus j is supposed as an inverse demand function:
λ j ¼ α j y j þ β j : ð6:24Þ
If there is no transmission congestion, the nodal price at bus n, λn and the price at
the reference bus λ are equal. By ignoring the transmission losses, the generation of
all GenCos and the aggregate demand are the same as described in (6.25)
X X X X
qi ¼ y j ¼ λ 1=α j þ β j =α j ¼ λ=αequal þ βequal =αequal ð6:25Þ
i2I j2J j2J j2J
where αequal and βequal are equivalent parameters of the aggregate demand. Based on
(6.25), the aggregate demand function can be expressed as an inverse demand
function as follows:
X
λ ¼ αequal y j þ βequal : ð6:26Þ
j2J
6.3.1.1 Market Equilibrium Without Considering Generation

and Transmission Limits
Each GenCo in the market competes with the others to maximize its own profit by
choosing its generation. The maximization problem of a GenCo at bus i can be
written as
max Ωi ¼ λi qi ci : ð6:27Þ
ISO also seeks to maximize social welfare. GenCos in a perfectly competitive

market cannot affect the market price, and there is no strategic behaviors. However,
the market price can be influenced by the capacity withholding of GenCos in an
oligopoly market. Then, social welfare can be decreased compared with that of
perfectly competitive market. In an oligopoly market with Cournot model, the of
market equilibrium problem can be described as a nonlinear programming (NLP)
problem as follows:
!
Xh i X X
max 1=2α j y2j þ β jy j 1=2ai qi þ bi qi
2
1=2αequal 2
qi
j2J i2I i2I
ð6:28Þ
X X
s:t: qi y j ¼ 0: ð6:29Þ
i2I j2J
Equation (6.28) shows the decreased social welfare in Cournot model due to the
GenCos’ capacity withholding and includes three terms:
• Demands’ cost
• GenCos’ cost
• Reduction in social welfare due to the capacity withholding of GenCos
The first two terms explain the social welfare in a perfectly competitive market.
By involving the third term, the market is cleared at the CNE point; otherwise, the
market is cleared at the perfect competition point. In fact, the third term explains the
effect of oligopolistic competition on the social welfare of perfectly competitive
market. It is to be considered that in the problem (6.28 and 6.29), the optimization
problems of each GenCo is implicit. By solving this single optimization problem, the
generations of all GenCos are determined [28, 29].
6.3.1.2 Market Equilibrium Model by Considering Generation

and Transmission Limits
Given the strategic decisions of GenCos, the ISO would determine the
corresponding generation of GenCos, the consumption of loads, and each node’s
price by solving the market clearing problem based on DC power flow. The ISO’s
optimization problem in Cournot model can be written in the vector form as
j k j k j k
max 1=2yT αy þ yT β 1=2qT aq þ qT b 1=2qT αequal q ð6:30Þ
s:t: eTI q eTJ y ¼ 0 ð6:31Þ

F T q y F ð6:32Þ
qmin q qmax ð6:33Þ
ymin y ymax : ð6:34Þ
in which
q=y Vector of generation/load, q ¼ ½q1 , q2 , . . . , qi , . . . , qI T ðMWÞ,

T
y ¼ y1 , y2 , . . . , y j , . . . , yJ ðMWÞ
qe =qp Vector of generation in oligopoly/perfectly competitive market (MW)
qpðeÞ Vector of generation in perfectly competitive market with the nodal
prices of oligopoly market (MW)
a=b Vector of slope/intercept of GenCos’ marginal cost function, a ¼
½a1 , a2 , . . . , ai , . . . , aI T ($/MW2), b ¼ ½b1 , b2 , . . . , bi , . . . , bI T ($/MW)
α=β Vector of slope/intercept of inverse demand functions, α ¼
T T
α1 , α2 , . . . , α j , . . . , αJ ($/MW2), β ¼ β1 , β2 , . . . , β j , . . . , βJ
($/MW)
a/α Diagonal matrix of the vector a=α
pmax =pmin Vector of maximum/minimum generation capacity constraints of
GenCos (MW)
ymax =ymin Vector of maximum/minimum demand capacity constraints of
consumers (MW)
F Vector of power transmission line flow constraints, F ¼
½F 1 , F 2 , . . . , F m , . . . , F M T (MW)
T Power transfer distribution factors (PTDFs) matrix
T TI =T TJ Rows of transposed T matrix for generation/consumption buses
Tmn Element of T matrix
eI =eJ Identity vector with I/J dimension
The key point of Cournot model is to solve the optimization problem of the social
welfare subject to generation and consumption balance, transmission network limits,
and capacity limits of GenCos. The Lagrangian function of the problem (6.30, 6.31,
6.32, 6.33, and 6.34) can be expressed as
h i h i
L q, y, λ, μþ , μ , θmax , θmin , ωmax , ωmin ¼ 1=2yT αy þ yT β 1=2qT aq þ qT b
h i h i h i
1=2qT αequal q þ λ eTI q eTJ y μTþ T q y F þ μT T q y þ F

θT max q qmax þ θT min q qmin ωT max y ymax þ ωT min y ymin :
ð6:35Þ
The Karush-Kuhn-Tucker (KKT) constraints that are significant for optimality of

(6.35) can be obtained by

∂L=∂q ¼ aq b αequal q þ λIq T TI μþ μ θmax θmin ¼ 0

∂L=∂y ¼ αy þ β λIy þ T TJ μþ μ ωmax ωmin ¼ 0
eTI q eTJ y ¼ 0

FT qy F
qmin q qmax
ymin y ymax
h i
μþ T q y F ¼ 0
h i
μ T q y þ F ¼ 0

θmax q qmax ¼ 0

θmin qmin q ¼ 0

ωmax y ymax ¼ 0

ωmin ymin y ¼ 0
μþ , μ , θmax , θmax , ωmax , ωmin 0:
ð6:36Þ
where
μþ =μ Vector of Lagrangian multiplier associated with positive/
negative direction of transmission line flow constraints ($/MW)
μeþ =μe Vector of Lagrangian multiplier in oligopoly market associated
with positive/negative direction of transmission line flow
constraints ($/MW)
μpþ =μp Vector of Lagrangian multiplier in perfectly competitive market
associated with positive/negative direction of transmission line
flow constraints ($/MW)
θmax =θmin Vector of Lagrangian multiplier associated with maximum/
minimum generation capacity constraints of GenCos ($/MW)
θe max =θe min Vector of Lagrangian multiplier in oligopoly market associated
with maximum/minimum generation capacity constraints of
GenCos ($/MW)
θp max =θp min Vector of Lagrangian multiplier in perfectly competitive market
associated with maximum/minimum generation capacity
constraints of GenCos ($/MW)
θpðeÞ max =θpðeÞ min Vector of Lagrangian multiplier in perfectly competitive market
with the nodal prices of oligopoly market associated with
maximum/minimum generation capacity constraints of GenCos

($/MW)
ωmax =ωmin Vector of Lagrangian multiplier associated with maximum/
minimum demand capacity constraints of consumers ($/MW)
0 Properly dimensioned zero vector
The first two terms of (6.36) show the derivatives of (6.35) with respect to
generation and consumption, respectively. Then, the generation of GenCo at bus
i and the consumption of load at bus j can be written as the following equations,
respectively:
P P
λ bi μþ
m T mi þ μ
m T mi θ i
max
þ θmin
i
m2M m2M
qi ¼ ð6:37Þ
ðai þ αequal Þ
P þ P
λ þ β j þ μm T mj þ μm T mj ωmax
j þ ωmin
j
m2M m2M
yj ¼ : ð6:38Þ
αj
The nodal price at bus n can be obtained by

X X
λn ¼ λ μþ
m T mn þ μ
m T mn : ð6:39Þ
m2M m2M
If there is no transmission congestion, the Lagrangian multipliers associated with

the transmission line flow constraints are zero. Then, according to (6.39), λn ¼ λ can
be achieved. However, when there is transmission congestion, the Lagrangian
multiplier at each node shows the nodal price λn.
6.3.2 Capacity Withholding Indices
Nodal capacity withholding index and NWSR can be used for analyzing GenCos’
market power. They compare the generations at oligopoly market and perfectly
competitive market with the nodal prices of oligopoly market. Nodal price distortion
indices and nodal capacity distortion can be used to analyze the distortions of the
oligopoly market from perfectly competitive market.
pðeÞ
Δqwithheld
i ¼ qi qei : ð6:40Þ
Δqdistortion
i ¼ qpi qei : ð6:41Þ
Δλdistortion
n ¼ λen λpn : ð6:42Þ
NWSRi ¼ Δqwithheld
i =qei : ð6:43Þ
According to (6.37) and (6.39), qei can be written as
λei bi θei max þ θei min

qei ¼ ð6:44Þ
ðai þ αequal Þ
where λei can be expressed as

X X
λei ¼ λe μeþ
m T mi þ μe
m T mi : ð6:45Þ
m2M m2M
The generation of each GenCo in perfectly competitive market can be determined

by neglecting the third term in (6.30). Therefore, qpi could be written as
λpi bi θpi max þ θpi min

qpi ¼ ð6:46Þ
ai
where λpi can be obtained by

X X
λpi ¼ λp μpþ
m T mi þ μp
m T mi : ð6:47Þ
m2M m2M
pðeÞ
Based on (6.46), qi can be determined by

pðeÞ λei bi θi þ θi
qi ¼ : ð6:48Þ
ai
Because of GenCos’ capacity withholding, the nodal prices and the generations of
GenCos deviate from their perfectly competitive levels to oligopoly. Then, the nodal
capacity distortion and the nodal price distortion can be formulated as Δqdistortion
i ¼
qpi qei and Δλdistortion
n ¼ λen λpn , respectively. It is to be noted that the withheld
capacity for each GenCo is not equal to qpi qei ; rather, it is formulated as
pðeÞ
Δqwithheld
i ¼ qi qei , indicating that the generation of GenCo at bus i with nodal
pðeÞ
price λei would be qi but, in fact, it was reduced to qei because of the capacity
pðeÞ
withholding of GenCo in the oligopoly market. It should be noted that qi is the
competitive generation, which could be generated at λi by GenCo at bus i. According
e
to (6.46) and (6.47), if GenCo at bus i does not hit its generation capacity constraints,
pðeÞ
qi gets larger than qpi ; then, Δqwithheld
i > Δqdistortion
i can be obtained.
By substituting (6.44) and (6.48) into (6.40), Δqwithheld
i can be determined by

Δqwithheld
i ¼ 1=ai αequal qei Δθwithheld
i ð6:49Þ

where Δθwithheld
i ¼ θi þ θei min θi θei max .
According to (6.49), NWSRi of GenCo at bus i can be obtained by

NWSRi ¼ 1=ai αequal Δθwithheld
i =qei : ð6:50Þ
Based on (6.50), it is possible to analyze the different parameters that influence

the capacity withholding of GenCos. If the GenCo at bus i does not exceed its
pðeÞ pðeÞ max
generation capacity limits (qmini < qei , qi < qmax
i ), that means θi ¼ θei min ¼
pðeÞ min
θi ¼ θi
e max
¼ 0, and then NWSRi would be equal to α equal
/ai ratio. Then, in
this case, the NWSRi of GenCo depends on the slope of its marginal cost function
and the slope of the aggregate inverse demand function. If the GenCo exceeds its
pðeÞ
maximum generation capacity limit (qi ¼ qmax i ), then Δθwithheld
i is positive.
According to (6.50), NWSRi decreases in a way depending on the values of
Δθwithheld
i and qei . The effect of network limits on the ability of GenCo for capacity
withholding is considered implicitly by qei. Also, NWSRi can be rewritten by (6.44)
and (6.50) with the below equation:
!
αequal Δθwithheld ai þ αequal
NWSRi ¼ e i
¼
ai ai λi bi θei max þ θei min
0 1
B equal C ð6:51Þ
Bα Δθwithheld ai þ αequal C
B !C
B ai C:
i
@ P P e min A
ai λ e μeþ
m T mi þ μe
m T mi bi θ i
e max
þ θi
m2M m2M
It can be seen that in addition to αequal/ai ratio, NWSRi could be influenced by μeþ m
,μe
m , Tmi, Δθwithheld
i , θei max , θei min , and bi in a transmission-constrained market.
In the same way, Δqdistortion
i can be obtained by using (6.41, 6.42, 6.44, 6.45, 6.46,
and 6.47) as follows:

Δqdistortion
i ¼ 1=ai αequal qei Δλdistortion
i Δθdistortion
i ð6:52Þ
where Δθdistortion
i ¼ θpi max þ θei min θpi min θei max .
Unlike Δqwithheld
i , which is positive, Δqdistortion
i can be negative or positive. A
positive (negative) value of Δqdistortion
i implies that the generation of GenCo has been
decreased (increased) with respect to the perfect competition model. The sign of
capacity distortion index depends on the slope of the marginal cost function of
GenCo i. For high-cost GenCos, capacity distortion index can be negative. It means
that these GenCos don’t have high market shares; however, because of the capacity
withholding of low-cost GenCos, which have high ability for the capacity
withholding, their market shares may be increased. Then, the high-cost GenCos in
oligopoly market may generate more than in perfectly competitive market, and
negative Δqdistortion
i may be achieved.
Based on (6.42), (6.45), and (6.47), Δλdistortion
i at the generator connected buses
can be obtained by
X
Δλdistortion
i ¼ Δλdistortion Δμdistortion
m T mi ð6:53Þ
m2M
where Δλdistortion ¼ λe λ p and Δμdistrotion

m ¼ μp
m þ μm μm μm .
eþ pþ e
According to (6.52), (6.53), and (6.26), Δqi distortion

of GenCo at bus i can be
obtained by
!
X X
Δqdistortion
i ¼ 1=ai αequal qei α equal
Δqi
distortion
Δθi
distortion
Δμm
distortion
T mi :
i2I m2M
ð6:54Þ
Another index to assess the capacity withholding may be defined by DWI, which
can be obtained by
X X
DWI ¼ Δqdistortion
i = Δqwithheld
i : ð6:55Þ
i2I i2I
As mentioned before, DWI gives the information about the potential ability of
GenCos in power markets for capacity withholding. Equation (6.56) provides
reformulation of this index by considering network and GenCos limits:
0 1
1
DWI ¼ @1 þ P αequal =a A
i
i2I
0P P P 1
αequal qei =ai Δθdistortion
i =ai Δμdistortion
m T mi =ai
Bi 2 I i2I m 2 M, i 2 I C
@ P equal e P A ¼ DWIUC CF
α qi =ai Δθwithheld
i =a i
i2I i2I
ð6:56Þ
0 1
1
where DWIUC ¼ @1 þ P αequal =a A and
i
i2I
0 P equal e P distortion P 1
α qi =ai Δθi =ai Δμdistortion
m T mi =ai
Bi 2 I i2I m 2 M, i 2 I C
CF ¼ @ P equal e P A.
α qi =ai Δθwithheld
i =a i
i2I i2I
The constraint factor (CF) shows the effect of the GenCos and transmission
network constraints on the potential ability in the market for capacity withholding.
If the GenCos and the transmission lines do not hit their constraints, therefore
Δθdistortion
i , Δθwithheld
i , and Δμdistortion
m are equal to zero. Then, CF is equal to 1 and
DWI is equal to the unconstrained DWI (DWIUC). According to (6.56), DWIUC
depends on αequal/ai ratio. CF becomes greater than 1, when the GenCos hit their
constraints. Then, DWI becomes greater than DWIUC. It means that the potential
ability in the market for capacity withholding is decreased because of the capacity
limits of GenCos. It can be seen that when the transmission line flow constraints are
exceeded, CF becomes lower than 1. The expected outcome with the increase of
transmission congestion, in terms of the capacity withholding of GenCos, could be
an increasing pattern along with the decreasing DWI.
6.3.3 Case Study
A modified IEEE 30-bus power system is studied to conceptually evaluate the

approach used for analyzing the transmission limits impacts via the defined indices.
The load parameters and transmission network data, power transfer distribution
factors (PTDFs), and topology are based on MatPower 5.1 Toolbox. There are six
generators in the power system, and it is supposed that each GenCo has only one
generator. The reference bus is bus 1. The GenCos’ parameters are shown in
Table 6.4. The inverse demand function of each load is λj ¼ αjyj + 45. The
slope of each inverse demand function is selected in a way that λj ¼ 35 $/MW for the
value of yj in MatPower 5.1 Toolbox. The aggregate demand’s parameters in (3) are
αequal ¼ 0.03528 $/MW2 and βequal ¼ 45 $/MW.
To analyze the capacity withholding indices, six simulation cases from A to F are
considered. To use the capacity withholding indices, it is necessary to calculate λpi, λei
pðeÞ
, qi , qpi, and qei for simulation cases, as listed in Table 6.5. The indices are listed in
Table 6.6.
In case A, the line flow constraints of transmission network are the same as the
data determined in MatPower 5.1 Toolbox. For transmission congestion modeling,
the flow constraints of transmission lines 2, 3, 6, 8, and 9 are decreased to 60 MW,
13 MW, 13 MW, 14 MW, and 15 MW in case B, respectively. For all buses in
Table 6.4 GenCos parameters for IEEE 30-bus test system

Bus number of GenCo ai ($/MW2) bi($/MW) pmin(MW) pmax(MW)
1 0.075 20 50 200
2 0.35 17.5 20 80
5 1.25 10 15 50
8 0.1668 32.5 10 35
11 0.5 30 10 30
13 0.5 30 12 40
Table 6.5 Simulation results of modified IEEE 30-bus test system for cases A to F
pðeÞ
Case Bus number of GenCo λpi ($/MW) λei ($/MW) qi (MW) qpi (MW) qei (MW)
A 1 34.559 36.028 200 193.549 145.328
2 34.559 36.028 52.937 48.740 48.088
5 34.559 36.028 20.822 19.644 20.250
8 34.559 36.028 21.151 12.291 17.456
11 34.559 36.028 12.056 10 11.261
13 34.559 36.028 12.056 12 12
B 1 27.299 29.403 125.38 97.323 85.263
2 25.189 27.648 28.994 21.968 26.338
5 38.355 38.416 22.733 22.684 22.108
8 39.210 39.311 35 35 33.702
11 38.590 38.795 17.591 17.181 16.432
13 36.990 37.464 14.930 13.981 13.945
C 1 26.106 28.842 117.886 81.415 69.116
2 24.128 27.255 27.871 20 24.211
5 35.921 36.404 21.123 20.737 20.265
8 37.271 37.795 31.749 28.604 24.1
11 36.690 37.330 14.66 13.381 13.256
13 35.190 36.127 12.254 12 12
D 1 29.742 30.996 146.620 129.896 118.694
2 27.545 29.026 32.931 28.700 31.350
5 41.540 41.558 25.246 25.232 24.895
8 42.141 42.117 35 35 35
11 41.496 41.538 23.078 22.993 22.291
13 39.831 40.044 20.090 19.662 19.405
E 1 30.724 32.067 80.444 71.492 65.125
2 29.242 30.795 37.986 33.548 34.508
5 38.553 38.799 23.039 22.842 22.406
8 39.088 39.243 35 35 33.365
11 38.653 38.869 17.740 17.306 16.570
13 37.529 37.905 15.812 15.059 14.769
F 1 23.422 26.340 151.196 112.289 102.823
2 20.575 24.020 20 20 20
5 38.355 38.416 22.733 22.684 22.108
8 39.490 39.431 35 35 34.300
11 38.654 38.750 17.502 17.308 16.348
13 36.496 36.992 13.984 12.992 13.063
case A, λpi , λei , and Δλdistortion

i are 34.559 $/MW, 36.028 $/MW, and 1.469 $/MW,
respectively. The nodal prices in case B are not the same because of the transmission
network congestion. In this case, the flow constraints of transmission lines 6 and
9 are limited in positive directions. It is to be noted that the positive directions
of transmission lines 6 and 9 are from bus 2 to bus 6 and from bus 6 to
Table 6.6 Capacity withholding indices of modified IEEE 30-bus test system for cases A to F
Δqwithheld
i Δqdistortion
i
Bus number of Δλdistortion
i
Case GenCo ($/MW) (MW) (MW) NWSRi DWI
A 1 1.469 54.672 48.221 0.3761 0.6472
2 1.469 4.849 0.652 0.1008
5 1.469 0.572 0.606 0.0282
8 1.469 3.695 5.165 0.2116
11 1.469 0.795 1.261 0.0705
13 1.469 0.056 0 0.0046
B 1 2.104 40.117 12.060 0.4705 0.2209
2 2.459 2.656 4.370 0.1008
5 0.061 0.625 0.576 0.0282
8 0.101 1.298 1.298 0.0385
11 0.205 1.159 0.749 0.0705
13 0.474 0.985 0.036 0.0705
C 1 2.736 48.77 12.299 0.7056 0.2106
2 3.128 3.66 4.211 0.1511
5 0.483 0.858 0.472 0.0423
8 0.524 7.649 4.504 0.3173
11 0.639 1.404 0.125 0.1059
13 0.936 0.254 0 0.0211
D 1 1.254 27.927 11.202 0.2353 0.3143
2 1.481 1.581 2.650 0.0504
5 0.018 0.352 0.337 0.0141
8 0.025 0 0 0
11 0.041 0.788 0.703 0.0353
13 0.213 0.686 0.257 0.0353
E 1 1.343 15.319 6.366 0.2352 0.3653
2 1.554 3.479 0.960 0.1008
5 0.246 0.633 0.436 0.0282
8 0.155 1.635 1.635 0.0490
11 0.216 1.170 0.736 0.0705
13 0.376 1.043 0.290 0.0705
F 1 2.918 48.373 9.466 0.4705 0.2246
2 3.446 0 0 0
5 0.061 0.625 0.575 0.0282
8 0.058 0.700 0.700 0.0204
11 0.096 1.154 0.961 0.0705
13 0.496 0.922 0.071 0.0705
bus 7, respectively. The flow constraints of transmission line 8 is limited in the

negative direction from bus 7 to bus 5. Because of the different locations in the
network, Δλdistortion
i decreases for some buses and increases for some others.
When evaluating cases A and B, it can be observed that because of the transmis-
sion network congestion, NWSRi for the GenCos at buses 1 and 13 increases to
0.4705 and to 0.0705, respectively. Moreover, the NWSRi of GenCo at bus 8 reduces
to 0.0385. In addition, the values of αequal/ai ratios for GenCos at buses 1, 8, and
13 are 0.4705, 0.0705, and 0.2116, respectively. Therefore, on the one hand,NWSRi
values of GenCos at buses 1 and 13 will become lower than their αequal/ai ratios
pðeÞ
when their constraints are bounded in case A (q1 ¼ qmax 1 ¼ 200 MW and qe13 ¼
q13 ¼ 12 MW). On the other hand, in case B, the generation capacity constraints of
min
pðeÞ
GenCos at buses 1 and 13 are not limited (q1 ¼ 125:38 MW and qe13 ¼
13:945 MW), and based on (6.50) and (6.51), NWSRi of GenCos grow up to the
points where NWSR1 ¼ αequal/a1 ¼ 0.4705 and NWSR13 ¼ αequal/a13 ¼ 0.0705. The
opposite outcomes are seen for the GenCo at bus 8. In case A, the generation
pðeÞ
capacity constraints of the GenCo are not limited (qmin 8 < qe8 , q8 < qmax
8 ) and
NWSR8 ¼ α equal
/a8 ¼ 0.2116. Due to the congestion of transmission network in
case B, the GenCo hits its upper generation capacity constraint
pðeÞ
(q8 ¼ qmax 8 ¼ 35 MW ), and NWSR8 reduces to 0.0385. The constant NWSRi
values of GenCos at buses 2, 5, and 11 in cases A and B show that the constraints
pðeÞ
of qi and qei for these GenCos are not limited. Then, their NWSRi for them are not
affected by the transmission network constraints.
Based on (6.56), unconstrained DWI of the system is 0.5122. In case A, DWI and
CF are 0.6472 and 1.2635, respectively. It means that the potential ability of GenCos
for capacity withholding is decreased because of the generation limits of GenCos
1, 11, and 13. Therefore, in case B, due to the congested lines 6, 8, and 9, CF reduces
to 0.4327 causing DWI to reduce to 0.2209. It implies that the GenCos’ ability for
market power is increased. Note that the reduced CF should be ascribed to the
transmission congestion.
The effect of demand elasticity on the GenCos’ capacity withholding is also
analyzed. Cases C and D are the same as case B except for αequal ¼ 0.05292 and
αequal ¼ 0.01764, respectively. These slopes are achieved from 50% increase to 50%
decrease in the inverse demand functions’ slopes of case B, respectively. As listed in
Table III, the indices show an improvement with the decrease of αequal. In addition,
in case D, DWI increases to 0.3143 and the NWSRi decreases.
Other conditions are studied for analyzing the effect of cost parameters on the
GenCos’ capacity withholding. Cases E and F are the same as case B except for
a1 ¼ 0.15 and b1 ¼ 15, respectively. By reducing of a1 in case E, NWSR1 reduces to
0.2352, showing that the GenCos’ potential ability reduces. However, NWSR8
increases to 0.0490; the growth of DWI to 0.3653 indicates that the GenCos’ ability
for market power decreases. Therefore, GenCo at bus 1 is more effective than GenCo
at bus 8, and a decrement in its capacity withholding can affect the market consid-
erably. The values of NWSR1 for GenCos at buses 1, 5, 8, 11, and 13 in case F are
the same as those in case B. Although, NWSR2 reduces to 0, leading DWI to
increase to 0.2246.
6.4 Capacity Withholding Mitigation and the Impact

of DGs on It
Market power and capacity withholding may be controlled by various programs,

which can be categorized in two classifications: the demand (consumer) side and the
supply (generator) side programs. Demand side programs are used to reduce the
consumption and increase the elasticity of demand. A significant subject in demand
side programs is interruptible load contracts. In fact, by increasing of the generations
of GenCos, specifically GenCos with low generation cost, the outcome in terms of
the capacity withholding of GenCos should be a decreasing trend along with the
decreasing market price. ISO by using the supply side programs motivates GenCos
to increase their generations up to the point where the market price is equal to their
marginal costs.
Distributed energy resources (DERs) can play an important role in capacity
withholding mitigation as supply side programs. For instance, by locating DGs in
congested areas of systems, capacity withholding of GenCos decreases. In fact, ISO
in hosting capacity analyzing can use the defined indices to select appropriate
programs for mitigation of capacity withholding as illustrated in Fig. 6.3. The left
and right panels in Fig. 6.3 are the market before and after capacity withholding
mitigation programs, respectively. Figure 6.3a, b is related to supply side programs
and Fig. 6.3c, d is based on demand side programs, and it is supposed that there is
one GenCo in the market. Based on Fig. 6.3a, the generation capacity constraint for
pðeÞ
GenCo i is greater than qi and DWI < 1 can be obtained. Therefore, this GenCo
can produce more than its oligopoly generation (qei ). As illustrated in Fig. 6.3b,
because of the eased decreased market power and increased generation, ΔQwithheld,
ΔQdistortion, and Δλdistortion are reduced. When evaluating Fig. 6.3a, b, it can be
observed that DWI is higher when a supply side program is used. Therefore, the
increased DWI should be ascribed to the mitigation of the capacity withholding. The
impacts of generation capacity limits on the GenCo i’s capacity withholding are
illustrated in Fig. 6.3c. In this figure, a deeper evaluation should be made to avoid
misunderstanding. Figure 6.3a, c is the same, except for considering the decreased
pðeÞ
maximum capacity constraint of GenCo i and setting qmax i ¼ qpi ¼ qi . Because of
the reduced maximum capacity constraint, ΔQ withheld
, ΔQdistortion
, and Δλ
distortion
are
decreased. Therefore, ΔQ withheld
and ΔQdistortion
are the same and then DWI equals
to 1. Then, the value 1 of DWI shows that the GenCo could not increase its
generation up to the perfect market output. Therefore, supply side programs cannot
provide a way to decrease capacity withholding, and demand side programs are
essential to improve the demand elasticity. An increase in demand elasticity is
coherent with lower level of capacity withholding and power consumption. The
demand curve with higher elasticity will be located in lower position with respect to
the original one, as illustrated in Fig. 6.3d. Because of the effect of price modifica-
tions on the optimal behavior of the generators, GenCo i will find it better to bid a
lower price to improve its market share. This is illustrated by the shift from the
strategic offer in Fig. 6.3c to the strategic offer in Fig. 6.3d, which is closer to the
marginal cost curve [30, 31].
Fig. 6.3 Assessment of

capacity withholding
mitigation programs.
(a) Unconstrained market.
(b) Effects of a supply side
program. (c) Constrained
market. (d) Effects of a
demand side program
In hosting capacity analysis, the maximum amount DERs that can be installed in
the power system is determined. In this study, the major concerns are technical
constraints such as adequacy and security limits of system. After this analysis, the
market operator or the ISO can evaluate the outcomes of the technical analysis from
market power and capacity withholding assessment. In other words, the ISO can use
the proposed global and nodal indices to measure the effects of DERs on capacity
withholding. In fact, ISO can use global index to analyze the effect of DERs on
capacity withholding in the whole system. In addition, by nodal indices the impacts
of DERs on capacity withholding at each bus of system can be analyzed. Then, the
ISO can use the optimum strategies to control the market power of GenCos. In
addition, ISO can propose new locations of DERs to decrease capacity withholding
in different buses of system by using nodal indices such as nodal price distortion
index and nodal capacity withholding index. The new locations should be analyzed
from technical constraints of system such as substations and feeder capacities. In
Fig. 6.4, the flowchart of this procedure is shown.
Fig. 6.4 Hosting capacity

analysis from capacity Start
withholding point of view
Finding the maximum amount of DERs in power system.
Calculating global and nodal indices in the presence of DERs.
The market is fun-

NO
damentally ineffi-
Is DWI cient and the ISO
acceptable? should use market
mitigation
programs.
Yes
Change the capacity
NO or location of DERs.
Are nodal
indices
NO
acceptable?
Yes Are nodal

Yes indices
acceptable?
Are Technical
Yes
constraints
exceeded?
End NO
6.5 Future Works
In this chapter, two categories of market power indices were presented to assess the
capacity withholding of GenCos in the presence of DERs. This indices are effective
and they can be used by the ISO besides the technical assessment of hosting
capacity. Future works are needed to evaluate the defined indices when maintenance
scheduling of generators and strategic transmission system operators (TransCo) are
considered.
References
1. Gan, D., & Bourcier, D. V. (2002). Locational market power screening and congestion
management: Experience and suggestion. IEEE Transactions on Power Systems, 17(1),
180–185.
2. Wang, P., Xiao, Y., & Ding, Y. (2004). Nodal market power assessment in electricity markets.
IEEE Transactions on Power Apparatus and Systems, 19(3), 1373–1379.
3. Bompard, E., Ma, Y. C., Napoli, R., & Jiang, C. W. (2006). Assessing the market power due to
the network constraints in competitive electricity markets. Electric Power Systems Research,
76, 953–961.
4. Bompard, E., Lu, W., & Napoli, R. (2006). Network constraint impacts on the competitive
electricity markets under supply-side strategic bidding. IEEE Transactions on Power Apparatus
and Systems, 21(1), 160–170.
5. Bompard, E., Ma, Y. C., Napoli, R., & Abrate, G. (2007). The demand elasticity impacts on the
strategic bidding behavior of the electricity producers. IEEE Transactions on Power Apparatus
and Systems, 22(1), 188–197.
6. Hesamzadeh, M. R., Biggar, D. R., & Hosseinzadeh, N. (2011). The TC-PSI indicator for
forecasting the potential ability for market power in wholesale electricity markets. Energy
Policy, 39, 5988–5998.
7. Lee, Y. Y., Hur, J., Baldick, R., & Pineda, S. (2011). New indices of market power in
transmission-constrained electricity markets. IEEE Transactions on Power Apparatus and
Systems, 26(2), 681–689.
8. Gabriel, S. A., Conejo, A. J., Fuller, J. D., Hobbs, B. F., & Ruiz, C. (2013). Complementarity
modeling in energy markets. New York: Springer Press.
9. Latorre, M. D. L., & Graville, S. (2003). The Stackelberg equilibrium applied to AC power
systems-a non-interior point algorithm. IEEE Transactions on Power Apparatus and Systems,
18(2), 611–618.
10. Chen, H., Wong, K. P., Chung, C. Y., & Nguyen, D. H. M. (2006). A coevolutionary approach
to analyzing supply function equilibrium model. IEEE Transactions on Power Apparatus and
Systems, 21(3), 1019–1028.
11. Hasan, E., & Galiana, F. D. (2008). Electricity markets cleared by merit order – Part II: Strategic
offer and market power. IEEE Transactions on Power Apparatus and Systems, 23(2), 372–379.
12. Ruiz, C., Conejo, A. J., & Bertand, R. G. (2008). Some analytical results pertaining to Cournot
models for short-term electricity markets. Electric Power Systems Research, 78, 1672–1678.
13. Bushnell, J. B. (2003). Looking for trouble competition policy in the US electricity industry.
Center for the Study of Energy Markets: Univ. California Energy Institute, California.
14. Baldick, R., Grant, R., & Kahn, E. (2004). Theory and application of linear supply function
equilibrium in electricity markets. Journal of Regulatory Economics, 25, 143–167.
15. Cunningham, L. B., Baldick, R., & Baughman, M. L. (2002). An empirical study of applied
game theory: Transmission constrained Cournot behavior. IEEE Transactions on Power Appa-
ratus and Systems, 17(1), 166–172.
16. Yang, Y., Zhang, Y., Li, F., & Chen, H. (2012). Computing all Nash equilibria of multiplayer
games in electricity markets by solving polynomial equations’. IEEE Transactions on Power
Apparatus and Systems, 27(1), 81–91.
17. Helman, U., & Hobbs, B. F. (2010). Large-scale market power modeling: Analyzing of the
U.S. eastern interconnection and regulatory applications. IEEE Transactions on Power Appa-
ratus and Systems, 25(3), 1434–1448.
18. Yu, C. W., Zhang, S. H., Wang, X., & Chung, T. S. (2010). Modeling and analysis of strategic
forward contracting in transmission constrained power markets. Electric Power Systems
Research, 354–361.
19. Xiaohong, G., Ho, Y. C., & Peypyne, D. L. (2001). Gaming and price spikes in electric power
markets. IEEE Transactions on Power Apparatus and Systems, 16(3), 402–408.
20. Tellidou, A. C., & Bakirtzis, A. G. (2007). Agent-based analysis of capacity withholding and
tacit collusion in electricity markets. IEEE Transactions on Power Apparatus and Systems, 22
(4), 1735–1742.
21. Hongyan, L. I., & Tesfatsion, L. (2009). Capacity withholding in restructured wholesale power
markets: An agent-based test bed study, Power Systems Conference and Exposition.
22. Mohtavipour, S. S., Haghifam, M. R., & Sheik-El-Eslami, M. K. (2011). Emergence of capacity
withholding: An agent based simulation of a double price cap electricity market. IET Genera-
tion Transmission and Distribution, 6(1), 69–78.
23. Chattopadhay, D. (2004). A game theoretic model for strategic maintenance and dispatch
decision. IEEE Transactions on Power Apparatus and Systems, 19(4), 2014–2021.
24. Wang, J., Shahidehpour, M., Li, Z., & Botterud, A. (2009). Strategic generation capacity
expansion planning with incomplete information. IEEE Transactions on Power Apparatus
and Systems, 24(2), 1002–1010.
25. Hesamzadeh, M. R., Hosseinzadeh, N., & Wolfs, P. J. (2010). Transmission system augmen-
tation based on the concepts of quantity withholding and monopoly rent for reducing market
power. IEEE Transactions on Power Apparatus and Systems, 25(1), 167–180.
26. Salarkheili, S., & Akbariforoud, A. (2012). Market power assessment in electricity markets:
Supply function equilibrium-based model. International Transactions on Electrical Energy
Systems, 23(4), 553–569.
27. Salarkheili, S., & SetayeshNazar. M. (2015). New indices of capacity withholding in power
markets. International Transaction on Electrical Energ Systems, 25(1), 180–196.
28. Salarkheili, S., & SetayeshNazar, M. (2016). Capacity withholding analysis in transmission-
constrained electricity markets. IET Generation Transmission and Distribution, 10(2),
487–495.
29. Salarkheili, S., & SetayeshNazar, M. (2017). Capacity withholding assessment in the presence
of integrated generation and transmission maintenance scheduling. IET Generation Transmis-
sion and Distribution, 11(16), 3903–3911.
30. Ameri, M., Rahimian, M., & Latifi, M. A. (2018). Capacity withholding constrained by
operational limits of generation under financial virtual divestiture in a day-ahead market.
IEEE Transactions on Power Apparatus and Systems, 33(1), 771–780.
31. Guo, H., Chen, Q., Xia, Q., & Kang, C. (2019). Market power mitigation clearing mechanism
based on constrained bidding capacities. IEEE Transactions on Power Systems, 34(6),
4817–4827.
Chapter 7
A Generalised Deterministic Approach
to Evaluate PV Hosting Capacity of LV
Distribution Networks Under Different
Operating Conditions
D. Chathurangi, U. Jayatunga, S. Perera, and A. Agalgaonkar
7.1 Introduction
Distributed generation (DG) in the form of renewable energy sources is gradually

becoming an integral part of many power systems around the world. Concurrently,
they are progressively replacing greenhouse gas-emitting technologies, accelerating
the transition towards an eco-friendly low-carbon environment.
Recent rapid growth of solar photovoltaics (PV) compared to other electricity
generation technologies is clearly evident. The total worldwide installed capacity of
solar PV generation has exceeded 505 GW in 2018 [1]. The year 2018 was a
landmark year for solar PV where 100 GW was installed worldwide for electricity
generation which is more than the cumulative sum of the capacities of other
generating technologies installed [1].
Challenges associated with solar PV are its intermittency and non-dispatchability,
rendering it hard to match supply and demand. Thus, there is a necessity for active
management of distribution networks which have not been previously designed to
D. Chathurangi (*)
Department of Electrical Engineering, University of Moratuwa, Moratuwa, Sri Lanka
Collaborated Research, School of Electrical, Computer & Telecommunications Engineering,
Faculty of Engineering and Information Science, University of Wollongong,
Wollongong, Australia
e-mail: dmcwl143@uowmail.edu.au
U. Jayatunga
Department of Electrical Engineering, University of Moratuwa, Moratuwa, Sri Lanka
e-mail: upuli@uom.lk
S. Perera · A. Agalgaonkar
School of Electrical, Computer & Telecommunications Engineering, Faculty of Engineering
and Information Science, University of Wollongong, Wollongong, Australia
e-mail: sarath@uow.edu.au; ashish@uow.edu.au

https://doi.org/10.1007/978-3-030-40029-3_7
150 D. Chathurangi et al.
accommodate DG. Bidirectional power flows are in reality with increased solar PV
penetration levels and can degrade the performance of LV distribution networks
resulting in a number of power quality and other related issues (such as overvoltage
limit violations, voltage unbalance, overloading of transformers and feeders, exces-
sive line losses, high harmonic distortion levels exceeding the stipulated limits and
protection mal-operation [2]). Degree of the occurrence of these problems increases
with the maximum permissible penetration level, especially when the hosting capac-
ity exceeds acceptable limits [3, 4].
Key negative impacts associated with high solar PV penetration levels may be
overcome by network reinforcements such as asset upgrades (conductors and trans-
formers) and advanced monitoring and protection schemes. However, such rein-
forcements are not economically feasible in general. Given the potential obstacles, it
is desirable to develop systematic approaches for new solar PV deployment from
network planning perspectives so as to optimally utilise existing assets instead of the
network augmentation.
Distribution network service providers (DNSPs) are required to ensure that the
connection of new solar PV systems does not violate any associated technical limits.
Technical issues that limit solar PV hosting capacity of LV distribution networks
include thermal overloading of the network components, overvoltage, voltage reg-
ulation, increasing short circuit levels and power quality issues, in addition to
islanding considerations and the possibility of reverse of power flows. At present,
most DNSPs follow simple practices and rules of thumb in order to assess the solar
PV hosting capacity of existing distribution networks. However, there is a necessity
for the development of systematic approaches for hosting capacity evaluation
considering influential factors such as geographical layout of networks as well as
electrical parameters.
Solar PV hosting capacity has to be determined using a transparent approach, that
is, to define an appropriate index or indices subject to well-defined performance
criteria or limits (e.g. feeder voltage, line loading level, voltage unbalance, harmonic
levels) [5]. However, such a hosting capacity can take multiple values when different
performance indices are utilised as constraints. Moreover, a series of constraining
performance indices can be utilised in order to assess the solar PV hosting capacities
of a given distribution network by accommodating stochastic solar PV generation at
different locations through the use of Monte Carlo methodologies. Accordingly, a
single value cannot be specified as the PV hosting capacity for a given network.
The work presented in this chapter covers a generalised deterministic approach to
evaluate the solar PV hosting capacity of low-voltage (LV) distribution networks
considering overvoltage as the constraining criterion and the performance parameter.
This drive to develop a deterministic approach is justified by considering the
drawbacks of a stochastic approach. Furthermore, the generalised framework of
the solar PV hosting capacity estimation aims at capturing all network and PV
plant characteristics to maximise the solar PV penetration level in LV distribution
networks.
7 A Generalised Deterministic Approach to Evaluate PV Hosting Capacity of LV. . . 151
7.2 High Penetration of Solar PV Generation in LV

Networks Resulting in Overvoltage Issues: A Case
Study from Sri Lanka
This section discusses the most common impacts of higher solar PV penetration
levels in distribution networks, based on a case study of an urban LV distribution
network in Sri Lanka which is reported in this section with the aid of feeders with the
highest levels of rooftop solar PV generation. This case study investigates overvolt-
age issues associated with high solar penetration levels in LV distribution networks.
The selected test network consists of a 250 kVA, 11 kV/400 V distribution
transformer with three radial feeders that supply a total of 336 customers including
253 single-phase customers and 83 three-phase customers. More than 90% of the
load on the system is shared by 315 residential customers, and the remaining is
classified as commercial consumers. The installed solar PV capacity in the system
associated with 24 solar PV systems (single-phase and three-phase) distributed along
the three feeders is 104 kW. The geographical layout of the network is shown in
Fig. 7.1, while Table 7.1 provides relevant network details. Measured voltage pro-
files of selected six customers in feeders 2 and 3 at different locations (marked as A,
B, C, D, E and F in Fig. 7.1) are analysed in Sect. 7.2.1 in relation to the overvoltage
problem.
Fig. 7.1 Geographical layout of the distribution network [6]

Table 7.1 Details of the test network [6]

Transformer rating 250 kVA, 11 kV/400 V
Present maximum demanda 154 kW (77%)
No. of customers served 336
No. of customers with rooftop solar PV generators 24
Solar penetration level (by transformer capacity) 40%
Conductor type 3 Phase Aerial Bundle Cable (ABC)
a
Recorded data for July 2017
Load profiles of more than 30% of the customers are derived based on the actual
15-minute energy data recorded remotely through smart meters. The remaining load
profiles are obtained by deriving an average load profile (using the measured data).
These average load profiles are used to generate load profiles of the remaining
customers (without smart meters) based on the monthly energy consumption data
acquired for individual customers. Further, the constant power factor of 0.8 lagging
is assumed for all the customers.
7.2.1 Impact of Solar PVs on LV Network Performance
Voltage rise in LV distribution networks in the presence of solar PVs is known to be

one of the main constraining factors that limit the maximum PV penetration levels
[6]. The voltage rise is highly sensitive to the location and capacity of PVs along the
feeder. Analysis of customer voltage profiles for the selected test network, obtained
from smart energy meter readings (at six different locations on feeders 2 and 3),
provides clear evidence of a violation of upper voltage limit during solar peak times
(i.e. between 10 am and 2 pm) as shown in Fig. 7.2a, b.
Figure 7.2a shows voltage profiles of a solar customer (at location F as shown in
Fig. 7.1) and a non-solar customer (at location E as shown in Fig. 7.1), located at the
middle of feeder 2, affected by overvoltage condition. Further, Fig. 7.2b shows four
selected customer’s voltage profiles for both solar and non-solar along feeder
3, where a solar customer (at location C as shown in Fig. 7.1) and a non-solar
customer (at location D as shown in Fig. 7.1) towards the feeder end are affected by
overvoltage condition during the solar peak times. However, customers who are
close to the transformer end (both solar customer at location A and non-solar
customer at location B) are not affected. It is to be noted that the upper voltage
limit stipulated by the concerned DNSP in Sri Lanka is in the range of +/ 6% of the
nominal voltage.
Fig. 7.2 Voltage profiles of solar/non-solar customers at different locations; (a) feeder 2 and
(b) feeder 3 [6]
7.2.2 Stochastic Approach for Determining Solar PV Hosting

Capacity
In this section, a stochastic analysis framework followed using Monte Carlo simu-
lation (MCS) method to assess the solar PV hosting capacity of the test system is
presented. The performance indices are selected as overvoltage and overloading
criterion with regard to the selection of an appropriate solar PV hosting capacity
level. Accurate analysis of solar PV hosting capacity requires the development of
solar deployment scenarios for all possible future PV locations and corresponding
PV installation size in the network. It is not feasible to simulate all possible PV
deployment scenarios for a given network. Thus, to represent the impact of the
location of a new solar PV connection, the MCS approach is used with a limited
number of PV deployment scenarios at a particular level of penetration.
For the test network, solar PV penetration level (N ) is defined as the ratio of the
number of customers with PV units and the total number of customers. With
increasing penetration levels, a number of stochastic solar PV deployment scenarios
(M) can be considered in such a way that in each scenario, the location of the PV
customers is randomly selected from the pool of solar customers in the network.
Furthermore, each simulated scenario is unique in order that PVs are deployed. The
overall framework developed for solar deployment scenarios for each of the pene-
tration level is illustrated in Fig. 7.3.
Since this systematic approach is applied to a practical distribution network, solar
PV connections are only allocated to customers with energy consumption greater
than 120 kWh per month, assuming that the solar PV installations are financially
feasible for these customers. Capacity of the solar PV system is determined based on
Fig. 7.3 Stochastic analysis framework for solar PV deployment

Table 7.2 Operational limits Criterion Definition Limit

for hosting capacity analysis
Overvoltage Feeder bus voltage 1.06 pu
Thermal loading Line and transformer loading 100%
the monthly energy consumption and assigned as a net metered1 customer. The total
customer demand is assumed to be constant for different solar PV penetration levels.
Following this method, several deployment scenarios are systematically developed,
until all financially feasible solar deployments are connected to the network. It is
assumed that all installed solar PV units are to provide active power at unity power
factor which is a common practice.
The influential factors governing the impact of solar PVs on the performance of
the LV distribution system include an accurate representation of the solar PV
systems, in particular the PV unit size and solar irradiance level which inherently
drives the solar PV output. Solar irradiance profiles for each PV units were devel-
oped using the System Advisor Model (SAM) software considering the irradiance
data applicable to Sri Lanka.
7.2.3 Overvoltage and Overloading Criteria
Maximum feeder voltage violation and line overloading are commonly reported as
the deciding and critical factors in assessing solar PV hosting capacity in distribution
networks. The considered test network is already associated with overvoltage vio-
lation issues as discussed in Sect. 7.2.1. With this in mind, solar PV hosting capacity
was evaluated for the test network by defining the maximum feeder voltage in
addition to considering the line and transformer loading violations. In each PV
deployment scenario, network power flows are monitored, and the corresponding
network voltages and line and transformer loading levels are calculated in order to
verify whether any performance standards are violated. The threshold values for
each performance index considered in this analysis are given in Table 7.2. It is
important to highlight that threshold value for a given performance index is mutually
exclusive of the rest of performance indices.
MCS results shown in Fig. 7.4a–c elaborate the resulting maximum feeder
voltage, maximum line loading and transformer loading levels, respectively, under
different solar penetration levels for a total number of 66,000 solar deployment
scenarios. Based on the MCS results, two levels of solar PV hosting capacity can be
identified as minimum hosting capacity (HCmin) and maximum hosting capacity
(HCmax) compliant with the given performance index, as shown in Fig. 7.4 and
summarised in Table 7.3. With the minimum hosting capacity level, none of the solar
penetration levels violates the criteria of the performance index, i.e. 48 kW of solar
1
This allows customers to be paid in cash for any surplus power generation from the solar PV
systems at the end of monthly billing cycles.
Fig. 7.4 Hosting capacity limits for performance index: (a) maximum feeder voltage criterion,
(b) maximum line loading criterion and (c) transformer loading criterion
Table 7.3 Solar deployment Criterion HCmin (kW) HCmax (kW)

limits for the test network
Overvoltage 48 178
Thermal loading
Line overloading 244 >400
Transformer overloading 250 332
PVs can be integrated to the test network not to violate the steady-state maximum
feeder voltage limits without any concern of the PV location. Similarly, considering
only the thermal overloading limits of line conductors and transformer, the same
network can withstand up to 244 kW solar PV units (note that the voltage violation
criterion is excluded in thermal loading analysis and vice versa). Solar penetration
levels in excess of the maximum hosting capacity limit, independent of the location
of the solar installation, will violate the relevant operational limits. Moreover, the
range in between minimum and maximum hosting capacity levels, certain solar
penetration levels which arise from random PV locations, may violate the network
constraints. Therefore, detailed studies are necessary with precise solar PV locations
to verify that a given level of penetration is safe.
Fig. 7.5 Variation of hosting capacity limit with the number of solar deployment scenarios for
overvoltage criterion; minimum hosting capacity and maximum hosting capacity
The dependency of the number of solar PV deployments on the accuracy of the

results obtained through the MCS method was tested by changing the number of
solar deployment scenarios and repeating the process for overvoltage criteria as
shown in Fig. 7.5a, b. All network modelling is carried out using DIgSILENT
PowerFactory.
7.2.4 Limitations with Stochastic Evaluation Framework
Evaluation of solar PV hosting capacity as discussed in Sect. 7.2.2 is a complex task

that requires accurate network modelling including extensive data. A major draw-
back of Monte Carlo-based studies is the higher computational time and storage
requirements in order to reach superior convergence of the final output. Furthermore,
stochastic assessment methodologies for addressing the randomness of the position
and rating of solar PVs are not easy to implement due to complexity and high
computational times in the evaluation procedures. In addition, solar PV hosting
capacity of a network depends on the characteristics of the network (feeder length,
type of conductor, level of loading, etc.), so each network will have a distinctive
solar PV hosting capacity value. From a network planning perspective, evaluation of
the PV hosting capacity of LV distribution networks through stochastic method is
unrealistic. Thus, there is a need for systematic approaches to assess the PV hosting
capacity with theoretical verifications wherever possible.
7.3 Sensitivity Analyses for Determining Solar PV Hosting

Capacities
PV hosting capacity depends on many factors including feeder characteristics and

solar PV system. Therefore, sensitivity analysis should be performed on individual
influential factors to determine their effects on the hosting capacity. The major
factors influencing the hosting capacity as identified are solar PV system rating
and its location [7] and feeder characteristics which include their operating voltage,
loading level, topology, conductor type, length and existing power/voltage control
mechanisms [6]. This section presents two sensitivity studies in detail on the solar
PV hosting capacity, location sensitivity and feeder characteristics.
7.3.1 Influence of PV Location on Hosting Capacity
Solar PV location and its aggregated capacity along the feeder are vital factors in
assessing hosting capacity for a given distribution network. A simple case study is
presented here to show the influence of solar PV system location and aggregated
capacity on maximum penetration level. A single feeder network which is 1 km long
(conductor type: AAC-Fly type (all aluminium conductor), R ¼ 0.4505 Ω/km,
X ¼ 0.292 Ω/km) is considered with the daily peak demand of 40 kW and
30 kVAr. The total load is assumed to be divided equally between three phases
and evenly distributed along the feeder. Figure 7.6 shows the voltage profile of the
feeder for maximum solar power that can be injected at three different locations; an
aggregated capacity is considered at a given location at a time (distance are measured
from the transformer end), not to violate overvoltage conditions. The results clearly
show how power injection levels tend to reduce when solar PV is connected towards
the feeder end.
7.3.1.1 Study Methodology
How solar PV location and aggregated capacity affect the PV hosting capacity levels
is assessed by considering a simple LV distribution network which has two feeders.
The study network comprises 11 kV/400 V, 100kVA transformer with a loading
level of 50%. Two different feeder lengths, 500 m and 1000 m (conductor type:
AAC-Fly type), are considered to analyse the effect of feeder length on the PV
hosting capacity. To assess the effect of load changes on PV hosting capacity, two
loading levels are considered—Case 1: 60% of total load is allocated to feeder 1, and
the rest (40%) is assigned to feeder 2, and Case 2: 40% of total load is allocated to
feeder 1, and the remaining (60%) is assigned to feeder 2. In addition, constant
power factor, 0.8 lagging, is assumed for all load points. For simplicity, balanced and
evenly distributed loads are considered. Each feeder length is analysed by dividing
Fig. 7.6 Feasible solar PV generation levels at different locations
Fig. 7.7 Feeder segments for hosting capacity analysis
into three equal-length segments as shown in Fig. 7.7 where the randomness of the
solar PV deployment is applied to each segment.
Accordingly, four solar deployment scenarios are considered by changing the
locations of the solar PV connections in relation to Cases 1 and 2:
Scenario 1: Solar PVs are distributed in the feeder end segment.
Scenario 2: Solar PVs are distributed in the feeder middle segment.
Scenario 3: Solar PVs are distributed in the feeder front segment.
Scenario 4: Solar PVs are distributed over all segments of the feeder.
As a specific case, solar PV hosting capacity at a predetermined location is also

investigated by evaluating the maximum connectable solar capacity in three phases
at the boundaries of each segment such that voltage limit is not violated.
7.3.1.2 Study Results
Initial values of maximum feeder voltage drop and line loading levels of the network
are given in Table 7.4 for both Case 1 and Case 2 (without any solar PV
connections).
The test network is analysed to examine the maximum connectable solar PV
capacity at different segments ensuring that the voltage upper limits are not violated.
Figure 7.8 shows the maximum connectable solar PV capacities at the boundaries
(the specific case of fixed locations of PV) of the three segments for both Case 1 and
Case 2 (hosting capacity values are given in per unit using a base Sb ¼ 100 kVA
which is the step down transformer rating, thus allowing the line loading level to be
compared against). Further, solar PV units are to provide only active power operat-
ing at unity power factor. The voltage at the secondary of the transformer was
maintained to a constant value of 1 pu.
The same two-feeder network is analysed using MCS method, and for illustrative
purposes, results obtained for Case 1 scenarios are shown in Figs. 7.9 and 7.10 with
Table 7.4 States of the feeder without solar PVs [6]

Case 1 (F1–60/F2–40) Case 2 (F1–40/F2–60)
Feeder 1 Feeder 2 Feeder 1 Feeder 2
Max. voltage drop 3% 4% 2% 6%
Max. line loading 0.3 pu 0.2 pu 0.2 pu 0.3 pu
Fig. 7.8 Maximum connectable solar capacity at different locations of two-feeder network [6]
Fig. 7.9 Variation of segment-wise hosting capacity levels for overvoltage criteria; Case 1, solar
PVs distributed in (a) feeder end segment, (b) feeder middle segment, (c) feeder front segment and
(d) over all segments of the feeder [6]
regard to overvoltage and line loading criterion, respectively. The limit for overvolt-
age exceedance is 1.06 pu, and the acceptable line loading limit is considered to be
100% of the thermal loading of the conductor. In this analysis, transformer loading
level was not considered as the aim was to consider the feeder level PV hosting
capacity only. However, it is vital that transformer loading levels are considered in
the application of the above methodology.
Referring to Fig. 7.10, it can be seen that when solar PVs are distributed in the
feeder front segment, the connectable maximum solar PV capacity level is essen-
tially limited by the line conductor and transformer overloading. Table 7.5 gives the
summary for minimum and maximum hosting capacity (HC) values (considering
overvoltage and line overloading criterion) derived using MCS for both Case 1 and
Case 2.
Referring to simulation outcomes presented with regard to fixed locations, as in
Fig. 7.8, it clearly indicates that the maximum connectable PV capacity increases as
the solar PVs are distributed towards the feeder front (close to the distribution
transformer). As shown in Fig. 7.8, in Case 1, for the end segment, the lowest
hosting capacity was obtained at the feeder end as 0.36 pu (at point A2 in feeder 2).
Further, for the same case, the summation of the hosting capacities at the nearest end
to the transformer of both feeders in the end segment is 1.42 pu (hosting capacities at
points B1 and B2). These values are equal to minimum and maximum solar PV
Fig. 7.10 Variation of segment-wise hosting capacity levels for maximum line loading criteria;
Case 1, solar PV distributed in (a) feeder end segment, (b) feeder middle segment, (c) feeder front
segment and (d) over all segments of the feeder
Table 7.5 Solar PV deployment limits for the two-feeder network [6]
Case 1 Case 2
HCmin HCmax HCmin HCmax
Scenario 1: End segment 0.4 1.4 0.45 1.4
Scenario 2: Middle segment 0.5 2.3 0.6 2.3
Scenario 3: Front segment 0.8 3.4 0.9 3.4
Scenario 4: Over all segments 0.7 2.5 0.9 3.0
hosting capacity values obtained from MCS method. Same argument is valid for all
three segments. Hence, for a multi-feeder network, minimum hosting capacity of the
entire network is the lowest of the minimum hosting capacities of all feeders. Thus,
considering a single feeder network, minimum hosting capacity can be defined as the
maximum connectable solar PV capacity at the far end of the given feeder, while
maximum hosting capacity is the maximum connectable solar PV capacity at the
nearest end to the transformer. The same argument is valid for segment-wise analysis
as well. In other words, the minimum and maximum hosting capacities are the
connectable maximum solar PV capacity at the boundaries of each segment, far
end of the feeder segment and nearest end to the transformer, respectively.
7.3.2 Influence of Feeder Characteristics
Multi-feeder LV distribution networks with different feeder characteristics were

simulated in DIgSILENT simulation platform in order to analyse the influence of
feeder characteristics on PV hosting capacity. In this study, all test networks
comprise a 11 kV/400 V, 100 kVA transformer with a transformer loading level of
50% (all multi-feeder scenarios are considered at the same transformer loading
level). In addition, a common network classification of urban and rural networks is
represented by selecting different length of feeders. Same analysis is repeated for
different types of conductors in order to analyse the effect of R/X ratios on hosting
capacity. For simplicity, balanced and evenly distributed loads are assumed. Fur-
thermore, it is assumed that the constant power factor of all load points is 0.8
lagging. Voltage at the transformer terminal is maintained at a constant value of
1 pu. This analysis is mainly focused on the maximum connectable solar PV capacity
at the feeder end (i.e. the minimum hosting capacity level) where the results are
given in the following sections.
7.3.2.1 Number of Parallel Feeders
To estimate the effect of number of parallel feeders on solar PV hosting capacity, the
analysis is carried out by increasing the number of parallel feeders. Total loading
level of the network (in terms of transformer loading level) is maintained to a
constant value (50% of the transformer capacity), and identical feeders are used to
observe the impact only of number of parallel feeders.
7.3.2.2 Feeder Length
A common LV network classification of urban and rural networks is represented by

selecting different length of feeders. In this work, 600-m-long feeders are used to
represent urban network, whereas 1200-m-long feeders are used to represent rural
networks. Total network loading levels are maintained at a constant value of 50% by
the transformer capacity.
7.3.2.3 Conductor Type
Analysis described in Sect. 7.3.2 is repeated for different conductor types, AAC-Fly
and ABC, in order to examine the effect of R/X ratio on solar PV hosting capacity.
The relevant conductor specifications are given in column one of Table 7.6.
Table 7.6 Solar deployment limits for two-feeder network per unit [6]
Number
of parallel Network 1 Network 2
Type of conductor feeders (feeder length – 600 m) kW (feeder length – 1200 m) kW
AAC-Fly 1 73 55
R ¼ 0.4505 Ω/km 2 55 36
X ¼ 0.292 Ω/km 3 50 30
R/X ¼ 1.5
4 46 28
ABC–70 mm2 1 64 45
R ¼ 0.441 Ω/km 2 51 32
X ¼ 0.08 Ω/km 3 47 27
R/X ¼ 5.5
4 45 25
ABC–50 mm2 1 51 38
R ¼ 0.641 Ω/km 2 38 26
X ¼ 0.08 Ω/km 3 34 21
R/X ¼ 8
4 32 19
7.3.2.4 Loading Level
Feeder loading level is one of the vital factors which govern the solar PV hosting
capacity. In the analysis of impact of multi-feeders on hosting capacity, each feeder
is applied with different loading levels while maintaining the same total transformer
loading level of 50% of the transformer rating. The corresponding results are
analysed to investigate the impact of loading level on the hosting capacity.
7.3.2.5 Study Results
The estimated PV hosting capacities in relation to influence factors discussed in

Sects. 7.3.2.1, 7.3.2.2, 7.3.2.3 and 7.3.2.4 are given in Table 7.6. As per the results,
since a number of factors are simultaneously influencing the hosting capacity, it is
difficult to derive a direct correlation between solar PV hosting capacity and
individual parameters. Nevertheless, it can be seen that for a given radial feeder,
the minimum hosting capacity decreases as the feeder length increases, and the R/X
ratio and the number of parallel feeders in the network increase. Furthermore, the
minimum hosting capacity of a given feeder increases with feeder loading level.
7.3.3 Feeder-Based Hosting Capacity Approach
The detailed analysis of location sensitivity and feeder characteristics on PV hosting

capacity shows that the PV hosting capacity essentially depends on the size and
location of a solar PV system, feeder loading level, feeder length and conductor type.
Thusly, for multi-feeder networks, individual feeders will exhibit different PV

hosting capacities. In addition, the minimum hosting capacity of a distribution feeder
is the maximum connectable solar PV capacity at the feeder end. Thus, for a given
LV distribution network, the minimum PV hosting capacity can be formulated as
given in (7.1).
8
>
>
HCF1,end
>
>
< HCF2,end
HCmin ¼ ð7:1Þ
>
> ⋮
>
>
:
HCFn,end
where HCFn, end is the maximum connectable solar PV capacity at the feeder end of
nth feeder (n ¼ 1, 2, 3,. . .. . .). Furthermore, maximum hosting capacity increases
with distribution of solar PV systems closer to the transformer. Comparison of
hosting capacity values obtained from spot simulations (hosting capacities at bound-
aries of the feeder segments) and MCS method has led to define segment-wise
minimum and maximum solar PV hosting capacities as:
• Minimum hosting capacity in a given segment is the connectable maximum PV
capacity at the furthest end (from the transformer) of that segment.
• Maximum hosting capacity in a given segment is the connectable maximum PV
capacity at the nearest end (to the transformer) of that segment.
It is evident that extensive analytical efforts are required for accurate hosting
capacity evaluation process because of the complexity associated with distribution
networks. Thus, there should be generalised approaches that are extendable to
address various aspects of concern. Thus, feeder-based hosting capacity approach
is the best as individual hosting capacity of each feeder has to be established in multi-
feeder networks. Such a methodology needs to be extended in order to better capture
the complexity of network modelling, constraints and technologies that will enable
the greatest potential in generalisation.
In essence, the development of a new deterministic approach for the evaluation of
solar PV hosting capacity in a feeder level is important from a distribution planning
perspective. Further, such an approach will facilitate the connection of solar PVs in a
cost-effective way, thus increasing the associated environmental and social benefits.
In this regard, Sect. 7.4 presents a generalised deterministic approach that can be
used to evaluate the feeder level solar PV hosting capacity levels. By implementing
such a methodology, distribution network operators can allow solar PV connections
up to the feeder minimum hosting capacity level. Further, detailed network analysis
will allow examination of whether new connections are permissible by increasing
PV capacities beyond the bounds of minimum PV hosting capacity level.
7.4 Devising Analytical Framework for Evaluating Solar

PV Hosting Capacity
This section explores a novel deterministic approach which facilitates the evaluation
of maximum solar PV penetration level governed by overvoltage limits which will
enable better understanding of the associated benefits.
7.4.1 Theoretical Background
This section first explains the principles underlying power flow and voltage calcu-
lations using a simple feeder model with balanced and uniformly distributed loads.
The principles are then extended to deal with distribution feeders embedded with
solar PV generation using examples.
Consider the single-line diagram of a single feeder network shown in Fig. 7.11.
The resistance and reactance of the line are R and X in ohm per unit length,
respectively, and the shunt capacitance is neglected. Vs is the supply voltage at the
secondary of the transformer.
The total voltage drop VDl in per unit (pu) of the feeder can be written as [8]
lðRPs þ XQs Þ lðXPs RQs Þ

VDl ¼ þj ð7:2Þ
2V 2b 2V 2b
where Vb is the nominal line-line voltage, Ps and Qs are the total real and reactive
power components of the total load that is uniformly distributed in the feeder, and l is
the total length of the feeder.
Active power and reactive power flow at a distance d from the transformer can be
written as (7.3).

d d
Pd ¼ PS 1 , Qd ¼ QS 1 ð7:3Þ
l l
where Pd and Qd are the active and reactive power flow at distance d from the
transformer.
Fig. 7.11 Simplified distribution feeder model

Active power Reactive power
Ps Qs
Pd Qd
0 d l Distance 0 d l Distance
Fig. 7.12 Active and reactive power profiles along the feeder (without solar PV)
nd
II Quadrant st Solar
Solar I Quadrant Photovoltaic
Photovoltaic
Distribution Distribution
Grid Vs POC Grid Vs R X POC
R X P Q
P Q PV , PV
PV , PV
P P
PL, QL Q PL, QL
Q
Solar Solar
Photovoltaic Photovoltaic
Distribution Distribution
Grid Vs R X POC Grid Vs POC
R X
P Q P Q
PV , PV PV , PV
P P
Q P QL Q PL, QL
L,
rd th
III Quadrant IV Quadrant
Fig. 7.13 Four-quadrant operation at POC of a distribution feeder [9]
Figure 7.12 illustrates PS and QS distribution in (7.3) graphically.

As the actual voltage of the feeder is affected by the connection of solar PV in the
feeder, (7.2) can be modified suitably. The next section discusses this aspect
subjected to different operating conditions of the solar PV inverter.
7.4.2 Modelling Regimes
Modern-day solar PV systems possess reactive power capability allowing inverters

to control the voltage at their point of connections (POC). In general, such an
operational capability can be referred to as four-quadrant operation as shown in
Fig. 7.13, where PPV and QPV are the real and reactive power outputs of the solar PV
system and PL and QL are the real and reactive power demands of the connected load
(this may include the battery). These operating scenarios of the distribution feeder
can be listed as follows:
• Ist Quadrant – POC operating at a lagging power factor while net P (PL PPV)
and net Q (QL QPV) are positive
• IInd Quadrant – POC operating at a lagging power factor while net P (PL PPV)
is negative and net Q (QL QPV) is positive
• IIIrd Quadrant – POC operating at a leading power factor while net P (PL PPV)
and net Q (QL QPV) are negative
• IVth Quadrant – POC operating at a leading power factor while net P (PL PPV)
is positive and net Q (QL QPV) is negative
In general, deterministic models should be developed in such a way that they can
represent the four-quadrant operation of the solar PV system (which may include a
battery) where possible. In specific terms, the operation at a unity power factor or at a
fixed leading or lagging power factor must be considered.
7.4.2.1 Solar PV Inverters Operating at Unity Power Factor
Suppose that a solar PV generator is installed at a distance d from the distribution

transformer and let PPV is the real power output of the PV system while reactive
power output, QPV, is zero as unity power factor operation. At the POC, the voltage
is VPV.
When PPV > Pd, active and reactive power flows along the feeder are illustrated in
Fig. 7.16. Zero crossing point denoted by d0 in Fig. 7.14 is where the transition to the
direction of active power flow occurs. Just to the right of the zero crossing point,
active power flows towards the transformer, and just to the left of the zero crossing
point, active power flows from the transformer. The zero crossing point can be
established as

0 PPV
d ¼ 1 l ð7:4Þ
PS
Reverse power flow due to the solar system causes a voltage rise at the POC.
Using (7.2), the new magnitude of the drop in the voltage from 0 to d0 can be written
as in (7.5):
d 0 RðPS PPV Þ þ X ðQS þ Q0 Þ d0

VD0d0 ¼ þ j
2 V 2b 2
X ðPS PPV Þ RðQS þ Q0 Þ
ð7:5Þ
V 2b
The magnitude of the rise in voltage from d0 to d can be written as in (7.6):

Fig. 7.14 Active and reactive power profiles along the feeder (with unity power factor solar
inverter)
ðd d 0 Þ RðPPV Pd Þ X ð Q0 þ Qd Þ ðd d 0 Þ
VRd0 d ¼ þ j
2 V 2b 2
X ðPPV Pd Þ þ Rð Q0 þ Qd Þ
ð7:6Þ
V 2b
0
where Q0 ¼ Qs 1 dl
The total voltage rise from distribution transformer to the POC can be obtained by
subtracting (7.5) from (7.6).

dRPPV re dXPPV im
VR0d ¼ 2λ λ 2
VD l þ j 2λ λ 2
VD l ð7:7Þ
V 2b V 2b
where λ ¼ dl , VDre im
l is the real component of voltage drop and VDl is the reactive
component of voltage drop.
The voltage rise in (7.7) is a linear function of the solar PV system as well as its
location on the feeder. Hence, the maximum connectable solar capacity at a fixed
location can be derived theoretically that can give rise to the maximum allowable
voltage rise in the feeder. For unity power factor operation of solar PV inverter, the
maximum connectable solar PV capacity in three phases at a distance d from the
distribution transformer can be formulated as given in (7.8). Here, VPV, VS and VDl
are in pu, and the phase angle deviations of VPV and VS have been neglected.
V 2b
PPV ¼ ðV PV V S Þ þ 2λ λ2 VDl ð7:8Þ
Rd
7.4.2.2 Solar PV Inverters Operating at Leading Power Factor

transformer and let PPV and QPV be the real and reactive power outputs of the PV
system.
This case is illustrated in Fig. 7.15 where the reversal of both active and reactive
power flows exists when solar inverter is operating at a leading power factor. Zero
crossing point of reactive power flow profile along the feeder can be formulated
similar to (7.4) leading to (7.9).

Q
d00 ¼ 1 PV l ð7:9Þ
QS
The total voltage rise from distribution transformer to the POC can be
decomposed using active and reactive power flows.

dPPV R þ X tan cos 1 pf pv
VR0d ¼ 2λ λ2 VDre
l
V 2b
! ð7:10Þ
dPPV X R tan cos 1 pf pv
þj 2λ λ VDim
2
l
V 2b
If the PV inverter operates at a fixed power factor, then the reactive power
becomes QPV ¼ PPV tan (cos1pfpv).
Maximum amount of solar power generation that will cause the voltage to reach
the overvoltage limit at a distance d from the transformer for a leading power factor
operation can be formulated as in (7.11). Here, VPV, VS and VDl are in pu, and the
phase angle deviations of VPV and VS have been neglected.
V 2b
d R þ X tan cos pf pv
1
Fig. 7.15 Active and reactive power profiles along the feeder (with leading power factor solar
inverter)
7.4.2.3 Solar PV Inverters Operating at Lagging Power Factor

transformer and let PPV and QPV be the real and reactive power outputs of the PV
system.
As shown in Fig. 7.16, reactive power supplied by the grid is increased due to
solar PV inverter operating as an inductive load. In this case, solar PV inverter will
cause an additional voltage drop in the feeder.
The total voltage rise from distribution transformer to the POC can be
decomposed using active and reactive power flows.

dPPV R X tan cos 1 pf pv
VR0d ¼ 2λ λ2 VDre
l
V 2b
!
dPPV X þ R tan cos 1 pf pv im
þj 2λ λ VDl
2
ð7:12Þ
V 2b
Similar to Sect. 7.4.2.2, maximum amount of solar power generation that will
cause the voltage to reach the overvoltage limit at a distance d from the transformer
for lagging power factor operation can be formulated as in (7.13). Here, VPV, VS and
VDl are in pu, and the phase angle deviations of VPV and VS have been neglected.
V 2b
d R X tan cos 1 pf pv
Table 7.7 summarises the mathematical models developed for evaluation of solar
PV hosting capacity at a given point in a feeder considering three different operating
conditions of a PV inverter. The simplified equation for the feeder end hosting
capacity (i.e. when d ¼ l) is also shown in the same table.
Fig. 7.16 Active and reactive power profiles along the feeder (with lagging power factor solar
inverter)
172
Table 7.7 Generalised mathematical models for different operation conditions of PV inverters
PV inverter operation mode At a distance d from the transformer At the feeder end (d ¼ l)
Unity power factor V 2b 2
V 2b
Rd ðV PV V S Þ þ 2λ λ VDl Rd fðV PV V S Þ þ VDl g
Leading power factor V 2b V 2b
d ðRþX tan ð cos 1 pf pv ÞÞ
ðV PV V S Þ þ 2λ λ2 VDl RþX tan ð cos 1 pf pv Þ
fðV PV V S Þ þ VDl g
Lagging power factor V 2b V 2b
d ðRX tan ð cos 1 pf pv ÞÞ
ðV PV V S Þ þ 2λ λ2 VDl RX tan ð cos 1 pf pv Þ
fðV PV V S Þ þ VDl g
D. Chathurangi et al.
Table 7.8 Safe limits of solar Safe limit for HC Safe limit for HC
PV hosting capacity for the calculated from model (kW) simulated (kW)
urban LV network
Feeder 1 32 30
Feeder 2 48 41
Feeder 3 37 38
7.4.3 Verification of the Proposed Analytical Framework
The proposed deterministic approach in Sect. 7.4.2 is used to evaluate the hosting
capacity of the practical distribution network discussed in Sect. 7.2 (named as Case
1) and of the simple distribution feeder (Case 2) with the aid of DIgSILENT
PowerFactory simulations for verification purposes.
Case 1: Practical LV distribution network
The total midday peak demand of the distribution network is (PS and QS) 82 kW
and 60 kVAr, respectively (40% by the transformer capacity). Based on the assump-
tion that all three feeders are loaded equally, the total voltage drop along each feeder
can be calculated using (7.2). These calculated total voltage drops of feeder
1 (ABC-50 mm2), feeder 2 (ABC-70 mm2) and feeder 3 (ABC-70 mm2) are about
3%, 1.7% and 2.6%, respectively, at the solar peak time. The safe limits of the solar
PV hosting capacities of each feeder are calculated using (7.8) where PV inverters
are operating at the unity power factor. Table 7.8 gives safe limits of hosting capacity
values calculated using the mathematical models and spot simulations constrained
by overvoltage for each feeder. As an example, detailed calculations relevant to
identifying safe limit of solar PV hosting capacity on feeder 1 are given below:
Example The total voltage drop in feeder 1 is approximately 3%. The secondary
line-line voltage at the distribution transformer is set at 1.04 pu to ensure that the
voltage is within the stipulated limits under maximum demand conditions at the
night peak. At this loading level (40%), supply voltage2 of the distribution trans-
former is about 1.024 pu.
From (7.8), the maximum connectable solar PV capacity at feeder end is
4002
PPV,Feeder 1 ¼ fð1:06 1:024Þ þ 0:03g 32 kW
0:641 0:51
Case 2: A single feeder test network

The network consists of a 100kVA, 11 kV/400 V transformer connected to a
1-km-long feeder as shown in Fig. 7.17. The total midday peak demand on the feeder
is (PS and QS) 40 kW and 30 kVAr (50% by the transformer capacity). Total load is
assumed to be balanced and uniformly distributed on ten nodes along the feeder.
Further, secondary of the transformer is assumed to be at a constant voltage of 1 pu.
2
Secondary voltage of the transformer depends on the transformer loading level.
Fig. 7.17 Single distribution feeder model
Table 7.9 Maximum connectable solar capacity when solar PV system operates at unity power
factor [9]
HC at feeder end HC at feeder middle
Simulated Calculated Simulated Calculated
AAC-Fly 0.58 0.57 0.96 0.96
ABC – 70 mm2 0.49 0.49 0.84 0.84
ABC – 50 mm2 0.41 0.41 0.69 0.69
Table 7.10 Maximum connectable solar capacity when solar PV system operates at 0.9 leading
power factor [9]
AAC-Fly 0.48 0.48 0.79 0.81
ABC – 70 mm2 0.50 0.50 0.84 0.85
ABC – 50 mm2 0.43 0.43 0.71 0.73
Table 7.11 Maximum connectable solar capacity when solar PV system operates at 0.9 lagging
power factor [9]
AAC-Fly 1.41 0.92 1.98 1.55
ABC – 70 mm2 0.63 0.59 1.06 1.02
ABC – 50 mm2 0.52 0.49 0.85 0.82
The test network is analysed for maximum connectable solar PV capacity at two
different locations so that the upper voltage limit stipulated by the utility (taken as
+6%) is not violated.
The analysis has been repeated for different types of conductors considered in
Sect. 7.3 in order to investigate the effect of R/X ratio on solar PV hosting capacity.
Accordingly, three cases are presented for different operating scenarios of the PV
inverters: unity power factor, 0.9 leading power factor and 0.9 lagging power factor.
Solar PV hosting capacity values which were obtained from the simulations and
mathematical model are given in Tables 7.9, 7.10 and 7.11 for the three scenarios. In
each case, hosting capacity values relevant to feeder end and middle of the feeder
were evaluated and compared against the results from the mathematical models. All
solar PV capacity values are given in per unit with transformer ratings as base values.
As shown in Tables 7.9, 7.10 and 7.11, solar PV hosting capacities obtained from
simulations and proposed deterministic method are in close agreement verifying the
accuracy of the deterministic approach.
The minimum hosting capacity of a given radial distribution feeder is the
maximum connectable solar PV capacity at the feeder end which can be defined as
the safe limit of the hosting capacity. Thus, the proposed mathematical model
enables the evaluation of a safe limit of the PV hosting capacity of distribution
networks without resorting to complex computer simulations. From a distribution
system planning perspective, the initial use of a deterministic approach is convenient
and more practical than using extensive simulations. Therefore, utilities can make
use of the proposed approach as a guide to evaluate the maximum connectable solar
capacity of a given distribution feeder using the data which are not overly
demanding.
7.4.4 Comparison of the Proposed Deterministic Approach

with Stochastic Approach
As stated in Sect. 7.3.3, the minimum PV hosting capacity (safe limit of the hosting
capacity) for a given radial distribution feeder is the maximum connectable solar PV
capacity at the feeder end (verified using stochastic Monte Carlo simulations). Thus,
the analysis presented in Sect. 7.4.3 (Case 2) justified that the PV hosting capacity
values obtained from both stochastic and deterministic approaches are in close
agreement. The pros and cons of stochastic and deterministic approaches in PV
hosting capacity analysis are shown in Table 7.12.
Table 7.12 Comparison of stochastic and feeder level deterministic methods in hosting capacity
analysis
Stochastic approach Deterministic approach
Pros Randomness of solar PV location and capacity of Feeder-based limitations and
PV system can be addressed easily diversity can be accommodated
Can be generalised
Improved accuracy
Easier and more practical
Whole network PV hosting
capacity can be maximised
Reduces computational effort
Cons Needs detailed network modelling Can be applied to one location at a
Increased computational burden time
Cannot be generalised Not able to capture all perfor-
Variability in accuracy depends on the number of mance criteria for hosing capacity
simulations and complexity of the network model
As per Table 7.12, the deterministic approach for feeder level PV hosting capacity
offers substantial benefits over a stochastic approach.
7.5 Conclusions and Future Works
In this chapter, a feeder based, deterministic approach was proposed for the evalu-
ation of the solar PV hosting capacity in LV distribution networks. The hosting
capacity was essentially constrained by overvoltage limits of many of the LV
distribution networks. The deterministic method was developed in order to evaluate
the maximum allowable solar PV capacity at a given point of distribution feeder
constrained by overvoltage limits. Related to PV hosting capacity evaluation, Monte
Carlo-type stochastic simulations are not practical in distribution network planning
environments due to extensive network modelling efforts and the complexity. Thus,
individual feeder-based analysis approach is recommended and considered to be
more effective.
Based on the findings, a safe limit of the solar PV hosting capacity (the minimum
hosting capacity) is defined as the maximum allowable solar PV capacity at the far
end of a feeder that can be evaluated using the proposed deterministic approach.
Under this framework, distribution network operators and planners can investigate
the capability and limitations of solar PV penetration.
The proposed feeder-based hosting capacity evaluation approach can be seen to
capture all influencing factors on solar PV hosting capacity. Thus, this can be used as
an approximate guide or a rule of thumb to evaluate solar PV hosting capacity at a
given point of LV feeders for overvoltage curtailment without using complex
stochastic techniques.
Further research is needed to confirm the applicability of the developed approach
in realistic network environments. Other network performance indices such as
voltage unbalance due to PV hosting have not been considered in the present work
and require investigation of whether such parameters become dominant or not.
Acknowledgments This work was supported by SRC grant SRC/LT/2017/16, University of

Moratuwa, Sri Lanka. The authors would like to thank Lanka Electricity Company Ltd., Sri
Lanka, and RMA Pvt. Ltd., Sri Lanka, for providing necessary data and field measurements for
this study.
References
1. REN 21 (2019). Renewables 2019 global energy report. Technical Report.

2. Smith, J., Rönnberg, S., Blanco, A., Bollen, M., Emin, Z., Ilisiu, D., et al. (2016). Power quality
aspects of solar power. Available via https://e-cigre.org/publication/672. Accessed Dec 2016.
3. Bollen, M., & Rönnberg, S. (2017). Hosting capacity of the power grid for renewable electricity
production and new large consumption equipment. Energies. https://doi.org/10.3390/
en10091325.
4. Jothibasu, S., Dubey, A., & Santoso, S. (2016). Integration photovoltaic generation. Research
Gate. https://doi.org/10.13140/RG.2.2.33939.37928.
5. Chathurangi, D., Jayatunga, U., Rathnayake, M., Wickramasinghe, A., Agalgaonkar, A., &
Perera, S. (2018a). Potential power quality impacts on LV distribution networks with high
penetration levels of solar PV. Paper presented at 18th international conference on Harmonics
and Quality of Power, Ljubljana, Slovenia.
6. Chathurangi, D., Jayatunga, U., Siyambalapitiya, T., Wickramasinghe, A., Perera, S., &
Agalgaonkar, A. (2018b). Connection of solar PV to LV networks: Considerations for maximum
penetration level. Paper presented at Australasian Universities Power Engineering conference on
transition to a low-carbon energy future, University of Auckland, New Zealand.
7. Ismael, S. M., Aleem, S. H. E. A., Abdelaziz, A. Y., & Zobaa, A. F. (2019). Review: State-of-the-
art of hosting capacity in modern power system with distributed generation. Renewable Energy.
https://doi.org/10.1016/j.renene.2018.07.008.
8. Kersting, W. H. (2012). Distribution system modelling and analysis (3rd ed.). London: Taylor
and Francis Group.
9. Chathurangi, D., Jayatunga, U., Perera, S., & Agalgaonkar, A. (2019). Evaluation of maximum
solar PV penetration: Deterministic approach for overvoltage curtailments. Paper presented at
9th IEEE international conference on Innovative Smart Grid Technologies, University of
Politehnica, Bucharest, Romania.
Chapter 8
Hosting Capacity Maximization Based
on Optimal Reconfiguration of Distribution
Networks with Optimized Soft Open Point
Operation
Ibrahim Mohamed Diaaeldin, Shady H. E. Abdel Aleem, Ahmed El-Rafei,

Almoataz Y. Abdelaziz, and Ahmed F. Zobaa
8.1 Introduction
Distribution network reconfiguration (DNR) is the change in the status of operating

lines in order to provide the resilience of distribution network topologies. DNR is
broadly used to minimize power losses, improve the voltage profile, and increase
network reliability. The requirements for applying DNR correctly hinges on
maintaining the radial topology of the distribution network, serving the connected
loads, keeping the branches within their specified thermal limits, and preserving the
network’s voltage limits. DNR was previously solved using various optimization
techniques. These optimization techniques are divided into three categories: heuris-
tic, metaheuristic, and mathematical techniques. From the mathematical technique
I. M. Diaaeldin (*) · A. El-Rafei

Engineering Physics and Mathematics Department, Faculty of Engineering,
Ain Shams University, Cairo, Egypt
e-mail: ibrahimmohamed@eng.asu.edu.eg; ibrahimmohamed@ieee.org;
ahmed.elrafei@eng.asu.edu.eg
S. H. E. Abdel Aleem
Mathematical, Physical and Engineering Sciences Department, 15th of May Higher Institute
of Engineering, Cairo, Egypt
e-mail: engyshady@ieee.org
A. Y. Abdelaziz
Faculty of Engineering and Technology, Future University in Egypt, Cairo, Egypt
e-mail: almoataz.abdealziz@ieee.org
A. F. Zobaa
College of Engineering, Design and Physical Sciences, Brunel University London,
Uxbridge, UK
e-mail: azobaa@ieee.org

https://doi.org/10.1007/978-3-030-40029-3_8
180 I. M. Diaaeldin et al.
point of view, it provides a near-global solution of the distribution system configu-

ration like mixed-integer nonlinear programming (MINLP) [1] and mixed-integer
second-order cone programming (MISOCP) [2]. From the heuristic and
metaheuristic viewpoint, it provides an acceptable local optimal solution, like
genetic algorithm [3], particle swarm optimization [4], Harris hawks optimization
algorithm [5], and others.
Meanwhile, DGs are widely integrated into the distribution network along with
fossil-fuel-based power stations. The wide integration of renewables has many
impacts on the system’s operational and performance indices, including total har-
monic distortion, aggregated voltage deviation, voltage profile, overall system’s
power factor, and others. As a result of these impacts as well as the emerging concepts
of smart grids that have made the power system more complex and unpredictable, the
hosting capacity (HC) limit of the distribution network is formulated to act as a simple
but reliable index that can help decide where and when to add more DGs into the
system as the HC provides an indication for network operators about the available DG
powers that can be accommodated by the distribution network.
In the literature, many perspectives were investigated to maximize HC [6] such as
power quality improvement, network reinforcement, DNR application, and soft open
point allocation. From the power quality perspective, studies were conducted in
[7, 8] to maximize constrained harmonic distortion HC using a C-type passive filter
to mitigate harmonic distortion and in turn maximize the harmonic constrained
HC. From the perspective of network reinforcement, a study was conducted in [9]
on an existing Egyptian distribution network to maximize its HC using network
reinforcement. From the perspective of DNR, a study was conducted in [10] on a
distribution network in Japan that contains 235 switches in order to maximize HC by
selecting the best configurations of the distribution network. Benefits of static and
dynamic reconfiguration were investigated in [11], in which a static reconfiguration
was found to be beneficial in the planning stage while a dynamic reconfiguration was
found effective for active distribution networks especially when a higher number of
remotely controlled switches are available. A minimum number of switching events
was done to maximize HC using multi-period reconfiguration in [12]. From the
perspective of SOP allocation, a strengthened second-order cone programming was
investigated on the IEEE 33-bus system for maximizing its HC using SOPs [13]; a
project was analyzed in [14] connecting island of Anglesey network with the
mainland in North Wales using an SOP to maximize HC of the network.
In this chapter, a strategy is proposed to maximize DG integration in the distri-
bution network without violating system operational limits. The strategy is based on
the available active and reactive powers provided by the substation. The proposed
strategy aims to prepare the distribution network using DNR and SOP installations to
host more DG installations in the distribution system. In order to solve the MINLP
problem under investigation, a new metaheuristic optimization technique named
expanded invasive weed optimization (exIWO), proposed by Josiński et al. [15],
was used.
This chapter is organized as follows: Sect. 8.2 presents a detailed problem
statement including power flow equations, DG and SOP models, and DNR
8 Hosting Capacity Maximization Based on Optimal Reconfiguration of Distribution. . . 181
methodology; Sect. 8.3 illustrates the problem formulation and a brief description of
the exIWO algorithm; Sect. 8.4 presents the results of the case studies under
investigation; and Sect. 8.5 is dedicated to the conclusions of this research.
8.2 Problem Statement
In this section, the power flow equations, DG and SOP modeling, and the proposed
DNR methodology are illustrated in detail. The distribution system model is illus-
trated in Fig. 8.1a in the presence of DGs and SOPs, and SOP-based IGBT
connection model is illustrated in Fig. 8.1b.
Fig. 8.1 Distribution system modeling with DG and SOP: (a) power injection model, (b) -
SOP-based IGBT connection model
8.2.1 Power Flow Equations
Power flow equations essential to solve the distribution system under study are
expressed as follows:
P2k þ Q2k
Pkþ1 ¼ Pk Pload
kþ1 r k,kþ1 ð8:1Þ
jV k j2
P2k þ Q2k
Qkþ1 ¼ Qk Qload
kþ1 xk,kþ1 ð8:2Þ
jV k j2
P2 þ Q2
jV kþ1 j2 ¼ jV k j2 2ðr k,kþ1 Pk þ xk,kþ1 Qk Þ þ r 2k,kþ1 þ x2k,kþ1 k 2 k ð8:3Þ
jV k j
where Pk and Qk are the injected active and reactive powers at the kth bus,
load
respectively; Pload
kþ1 and Qkþ1 are the active and reactive powers of loads attached to
bus k + 1 loads, respectively; |Vk| is the magnitude of the kth bus voltage; and rk, k + 1
and xk, k + 1 are the branch resistance and reactance, joining between buses k and
k + 1, respectively.
8.2.2 DG Modeling
In this study, DG is modeled as a generator with a unity power factor. A constraint

regarding the DG installed capacity limit is given in (8.4). It should be mentioned
that other DG sets with smart inverters can be used; however, there is no need for
them as the SOP can regulate both active and reactive powers in the system.
0 PDG
k DGcap , 8k 2 N bus ð8:4Þ
where PDG
k is the injected DG active power at bus k, DGcap is the maximum DG
capacity limit, and Nbus is the number of buses in the distribution system.
8.2.3 SOP Modeling
Soft open points were first proposed by Bloemink and Green [16]. Three different
topologies were employed to integrate SOPs, and they are composed of back-to-back
(B2B) voltage source converter (VSC), static synchronous series compensator, and
unified power flow controller [17]. The B2BVSC is used in this study because of its
capabilities to enhance power quality. An SOP is modeled using its injected powers
(active and reactive) at its terminals as shown in Fig. 8.1. In addition, the algebraic
sum of the SOP active powers injected at its terminals and its internal active losses
by its VSCs must equal to zero, as expressed in Eq. (8.5). Thus,
PSOP
I þ PSOP
J þ PSOPloss
I þ PSOPloss
J ¼0 ð8:5Þ
The reactive power limits are provided in Eq. (8.6), and the SOP size limit is
presented in Eq. (8.7). Thus,
QSOP
I
min
QSOP I QSOP I
max
, 8I, J 2 N f ð8:6Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 SOP 2 ffi
PSOP
I þ QI SSOP
I , 8I 2 N f ð8:7Þ
where Nf is the number of feeders, PSOP I is the SOP’s active power transferred to the
Ith feeder, PSOP
J is the SOP’s active power transferred to the Jth feeder, PSOPloss
I is
the internal power loss of the VSC coupled to the Ith feeder, PSOPlossJ is the internal
active loss of the VSC coupled to the Jth feeder, QSOP I is the SOP’s reactive power
transferred to the Ith feeder, QSOP
J is the SOP’s reactive power transferred to the Jth
feeder, QSOP
I
min
and Q SOP max
I are the lower and upper limits of the SOP’s reactive
power transferred to the Ith feeder, and SSOP I is the maximum size of the planned
SOP. Further, the active loss of each VSC (PSOPloss I and PSOPloss
J ) is expressed in
Eq. (8.8):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SOP 2 SOP 2ffi
PSOPloss
I ¼ ASOP
loss PI þ QI , 8I 2 N f ð8:8Þ
where ASOP
loss is the loss coefficient associated to converter internal power loss [18],
which signifies the percentage of leakage in the delivered power between feeders.
8.2.4 Network Reconfiguration
In this study, the procedure of DNR is employed as follows [19]:

(i) An individual sectionalized line is chosen inside each loop in order to be
changed to a tie-line line.
(ii) The previous tie line in this loop changes to a sectionalized line, as described in
Fig. 8.2. Thus, there is no need to check radiality as the exchange between
switches was done on the same loop without violating the network’s radial
structure.
(iii) The procedure is repeated while keeping the number of tie lines fixed before
and after reconfiguration.
Fig. 8.2 DNR procedure: (a) before DNR, (b) after DNR
8.3 Problem Formulation

8.3.1 Objective Function
The overall hosting capacity (HC) is the objective function, needed to be maximized.
HC is formulated as follows:
PN DG DG
Pi
Max HC ð%Þ ¼ PN bus i¼1
loadnormal
100 ð8:9Þ
k¼1 k P
where Ploadnormal
k is the active demanded power at bus k at the normal loading
conditions.
8.3.2 Constraints
In addition to the DG capacity limit expressed in (8.4), SOP reactive power con-
straints given in Eq. (8.6), and SOP size limit given in Eq. (8.7), constraints
concerning slack bus power factor (PFslack), voltage magnitudes, and lines thermal
capacities are formulated as follows:
min PF
PFslack PFslack ð8:10Þ
slack
max
V min jV k j V max , 8k 2 N bus ð8:11Þ

jI k,kþ1 j I rated
k,kþ1 , 8k 2 N bus ð8:12Þ
where PFslack slack

min and PFmax are the minimum and maximum slack power factor limits,
respectively; Ik, k + 1 is the current flowing in the branch joining buses k and k + 1;
and I rated
k,kþ1 is the rated current of this branch. Besides, the reverse power due to SOP
and DG installations is not permitted in this study as expressed in Eq. (8.13).
Moreover, reactive power provided by SOPs is limited as shown in Eq. (8.14) to
avoid overcompensation of the network.
PLi α PDG
i β PSOP
I γ PSOP
J 0, 8i 2 N bus ð8:13Þ
X XN bus
QSOP
I þ QSOP
J Qload
p ð8:14Þ
p¼1
where α equals 1 when a DG unit is connected to bus i, β equals 1 when an SOP is

connected to feeder I of bus i, and γ equals 1 when an SOP is connected to feeder J of
bus i; otherwise, α ¼ β ¼ γ ¼ 0.
8.3.3 Expanded Invasive Weed Optimization
The invasive weed optimization (IWO) algorithm was designed by Mehrabian and
Lucas [20] to deal with nonlinear optimization problems (Fig. 8.3). The motivation
of the IWO is based on the colonization of weed and was further upgraded in 2014
[15] to deal with MINLP problems including discrete and continuous variables. The
effectiveness of the exIWO arises in its fast convergence toward finding the global
optimum and also its ability to solve mixed continuous-discrete problems. The
exIWO is based on the propagation of weeds and their compliance with ecological
conditions. The population is divided into individuals and seeds. First, the popula-
tion assigned by random values. Further, each individual i produces a bunch of seeds
(Si) based on its objective function value (Fi) as follows:

S Smin
Si ¼ Smin þ ðF i F min Þ max ð8:15Þ
F max F min
where Smax and Smin are the upper and lower number of seeds, respectively. Fmax and
Fmin represent the best and worst objective function value of the population. Second,
the exploration phase includes three methodologies: spreading, dispersing, and
rolling down. The spreading hinges on seed dissemination, in which a set of new
individuals is produced by each seed randomly. Besides, the dispersing is known as
the divergence among individuals and their offspring. It signifies the distance
covered by the seed from the parent plant to its final placement. The distance
formulated as a normal distribution function in which its standard deviation is
expressed Eq. (8.15).
Fig. 8.3 Proposed problem formulation using exIWO algorithm
m
itermax iter
σ iter ¼ ðσ init σ fin Þ þ σ fin ð8:16Þ
itermax
where iter is the present iteration, itermax is the iterations’ maximum limit, σ init and
σ fin are the initial and final values of the standard deviation, and m is a nonlinear
modulation factor. Further, the rolling down resembles the placement of a seed to a
new location based on its objective function value. Finally, the procedure is then
started over again from the exploration phase until the termination criteria are met.
Termination is fulfilled if iter is higher than itermax or the difference between
individuals’ objective function values was lower than a predefined value.
8.3.4 HC Maximization Strategy
Starting from the aim of this study, i.e., HC maximization resulting from the
installation of more DGs, the main obstacle facing maximization of HC is PFslack
as a result of injecting active power from DGs; the PFslack will decrease, and it will
be limited by PFslack
min . Therefore, increasing DG penetration depends on the reactive
power provided by the slack bus (Qslack), where the initial injected active, reactive,
and apparent powers by the slack bus are Pslack
0 , Qslack
0 , and Sslack
0 , respectively. In
order to increase DG penetration, the following strategy is deployed as follows:
Step 1: DNR is done first to provide the resilience of the network topology.
Step 2: SOPs are installed instead of tie lines after DNR is done to inject reactive
powers to the system, and as a result, the reactive power injected by the slack bus
Qslack
0 will decrease to Qslack
1 , and also PFslack will increase until it reaches its near
maximum/maximum limit PFslack max .
Step 3: DGs are then installed, Pslack0 decreased to Pslack
2 , and PFslack will decrease
again until it reaches its near minimum/minimum limit PFslack min . The proposed
strategy will be tested in the following section.
8.4 Simulation Results and Discussions
In this study, the 83-bus distribution system [21] is the system investigated to
maximize its HC. The topology of this system consists of 83 nodes and 96 lines,
as shown in Fig. 8.4. These lines are composed of 96 lines including sectionalized
and tie lines. The voltage limits Vmin and Vmax are set to 0.95 and 1.05 p.u.,
respectively, and I rated
k,kþ1 is set to 310 A. The number of installed SOPs ranges from
one to five, and SSOP
I ¼ SSOP
J are set to 5 MVA and ASOPloss is set to 2% [18]. The results
obtained after installing DGs along with DNR with and without SOPs for the 83-bus
distribution system are shown in Tables 8.1, 8.2, 8.3, and 8.4.
Results showed that applying DNR only did not guarantee a solution due to the
violation of PFslack
min limit at different loading levels. However, installing SOPs
allowed DG installations as it injects reactive power in order to improve PFslack to
be greater than PFslack
min . Besides, it is notable that HC increases proportionally as the
number of SOPs increases. The best HC values resulted from simultaneous DNR
along with DG and SOP installations, displayed in bold in Table 8.1, where five
SOPs are installed at the three loading levels.
Optimal network configuration, capacity, and sites of SOPs and DGs for five
installed SOPs at normal loading conditions are given in Table 8.5. Branch current
and voltage profiles at normal loading conditions for five installed SOPs at normal
loading conditions are given in Figs. 8.5 and 8.6, respectively. An 83-bus compen-
sated system at normal loading condition for five installed SOPs is given in Fig. 8.7.
Fig. 8.4 83-bus distribution system
Table 8.1 HC results at different loading levels

Simultaneous DNR with SOP and DG allocation
Loading level DNR only 1 SOP 2 SOPs 3 SOPs 4 SOPs 5 SOPs
Light (50%) NAa 29.1164 29.4533 29.8589 29.8059 30.7760
Normal (100%) NA 9.1711 53.9682 56.1905 56.7901 59.0829
Heavy (130%) NA NA NA 71.0758 72.0031 73.2427
a
NA means not applicable
Table 8.2 NDG at different loading levels

Light (50%) NA 58 14 59 14 60
Normal (100%) NA 2 14 59 14 24
Heavy (130%) NA NA NA 14 59 59
Table 8.3 Total active losses in kW at different loading levels

Light (50%) NA 290.6481 318.8077 258.2105 256.0974 266.5299
Normal (100%) NA 728.9004 627.7467 725.2993 595.8846 645.9454
Heavy (130%) NA NA NA 914.4549 860.9542 904.6237
Table 8.4 PFslack at different loading levels

Light (50%) NA 0.9017 0.9031 0.9022 0.9009 0.9000
Normal (100%) NA 0.9021 0.9089 0.9080 0.9096 0.9009
Heavy (130%) NA NA NA 0.9116 0.9126 0.9105
Table 8.5 Optimal network configuration, capacity, and sites of SOPs and DGs for five SOPs at
normal loading condition
SOP SOP sizing DG DG

Configuration locations PSOP
I PSOP
J QSOP
I QSOP
J location sizing
(tie lines) (tie lines) (MW) (MW) (MVAr) (MVAr) (bus) (MW)
16-42-72-84-85- 85 0.4000 0.540 3.2806 3.6594 28 1.8000
86-88-89-91-92- 31 1.8000
93-94-96 72 0.3290 0.2000 3.2128 1.7144 19 1.2000
6 1.1000
83 0.4000
80 0.2000
79 2.0000
78 0.4000
41 0.2000
40 0.0200
93 0.2000 0.2193 0.5545 0.3055 36 0.1000
53 0.5000
52 0.5000
32 0
92 0.3084 0.2000 2.6975 0 75 1.2000
12 1.2000
27 0.1000
46 0.2000
45 0.8000
10 0.3000
94 0 0 0 0 8 0.3000
44 0.0300
68 0.4000
71 2.0000
Fig. 8.5 Branch current profile at normal loading condition
Fig. 8.6 Voltage profile at normal loading condition

84
8
1 53 47
A 9 G
5 7 55
10 96
11 12 13 60 56
B 88 64 H
89 ~ ~
14 22
20 23 24 72 65
C ~ ~ I
18 21
16
16 91
86
28 29 76 73
D J
26 40
~
~ ~ 39
~
34 83 77
E 42 42 K
32 38 41
~
DG node
~
F ~ ~ SOP
43 46
Fig. 8.7 An 83-bus compensated system at normal loading condition for five installed SOPs
8.5 Conclusion
Feeder reconfiguration is an integral component of smart distribution strategy in

modern active distribution networks. Also, much attention has been paid to the
concept of the HC approach to determine how much of new renewable energy
resources or DG units can be connected to the network at a certain feeder based on
the comparison of a set of performance indices with a known index limit for each of
them. Once any of those indices exceeds its limit, the HC is determined. If more than
a value for each index is determined, the minimum (safer) value is then chosen as the
overall HC value of the network [22]. In this regard, this chapter presents an integral
component strategy to optimally allocate DGs using DNR and SOP allocation in
order to maximize the HC of an 83-distribution network. The exIWO metaheuristic
algorithm was employed to solve the MINLP problem under study. A case study was
conducted to increase the HC of the 83-bus distribution system. The proposed
strategy has proven its ability to maximize the HC of the distribution network at
three loading levels.
References
1. Paterakis, N. G., et al. (2016). Multi-objective reconfiguration of radial distribution systems

using reliability indices. IEEE Transactions on Power Apparatus and Systems, 31, 1048. https://
doi.org/10.1109/TPWRS.2015.2425801.
2. Taylor, J. A., & Hover, F. S. (2012). Convex models of distribution system reconfiguration.
IEEE Transactions on Power Apparatus and Systems, 27, 1407–1413.
3. Eldurssi, A. M., & O’Connell, R. M. (2015). A fast nondominated sorting guided genetic
algorithm for multi-objective power distribution system reconfiguration problem. IEEE Trans-
actions on Power Apparatus and Systems, 30, 593. https://doi.org/10.1109/TPWRS.2014.
2332953.
4. Sulaima, M. F., Shamsudin, N. H., Jaafar, H. I., Dahalan, W. M., & Mokhlis, H. (2014). A DNR
and DG sizing simultaneously by using EPSO. In 2014 5th International conference on
intelligent systems, modelling and simulation (pp. 405–410). IEEE. https://doi.org/10.1109/
ISMS.2014.75.
5. Abdel Aleem, S. H. E., Zobaa, A. F., Balci, M. E., & Ismael, S. M. (2019). Harmonic
overloading minimization of frequency-dependent components in harmonics polluted distribu-
tion systems using Harris Hawks optimization algorithm. IEEE Access, 7, 100824–100837.
6. Ismael, S. M., Abdel Aleem, S. H. E., Abdelaziz, A. Y., & Zobaa, A. F. (2019). State-of-the-art
of hosting capacity in modern power systems with distributed generation. Renewable Energy,
130, 1002–1020.
works. In 2016 IEEE 16th International conference on environment and electrical engineering
(EEEIC) (pp. 1–6). IEEE. https://doi.org/10.1109/EEEIC.2016.7555515.
8. Sakar, S., Balci, M. E., Abdel Aleem, S. H. E., & Zobaa, A. F. (2017). Increasing PV hosting
capacity in distorted distribution systems using passive harmonic filtering. Electric Power
Systems Research, 148, 74–86.
9. Ismael, S. M., Abdel Aleem, S. H. E., Abdelaziz, A. Y., & Zobaa, A. F. (2018). Practical
considerations for optimal conductor reinforcement and hosting capacity enhancement in radial
distribution systems. IEEE Access, 6, 27268–27277.
10. Takenobu, Y., Yasuda, N., Minato, S., & Hayashi, Y. (2019). Scalable enumeration approach
for maximizing hosting capacity of distributed generation. International Journal of Electrical
Power & Energy Systems, 105, 867–876.
11. Capitanescu, F., Ochoa, L. F., Margossian, H., & Hatziargyriou, N. D. (2015). Assessing the
potential of network reconfiguration to improve distributed generation hosting capacity in active
distribution systems. IEEE Transactions on Power Apparatus and Systems, 30, 346–356.
12. Fu, Y. Y., & Chiang, H. D. (2018). Toward optimal multi-period network reconfiguration for
increasing the hosting capacity of distribution networks. In IEEE power and energy society
general meeting. https://doi.org/10.1109/PESGM.2017.8274614.
13. Qi, Q., et al. (2017). Using an MVDC link to increase DG hosting capacity of a distribution
network. Energy Procedia, 142, 2224–2229.
14. Ji, H., et al. (2017). A strengthened SOCP-based approach for evaluating the distributed
generation hosting capacity with soft open points. Energy Procedia, 142, 1947–1952.
15. Josiński, H., Kostrzewa, D., Michalczuk, A., & Świtoński, A. (2014). The expanded invasive
weed optimization Metaheuristic for solving continuous and discrete optimization problems.
Scientific World Journal, 2014, 1–14.
16. Bloemink, J. M., & Green, T. C. (2011). Increasing photovoltaic penetration with local energy
storage and soft normally-open points. In 2011 IEEE power and energy society general meeting
(pp. 1–8). IEEE. https://doi.org/10.1109/PES.2011.6039561.
17. Bloemink, J. M., & Green, T. C. (2013). Benefits of distribution-level power electronics for
supporting distributed generation growth. IEEE Transactions on Power Delivery, 28, 911–919.
18. Diaaeldin, I., Abdel Aleem, S., El-Rafei, A., Abdelaziz, A., & Zobaa, A. F. (2019). Optimal
network reconfiguration in active distribution networks with soft open points and distributed
generation. Energies, 12, 4172.
19. Chiou, J.-P., Chang, C.-F., & Su, C.-T. (2005). Variable scaling hybrid differential evolution for
solving network reconfiguration of distribution systems. IEEE Transactions on Power Appa-
ratus and Systems, 20, 668–674.
20. Mehrabian, A. R., & Lucas, C. (2006). A novel numerical optimization algorithm inspired from
weed colonization. Ecological Informatics, 1, 355–366.
21. Mohamed Diaaeldin, I., Aleem, S. H. E., El-Rafei, A., Abdelaziz, A. Y., & Zobaa, A. F. (2019).
A novel graphically-based network reconfiguration for power loss minimization in large
distribution systems. Mathematics, 7, 1182.
22. Bollen, M., et al. (2017). Consequences of smart grids for power quality: Overview of the
results from CIGRE joint working group C4.24/CIRED. In 2017 IEEE PES innovative smart
grid technologies conference Europe (ISGT-Europe) (pp. 1–6). IEEE. https://doi.org/10.1109/
ISGTEurope.2017.8260116.
Chapter 9
Generation Regulation Control Systems
Krishnan Manickavasagam, Ilango Karuppasamy, and Vineetha Puttaraj
9.1 Introduction
Power systems are steadily growing to meet the power demand. Increase in demand
is a challenging concern in developing countries. To meet this ever-increasing
demand, enhancement of power generation is required in conventional as well as
non-conventional form. Conventional power generation has its own limitations such
as land requirement and environmental concern for future enhancement. Conven-
tional power generation has its own limitations such as land requirement, environ-
mental associated pollution and exploitation of fossil fuel. At the same time,
worldwide, distributed generations (DGs) are contributing more due to sustainabil-
ity, pollution free environmnet, ease of energy, and load management. It is expected
that the renewable energy capacity will reach 30 GW over the next two decades,
merely 10% new capacity addition of DG during this period [1]. Integration of DG
into grid is in practice for the past few decades in many countries.
Conventional power system consists of thermal, hydro, and nuclear power plants
interconnected with grid, transmission system, and distribution system. The purpose
of controlling generation is to track load change by adjusting the generation auto-
matically to restore the frequency and tie-line power flow within the prescribed limit
using automatic generation control (AGC) in a conventional power system. Under
normal conditions, generators are operated at a scheduled output, and system
frequency is maintained constant [2]. The primary objective of AGC is to minimize
K. Manickavasagam · V. Puttaraj
Department of Electrical Engineering, M. S. Ramaiah University of Applied Sciences,
Bangalore, Karnataka, India
I. Karuppasamy (*)
Department of Electrical and Electronics Engineering, Amrita School of Engineering, Chennai
Campus, Amrita Vishwa Vidyapeetham, Chennai, Tamil Nadu, India

https://doi.org/10.1007/978-3-030-40029-3_9
196 K. Manickavasagam et al.
the frequency deviation and tie-line error to zero such that operating costs of power
system are at a minimum [2].
In a power system, change in load is reflected throughout the system as a change
in frequency. In response to change in system frequency, load frequency control
(LFC) is regulating generator output within a prescribed limit. LFC controls the
frequency deviation, as well as tie-line power flow for maintaining system frequency
and scheduled interchange of power with inter-areas within prescribed limits [2].
Penetration of distributed generation (DG) into the power system is quantified by
hosting capacity approach. This approach aids in better understanding the customer
requirement on the system and requirements of the system operator on customer for
reliable operation. The maximum value of DG penetration up to which power system
operates smoothly without any disturbance is termed hosting capacity. This is
determined by performance index of DG penetration limit. Power system operates
satisfactorily if the performance index is less than hosting capacity.
The European Distributed Energy Partnership (EU-DEEP) project had been
started in 2004 for removing technical and non-technical barriers that prevent a
massive deployment of distributed energy resources (DER) in Europe. Hosting
capacity concept was introduced in a systematic approach starting with the definition
of a number of performance indices and performance objectives. Hosting capacity
calculation varies differently based on power quality events. Hosting capacity
calculation of voltage variations is different from hosting capacity calculation of
frequency variations. The hosting capacity varies depending on structure of the
network, type of DG unit, load, climate, and type of power quality event. There is
no existence of general objectives for power quality events. There are no general
accepted limits for interruptions and dips. Some countries use maximum number of
interruption and maximum duration of interruption as indices and limits to determine
the hosting capacity of their networks [28]. The main purpose of this chapter is to
present the mathematical model for incorporating hosting capacity using power flow
equations in large signal and transfer function model in small signal analysis.
9.2 Regulation Control Schemes
In general, power system dynamics is classified as large signal (static) and small
signal (dynamic) analysis [7]. Large signal analysis is used to study major distur-
bances, in which the magnitude of voltage and power may suddenly increase up to
100% of normal operating values [6–10]. Static equations are used in steady-state
power flow analysis to control existing power system by calculating steady-state
powers and voltages at various buses [10–12]. Small signal analysis is used when
minor disturbances occur, in which the magnitude of voltage and power may increase
few percent of normal operating values. Linear differential equations and Laplace
transform analysis are used for finding the solution in small signal analysis. The
renewable energy generation is usually analyzed by small signal analysis which is
dynamic in nature, since the electrical parameters pertaining to DG such as voltage,
9 Generation Regulation Control Systems 197
Fig. 9.1 Control strategies for DG connected to microgrid
current, and frequency are always continuously varying in nature and time depen-
dent. The control strategies for DG connected to microgrid are shown in Fig. 9.1.
DG operations in microgrid are classified as (i) grid-connected mode and
(ii) island or autonomous mode [3]. In grid-connected mode, main grid is absorbing
or delivering power to the microgrid similar to an infinite bus [4]. In island mode,
supply and demand have to be managed by DG connected to microgrid without
power from the main grid [5]. The control strategies for island mode are classified
further into two categories: communication based and droop based.
9.2.1 Large Signal Analysis
In large signal analysis, speed and voltage control of a generator is achieved by

automatic generation control (AGC) in a power system which consists of conven-
tional energy sources. AGC controls the governor, based on the frequency deviation
for adjusting the generation according to the load to minimize frequency deviation
and tie-line error to the prescribed limit when disturbance occurs. The real power and
system frequency of power system are strengthened depending on the amount of DG
incorporated at the same time increasing hosting capacity issues.
In developed countries, deregulation of power system is in practice, and sufficient
generating units are always kept under reserve. Load sharing is carried out by
economic dispatch controller (EDC), and the selection of utility is decided based
on market price. Developing countries follow vertically integrated utility (VIU)
where the supply-demand gap always exists. The majority share of supply-demand
gap balancing is achieved by controlling the output of conventional generation
resources. In addition, with conventional energy sources, DG can be integrated
using AGC. Hence, modification is required in AGC to integrate DG in addition to
conventional energy sources. Research challenges are involved to integrate DG into
the conventional power systems [6]. In practice, frequency deviation is measured in
the power system network by using frequency meter, and the measured data such as
real power, reactive power, system frequency, and voltages at various buses are sent
to the energy control center (ECC). AGC will adjust the governor according to the
measured frequency deviation and tie-line error to generate required power to match
the load in economic way using EDC.
Solar power penetration into the grid is achieved by grid tie inverter. Wind power
generation is used to inject power into the grid when the induction motor speed is
more than synchronous speed. Thus, the DG penetration into the grid is achieved
based on the availability of natural resources. At present, the contribution of DG is
tremendously increasing worldwide due to their inherent advantages [13]. At the
same time, integration of DG into a power system is a challenging task because of
the dynamic nature of the resources [13, 14]. Setting limits to the introduction of DG
to power system network is only possible when there is agreement on performance
indices and objectives. If the power quality requirements on the network operator are
too narrow, then lesser DG can only be connected to power system network. It
necessitates the need of a wider range of performance indices and objectives [29]. In
this section, method of computing frequency deviation, voltage magnitude variation,
and tie-line error between areas from load flow equations have been explained by
mathematical model.
9.2.1.1 Generator Model
Generator’s prime mover responses according to the load change which has been
initiated by governor action of AGC. The generator model can be represented by the
mathematical model with the following nomenclature:
αi Participation factor of ith generation
B Bias factor setting of AGC regulator constant for area load-frequency
characteristics
F0 Scheduled system frequency in p.u.
ΔF Frequency deviation
Pgi Active power generation
Pgseti Active power generation schedule
ΔPgi Active power generation due to primary and secondary control
Pgmin i Minimum active power generation
Pgmax i Maximum active power generation
PT Actual tie-line power flow
PT0 Scheduled tie-line power flow
ri Speed-droop setting on turbine governor of ith generating unit
The nomenclatures are mentioned for ith bus in p.u.

Active power generation Pgi is obtained by
Pgi ¼ Pgseti þ ΔPgi ð9:1Þ
where Pgmax i > Pgi > Pgmin i
ΔPgi ¼ ð1=r i ÞΔF þ αi ðΔPtie þ BΔF Þ ð9:2Þ

ΔF ¼ F F 0 &ΔPtie ¼ PT PTO ð9:3Þ
The generation in each area is to be controlled to maintain the scheduled power

interchange [14] when an interconnected power system consists of two or more
independent control areas in addition to control of frequency within the area. The
effect of DG on tile-line power flow is ignored in this chapter. By ignoring tile-line
power flow, Eq. 9.2 becomes
ΔPgi ¼ ð1=r i ÞΔF ð9:4Þ
9.2.2 Mathematical Model for Regulation Characteristics
The mathematical modelling of regulation characteristics is obtained by power flow

equations with the following nomenclature:
N Number of buses
M Voltage controlled buses
(N 1) Number of voltage phase angles except the reference bus
For a N bus system, bus 1 is considered as reference bus for the purpose of voltage
and phase angle calculations of other buses. (N M) number of voltage magnitudes
are unknowns at (N M ) buses. From complex power balance equations at N buses,
2N number of nonlinear equations are obtained by separating real and imaginary
parts. This is solved by the application of decoupled Newton-Raphson (NR) method
as follows:
Si ¼ Pi þ jQi ð9:5Þ

Pi ¼ Pgseti þ Pgi PLi ð9:6Þ
Pgseti ¼ Pgi þ PLi þ Pi ð9:7Þ
Qi ¼ Qgi QLi ; Qgi ¼ QLi Qi
With bus 1 as reference, the equations for computing the voltage and phase angle
of other buses using decoupled Newton-Raphson iterative solution method are
given:
2 3 2 3
ΔP1 2 3 Δx
6 7 ∂P1 =∂x ∂P1 =∂δ2 ∂P1 =∂δN 6 7
6 ΔP 7 6 76 Δδ 7
6 2 7 ∂P =∂x ∂P =∂δ ∂P =∂δ 6 7
7¼6 N 76 2
6 6
2 2 2 2
7 7 ð9:8Þ
6 7 4 56 7
6
6 7 7
4 5 4 5
∂PN =∂x ∂PN =∂δ2 ∂PN =∂δN
ΔPN ΔδN
2 3
ΔQ1 2 3
6 7 ∂Q1 =∂V 1 ∂Q1 =∂V 2 ∂Q1 =∂δNM
6 7
ΔQ2 7 6
6
6 7¼6 ∂Q2 =∂V 1 ∂Q2 =∂V 2 ∂Q2 =∂δNM 7 7
6 7 4 6 7
6 7
5 5
4
∂QNM =∂V 1 ∂QNM =∂V 2 ∂QNM =∂δNM
ΔQN
2 3
ΔV 1
6 7
6 ΔV 7
6 2 7
66
7
7 ð9:9Þ
6 7
4 5
ΔV NM
The above relations can be written as

Δx
½ΔP ¼ ½J 1 ð9:10Þ
Δδ
½ΔQ ¼ ½J 2 ½ΔV ð9:11Þ
where the size of J1 is [N, N] and J2 is [(N M), (N M)].

In Eq. (9.10), Δx will be replaced by ΔF to calculate frequency deviation. The
same procedure is repeated to find tie-line error by replacing Δx by ΔPtie. The active
and reactive powers are calculated using Eqs. (9.10) and (9.11) by assuming that
initial bus voltages are 1 ∠ 0. The differences between the specified and calculated
values give the changes in power.
ΔPK ¼ P specified P calculated

ΔQK ¼ Q specified Q calculated
Elements of the Jacobian matrices J1and J2 are obtained from estimated bus
voltages, x, and calculated powers. ΔQ and ΔV are obtained by the triangularization
method from the set of linear Eqs. (9.10) and (9.11) that are solved. Frequency
deviation and tie-line error are treated as Δx is in this work. To observe range of
ΔF and ΔV from minimum to maximum, generation and load are varied accordingly,
and economic allocation of generations is stored in a lookup table.
9.2.2.1 Algorithm for AGC in Large Signal Analysis
The following steps are used to calculate the steady-state frequency deviation and
tie-line exchange deviation using decoupled NR method [8, 16, 17] which is one of
the inputs to the fuzzy logic controller (FLC). Flat frequency control (FFC), flat
tie-line control (FTC), and bias tie-line control (BTC) are different types of AGC
control strategies. This analysis is assumed in an FFC environment to integrate DG.
I. Calculation of AGC Components
1. Input power system data related to:
(a) Network, scheduled loads, and primary and secondary control
characteristics
(b) Type of control strategy used (in this case FFC)
2. Form the bus admittance matrix.
3. Evaluate steady-state frequency deviation ΔF.
Subroutine:
(i) Calculate active power mismatch vector [ΔP].
(ii) Check active power mismatch is within specified tolerance (ε) at all the
buses; if [ΔP < ε], then go to step (iv); else, continue.
(iii) Calculate the Jacobian matrix [J1], vector of unknowns [Δx], and update
voltage [Δx] and phase angles [Δδ].
(iv) Calculate the reactive power mismatch vector [ΔQ].
(v) Check reactive power mismatch is within specified tolerance (ε) at all the
buses; if [ΔQ < ε], then go to main routine; else, continue.
(vi) Compute the Jacobian matrix [J2] and vector of unknowns [ΔV], update
the estimated voltage magnitude [V], and go to step (i).
4. Check whether [ΔF < ε]. If ΔF lies within specified limits, print the results.
5. If ΔF is not less than ε, the corrective action ΔPgi is determined by conven-
tional controller, FLC, and artificial neural network controller (ANNC). ΔPgi
is added with existing generation Pgi and repeat the steps from number 3.
In large signal analysis, frequency deviation and voltage deviations are consid-
ered for performance indices. The allowable limit of frequency deviation is
0.05 Hz and voltage magnitude 0.05 p.u. as shown in Fig. 9.2. If hosting capacity
exceeds performance indices, then power system will not perform satisfactorily.
Control action of AGC is performing in between negative to positive limit for
satisfactory action of power system.
Fig. 9.2 Holding capacity versus performance indices
9.2.2.2 Assumptions Made for Large Signal Analysis
Performance indices are assumed as [ΔF] and [ΔV].

The values are assumed [ΔF] as 0.05 Hz and [ΔV] 0.05 p.u.
9.2.2.3 Flowchart for AGC
Flowchart is shown in Fig. 9.3.
9.3 Design of Controllers
9.3.1 Conventional Controller Model
The design of conventional controller is obtained by considering system frequency

and tie-line power flow as feedback signals for the controller. These signals are
obtained from regulating characteristics (ri) and participation factors (αi) of the
generator units using Eq. (9.2) to evaluate the required change in generation. The
designed conventional controller is represented in Fig. 9.4.
αi ¼ ΔPdi =ΔPD ,
ΔPdi¼ unit load change

ΔPD¼ sum of individual unit load changes
B ¼ (1/ri) + D is constant [8]
ri – p.u. change in frequency in load/p.u. change in unit output ¼ 0.05 as in [8]
D percentage change in load/percentage change in frequency ¼ ΔPL/Δω
Fig. 9.3 Flowchart for AGC

Fig. 9.4 Conventional controller
9.3.2 Fuzzy Logic Controller
Fuzzy logic controller is used for decision-making based on the availability of wind
velocity, solar irradiation, and the calculated steady-state frequency deviation of
power system. The design procedure of FLC is discussed in the following section.
9.3.2.1 Selection of Input and Output Variables
Three inputs have been considered to design FLC, i.e., ΔF calculated from power
system network data using decoupled NR method, solar irradiation, and wind speed.
The output of FLC is the available amount of power generation from solar and wind
resources.
9.3.2.2 Ranges of Input and Output Variables
The input ΔF varies from 0.025 to 0 Hz, solar irradiation varies from 0 to 1500
W/m2, and wind velocity varies from 0 to 20 m/s.
9.3.2.3 Knowledge Base
The input and output variables are divided into a suitable number of linguistic
variables [17]. Similarly, the output solar and wind power generated is divided
into a suitable number of linguistic variables. Based on the number of linguistic
variables, the rule base is formed. The input and output conditions have been
described by fuzzy sets. For example, “if solar irradiation is very high, wind speed
is very high, and ΔF is zero, then SPV is no power and wind power (WP) is excess.
When this rule is executed power will not be generated from DG.” These rules can be
developed from operator’s practical experience, knowledge, survey results, general
principles, etc.
9.3.2.4 Defuzzification
The process of converting a fuzzified value into actual crisp value is known as
defuzzification. A centroid defuzzification method is normally used to represent the
center of gravity as in (9.12) of the fuzzy set.
PR 0
μr H r
u ¼ Pr¼1
R ð9:12Þ
r¼1 H r
The defuzzified values from FLC are the required amount of power generation
from DG to minimize the AGC component.
9.3.3 Artificial Neural Network Controller (ANNC)
A multilayer perceptron (MLP) network is typically composed of input layer, hidden

layer, and output layer. An input layer is receiving external information, and output
layer is presenting the solution. Hidden layers separate the input layer and output
layer. Hidden layer has no direct contact with the external environment, and it is
internal to network. Every output layer is connected to every node of the next
network layer and is said to be fully connected. Figure 9.5 shows a fully connected
multilayer feed-forward neural network.
ANN must be trained before performing any task. The process of determining and
updating the connecting (synaptic) weights is known as training of neural network.
Training is the key element of an ANN. The learning of the network is stored in the
synapses in the form of synaptic weights and nodes in the form of node biases. An
ANN is trained for complex nonlinear mapping through connecting weights from
input nodes to output nodes. The desired output (target value) of the network for
each input pattern is always available for supervised MLP training.
Input variables or training patterns are the training input data in the form of
vectors. Input vector of each element is connected to an input node of input layer in
Fig. 9.5 A typical feed-

forward neural network
the network, and so the input vectors are equal to the number of input nodes.
Estimation of connection (synaptic) weights is done by the training set. Measure-
ment of the generalization ability of the network has been performed by the test set.
The synaptic weights are considered as free parameters. Hidden layers are using
nonlinear activation functions such as logistic sigmoid and hyperbolic tangent.
9.3.4 Incorporation of DG for Large Signal Analysis

with Centralized RES
In the proposed method, the hosting capacity of the system [ΔF and ΔV] is
calculated from the steps given in flow chart Fig. 9.3 using load flow equations.
Performance indices are assumed as 0.05 Hz and 0.05 p.u. After penetration of
DG, sudden changes in input parameters such as solar irradiation wind velocity lead
to unwanted fluctuations in voltage, frequency, and power. The hosting capacity
exceeds performance indices and creates a power system stability problem. Sudden
decrease or increase in load also results in achieving higher value of hosting capacity
than performance indices and causes problem in power system. The operator guide-
line has to be fixed by company authories as a policy to choose of their own
performance indices.
9.4 Small Signal Analysis
Small signal analysis is performed by modelling state space equations of single-area,

two-area, and multi-area power system. Small signal analysis is dealt with variables
such as deviations of frequency, angle, and voltage from its normal operating
value, ΔF, Δδ, and ΔV, respectively. Laplace transformation has been used for
solving small signal model analysis.
9.4.1 Conventional Generator Control Loops
All generators are installed with LFC and automatic voltage regulator (AVR) in an
interconnected power system. The controllers are designed to take care of sudden
disturbances in load to maintain frequency and voltage magnitude within the spec-
ified limits. Changes in rotor angle cause disturbance in frequency as well as real
power. Voltage magnitude affects the reactive power in the power system. When
time constant of prime mover and excitation system is compared, prime mover time
constant is larger than excitation system time constant. Hence, comparing the effect
of LFC loop with AVR loop, AVR loop is neglected. In practice, AVR loop and LFC
loop are analyzed independently.
9.4.2 Single-Area System
Turbine, generator, governor, and load are connected to feedback loop as is shown in
Fig. 9.6 to represent in block diagram an isolated power system. In block diagram,
K and T represent gain and time constant of main component governor (Kg and Tg),
turbine (Kt and Tt), generator ((Kg and Tg), and sudden load change (ΔPD).
9.4.3 Design of Controllers
Conventionally, PID controllers are used for small signal analysis in AGC. PID
controllers consist of the following elements: The different combinations of P, I, D,
PI, PD, and PID controllers are possible to implement in AGC. The integral
controller is an effective controller for AGC as per the literature. FLC and ANNC
can be designed for controlling AGC of small signal analysis with the same
techniques discussed in Sects. 9.3.2 and 9.3.3.
9.4.4 Incorporation of DG for Small Signal Analysis
The corrective action determined by integral controller, FLC, and ANNC has to be
checked with availability of solar and wind power generation. The available DG can
be incorporated with various types of droop control and communication control
strategy as discussed in Sect. 9.2.
Fig. 9.6 Single-area system without controller

9.4.5 Droop-Based Control Strategy
9.4.5.1 Conventional Droop Control Strategy
AC and DC microgrids are widely controlled by droop control scheme [5]. In AC

microgrid, voltage magnitude V and frequency f are determined from the measure-
ment of real power P and reactive power Q using the following equations:
f y ¼ f x þ mx Px ð9:13Þ
V y ¼ V x þ nx Qx ð9:14Þ
where Vx is the maximum voltage at no load, fx is the maximum frequency at no load,

and mx, nx are negative droop coefficients.
In DC microgrid, reactive power and frequency terms are not present, and active
power P is only considered. The DC droop equations are written using the following:
V dc,y ¼ V dc,x þ kx Pdc,x ð9:15Þ

k1 Sdc,1 ¼ k 2 Sdc,2 ¼ . . . ¼ k x Sdc,x ð9:16Þ
9.4.5.2 Angle Droop Control
Real and reactive power flow control is obtained from an angle of voltage in an angle
droop control.
Equations (9.19) and (9.20) give the relation of P–δ and Q–V.

δj ¼ δrat ma Pj P j,rat ð9:17Þ

V j ¼ V rat na Qj Q j,rat ð9:18Þ
where δrat is the rated voltage angle, Vrat is the rated voltage magnitude, and ma, na
are angle droop coefficients.
Angle droop control is utilized in scenario such that DG coupled converters are
used to form an AC microgrid. In angle droop control, output voltage angle of VSC
is instantaneously changed to reach steady state. Droop coefficient values are
determined by DG ratings for voltage regulation.
9.4.5.3 Droop-Based Voltage and Frequency Control Strategy
An active power-frequency (P f ) droop is used for frequency stability by

balancing active power, and reactive power-voltage (Q V ) droop is used for
voltage stability by balancing reactive power. In practice, Battery Energy Manage-

ment System (BEMS) is built using bidirectional voltage source converter (VSC)
that consists of external and internal embedded cascaded control loops.
An external loop is controlled (P f ) droop and (Q V ) droop. Battery Energy
Storage active and reactive power is given by:
PB ¼ PL PðPVþWÞ ð9:19Þ
QB ¼ QL QðPVþWÞ ð9:20Þ
Frequency droop ( fdroop) is derived from active power of battery PBattery using
Eq. (9.19). Voltage droop (Vdroop) is derived from reactive power of battery QBattery
using Eq. (9.20). Control relationships for fdroop and Vdroop from Eqs. (9.19) and
(9.20) can be written as

f droop ¼ f at noload m PBðat fullloadÞ PBðat noloadÞ ð9:21Þ
h i
V droop ¼ V at noload n QBðat fullloadÞ QBðat noloadÞ ð9:22Þ
System stability is controlled within their acceptable limits by choosing droop

control coefficients m and n using f droop and V droop from Eqs. (9.21) and (9.22).
Droop characteristics for sharing the power under various operating conditions of
standalone hybrid renewable energy system (SHRES) are illustrated in Table 9.1.
An internal control loop is controlled in three-phase voltage reference frame
based on f droop and V droop with the following equations:
8
< V aref ¼V droop sin ðωt þ θÞ
>

V bref ¼V droop sin ωt 120 þ θ ð9:23Þ
>
:
V cref ¼V droop sin ωt 240 þ θ
By integrating f droop in Eq. (9.23), the phase angle θ is derived. Voltage and
current controllers are used to configure the internal control loop.
Droop control can be implemented even sources are separated by long distances
with low bandwidth data lines. However, for nonlinear loads, droop control scheme
does not consider current and voltage harmonics. The presence of circulating
currents makes reactive power control difficult. The major disadvantage of using
droop control is the need for a secondary control to compensate steady-state error.
During operation of the system, droop coefficients are normally constant, and
occasionally, re-set will be done according to the system requirement.
Table 9.1 Droop characteristics of power sharing

Voltage
Operating and
conditions frequency Droop characteristics Power sharing
No DG Droops At steady-state value, the The total power demand (PD) is
generation decrease in voltage and fre- supplied by the battery energy
quency is regulated storage
DG gener- Decreases Within acceptable limits, the Battery discharge power proportional
ation <PD decrease in voltage and fre- to decrease in bus voltage and fre-
quency is regulated quency up to minimum discharging
limit
DG gener- Stable At the nominal value, the The total PD is supplied by PV and
ation ¼PD voltage and frequency values wind sources while battery energy
are maintained stable storage is operated at idle state
DG gener- Increases Within acceptable limits, the Battery absorbs excess power pro-
ation >PD increase in voltage and fre- portional to increase in bus voltage
quency is regulated and frequency up to the maximum
charging limit
9.4.6 Incorporation of Hosting Capacity in Small Signal

Analysis with Decentralized RES
In small signal analysis, frequency and voltage deviations [ΔF and ΔV] are consid-
ered for hosting capacity, and performance indices are assumed with allowable limit
of 0.05 Hz and 0.05 p.u. Implementation of FLC by replacing integral controller
for area 1 is shown in Fig. 9.7.
In the same way, multi-area can be controlled by replacing conventional control-
ler shown in Fig. 9.8 for a multi-area system. Individual FLC and ANNC for each
area or single FLC and ANNC for three areas can be developed for a decentralized
multi-area system [30]. If hosting capacity exceeds performance indices, then power
system will not perform satisfactorily. The important task of various controllers is to
maintain performance indices within the hosting capacity.
9.4.6.1 Assumptions Made for Small Signal Analysis
Performance indices are assumed as [ΔF] and [ΔV].

The values are assumed [ΔF] as 0.05 Hz and [ΔV] 0.05 p.u.
Fig. 9.7 Incorporation of hosting capacity in small signal analysis in area 1
Fig. 9.8 Incorporation of hosting capacity in small signal analysis in three area system [30]
9.5 Case Study
9.5.1 Case Study 1
A stand-alone 95-kW capacity microgrid is installed in Gasa Island, South Korea

[25]. This microgrid is considered for case study of DG and power demand. Korea
Electric Power Corporation (KEPCO) constructed a microgrid in Gasa Island. An
automated energy management system (EMS) is developed to forecast demand,
track electricity consumption and decision-making of power delivers to load or
store energy in Battery Energy Storage System (BESS). Performance analysis of
connected loads to the microgrid is an important design parameter in stand-alone
microgrid. Connected load in Gasa Island consists of a school, a town hall, a

lighthouse, and seaweed and fish farms. An average monthly load of Gasa Island
is shown in Fig. 9.9. Monthly average global horizontal irradiance (GHI) for a 1-year
period is shown in Fig. 9.10. Monthly average wind speed for a 1-year period is
shown in Fig. 9.11. DER locations and single line diagram of GASA Island as in [25]
are shown in Figs. 9.12 and 9.13, respectively. DG and load demand data are given
below as per NASA Surface Meteorology and Solar Energy website:
• Annual average solar radiation is 4.01 kW/m2/day.
• Annual average wind speed is 6.07 m/s at 50 m above the sea level.
• Average load demand is 2164 kWh/day with a load factor of 0.48.
Fig. 9.9 Average daily load

profile as in [25]
Fig. 9.10 Month-wise

average GHI as in [25]
Fig. 9.11 Month-wise

average wind speed as in
[29]
Fig. 9.12 DER locations in

GASA Island as in [25]
9.5.2 Case Study 2
Seoul National University (SNU), South Korea [26], is planning to implement DC

microgrid in campus. Rooftop 36-kW solar PV panels are connected through
converters to form DC grid. EVs, AC utility, AC and DC hybrid distribution system,
and building integrated photovoltaic (BIPV) cells are connected to the DC
Fig. 9.13 Single line diagram of electric circuit model of Gasa Island microgrid
microgrid. DC microgrid is controlled in two ways as shown in Fig. 9.14. Overview

of Korea Micro Energy Grid (K-MEG) is shown in Fig. 9.15.
The advantages and disadvantages of DC microgrid control strategy are as
follows:
• Centralized control
Advantage: Simple implementation and improved regulation
Disadvantage: Communication burden and single point of failure
• Decentralized control
Advantages: Autonomous operation, high reliability, and high expandability
Disadvantages: Low control flexibility and regulation trade-off
• A low-voltage DC distribution has the following advantages:
• Reduction in converters between DG and loads.
• Increase in system efficiency due to reduction in converters.
• Less number of converters will reduce harmonics in power supply.
• Improvement in power quality.
• No need of synchronization.
• No need of reactive power compensation.
• Simplified converter design.
The main motivations of SNU to implement DC distribution are as follows:
• Increase the usage of DC loads such as consumer electronics, digital, LED,
electric vehicles (EVs), and inverter-based loads
• Penetration of DG into mail grid if surplus power is available
• Development of power electronics and device technology
Theoretically, combining conventional energy sources and renewable energy
sources is a difficult task because of their static and dynamic nature. Automatic
generation control including renewable energy sources is successfully implemented
in this algorithm. Similar type of work is carried out in Canada.
Central Decentral
Central
High band Controller Local Local Local
communication link Controller Controller Controller
Distributed Distributed Distributed

Distributed Distributed Distributed
Generation Generation Generation
Generation Generation Generation
DC bus Power link DC bus Power link
DC loads
DC loads
Fig. 9.14 Representation of centralized and decentralized DC microgrid [25]
Fig. 9.15 Overview of Korea Micro Energy Grid (K-MEG)
9.6 Communication-Based Control
Communication-based control uses communication lines with high bandwidth for

sharing information between sources and loads. Communication-based control is
exchange of information within the modules using high bandwidth [19–23]. Actual
information can be communicated with less delay as compared to autonomous
strategies for achieving high accuracy. However, there are chances of loss of
communication link which causes system stability, security, and reliability issues.
These problems can be avoided by combining droop and communication-based

methods using the hierarchical control structures proposed in [24–27]. Both control
strategies are used to achieve localized device-level control and system-level coor-
dination simultaneously.
9.6.1 Software and IoT Applications
The Internet of Things (IoT) is a platform to connect physical world to digital world.
Any objects physically not connected with any type of network can be connected by
IoT. Various sensors are used to quantify physical parameters, and actuators are used
to change the condition of the Internet. Radio-frequency identification (RFID) is
encoded in digital data tags to find, distinguish, compute, and store data using cloud
technology in IoT. The term “Things” refers to the everyday physical objects, and
the term “Internet” represents the interconnection between networks of heteroge-
neous objects. There are many potential research areas of interest, and one of them is
software architecture and solution. A class is formed for a set of IoT solutions to
illustrate the solution space by investigating the similarities and differences of few
IoT solutions. At the same time, it is not possible to identify a complete set of classes
for achieving such a one-size-fits-all architecture.
In literature, most importantly, the following two qualities are discussed by
various authors:
• IoT is a network which referred to singular form with only one IoT.
• Characteristic definitions refer to the things in the IoT.
• These physical objects are installed by innovation for.
• Associate with the IoT.
• State measurement and their changes.
• Communicate with each other for greater value and service.
Since the IoT applications are enormous variety, accomplishing greater value and
service is extremely broad. Greater value and service is achieved by:
• Formation of new functionality, previously not possible without physical network
• Increasing the quality of existing processes with the help of the IoT
The requirement of IoT should have the following property:
• Self-conjuring, always adaptable for change
• Accessible everywhere
• Seamless communication infrastructure with the following:
(i). One-to-one relationship
(ii). One-to-many relationship
(iii). Many-to-many relationships
The present state of the IoT can be categorized into connected home, connected
body, associated retail, associated transportation, smart city, industrial application,
E-Health, and smart energy.
Methodology for analyzing the solutions is performed by identifying the follow-
ing variables:
• Physical Entity being measured
• Attribute(s) of the physical element being measured
• Type of IoT connector(s) used in the solution
• Topology of diagram showing the relationship between parts
• Direction of messages passed between the components in a diagram
• Area of application with logic and data storage
• Possibilities of client interaction
• Estimation of the scalability requirement of each component with the following:
(i). Fixed scale
(ii). Potential increase
9.6.2 Implementation of IoT for DG

9.6.2.1 Case Study 1
Implementation of IoT-based monitoring system consists of identification of system

architecture, program analysis, and implementation method of DG. The open IoT
platforms available are Arduino, Raspberry Pi, and low-cost Long Range (LoRa)
network [28]. Architecture of energy monitoring system is shown in Fig. 9.16, and
its purposes are given below:
• Energy IoT node: Power generation data collection status from energy device
• IoT gateway: Data capturing and storing from nodes at remote site
• Low cost LoRa network: Wide area networking (WAN) for low-cost wireless
communication
Serial interface is used to connect energy device controller and Arduino-based
IoT node. Voltage, current, temperature, and battery current status information are
collected and sent to IoT. The LoRa modem and Arduino are connected in serial
interface. The application program interface (API) function from the LoRa protocol
stack will be called by the embedded application. IoT node utilizes the power level
sleep function to support a low-power mode and is configured by periodically wake-
up function. Arduino-based IoT sensor platform is shown in Fig. 9.17.
Raspberry Pi operates as a server of energy IoT gateway for monitoring the
system accessed by web protocol. The transfer is done by LoRa modem for sending
and receiving energy data to Raspberry Pi through serial interface as shown in
Fig. 9.16 Energy monitoring system architecture
Fig. 9.17 Arduino-based IoT sensor platform
Fig. 9.18. Control, radio, and timestamp information are also carried out by Rasp-
berry Pi. Mongo data base (DB) stores data received through the LoRa modem and
works with NoSQL method to support the structure suitable for large-capacity big
data.
9.6.2.2 IoT Impact on the HC
IoT-based monitoring and control of microgrid required a reliable router. Router

performance is based on the number of IoT devices connected and the number of
video streams activated at once. The user intends to add high-bandwidth streams;
even though the total value router capacity is less than theoretical throughput, a
higher quality of router is required. Implementation of IoT in microgrid is creating
limitation on hosting capacity.
9.6.2.3 Future Work
In this chapter, mathematical modelling of incorporating hosting capacity using

power flow equations in large signal and transfer function model in small signal
Fig. 9.18 Raspberry Pi monitoring system
analysis are discussed. A mathematical model including hosting capacity with all
possible variables can be developed in the future. The requirement of bandwidth of
IoT may be increasing in the future. Multi-agent system (MAS) based control
scheme is used to monitor, coordinate, control and display data of the system
[6]. These tasks are assigned with various agents such as DER agent, control
agent, database agent, and visualizer or user agent in energy control center. Imple-
mentation of MAS will replace the burden on IoT; in turn, hosting capacity may
increase to higher value.
References
1. Basso, T., & Friedman, N. R. (2003). IEEE 1547 national standard for interconnecting
distributed generation: How could it help my facility? Preprint (No. NREL/JA-560-34875).
National Renewable Energy Lab., Golden, Co.(US).
2. Kanniah, J., Tripathy, S. C., Malik, O. P., & Hope, G. S. (1984). Microprocessor-based adaptive
load-frequency control. IEE Proceedings C Generation, Transmission and Distribution, 131(4),
121–128.
3. Radwan, A. A. A., & Mohamed, Y. A. R. I. (2014). Bidirectional power management in hybrid
AC-DC islanded microgrid system. In 2014 IEEE PES General Meeting Conference & Expo-
sition (pp. 1–5). IEEE.
4. Nejabatkhah, F., & Li, Y. W. (2014). Overview of power management strategies of hybrid
AC/DC microgrid. IEEE Transactions on Power Electronics, 30(12), 7072–7089.
5. Majumder, R., Chaudhuri, B., Ghosh, A., Majumder, R., & Ledwich, G. F. (2010). Improve-
ment of stability and load sharing in an autonomous microgrid using supplementary droop
control loop. IEEE Transactions on Power Systems, 25(2), 796–808.
6. Manickavasagam, K. (2014). Intelligent energy control center for distributed generators using
multi-agent system. IEEE Transactions on Power Systems, 30(5), 2442–2449.
7. Elgerd, O. I. (1982). Electric energy systems theory: An introduction. McGraw-Hill,
New York, US.
8. Subbaraj, P., & Manickavasagam, K. (2007). Generation control of interconnected power

systems using computational intelligence techniques. IET Generation, Transmission & Distri-
bution, 1(4), 557–563.
9. Michael, N. (2005). Artificial intelligence a guide to intelligent systems, Pearson Education
Limited, New Delhi, India
10. Dos Santos, M. J., Pereira, J. L. R., De Oliveira, E. J., & da Silva, I. C. (2004). A new approach
for area interchange control modeling. IEEE Transactions on Power Systems, 19(3),
1271–1276.
11. Chang, S. K., & Brandwajn, V. (1988). Adjusted solutions in fast decoupled load flow. IEEE
Transactions on Power Systems, 3(2), 726–733.
12. Saikia, L. C., Nanda, J., & Mishra, S. (2011). Performance comparison of several classical
controllers in AGC for multi-area interconnected thermal system. International Journal of
Electrical Power & Energy Systems, 33(3), 394–401.
13. Moreno-Munoz, A., De La Rosa, J. J. G., Lopez, M. A., & De Castro, A. G. (2009). Grid
interconnection of distributed generation: The Spanish normative. In 2009 International Con-
ference on Clean Electrical Power (pp. 466–470). IEEE.
14. Dugan, R. C., McDermott, T. E., & Ball, G. J. (2001). Planning for distributed generation. IEEE
Industry Applications Magazine, 7(2), 80–88.
15. Logenthiran, T., Naayagi, R. T., Woo, W. L., Phan, V. T., & Abidi, K. (2015). Intelligent
control system for microgrids using multiagent system. IEEE Journal of Emerging and Selected
Topics in Power Electronics, 3(4), 1036–1045.
16. Thukaram, D., Iyengar, R., Khincha, H. P., & Parthasarathy, K. (1984). Steady state power flow
analysis incorporating load and generation regulation characteristics. Journal of the Institution
of Engineers (India), 64(5), 274–279.
17. https://www.springer.com/productFlyer_978-3-540-76283-6.pdf?SGWID¼0-0-1297-
173779129-0
18. Eid, B. M., Rahim, N. A., Selvaraj, J., & El Khateb, A. H. (2014). Control methods and
objectives for electronically coupled distributed energy resources in microgrids: A review.
IEEE Systems Journal, 10(2), 446–458.
19. Ren, W., & Beard, R. W. (2008). Distributed consensus in multi-vehicle cooperative control
(pp. 71–82). London: Springer.
20. Behjati, H., Davoudi, A., & Lewis, F. (2014). Modular DC–DC converters on graphs: Coop-
erative control. IEEE Transactions on Power Electronics, 29(12), 6725–6741.
21. Nasirian, V., Moayedi, S., Davoudi, A., & Lewis, F. L. (2014). Distributed cooperative control
of DC microgrids. IEEE Transactions on Power Electronics, 30(4), 2288–2303.
22. Baghaee, H. R., Mirsalim, M., & Gharehpetian, G. B. (2016). Performance improvement of
multi-DER microgrid for small-and large-signal disturbances and nonlinear loads: Novel
complementary control loop and fuzzy controller in a hierarchical droop-based control scheme.
IEEE Systems Journal, 12(1), 444–451.
23. Zhao, Z., Yang, P., Guerrero, J. M., Xu, Z., & Green, T. C. (2015). Multiple-time-scales
hierarchical frequency stability control strategy of medium-voltage isolated microgrid. IEEE
Transactions on Power Electronics, 31(8), 5974–5991.
24. Wei, B., Yuehao, Y., Bo, K., Yuanhong, C., & Xin, L. (2015). An novel hierarchical control of
microgrid composed of multi-droop controlled distributed power resources. In 2015 5th inter-
national conference on Electric Utility Deregulation and Restructuring and Power Technolo-
gies (DRPT) (pp. 2173–2178). IEEE.
25. Husein, M., Hau, V. B., Chung, I. Y., Chae, W. K., & Lee, H. J. (2017). Design and dynamic
performance analysis of a stand-alone microgrid. Journal of Electrical Engineering & Tech-
nology, 12(5), 1777–1788.
26. http://microgrid-symposiums.org/wpcontent/uploads/2014/12/tianjin_bo-hyung-cho.pdf
27. Choi, C. S., Jeong, J. D., Lee, I. W., & Park, W. K. (2018). LoRa based renewable energy
monitoring system with open IoT platform. In 2018 international conference on Electronics,
Information, and Communication (ICEIC) (pp. 1–2). IEEE.
28. Nayanatara, C., Divya, S., & Mahalakshmi, E. K. (2018). Micro-Grid management strategy
with the integration of renewable energy using IoT. International conference on Computation of
Power, Energy, Information and Communication (ICCPEIC) (pp. 160–165), IEEE.
29. Yang, Y., & Bollen, M. H. J. (2008). Power quality and reliability in distribution networks with
increased levels of distributed generation’ (Report Elforsk, Stockholm, Sweden, 2008).
30. Saha, A., Saikia, L. C., Tasnin, W., Rajbongshi, R., & Saha, D. (2018). Automatic generation
control of multi-area multisource system incorporating distributed generation units and RFB.
2nd international conference on Power, Energy and Environment: Towards Smart Technology
(ICEPE) (pp. 1–6). Shillong, India.
Chapter 10
Exploring the Concept of Hosting Capacity
from an Electricity Market Perspective
Elias Valenzuela, Rodrigo Moreno, Dimitrios Papadaskalopoulos,

Francisco D. Muñoz, and Yujian Ye
Nomenclature
In this section, the notation of the most important terms used throughout this chapter
is presented for a quick reference.
A. Index and Sets
i2I Index and set of conventional generation units
In Subset of conventional generation units connected to node n
n, m 2 M Indexes and set of nodes
Mn Set of nodes connected to node n through a line
t2T Index and set of hours
V Set of decision variables of MPEC model
VLL Set of decision variables of lower level problem
E. Valenzuela
University of Chile, Santiago, Chile
R. Moreno
University of Chile, Santiago, Chile
Imperial College London, London, UK
Instituto Sistemas Complejos de Ingeniería (ISCI), Santiago, Chile
D. Papadaskalopoulos (*) · Y. Ye
Imperial College London, London, UK
e-mail: d.papadaskalopoulos08@imperial.ac.uk
F. D. Muñoz
Instituto Sistemas Complejos de Ingeniería (ISCI), Santiago, Chile
Universidad Adolfo Ibáñez, Santiago, Chile

https://doi.org/10.1007/978-3-030-40029-3_10
224 E. Valenzuela et al.
B. Parameters
ai Quadratic coefficient of operating cost of conventional generation unit
i ($/MW2)
bi Linear coefficient of operating cost of conventional generation unit i ($/MW)
dn, t Demand at node n and time period t (MW)
τt Weighting factor of hour t
n,m
F Capacity of line (n, m) (MW)
g Minimum output of conventional generation unit i (MW)
i
gi Maximum output of conventional generation unit i (MW)
K Annuitized investment cost of renewable generation ($/MW/year)
xn, m Reactance of line (n, m) (p.u.)
βn, t Normalized output of renewable generation at node n and hour t
γ Price of renewable energy certificate ($/MWh)
C. Decision Variables
gi, t Output of conventional generation unit i at hour t (MW)
rn, t Actual output of renewable generation at node n and hour t (MW)
Rn Capacity of renewable generation at node n (MW)
θn, t Voltage angle at node n and period t (rad)
λn, t Market clearing price at node n and hour t ($/MWh)
10.1 Introduction
10.1.1 Background
Electricity systems are currently facing fundamental challenges associated with the
necessary decarbonization measures to tackle climate change. Such measures need
to consider the deregulated setting of the electricity market, where infrastructure’s
owners, particularly in the generation sector, need clear economic incentives to
deploy future expansions.
The continuously increasing levels of greenhouse gas emissions and the associated
environmental concerns have driven governments worldwide to take significant
initiatives for the large-scale integration of renewable generation. In Europe, the
European Commission (EC) has put forward a legally binding target for renewable
energy sources to cover 20% and 27% of the total energy consumption in the European
Union (EU) by 2020 and 2030, respectively [1, 2]. In the USA, many states have opted
for different kinds of policies that promote renewable energy generation. For example,
in the state of Oregon, the Renewable Portfolio Standards (RPS) require that large
utilities supply 25% of electricity from renewable resources by 2050, whereas smaller
utilities must provide between 5% and 10% by 2025 [3]. In Latin America, there are
countries such as Chile, Peru, Colombia, Brazil, and Mexico with specific targets for
renewable generation and other regulatory policies to foster its integration. In Chile,
10 Exploring the Concept of Hosting Capacity from an Electricity Market Perspective 225
for instance, there is a 20% target (i.e., percentage of the electricity demand supplied
by renewables) by 2025 and a 70% target by 2050 [4].
However, this large-scale integration of renewable generation introduces signif-
icant techno-economic challenges to the operation and planning of electricity grids.
First of all, the inherent variability and limited predictability and controllability of
renewable generation imply that conventional generators need to remain in the
system and operate in an inefficient fashion (part loaded and with more frequent
start-up and shut-down cycles) to provide the required balancing services to the
system [5]. Furthermore, integration of inverter-connected renewable generation
reduces the system inertia (which is provided by the kinetic energy stored in the
rotating mass of the conventional generators), which implies that imbalances
between supply and demand change system frequency more rapidly, challenging
the stability of the system [6]. Finally, the large-scale connection of renewable
generation to transmission and distribution grids creates certain network challenges,
such as congestion (due to thermal and stability limits), increased voltage levels, and
increased short-circuit current levels, which threaten the security of these grids [7].
In order to capture and quantitatively analyze these challenges, the concept of
hosting capacity was firstly introduced to determine the maximum limits of distrib-
uted energy resources (DER) that can be connected to a distribution network without
harming power quality [8, 9]. In the same vein, references [10, 11] establish maxi-
mum quantities of distributed generation (DG) capacity, especially PV, that respect
network’s voltage limits. Other relevant papers have investigated the hosting capacity
of distribution grids beyond voltage issues, considering thermal, short-circuit current
and harmonic distortions [12–15]. Apart from the calculation of the hosting capacity,
measures to enhance it have been proposed too in [16–18], including advanced
control and operational measures, network reinforcements, and the installation of
storage facilities and reactive power equipment. Although the vast majority of the
relevant literature is focused on distribution networks [12, 19–24], at a fundamental
level, the concept can be applied to transmission networks too as in [25]. Hence, the
concept of hosting capacity is useful for governments, regulators, and central policy
makers as it expresses the maximum capacity of renewable generation that can be
integrated to electricity grids without breaching its technical limits and threatening its
secure operation and/or without surpassing certain system cost thresholds.
In combination with the abovementioned challenges of decarbonization, we need
to consider that the electricity industry has been deregulated in many countries,
allowing market forces to guide future investments. This deregulation has led to
unbundling of vertically integrated monopoly utilities and the introduction of com-
petition in the generation sector [26]. Chile and the UK are good examples of early
deregulation of electricity markets in the beginning of the 1980s and 1990s, respec-
tively. The deregulation of electricity markets, however, has not been implemented
everywhere, but more and more countries are adhering to this practice since dereg-
ulation has been key to promote cost efficiency and investments [27].
In this context, generation investment planning is not anymore carried out by a
central regulated utility aiming to minimize system costs but relies on profit-driven
decisions of self-interested generation companies, operating within a competitive
electricity market. In this setting, the large-scale integration of renewable generation
introduces a fundamental market challenge: given that operating costs of renewables

are very low, the electricity prices are significantly reduced, threatening the recovery
of their investment costs and therefore their profitability [28–30].
10.1.2 Motivation and Contributions
The motivation behind this work lies in the fact that previous work has investigated
the renewable generation hosting capacity of electricity systems from a technical
perspective, but no work has been carried out in exploring and analyzing the same
concept from a market perspective, considering the abovementioned challenges of
investment cost recovery and overall profitability.
This chapter makes the first attempt to fill this knowledge gap by proposing a
new modeling approach aiming to quantitatively determine the maximum renew-
able generation capacity that can be integrated in electricity grids, while respecting
long-term profitability constraints, which we define as the market hosting capacity
(MHC) of renewable generation. This approach involves a bi-level optimization
model. The upper level aims at optimizing renewable generation investments in
different locations of the network so as to maximize the total renewable generation
capacity integrated, subject to the non-negative long-term profit constraint imposed
by generation owners. The lower level represents endogenously the market clearing
process and determines the generation dispatch and electricity prices.
The proposed model is validated through case studies on a simple 2-node system,
the IEEE 24-node system, and a 42-node model of the Chilean electricity system.
These case studies quantitatively analyze the dependence of the MHC on a number
of factors, including the network capacity, the flexibility of the conventional gener-
ation fleet, and renewable generation subsidies. Interestingly enough, our results
demonstrate that a stronger network does not always increase the MHC.
10.1.3 Chapter Organization
The rest of this chapter is organized as follows. Section 10.2 details the proposed
modeling approach. Case studies and quantitative results are presented in Sects. 10.3
and 10.4. Finally, Sect. 10.5 discusses conclusions and future extensions of this work.
10.2 Proposed Modeling Approach
10.2.1 General Description
As mentioned earlier, the MHC concept introduced in this work aims to identify
the maximum renewable generation capacity that an electricity network can host
from a market perspective. This maximum value exists since the more the
renewable capacity connected to a network, the lower the electricity prices.

Evidently, if prices drop too much, there will be no incentives for investors to
integrate more renewables. Thus, there is a maximum limit of renewable genera-
tion capacity that can be integrated in a profitable fashion. Note that the impact of
new generation capacity on prices is not uniform across the network, meaning that
zones with higher energy prices may present better opportunities for connecting
more renewables than zones with lower prices. Thus, we aim to identify those
promising areas for integrating large amounts of renewables. We also determine
amounts of renewables to be connected in other less promising areas. Overall, we
attempt to maximize the total installed capacity of renewables integrated in a given
power grid, without affecting the ability of owners to recover their capital invested
in renewable generation.
In order to formulate the abovementioned problem, we use a bi-level optimiza-
tion approach, the structure of which is illustrated in Fig. 10.1. Since we need to
represent the market clearing process to obtain prices and generation dispatches,
we use a classical dispatch model that minimizes total operational costs for a given
network and set of generation capacities, assuming that the entire operation of the
network is resolved by a single market operator at the transmission or distribution
network level. It should be noted that this model can deliver both dispatches
(as primal variables) and energy prices per node (as dual variables). As shown in
Fig. 10.1, we use this model as the lower level problem within our bi-level
optimization framework. As we aim to maximize integration of new renewable
generation so as to determine the MHC, we establish this objective in an upper
level, which is subject to the previous lower level, market clearing problem and an
additional constraint that avoids negative profits. This bi-level problem is solved
after converting it to a Mathematical Program with Equilibrium Constraints
(MPEC) and subsequently to a mixed-integer linear problem (MILP). The details
of the model assumptions, formulation, and solution strategy are described next.
Fig. 10.1 Structure of

proposed bi-level
optimization model
10.2.2 Model Assumptions
For clarity reasons, next, we declare the main assumptions behind the proposed
model:
• The model assumes a static planning approach with a yearly operation horizon,
implying that the renewable generation capacity is maximized considering a
single, future target year. This yearly operation horizon is discretized and divided
into a number of representative hours. Both investment and operating costs and
revenues are calculated at the same yearly basis.
• The considered electricity market is a pool-based energy-only market which is
cleared by the market operator through the solution of a short-term social welfare
maximization problem. In order to account for the effect of the network, the
market clearing process incorporates a DC power flow model and yields loca-
tional marginal prices (LMP) λn, t for each node n and time period t. Note that the
LMP system can be applied on both transmission and distribution networks [31],
and thus, our results are equally applicable to both as long as energy is priced in
every node following the LMP principles.
• For simplicity reasons and without loss of generality, a single type of renewable
generation technology is considered, but its normalized output profile is generally
different in different nodes of the network in order to account for the locational
availability of renewable energy sources. Its operating costs are assumed zero,
and their output can be curtailed if required.
• Each conventional generation technology is characterized by a different operating
cost as well as different maximum output and minimum output constraints, with
the latter representing the operation of must-run generation technologies.
• The demand side is assumed inelastic.
10.2.3 Bi-level Optimization Model
The proposed bi-level optimization model (Fig. 10.1) aims at maximizing the total
renewable generation capacity subject to long-term profitability constraints and is
formulated as follows:
(Upper Level)
X
max Rn ð10:1Þ
n
subject to
X X
τt ðλn,t þ γ Þr n,t KRn 0 ð10:2Þ
n, t n
(Lower Level)
X
min V LL ai gi,t þ bi g2i,t ð10:3Þ
i, t
where

V LL ¼ gi,t , r n,t , θn,t ð10:4Þ
subject to
X X θn,t θm,t
d n,t gi,t r n,t þ ¼ 0 : λn,t , 8n, 8t ð10:5Þ
i2I m2M
xn,m
n n
gi gi,t gi : μ þ
i,t , μi,t , 8i, 8t ð10:6Þ
0 r n,t βn,t Rn : ν þ
n,t , νn,t , 8n, 8t ð10:7Þ
n,m θn,t θm,t

F F n,m : π þ
n,m,t , π n,m,t , 8n, 8m 2 M n , 8t ð10:8Þ
xn,m
π θn,t π : ρ þ
n,t , ρn,t , 8n, 8t ð10:9Þ
θ1,t ¼ 0 : φt , 8t ð10:10Þ
The upper level (UL) problem determines the optimal renewable generation
investments at the different nodes of the network so as to maximize the total
renewable generation capacity (10.1). This problem is subject to the renewable
generation’s long-term profitability constraint (10.2), which ensures that its revenue
from the energy market and potential renewable energy certificates [32] is higher
than or equal (i.e., it at least recovers) to its investment cost, as well as the lower level
(LL) problems (10.3), (10.4), (10.5), (10.6), (10.7), (10.8), (10.9), and (10.10). The
latter represents the market clearing process at each representative hour, minimizing
the total cost of the conventional generation units (10.3), subject to nodal demand-
supply balance constraints (10.5) (the Lagrangian multipliers of which constitute the
LMP), the operating constraints of conventional generators (10.6) and renewable
generation (10.7), and network constraints (10.8), (10.9), and (10.10).
These two problems are coupled, since the renewable generation capacity deter-
mined by the UL problem affects the constraints (10.7) of the LL problem, while the
renewable generation dispatch and the LMP determined by the LL problem affect
the constraint (10.2) of the UL problem.
10.2.4 MPEC Formulation
In order to solve the abovementioned bi-level optimization problem, the LL problem

is replaced by its Karush-Kuhn-Tucker (KKT) optimality conditions, a replacement
enabled by the continuity and convexity of the LL problem. This converts the
bi-level problem to a single-level MPEC which is formulated as
X
max V Rn ð10:11Þ
n
where

V ¼ Rn , V LL , λn,t , μ þ þ þ þ
i,t , μi,t , νn,t , νn,t , π n,m,t , π n,m,t , ρn,t , ρn,t , φt ð10:12Þ
subject to
(10.2), (10.5), (10.10),
ai þ bi gi,t λðn:i2I n Þ,t μ þ

i,t þ μi,t ¼ 0, 8i, 8t ð10:13Þ
λn,t ν þ
n,t þ νn,t ¼ 0, 8n, 8t ð10:14Þ
X λn,t λm,t X ρþ þ
n,m,t ρm,n,t
X ρ
n,m,t ρm,n,t
þ þ σþ
n,t σ n,t þ ð10:15Þ
m2M
xn,m m2M
xn,m m2M
xn,m
n n n
ðφt Þn¼1 ¼ 0, 8n, 8t

0 μ i,t ⊥gi,t 0, 8i, 8t ð10:16Þ

0 μþ
i,t ⊥ gi gi,t 0, 8i, 8t ð10:17Þ
0 ν n,t ⊥r n,t 0, 8n, 8t ð10:18Þ

0 νþn,t ⊥ β n,t Rn r n,t 0, 8n, 8t ð10:19Þ

θn,t θm,t
0 ρn,m,t ⊥ F n,m þ 0, 8n, 8m 2 M n , 8t ð10:20Þ
xn,m

þ θn,t θm,t
0 ρn,m,t ⊥ F n,m 0, 8n, 8m 2 M n , 8t ð10:21Þ
xn,m
0 σ
n,t ⊥ðπ þ θ n,t Þ 0, 8n, 8t ð10:22Þ
0 σþ
n,t ⊥ðπ θ n,t Þ 0, 8n, 8t ð10:23Þ
The objective function of the MPEC formulation (10.11) coincides with the
objective function of the UL problem. The set of decision variables of the MPEC
formulation (10.12) includes the decision variables of the UL problem and the LL
problem as well as the Lagrangian multipliers associated with the constraints of the
LL problem. The KKT optimality conditions of the LL problem correspond to
Eqs. (10.13), (10.14), (10.15), (10.16), (10.17), (10.18), (10.19), (10.20), (10.21),
(10.22), and (10.23).
10.2.5 MILP Formulation
The abovementioned MPEC formulation is nonlinear implying that any solution

obtained by commercial optimization solvers is not guaranteed to be globally
optimal. In order to address this challenge, this MPEC is converted to a regular
MILP, which commercial branch-and-cut solvers can efficiently solve to global
optimality [33].
Specifically, the abovementioned MPEC formulation includes two types of non-
linearities. The first one involves the bilinear terms ∑n,trn,tλn,t in constraint (10.2). In
order to linearize it, we adopt the linearization approach proposed in [34] which uses
some of the KKT conditions.
First of all, by making use of the nodal demand-supply balance constraints (10.5),
and summing for every n and t, we obtain
X X X X θn,t θm,t
r n,t ¼ dn,t gði2I n Þ,t þ ð10:24Þ
xn,m
n, t n, t n, t n, ðm2M n Þ, t
By multiplying both sides of (10.24) by λn,t, the nonlinear term becomes equal to
X X X X ðθn,t θm,t Þλn,t

r n,t λn,t ¼ dn,t λn,t gði2I n Þ,t λn,t þ ð10:25Þ
xn,m
n, t n, t n, t n, ðm2M n Þ, t
By multiplying both sides of (10.13) by gi,t, summing for every i and t, and
rearranging some terms, we obtain
X X X
λðn:i2I n Þ,t gi,t ¼ ai þ bi gi,t gi,t þ μþ
i,t gi,t μi,t gi,t ¼ gði2I n Þ,t λn,t ð10:26Þ
i, t i, t n, t
By making use of (10.16) and (10.17), Eq. (10.26) becomes

X X
gði2I n Þ,t λn,t ¼ ai þ bi gi,t gi,t þ μþ
i,t gi ð10:27Þ
n, t i, t
By multiplying both sides of (10.15) by θn, summing for every n and t, making
use of (10.10), and rearranging some terms, we obtain
X ðθn,t θm,t Þ X ðθn,t θm,t Þ

λn,t ¼ ρ þ
n,m,t þ ρn,m,t ð10:28Þ
xn,m xn,m
n, ðm2M n Þ, t n, ðm2M n Þ, t
By making use of (10.20) and (10.21), Eq. (10.28) becomes
X ðθn,t θm,t Þ X
λn,t ¼ ρn,m,t þ ρþ
n,m,t F n,m ð10:29Þ
xn,m
n, ðm2M n Þ, t n, ðm2M n Þ, t
By substituting (10.27) and (10.29) into (10.25), this bilinear term is replaced by
the following quadratic expression:
X X X
r n,t λn,t ¼ d n,t λn,t ai þ bi gi,t gi,t þ μþ
i,t gi
n, t n, t i, t
X
ρ
n,m,t þ ρþ
n,m,t F n,m ð10:30Þ
n, ðm2M n Þ, t
The second type of nonlinearity involves the bilinear terms in constraints (10.16),
(10.17), (10.18), (10.19), (10.20), (10.21), (10.22), and (10.23), which can be written
in the generic form 0 μ ⊥ p 0, with μ and p representing generic dual and primal
terms, respectively. The linearization approach proposed by Fortuny-Amat and
McCarl in [35] replaces each of these conditions with the set of mixed-integer linear
conditions μ 0, p 0, μ ωMD, p (1 ω)MP, where ω is an auxiliary binary
variable, while MD and MP are large positive constants. The set of decision variables
of the MILP formulation includes the set (10.12) as well as the auxiliary binary
variables introduced for linearizing (10.16), (10.17), (10.18), (10.19), (10.20),
(10.21), (10.22), and (10.23).
10.3 Theoretical Case Studies

10.3.1 Small-Scale Study: Illustration, Validation,
and Analysis over 2-Bus Network
This section studies the impacts of maximizing renewable generation capacity on a

given illustrative network, calculating the so-called market hosting capacity. We run
several sensitivity analyses for different network capacities, focusing on how the
market hosting capacity varies. This small network also serves to illustrate and
validate the model proposed in this chapter.
10.3.1.1 Input Data
Figure 10.2 shows the configuration of the 2-bus network with two conventional
units, two solar power units, two demands, two buses, and one line. This network is a
modified version of the 2-bus Borduria (node 1) and Syldavia (node 2) example
described in [26]. The specific data used in this chapter is shown in Table 10.1. Note
that while demand is the highest in node 2, the variable cost of thermal generation is
the lowest in node 1. Also, node 1 exhibits the highest availability of the solar power
resource. With all of the above, we can anticipate that power transfers will go from
node 1 to node 2.
Fig. 10.2 Two busbar example
Table 10.1 Input data of illustrative example

Parameter Node 1 Node 2
Demand (MW) 500 1500
Installed capacity of thermal 2000 2000
generation (MW)
Variable cost of thermal gen- 10 P1 þ 0:01
2 P1
2
13 P2 þ 0:02
2 P2
2
eration ($/MWh) P1 represents the output of P2 represents the output of

thermal generation in node 1 thermal generation in node 2
Normalized availability of 0.28 0.15
solar power resource (p.u.)
Investment cost of solar gen- 5 5
eration ($/MW/h)
For illustrative purposes, the following results are obtained assuming a single
operating hour; hence, both operating and investment costs are calculated hourly.
Overall, we study 2400 cases with different network capacities ranging from 0 to
1200 MW with a step of 0.5 MW, in which we seek to calculate, for each of them, the
maximum capacity of solar power generation that can be installed in a profitable
fashion in both nodes.
10.3.1.2 Results and Discussion
Figure 10.3 shows the market hosting capacity of the 2-bus network as a function of
the line capacity. Interestingly, for smaller line capacity values (i.e., between 0 and
400 MW), the hosting capacity decreases, while the line capacity increases. Note this
result is very counterintuitive since, in practice, we usually expand networks to
increase HC. This counterintuitive result is justified because demand in node 2 can
be supplied by higher volumes of low-cost generation in node 1 (thermal and solar
power generation) when network capacity expands, which replaces capacity of solar
power generation in node 2. Note that building more line capacity clearly hampers
business opportunities for investors in node 2, as node 1 features lower fuel costs and
higher availability factors for solar power generation, making these resources more
Fig. 10.3 MHC versus network capacity
Fig. 10.4 Power generation versus network capacity
preferable. In this context, Fig. 10.4 shows how power generations change with
network capacity expansions. Note also that the lower investment levels in solar
power generation in node 2 are not replaced by more investments in solar power
generation in node 1 but are rather replaced by existing low-cost thermal generation
in node 1 that can be exported due to network expansions. This clearly reduces the
Fig. 10.5 Locational marginal price versus network capacity
overall MHC since the utilization of existing generation capacity is favored over new
investments in renewables.
However, as existing low-cost thermal generation features a limited capacity, for
larger network expansions (above 400 MW), generation in node 2 is replaced by
investments in new solar power generation capacity in node 1 (while thermal
generation in node 1 is producing at its maximum output). Hence, it is demonstrated
that network expansions above 400 MW unlock the significant solar power gener-
ation potential in node 1, increasing the overall MHC. This is also shown in
Fig. 10.4, where, for power transfers above 400 MW, both thermal and solar
power generations in node 2 decrease due to line capacity expansions that facilitate
the installation of new solar power generation capacity in node 1.
Figure 10.5 shows the increase in the price of node 1 and the decrease in the price
of node 2, as the line capacity increases. This is because the output of the low-cost
thermal generator located in node 1 can be increased while network capacity
expands, exporting power to node 2. This low-cost energy produced in node 1, in
turn, displaces costly energy production in node 2, reducing its locational marginal
price. Therefore, the increase in line capacity constantly reduces the economic
attractiveness of node 2 for investors and increases that in node 1.
All of the abovementioned arguments demonstrate that increasing the capacity of
the network does not necessarily allow private investors to integrate higher volumes
of renewables. On the contrary, increasing network capacity can even reduce the
business opportunities for new renewable generation developments, especially in
those areas close to the load centers and with higher electricity prices.
10.3.2 IEEE RTS Study: Analyzing the Effect of Network

Capacity, Generation Flexibility, and Subsidies
on MHC
This section determines and analyzes the market hosting capacity of different
network configurations, carrying out sensitivity analyses around the capacity of
the network, the flexibility of conventional generation, and the potential subsidies
that may exist to foster a larger penetration of renewable generation.
10.3.2.1 Input Data
We have modified the IEEE RTS described in [36] and shown in Fig. 10.6 by
carrying out the following changes:
(a) Reduction of all lines and transformers’ capacities by 50%
(b) Increase of thermal generation capacity in nodes 1 and 2 by 10 (MW)
(c) Reduction of thermal generation capacity in node 7 by 50 (MW)
These changes (a–c), implemented in our base case, are aimed to enhance
congestion in the system and increase prices in the low voltage area. Also, we
consider that solar power generators can be installed in every node at an annuitized
investment cost of 110 k$/MW.year, and all nodes present the same availability
profile of renewable sources across different time periods. To represent the multiple
market clearing processes that occur across a year, various combinations of different
demand and solar power levels are selected through standard clustering techniques,
e.g., K-means (following [37]). For this particular case, we use demand and solar
profiles observed in Chile. The relevant operational cost data is shown in Table 10.2.
10.3.2.2 Case Studies
We analyze three case studies which demonstrate how the market hosting capacity
changes for:
(a) Four different network capacity levels
(b) Four different levels of flexibility in conventional generation technologies,
which is varied by changing their must-run output
(c) Four different levels of subsidies in the form of a renewables certificate
10.3.2.3 Results and Discussion
Table 10.3 shows the market hosting capacity levels of various modified versions of
the IEEE RTS system when line capacities are increased by 10%, 20%, and 30%
with respect to the base case. Interestingly, the most attractive case to maximize
Fig. 10.6 IEEE RTS system
Table 10.2 Variable cost per technology in the IEEE RTS

Type Technology Variable cost ($/MWh)
U12 Oil/steam 56.56
U20 Oil/CT 130.00
U76 Coal/steam 16.08
U100 Oil/steam 43.66
U197 Oil/steam 45.58
U400 Nuclear 4.42
Table 10.3 MHC sensitivity to network capacity

Case MHC (MW)
Base case 617
Base case + 10% increase in line capacity 524
Base case + 1 increase in line capacity 0
Table 10.4 Hosting capacity sensitivity operational flexibility

Case MHC (MW)
Base case (no must-run generation) 617
Base case + must-run generation of 10%a 508
a
Percentage with respect to maximum capacity
renewables penetration in a profitable fashion is that of minimum network capacity,

and this is so because of the higher LMPs present in the base case due to increased
network congestion. Furthermore, in the case where the network capacity is
increased by 30%, the market hosting capacity of the network is zero. A key point
in this example is that the load factors of renewables are the same across the whole
network. This makes less attractive the usual business case (absent in this particular
example but present in the previous 2-bus example) to expand network capacity in
order to access remote renewable resources with higher load factors. In other words,
network capacity expansions may impose a barrier for the development of new
renewable generation in the presence of equally attractive renewable resources
across a power system, and this is a key result to be considered by planners and
policy makers.
We also analyze how operational flexibility may support higher renewables
integration levels. Hence, Table 10.4 shows the market hosting capacity levels of
various cases for different levels of must-run generation limits. The results demon-
strate that lower levels of operational flexibility (i.e., higher levels of must-run
generation) lead to smaller levels of renewables integration in a profitable fashion.
This is so since higher must-run generation increases the level of reviewable
generation curtailment and thus decreases average energy prices.
Finally, we study the effect of subsidies, which in this study involve a certificate
type payment for each MWh of renewable energy produced, on top of the energy
price. As expected, a higher price of this certificate increases the amount of
renewables that can be integrated profitably. For example, Table 10.5 shows that
a subsidy of 12 $/MWh can achieve more than 50% increase in MHC with respect
to the base case. This is so since potentially negative profits can be cancelled due to
the presence of extra revenue streams from subsidies. Furthermore, by running
Table 10.5 Hosting capacity sensitivity to certificate prices

Case MHC (MW)
Base case (no subsidy/certificate) 617
Base case + subsidy/certificate of 3 $/MWh 687
several sensitivities, we can demonstrate that the efficacy of these subsidies may
depend on network capacity and generation flexibility. This is an important piece
of information for policy makers since subsidies may not present a significantly
positive impact on promoting renewables due to other inefficiencies in terms of
network configuration and level of (in)flexibility of generation plants.
10.4 Practical Case Study: Investigating the Impact

of Massive HVDC Line Investment on MHC
on the Chilean Power Sector
10.4.1 Background
The planning authority in Chile (which is within the regulatory office) has recently
approved a $1.3 billion investment in a new high voltage direct current (HVDC) line
that will interconnect the north of the country, particularly the Atacama Desert
(location with one of the highest solar irradiation levels in the world), with the
load center of the electricity network located in Santiago, where about 30% of the
national electricity demand is placed. This massive transmission network invest-
ment, with a capacity of 2000 MW and a length of 1500 km, has been justified for
shipping low-cost solar power that can be produced in the Atacama Desert to the
main load center of the country, without originating congestions and without the
need for further network investments in the rest of the AC power network corridor.
This important transmission project, however, received several objections from a
number of market participants, who remained concerned about the benefits of the
new asset in question. One of the strongest arguments used against the project was
that solar resources (and other renewable resources) in Chile are present all along the
country, not only in Atacama, and thus, renewable generation can be perfectly
connected closer to the load center without escalating network investment costs.
Furthermore, these market participants officially took a stand against this project,
bringing their arguments to the Expert Panel. The Expert Panel is a special tribunal
created in Chile to resolve disputes either among market participants or between
market participants and the regulator or planning authority [38]. The first time this
project was proposed by the regulator, the Expert Panel favored the position of those
market participants who opposed the project. In the second time, nevertheless, the
regulator faced no opposition, and hence, the project should go forward. Despite this,
there are currently various concerns regarding the effectiveness of this new network
investment option, especially in terms of supporting cost-effective integration of
renewable generation.
In this context, we aim to investigate the effect of the proposed HVDC line on
MHC, particularly analyzing whether (or not) more network capacity really drives
higher MHC in a realistic setting like the Chilean electricity market. Note that the
effect of network investments on the ability of the electricity grid to absorb more
renewable generation capacity is unclear due to the changes driven by network
investments on LMPs. So, we use the concept of MHC to determine whether, in
this case, more network investments may lead to a higher penetration of
renewables.
10.4.2 Brief Description of the Chilean Electricity Market

and Transmission Network
Figure 10.7 illustrates the Chilean transmission system, which is modeled by using a
40-node representative network with 56 transmission corridors. We focus our
analysis on the year 2018 with a total demand equal to 76 TWh/year and 13 GW
in terms of energy consumption and peak power, respectively. Its generation fleet
contains hydropower (28%), coal (22%), gas (19%), and variable wind and solar
resources (16%), with a total installed generation capacity of 24 GW. Regarding the
market organization, as mentioned in Sect. 10.1, Chile was deregulated in the early
1980s, unbundling generation, transmission, and distribution and establishing gen-
eration as a competitive market while networks remain as regulated monopolies.
Despite this early deregulation, the market clearing in Chile is not bid based but
rather cost based with audited costs. Under this cost-based approach, it is clear that
zero variable cost generation, such as wind and solar power, will drive energy prices
down if their penetration is too large.
10.4.2.1 Case Studies
We aim to investigate the effects of the HVDC line on MHC in Chile. In Fig. 10.7,
the HVDC line (indicated as a dotted line) goes from node 4 to node 29, bypassing
all existing electrical infrastructure between the Atacama Desert (node 4) and
Santiago (node 29). As the line is of HVDC type, we neglect the effect of the second
Kirchhoff’s law in the model for this particular line, considering the ability to fully
control the power flows through this asset due to the associated flexible converter-
based substations at its both ends. For this model, an additional constraint (to those
Fig. 10.7 Chilean transmission network in 2018, indicating topology, generation and demand
nodes, and number of circuits per corridor. The HVDC link is depicted as a dotted line
Table 10.6 Investment cost of renewables in Chile

Case Investment cost – solar power ($/kW) Investment cost – wind power ($/kW)
Base case 1000 1300
Sensitivity 1 950 1300
indicated in Sect. 10.2) has been added to the hydro reservoirs to make water
management more efficient, limiting the available energy generated per hydropower
plant in addition to the maximum power that each unit can generate, referring to the
limited amount of water available to generate energy.
Apart from applying the proposed model (in which the summation of the wind
and solar power generation capacity invested across the entire system is maximized)
with and without the HVDC link, we run three sensitivity scenarios with different
investment costs for wind and solar power generation, presented in Table 10.6.
These values are annuitized by using a discount rate of 10% and a life span of
25 years.
Finally, the cost data used in this study can be found in [39, 40], and the
availability profiles of renewable generation are specified in [41, 42]. As mentioned
earlier, the richest solar power resources are located in the north of Chile, specifically
in the Atacama Desert.
10.4.2.2 Results
Table 10.7 shows the MHC for the four examined cases, demonstrating that MHC
increases when the HVDC line is built. Furthermore, the MHC is clearly more
sensitive to variations in the investment cost of solar rather than wind power
generation, and this is because the new generation capacity built by the model
corresponds mostly to solar power generation as shown in Fig. 10.8. Indeed, with
the HVDC line, new solar power generation can significantly increase its market
position from 4.7% to 12.7% of the total energy produced. This significant increase
in the penetration of solar power generation due to the installation of the HVDC line
displaces mainly thermal generation from a participation of 40% in the case without
the HVDC line to 34% in the case with the HVDC line. There is also 1.5% decrease
in the participation of hydropower generation when the HVDC line is built, and this
is because the model attempts to maximize new renewable generation capacity,
regardless of the effects of doing so on existing generation capacity.
As demonstrated above, at least for this real case study, expanding network
capacity will increase MHC due to the need to export higher amounts of solar
power generation from the Atacama Desert to the rest of the country, where the
use of more expensive thermal generation will be reduced. In fact, this result is in
line with our previous finding (particularly that on the 2-bus network) that deter-
mines the value of large network capacity expansions in terms of increasing MHC
Table 10.7 MHC in Chile

Case MHC without HVDC link (MW) MHC with HVDC link (MW)
Base case 1075 2925
50%
45%
40%
Energy partication [%]
35%
30%
25%
20%
15%
10%
5%
0%
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c)
Thermal Hydro Existing wind Existing solar New wing New solar
Fig. 10.8 Energy participation per technology in the base case and (a) without generation
expansion, (b) with generation expansion and no HVDC line, and (c) with generation expansion
and HVDC line
when there are significantly richer renewable resources in a specific area that need to
be exploited. Hence, extra network capacity will favor an increase in MHC in order
to facilitate the shipping of valuable renewable resources to other locations, partic-
ularly large load centers.
10.5 Conclusions and Future Work
This chapter has introduced for the first time the concept of market hosting capacity,
expressing the maximum amount of renewables that can be connected to a given
power network while preserving their profitability in a deregulated environment.
This chapter has also developed a suitable mathematical model to quantify this
market hosting capacity, founded on bi-level optimization principles. This is a new
measure for hosting capacity from a market point of view, which is useful to
understand how various factors, such as network investments, subsidies, and
generation flexibility, can effectively support higher investments in renewable

generation. This new measure can be applied to both distribution and transmission
networks as long as the network in question uses an LMP system to price energy. In
this context, we demonstrate (through two theoretical studies and one practical
study) that higher network capacity does not necessarily drive higher renewable
generation investment since prices may drop in key areas of the network. We also
demonstrate how system flexibility and subsidies can significantly support higher
levels of MHC.
Two key results with important implications from a policy and regulatory per-
spective are that (i) more network investments may not necessarily drive higher
levels of MHC and that (ii) efficacy of subsidies to increase MHC is subject to
network configuration and levels of flexibility of the generation fleet. Regarding the
former, we demonstrated that, indeed, network expansions may threaten investment
opportunities in renewable generation as network congestion can originate high
prices in key importing areas, which can be attractive to invest in new renewable
generation. Regarding the latter, it is important to understand up to which point
subsidies can be an effective manner to increase MHC or whether there are other
more effective means to achieve such MHC increase.
Future work aims at enhancing the presented model in two directions. First of all,
the proposed bi-level optimization model to quantify the MHC, as well as all bi-level
optimization models employed in electricity market modeling, neglects the
non-convex unit commitment constraints of conventional generators, due to their
inherent inability to capture binary decision variables in their representation of the
market clearing process. However, these complex operating properties affect the
market outcome and subsequently the MHC, as they encapsulate the actual flexibil-
ity characteristics of the generation side. In this context, future work will explore
mathematical approaches enabling incorporation of these complex constraints in the
developed model without deteriorating significantly its computational performance.
Furthermore, the inherent variability and limited controllability of renewable gener-
ation have lately attracted significant research interest around the value of flexible
technologies, in particular energy storage and demand response, in enabling a higher
integration of renewables. In this context, future work will investigate the impact of
these technologies on the MHC by properly incorporating a representation of their
time-coupling operating characteristics in the developed model.
Acknowledgement The research presented in this chapter was supported by grants ANID/
FONDECYT/1181928 and FONDECYT/1190228, the Complex Engineering Systems Institute
(ANID PIA/BASAL AFB180003), SERC-CHILE (ANID/FONDAP/15110019), ANID-Basal Pro-
ject FB0008, and ANID/PIA/ACT192094.
References
1. European Commission. 2020 climate & energy package [Online]. Available: https://ec.europa.
eu/clima/policies/strategies/2020_en.
2. European Commission. 2030 climate & energy package [Online]. Available: https://ec.europa.
eu/clima/policies/strategies/2030_en.
3. Delmas, M., & Montes-Sancho, M. (2011). U.S. state policies for renewable energy: Context
and effectiveness. Energy Policy, 39(5), 2273–2288.
4. Ministerio de Energia. 2015. Energy 2050: Chile’s energy policy [Online]. Available: http://
www.energia2050.cl/en/energy-2050/energy-2050-chiles-energy-policy/.
5. Dale, L., Milborrow, D., Slark, R., & Strbac, G. (2004). Total cost estimates for large-scale wind
scenarios in UK. Energy Policy, 32(17), 1949–1956.
6. Troy, N., Denny, E., & O’Malley, M. (2010). Base-load cycling on a system with significant
wind penetration. IEEE Transactions on Power Systems, 25(2), 1088–1097.
7. von Appen, J., Braun, M., Stetz, T., Diwold, K., & Geibel, D. (2013). Time in the sun: The
challenge of high PV penetration in the German electric grid. IEEE Power and Energy
Magazine, 11(2), 55–64.
8. Bollen, M. H. J., & Häger, M. (2005). Power quality: Interactions between distributed energy
resources, the grid, and other customers and with distributed energy. Electrical Power Quality
and Utilisation, Magazine, I, 51–61.
9. Papathanassiou, S., Hatziargyriou, N., Anagnostopoulos, P., Aleixo, L., Buchholz, B., Carter-
Brown, C., et al. (2014). Capacity of distribution feeders for hosting DER (Vol. 24). Working
Group C6.
10. Bollen, M., & Hassan, F. (2011). Integration of distributed generation in the power system.
Hoboken: Wiley/IEEE Press. https://doi.org/10.1002/9781118029039.
11. Walla, T. (2012). Hosting capacity for photovoltaics in swedish distribution grids (M.Sc. thesis,
Uppsala University, Uppsala, Sweden).
12. Deuse, J., Grenard, S., Bollen, M. H. J.,Hager, M., & Sollerkvist, F. (2007, May). Effective
impact of DER on distribution system protection. Presented at the CIRED 2007-19th Interna-
tional Conference on Electricity Distribution, Vienna.
13. Etherden, N., & Bollen, M. (2011). Increasing the hosting capacity of distribution networks
by curtailment of renewable energy resources. In 2011 IEEE Trondheim PowerTech. Trond-
heim https://ieeexplore.ieee.org/abstract/document/6019292.
14. Capitanescu, F., Ochoa, L. F., Margossian, H., & Hatziargyriou, N. D. (2015). Assessing the
potential of network reconfiguration to improve distributed generation hosting capacity in active
distribution systems. IEEE Transactions on Power Systems, 30(1), 346–356.
15. Bollen, M. H. J., Sollerkvist, F., Larsson, A., & Lundmark, M. (2006). Limits to the hosting
capacity of the grid for equipment emitting high-frequency distortion. In Proceedings Nordic
distribution and asset management conference, NORDAC 2006. Stockholm, Sweden.
16. Etherden, N. (2014). Increasing the hosting capacity of distributed energy resources using
storage and communication (Ph.D. thesis, Lulea University of Technology, Lulea, Sweden).
17. Sun, W. (2015). Maximizing renewable hosting capacity in electricity networks (Ph.D. thesis,
The University of Edinburgh, Edinburgh, Scotland).
18. AlAlamat, F. (2015). Increasing the hosting capacity of radial distribution grids in Jordan
(B.Sc. thesis, Uppsala University, Uppsala, Sweden).
19. Bollen, M., & Rönnberg, S. (2017). Hosting capacity of the power grid for renewable electricity
production and new large consumption equipment. Energies, 10(9), 1325.
20. Bollen, M., & Hassan, F. (2011). IEEE press series on power engineering. In Integration of
distributed generation in the power system (pp. 88–91). Hoboken: Wiley.
21. Deuse, J., Grenard, S., & Bollen, M. (2008). EU-DEEP integrated project - technical implica-
tions of the “hosting-capacity” of the system for DER. International Journal of Distributed
Energy Resources, 4(1), 17–34.
22. Bollen, M., Ribeiro, P., Gu, I., Duque, Y. H., & Carlos, A. (2010). Trends, challenges and
opportunities in power quality research. European Transaction on Electrical Power, 20(1),
3–18.
23. Haesen, E., Minne, F., Driesen, J., & Bollen, M. (2005). Hosting capacity for motor starting in
weak grids. In International conference on future power systems. Amsterdam https://ieeexplore.
ieee.org/abstract/document/1600564.
24. Bollen, M., & Häger, M. (2005). Power quality: Interactions between distributed energy
resources, the grid, and other customers. Electrical Power Quality and Utilisation Magazine,
1(1), 51–61.
25. Sáiz-Marín, E., Lobato, E., & Egido, I. (2018). New challenges to wind energy voltage control.
Survey of recent practice and literature review. IET Renewable Power Generation, 12(3),
267–278.
26. Kirschen, D., & Strbac, G. (Eds.). (2004). Fundamentals of power system economics,. Univer-
sity of Manchester Institute of Science & Technology (UMIST), UK (p. 150). Hoboken: Wiley.
27. Joskow, P. (2003). Electricity sector restructuring and competition: Lessons learned. Cuadernos
de Economia, 40(121), 548–558.
28. Green, R., Vasilakos, N., & Kensington, S. (2011). The long-term impact of wind power on
electricity prices and generating capacity. In 2011 IEEE power and energy society general
meeting. Detroit https://ssrn.com/abstract=1851311.
29. Ketterer, J. (2012). The impact of wind power generation on the electricity price in Germany.
Energy Economics, 44, 270–280.
30. Pollitt, M., & Anaya, K. (2015). Can current electricity markets cope with high shares of
renewables? A comparison of approaches in Germany, the UK and the State of New York.
EPRG Working Paper 1519, University of Cambridge.
31. Sotkiewicz, P., & Vignolo, M. (2006). Nodal pricing for distribution networks: Efficient pricing
for efficiency enhancing DG. IEEE Transactions on Power Systems, 21(2), 1013–1014.
32. Johnstone, N., Haščič, I., & Popp, D. (2010). Renewable energy policies and technological
innovation: Evidence based on patent counts. Environmental and Resource Economics, 45(1),
133–155.
33. Pozo, D., Sauma, E., & Contreras, J. (2017). Basic theoretical foundations and insights on
bilevel models and their applications to power systems. Annals of Operations Research, 254
(1–2), 303–304.
34. Ye, Y., Papadaskalopoulos, D., Moreira, R., & Strbac, G. (2017, Sep). Strategic capacity
withholding by energy storage in electricity markets. In 2017 IEEE PowerTech conference.
Manchester https://ieeexplore.ieee.org/abstract/document/7981200.
35. Fortuny-Amat, J., & McCarl, B. (1981). A representation and economic interpretation of a
two-level programming problem. Journal of the Operational Research Society, 32(9), 783–792.
36. Grigg, C., Wong, P., Albrecht, P., Allan, R., Bhavaraju, M., Billinton, R., Chen, Q., Fong, C.,
Haddad, S., Kuruganty, S., Li, W., Mukerji, R., Patton, D., Rau, N., Reppen, D., Schneider, A.,
Shahidehpour, M., & Singh, C. (1999). The IEEE reliability test system-1996. A report prepared
by the reliability test system task force of the application of probability methods subcommittee.
IEEE Transactions on Power Systems, 14(3), 1010–1020.
37. Lara, C., Mallapragada, D., Papageorgiou, D., Venkatech, A., & Grossmann, I. (2017). Electric
power infrastructure planning: Mixed integer programming model and nested decomposition
algorithm. European Journal of Operational Research, 271(3), 1037–1054.
38. Rudnick, H., Araneda, J., & Mocarquer, S. (2009, Dec). Transmission planning-from a market
approach to a centralized one-the Chilean experience. In IEEE power and energy society
general meeting. Calgary https://ieeexplore.ieee.org/abstract/document/5275552.
39. Comisión Nacional de Energía. Informe de costos de tecnologías de generación [Online].
Available: https://www.cne.cl/wp-content/uploads/2017/12/Res-Ex-CNE-55-2018.pdf.
40. Diaz, G., Inzunza, A., & Moreno, R. (2019). The importance of time resolution, operational
flexibility and risk aversion in quantifying the value of energy storage in long-term energy
planning studies. Renewable and Sustainable Energy Reviews, 112, 797–812.
41. Ministerio de Energia. Explorador solar [Online]. Available: http://www.minenergia.cl/

exploradorsolar/.
42. University of Chile. Explorador eolico [Online]. Available: http://walker.dgf.uchile.cl/
Explorador/Eolico2/.
Index
A economic feasibility, 114

Active power curtailment, 102 installation constraints, 103
Actual distribution network, 25 network constraints, 104
Agricultural feeder, 109 operation constraints, 103
Angle droop control, 208 optimal allocation, 114
Annual capital cost, 101 optimal size and location, 102
Application program interface (API) parameters, 111
function, 217 PV and wind generation, 112
Approximation method, 97 Betz limit, 52
Arduino-based IoT sensor platform, 217, 218 Bidirectional power flow, 5
Artificial Neural Network Controller Bi-level optimization approach, 227
(ANNC), 205 Bi-level optimization model, 228, 229
Automatic generation control (AGC), 195, 197 Biofuel, 55
components, 201 Biofuel-powered generators, 74, 82
flowchart, 202 2-bus network, 232, 233
Automatic Q Control (AQC), 11 5-bus distribution system, 64
Automatic voltage regulator (AVR), 206 DG connections, 65
line and load data, 64
MATLAB, 69
B solar photovoltaic sources, 70
Background harmonic distortion (BHD) wind power plant, 73
electrical networks, 37 11 bus feeder, 28
knowledge, 35 83-bus compensated system, 191
power system, 36 83-bus distribution system, 187, 188, 191
primary emission, 36
secondary emission, 36
television viewing, 38 C
voltage distortion, 36 California electricity crisis, 121
Balanced distribution system, 108 Capacity distortion index, 125, 126
Battery Energy Management System Capacity withholding, 122, 125
(BEMS), 209 electricity markets, 130
Battery energy storage systems (BESSs), transmission congestion, 130
87, 211 Capacity withholding assessment, 130
annual costs, 116 Capacity withholding indices, 124, 125, 128,
charge and output power, 104 129, 141

https://doi.org/10.1007/978-3-030-40029-3
250 Index
Capacity withholding mitigation, 143, 145 E

Chilean transmission system, 240 Economic assessment
Clearness index (CI), 16, 17, 56 annual cost, 99
Communication-based control, 215 capital cost, 99
Concentrated solar power (CSP), 53 maintenance and operation, 100
Constraint factor (CF), 139 Economic dispatch controller (EDC), 197
Conventional controller, 202 Electrical grid, 38
Conventional electrical networks, 5 Electricity market, 228
Conventional generation technology, 228 Electricity network, 226
Conventional power generation, 195 Electricity price data, 111
Conventional power system, 48, 195 Electricity systems, 224
Cost and development status indexes, 107 Electrochemical battery technologies, 91
Cost/benefit assessment, 106 Electrochemical storages, 90
Cournot model, 131–133 Energy control center (ECC), 198
Cournot-Nash equilibrium (CNE), 122, 130 Energy management system (EMS), 211
Cournot-type competition, 128 Energy monitoring system architecture, 218
Energy storage (ES), 2, 6
Energy storage technologies
D classification, 89
Decarbonization, 224, 225 electric field, 90
Defuzzification method, 205 electrochemical, 89
Demand elasticity, 122, 128, 129 mechanical forms, 89
Demand side programs, 143 steam turbines, 90
Deregulation, 225 Enhanced hosting capacity (EHC), 83
Design of controllers, 202 European Distributed Energy Partnership
Deterministic approach, 150 (EU-DEEP), 196
Distributed energy resources (DERs), 4, 48, 56, exIWO algorithm, 186
87, 143, 196, 225 Expanded invasive weed optimization
optimal sizing, 57 (exIWO), 180
penetration, 5
Distributed energy technology, 56
Distributed generation (DG), 2, 40, 149, 195, F
196, 225 Feeder characteristics
owners/investors, 2 conductor types, 163
penetration problems, 2 length, 163
Distribution feeder, 26 loading level, 164
Distribution network reconfiguration location sensitivity, 164
(DNR), 179 parallel feeders, 163
Distribution network service providers Feeder reconfiguration, 191
(DNSPs), 150 Financial withholding, 122
Distribution system, 97, 98 Flat tie-line control (FTC), 201
Distribution system model, 181 FPGA-based HIL platform, 58
Distribution system operators (DSOs), 2 Frequency droop, 209
Droop control, 209 Fuzzy logic controller (FLC), 201, 204
Droop control scheme, 208 ANN, 205
Dynamic hosting capacity (DHC), 3, 51 defuzzification method, 205
analysis, 4 inputs, 204
characteristic, 3 output, 204
descriptive graph, 4 solar and wind power, 204
and DSTHC concept, 7
PCC, 4
profile, 3 G
Dynamic-storage-hosting capacity (DSTHC), 7 Game theory-based formulation, 130
Index 251
Gaussian-shape distribution, 112 line and load data, 75

GenCo’s cost function, 123 MATLAB, 76
GenCos’ marginal costs, 122 solar photovoltaic, 78
Generation companies (GenCos), 121 wind speed, 79
ability, 122 IEEE 519, 57
capacity withholding, 121, 122, 126, 128 IEEE RTS study, 236
cost function, 122, 126, 131 hosting capacity levels, 236
cost parameters, 128 variable cost, 237
Cournot model, 123 Integrated hosting capacity
generation, 124 PCC, 43
market performance, 121 RMS, 41
market power, 121 sensitivity analysis, 44
oligopoly market, 123 voltage rise, 43
parameters, 139 Interconnected power system, 199
Generation investment planning, 225 Internal control loop, 209
Geographical information system (GIS) data, Internet of Things (IoT), 216
64 cloud technology, 216
Global horizontal irradiance (GHI), 212 implementation, 217
Global information system (GIS), 50 microgrid, 218
requirement, 216
Invasive weed optimization (IWO), 185
H Inverse demand function, 139
Harmonic sources, 39 Investment cost recovery, 226
Harmonics
MATLAB, 57
Typhoon HIL emulator, 58 J
HC Maximization strategy, 187 Jacobian matrices, 200
High-carbon energy sources, 2
High PV penetration, 32
Hosting capacity estimation, 92 K
distribution system, 93 Karush-Kuhn-Tucker (KKT), 133, 229
framework, 92 Kirchhoff’s voltage law (KVL), 51
Hosting capacity (HC), 12, 49, 87, 121 Korea Micro Energy Grid (K-MEG), 214, 215
analytical analysis, 40, 145, 155, 159 Kronrod-Patterson rule, 20
approach, 39
calculation, 196
concepts and applications, 39 L
definition, 12 Lagrangian function, 123, 133
determination method, 15 Lagrangian multiplier, 134, 135
and DG impacts, 12 Linearization technique, 106
LV grid, 12 Locational marginal prices (LMP), 228
science and technology, 50 Low-voltage (LV) distribution networks,
studies, 13 150, 173
x-axis, 15
Holding capacity vs. performance indices, 202
Hypothesis testing, 19 M
Maiming power system, 36
Market equilibrium, 131
I Market equilibrium model, 130
IEEE 30-bus power system, 139 Market hosting capacity (MHC), 226, 232, 238
IEEE 30-bus test system, 140 Market power and capacity withholding, 143
IEEE 33-bus distribution system, 76, 78 Mathematical program with equilibrium
biofuel-powered generator, 82 constraints (MPEC), 227, 230
252 Index
Maximum connectable solar capacity, 174 Power flow equations, 182

MHC sensitivity, 238 Power generation vs. network capacity, 234
Microgrids Power market modeling, 122
biofuel, 55 transmission congestion, 130
solar energy, 53 Power quality, 4
structure, 51 Power systems, 195
supercapacitors, 55 Power transfer distribution factors
WECS, 52 (PTDFs), 139
wind power plant, 53 Probability-consequence diagrams, 22
Mixed-integer linear problem (MILP), 227, Probability density functions (PDFs), 17
231, 232 Problem formulation
Modern-day solar PV systems, 167 constraints, 184
Monte Carlo methodologies, 150 distribution function, 185
Monte Carlo simulation (MCS) method, 154 exIWO, 185
Monte Carlo technique, 19, 23, 24 IWO, 185
Multi-agent system (MAS), 219 objective function, 184
Multi-feeder LV distribution networks, 163 reactive power, 185
Multilayer perceptron (MLP), 205 Problem statement
DG and SOP modeling, 181
DG modeling, 182
N network reconfiguration, 183
Network reconfiguration, 183 power flow equations, 182
Network reinforcement, 57 SOPs, 182, 183
Newton-Raphson iterative solution
method, 200
Nodal capacity distortion index, 130 Q
Nodal indices, 124 Quantitative risk assessment (QRA), 20
Nodal price distortion indices, 135
Nodal withholding-supply ratio (NWSR), 130
Nonintrusive method, 16, 17 R
Nonlinear programming (NLP) problem, 131 Radio-frequency identification (RFID), 216
Numerical simulation, 16 Raspberry Pi monitoring system, 219
Regulation control schemes
AGC, 197
O DG, 196
Oligopoly market, 135 generator prime, 198
On-line tap changer (OLTC), 23 island mode, 197
On-line voltage regulator (OLVR), 11 large signal analysis, 197
On-load tap changer (OLTC), 2 mathematical model, 198
Optimal allocation, 114 microgrid, 197
Optimal network configuration, 187 regulation characteristics, 199
Optimal power flow (OPF), 88 signal analysis, 196
Optimization model, 97 Renewable energy, 1, 2, 8, 191
Over-voltage, 88, 97 Renewable generation, 225
Renewable Portfolio Standards (RPS), 224
Reverse power flow, 168
P Risk analysis approach
Perfect competition, 128 definition, 20
Performance indices, 206 distribution power, 21
Photovoltaics (PVs), 11, 54 operational performance, 21
Point of common coupling (PCC), 35, 57 stochastic process, 21
Point of connections (POC), 167 Risk-based analysis method, 23
Index 253
Risk indices, 25 stochastic analysis framework, 154

Risk visualization, 22 test network, 152, 160
Root mean square (RMS), 41 two-feeder network, 160
voltage drop, 171
Solar power penetration, 198
S Solar PV hosting capacity, 150
Segment-wise hosting capacity levels, 162 Sparse grid technique, 19, 20
Semiconductor switches, 49 Static hosting capacity, 51
Sensitivity analysis, 158 Static planning approach, 228
Seoul National University (SNU), 213 Steady-state frequency deviation, 201
Simplified distribution feeder model, 166 Stochastic and feeder level, 175
Simulation results, 129 Stochastic evaluation, 157
Simulation test, 24 Stochastic parameter variation, 56
Single distribution feeder model, 174 Stochastic solar PV generation, 150
Single feeder test network, 173 Storage sizing framework
Small signal analysis, 206, 210, 211 approximation method, 97
DG, 207 BESS, 95
PID controllers, 207 BESS capacity, 94
single area system, 207 optimization models, 95
voltage magnitude, 206 valid generated scenarios, 97
Smart grids, 2 Storage system, 6
Social welfare, 131, 132 Supervisory control and data acquisition
Soft open points (SOP), 182 (SCADA), 62
Solar deployment, 156 Supply function equilibrium (SFE), 122
Solar energy, 53 System Advisor Model (SAM) software, 155
Solar irradiation, 64 System evaluation
Solar irradiation data, 63 analytical solution, 16
Solar irradiation values, 79 classical tensor, 19
Solar photovoltaic sources, 70 hypothesis testing, 19
Solar photovoltaic system, 54 nonintrusive method, 16, 17, 19
Solar photovoltaics (PV), 53 procedures, 16
connections, 159 uncertain quantities, 16
distribution network, 151 uncertainties, 16
and DNSPs, 150
feeder, 160
feeder voltage violation, 155 T
hosting capacity, 150 Total annual cost (TAC), 101
linear function, 169 Total harmonic distortion (THD), 49
load profiles, 152 analysis, 82
location, 158 biofuel-powered generator, 82
LV distribution networks, 150 33-bus distribution system, 77, 82
mathematical models, 172 buses, wind speed, 71
MCS method, 157 estimation, 56
network planning perspective, 157 injection condition, 70
and non-dispatchability, 149 values, solar irradiation and load, 70
penetration levels, 152 wind speed, 74, 81
POC, 168 Transmission congestion, 131
power generation, 170 Transmission-constrained market, 137
principles, 166 Transmission network congestion, 130
radial feeders, 151 Transmission system operators
solar customer, 152 (TransCo), 146
solar penetration levels, 155 Two-feeder network, 162, 164
254 Index
Typhoon hardware in loop (HIL) V

application, 58 Valve-regulated lead-acid (VRLA), 114
CLOSE mode, 62 Voltage harmonic distortion, 37
compilation and model execution, 61 Voltage phasor diagram, 37
configuration, 60 Voltage rise, 5–8, 38, 41, 152, 168
emulator, 60 Voltage source converter (VSC), 209
FPGA-based, 61
library, 61
model settings, 62 W
parameters, 58 Weibull distribution function, 56
SCADA, 62 Wind energy conversion system (WECS), 52,
switching element, 59 56
Typhoon HIL emulator, 59 Wind speed data, 64
Typhoon HIL simulation, 50 Wind technology, 114
Typhoon HIL simulators, 58 Wind turbine, 52
Typhoon HIL402 emulator, 59
Z
U ZnBr battery technology, 116
Upper level (UL) problem, 229

Hosting Capacity For Smart Power Grids

Uploaded by

Copyright:

Available Formats

Hosting Capacity For Smart Power Grids

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

Hosting Capacity For Smart Power Grids

Uploaded by

Copyright:

Available Formats

What is the purpose of the book?

What is the purpose of the book?

What topics are covered in the book?

What topics are covered in the book?

Ahmed

Hosting Capacity for Smart

Sherif M. Ismael Paulo F. Ribeiro

ISBN 978-3-030-40028-6 ISBN 978-3-030-40029-3 (eBook)

© Springer Nature Switzerland AG 2020

capacity concepts, calculation methods, performance limits, enhancement tech-

Chapter 7: This chapter elaborates a generalized deterministic approach to eval-

1. Lewis, C. S. (1961). An experiment in criticism. Cambridge: Cambridge Univer-

1 An Overview of Hosting Capacity for Modern Power Grids . . . . . . 1

8 Hosting Capacity Maximization Based on Optimal

Tiago E. Castelo de Oliveira, Paulo F. Ribeiro, Ahmed F. Zobaa,

Conventional fossil fuel-based resources are affected by several problems, such as

T. E. Castelo de Oliveira (*)

© Springer Nature Switzerland AG 2020 1

HC enhancement tool. ES systems help in overcoming the overvoltage resulting

1.2 Dynamic Hosting Capacity Approach

Fig. 1.1 Dynamic hosting Unacceptable

where Pmaxg is the system’s HC; Pmax

Fig. 1.2 Conventional

Fig. 1.3 Bidirectional P Pg DERs

1.3 Dynamic Hosting Capacity Considering Storage

Fig. 1.4 The storage P’ Pg DERs

Fig. 1.5 The DHC with and

1.4 Formulation of the DHC with Energy Storage Systems

In the following sections, the importance of considering the maximum amount of

This method represents an alternative to analyze the storage system behavior

Taking advantage of DSTHC, we can hypothetically propose that the maximum

It is easily veriﬁable that the maximum dynamic HC can be extended by (1.3) in

It is important to highlight that the local DER system, considering an ES system,

1.5 Further Remarks

distribution systems. IEEE Access, 6, 27268–27277. https://doi.org/10.1109/ACCESS.2018.

The high-power penetration into a distribution network exported from photovoltaics

© Springer Nature Switzerland AG 2020 11

2.2 Recent Methods on Hosting Capacity Determination

This section provides a quick summary of different techniques used to determine HC

Case study Criteria HC and penetration method Grid type

Fig. 2.1 HC determination method. Adapted from [4]

2.3 System Formulation and Evaluation

2.3.1 System Evaluation

2.3.1.1 Step A: Uncertainty Characterization

This step characterizes the involved uncertainties by scenario-based method or

Fig. 2.2 System evaluation using nonintrusive method

PPV ðkt Þ ¼ A ∙ η ∙ I βt ðkt Þ ð2:1Þ

2.3.1.2 Step B: System Evaluation via Nonintrusive Method

Fig. 2.3 The probabilistic representation of output power produced by grid-connected PV in

2.3.1.3 Step C: Hypothesis Testing

2.3.2 Sparse Grid Technique

In multidimensional environment, the univariate quadrature rule can be employed to

where the index li denotes the accuracy’s level for a dimension i, i ¼ 1, 2, . . ., n,

with the ﬁnest number of nodes, letting the degree of computation to be O

2.4 Proposed Risk Analysis Approach for Hosting Capacity

2.4.1 Risk Deﬁnition

2.4.2 Deterioration Indices and Hosting Capacity

2.4.2.1 Risk Indices

and the expectation of the severity is estimated as follows: