Latency and Bandwidth Analysis of LTE

for a Smart Grid

XU, YUZHE

Master Thesis
Stockholm, Sweden 2011

XR-EE-RT 2011:018
Latency and Bandwidth Analysis of LTE
for a Smart Grid

XU, YUZHE

Master’s Thesis at Electrical Engineering

Supervisor: Fischione, Carlo; Johansson, Karl Henrik
Examiner: Fischione, Carlo
 ii

Abstract

Smart grid has been proposed as an alternative to the traditional electric-

ity grid recently thanks to its advantages of real time control on consumption
demands. The latest wireless network, 3GPP Long Term Evolution (LTE),
is considered to be a promising solution to interconnecting the smart objects
in a smart grid because LTE provides both low latency and large bandwidth.
However, the theories and standards for deploying a smart grid are still under
study. Furthermore, the performance of LTE network depends on the user
devices’ conditions as well as the network service operators. Therefore it is
important to analyze the performance of LTE network experimentally. In this
master thesis report, the specific requirements in terms of latency and band-
width were first determined for a hypothetical smart microgrid which consists
of several key components, such as one substation/distributed generation, Pha-
sor Measurements Units (PMU) and Advanced Meter Infrastructures (AMI).
Then the latency and the peak data rats of the LTE networks provided by two
service operators TELE2 and TELIA were investigated and compared. The
experimental results show that both latency and peak data rate of the LTE
network provided by TELE2 fulfil the requirements for the smart microgrid,
while the LTE network provided by TELIA gives a little longer latency. In ad-
dition, the simulation results based on a proposed scheduler indicate that the
latency can be improved by an appropriate scheduler. This study has proven
that LTE network is a promising solution for smart grids.
 iii

Acknowledgement

I would like to thank my examiner Carlo Fischione for his guidance and
patience throughout this thesis. I would also like to express my gratitude to
professor Karl Henrik Johansson for all his guidance and valuable suggestions.
I would also like to thank Yong Wang for providing useful research materials.
I reserved the most special gratitude for my parents in Shanghai. Without
your unconditional support and love, this could have been impossible.
Finally, a special thanks to my wife Yi, for her love, patience and support.
Contents

Contents iv

List of Figures vi

List of Tables viii

List of Abbreviations ix

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Smart Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 LTE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Problem Formulation 6
2.1 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Model for Communication Requirement . . . . . . . . . . . . . . . . 7
2.3 Experiment Configurations for LTE Performance Analysis . . . . . . 7
2.4 Scheduler Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Communication Requirements for a Smart Grid 9

3.1 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1 PMU . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.2 AMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.3 Distributed Power . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.4 Remote Sensing: Monitoring and Control . . . . . . . . . . . 13
3.2 Communication Requirement . . . . . . . . . . . . . . . . . . . . . . 14
3.2.1 Communication in A Substation/DG . . . . . . . . . . . . . . 14
3.2.2 Collecting and Dissemination of Phasor Data . . . . . . . . . 15
3.2.3 Collecting and Dissemination of Consumption Data . . . . . 16

4 A Hypothetical Smart Microgrid 17

4.1 Assumption and Structure of Model . . . . . . . . . . . . . . . . . . 17

iv
CONTENTS v

4.2 Communication Requirement for LTE Connection . . . . . . . . . . 17

4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5 Analysis of Latency and Bandwidth of LTE 20

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.1.1 Main Techniques in LTE . . . . . . . . . . . . . . . . . . . . . 20
5.2 Latency Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.2.1 FDD Mode Analysis . . . . . . . . . . . . . . . . . . . . . . . 22
5.2.2 TDD Mode Analysis . . . . . . . . . . . . . . . . . . . . . . . 22
5.3 Bandwidth Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

6 Experiment Results 26
6.1 Loss Rate via LTE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.2 Latency Analysis via LTE . . . . . . . . . . . . . . . . . . . . . . . . 26
6.2.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.2.2 Maximum Likelihood & Least Squares Estimations . . . . . 30
6.2.3 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.3 Throughput via LTE . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

7 Scheduling 34
7.1 Scheduler Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.1.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . 34
7.1.2 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.1.3 Utility Function Design . . . . . . . . . . . . . . . . . . . . . 38
7.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

8 Conclusion & Future Work 42

8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Reference 44

A Ping Introduction 46
List of Figures

1.1 LTE development Timeline . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 EUTRAN architecture diagram [2] . . . . . . . . . . . . . . . . . . . . . 4

2.1 Experiment Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3.1 General WAMS Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.1 A hypothetical smart microgrid . . . . . . . . . . . . . . . . . . . . . . . 18

5.1 Construction of a multicarrier OFDM signal . . . . . . . . . . . . . . . . 21

5.2 Physical layer structure of LTE . . . . . . . . . . . . . . . . . . . . . . . 21
5.3 User plane latency components for FDD . . . . . . . . . . . . . . . . . . 23
5.4 User plane latency components for TDD . . . . . . . . . . . . . . . . . . 24
5.5 Basic time-frequency resource structure of LTE (normal CP case) . . . . 25

6.1 Mean values and Standard variance of RTT for small data packets. . . . 27
6.2 Mean values and Standard variance of RTT for large data packets. . . . 28
6.3 Data Pre-processing. The data set used in this figure is 1000 RTT values
collected with 100 bytes data packets via TELE2 LTE network. The left
figure shows the point of RTT values. The middle one illustrates the
histogram, while the right one shows the probability density function of
RTT values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.4 Probability density function with 100 bytes data packets via TELE2 and
TELIA LTE network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.5 RTT values distribution. The data set is collected via TELE2 LTE
network with 100 bytes data packets. . . . . . . . . . . . . . . . . . . . . 32
6.6 RTT values distribution. The data set is collected via TELIA LTE
network with 100 bytes data packets. . . . . . . . . . . . . . . . . . . . . 32

7.1 Downlink and uplink scheduling process . . . . . . . . . . . . . . . . . . 35

7.2 Resource allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.3 Allocation example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7.4 Latency simulation of PMU data messages . . . . . . . . . . . . . . . . . 41

vi
List of Figures vii

7.5 Simulation allocation result for 1 PMU and 10 other UEs. Different
colours represent different UEs, besides red colour illustrates PMU data
packet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

A.1 ICMP Packet Example Captured by Microsoft Network Monitor 3.4 . . 47

List of Tables

2.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 Experiment hardware List . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3.1 Comparison of a smart grid with the existing traditional grid [5] . . . . 10
3.2 Data message example . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Four Type messages for AMI . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Messages defined by IEC 61850 for a distribution substation . . . . . . . 15

4.1 Main parameters in the hypothetical smart microgrid . . . . . . . . . . 18

4.2 Communication requirements for worst case . . . . . . . . . . . . . . . . 19

5.1 IMT-A Latency Requirement . . . . . . . . . . . . . . . . . . . . . . . . 22

5.2 User plane latency analysis for FDD . . . . . . . . . . . . . . . . . . . . 23
5.3 User plane latency analysis for TDD in uplink . . . . . . . . . . . . . . . 24

6.1 Loss rates of LTE communication network . . . . . . . . . . . . . . . . . 26

6.2 RTT time for different data packets lengths . . . . . . . . . . . . . . . . 28
6.3 Summary fits of RTT values for both TELE2 in Figure 6.5 and TELIA
in Figure 6.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.4 Peak data rates of LTE communication network . . . . . . . . . . . . . 33

7.1 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

8.1 Comparison between communication requirements and experiment results. 43

A.1 ICMP Packet Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

viii
List of Abbreviations

3GPP Third Generation Partnership Project

AMC Adaptive Modulation and Coding

AMI Advanced Meter Infrastructure

AMR Automatic Meter Reading

CP Cyclic Prefix

DFT Discrete Fourier Transform

DG Distributed Generation

eUTRAN evolved UMTS Terrestrial Access Network

FDD Frequency Division Duplex

GPS Global Position Satellite

HARQ Hybrid Automatic Repeat Request

ICMP Internet Control Message Protocol

IFFT Inverse Fourier Transform

ITU International Telecommunication Union

LSE Least Squares Estimation

LTE Long Term Evolution

MCS Modulation and Coding Scheme

MIMO Multiple Input and Multiple Output

MLE Maximum Likelihood Estimation

MTU Maximum Transmission Unit

OFDM Orthogonal Frequency-Division Multiple

ix
List of Tables x

OFDMA Orthogonal Frequency-Division Multiple Access

PAPR Peak-to-Average Power Ratio

PDC Phasor Data Concentrator

PMU Phasor Measurement Unit

RB Resource Block

RE Resource Element

RRC Radio Resource Control

RTT Round-Trip Time

SC-FDMA Single-Carrier Frequency-Division Multiple Access

SCADA Supervisory Control and Data Acquisition

TDD Time Division Duplex

TTI Transmission Time Interval

UE User Equipment

WAMS Wide-Area Measurement System

Chapter 1

Introduction

1.1 Background
1.1.1 Smart Grid
The term “Smart Grid” refers to two-way communicational electricity grid. A smart
grid is expected to be capable of remotely detecting statuses of electricity genera-
tions, transmission lines and substations; of monitoring consumption of user elec-
tricity usage; of adjusting the power consumption of household applications in order
to conserve energy, reduce energy losses and increase electricity grid reliability. In
principle, a smart grid is a upgrade of the 20th century power grid which supplies
power from a few central power generations to a large number of users. The current
power grid was mainly developed under parts of Nikola Tesla’s design which was
published in 1888. Many implementation decisions that are still in use were made
for the first time using the limited emerging technology available 120 years ago.
Compared with these traditional power grids, the topology of a smart grid is more
optimized to meet various electricity need conditions. For instance, decentralized
or distributed generations, renewable energy and battery would be widely used in a
smart grid in order to supply reliable and clean energy. Through using smart grid,
a wider variety of power operations would come into practice, such as recharging
the battery in the low power consumption period and providing power from that
battery in the consumption peak period.

1.1.2 LTE
The term “LTE” is the abbreviation of 3GPP Long Term Evolution which is the
latest standard in use in the mobile communication network [1]. It was developed
to fulfil mobile users’ demands for higher data rates and stabler service perfor-
mance. Its main targets were to provide average users with three to four times the
throughput of the Release 6 HSDPA levels in the downlink (100Mbps), and two to
three times of the HSUPA levels in the uplink (50Mbps). Furthermore, a simple
architecture and backwards compatibility was also required for cost and complexity

1
CHAPTER 1. INTRODUCTION 2

reduction. 3GPP summarizes the motivation to develop LTE network.

• Competitiveness of the 3G system
• User demand for higher data rates and service quality
• Packet Switch optimized system
• Continued demand for cost reduction (Capital and Operating Expenditure)
• Low complexity
• Avoid unnecessary fragmentation of technologies for paired and unpaired band
operation
In order to meet all these requirements in the near future, in 2004 NTT Do-
CoMo of Japan proposes LTE as the latest standard. It was first time that LTE
appeared as a communication project. After 4 years by December 2008, 3GPP
frozen Release 8 code and published it as the basic version standard for developing
first wave LTE equipments, such as chipsets, testing devices and base stations. In
this specification, several key techniques were determined. For instance, Orthog-
onal Frequency Division Multiplexing(OFDM) was selected for the downlink and
Single Carrier-Frequency Division Multiple Access(SC-FDMA) for the uplink. Data
modulation schemes QPSK, 16QAM, and 64QAM were supported by downlink and
BPSK, QPSK, 8PSK and 16QAM by uplink. About three-forth year later, the final
version of the LTE Release 8 specifications was frozen in September 2009. In the
same year December, Release 9 was functionally frozen with small enhancements.
In the same month, the first publicly available LTE service of world was opened by
TeliaSonera in Stockholm and Oslo. The LTE development time line is summarized
in Figure 1.1.
After being developed for several years, the main advantages with LTE have
been proven to provide high throughput, low latency, plug and play, FDD and TDD
in the same platform, an improved end-user experience and a simple architecture
resulting in low operating costs. Evolved UMTS Terrestrial Radio Access Network
(eUTRAN) is chosen as the air interface of LTE. It consists only of eNodeBs on the
network side aiming to reduce the latency of all radio interface operation. eNodeBs
are connected to each other via the X2 interface, and they connect to the packet
switched core network via the S1 interface as shown in Figure 1.2. The data packets
are sent from the end terminal, the UEs, to eNodeBs via radio signals. Such an
end terminal is connected to a smart grid to make measurements of its electricity
status.
To summarize, LTE key features are listed below:
• High spectral efficiency
- OFDM in downlink, Robust against multipath interference & High affin-
ity to advanced techniques such as Frequency domain channel-dependent
scheduling & MIMO (Multiple-Input and Multiple-Output)
CHAPTER 1. INTRODUCTION 3

Rel-7 Study Phase

Rel-8 Work Phase
Test Specs
Rel-9 Work

Core specs First UE Commercial

drafted certification developing

First test First LTE

specs service
drafted
2005 2006 2007 2008 2009 2010 2011

Figure 1.1. LTE development Timeline

- DFTS-OFDM (“Single-Carrier FDMA”) in uplink, Low PAPR (Peak-to-

Average Power Ratio), User orthogonality in frequency domain
- Multi-antenna application

• Very low latency

- Short setup time & Short transfer delay

- Short handover latency and interruption time; Short TTI (Transmission
Time Interval), RRC (Radio Resource Control) procedure, Simple RRC
states

• Support of variable bandwidth

- 1.4, 3, 5, 10, 15 and 20 MHz

• Simple protocol architecture

- Shared channel based

- PS mode only with VoIP capability

• Simple Architecture

- eNodeB as the only E-UTRAN node

- Smaller number of Radio Access Network (RAN) interfaces

• Compatibility and inter-working with earlier 3GPP Releases

• Inter-working with other systems, e.g. cdma2000

CHAPTER 1. INTRODUCTION 4

S1
UEs
eNodeB

X2

MME / S-GW /
P-GW

UEs
eNodeB

UEs
eNodeB

Evolved
EUTRAN
Packet Core

Figure 1.2. EUTRAN architecture diagram [2]

• FDD and TDD within a single radio access technology

• Efficient Multicast/Broadcast

- Single frequency network by OFDM

• Support of Self-Organising Network (SON) operation

The next standard after LTE is LTE Advanced which is currently being stan-
dardized in 3GPP Release 10 [3][4]. LTE Advanced is expected to meet the re-
quirements for 4G which is also called IMT Advanced defined by the International
Telecommunication Union. For example the peak data rate is up to 1Gbps.
CHAPTER 1. INTRODUCTION 5

1.2 Motivation
For more efficient use of electricity, a smart grid should monitor and control more
status of itself than the traditional grid. To achieve this property, a smart grid needs
various sensors, controllers, actuators and the communication infrastructures used
for data transmission. So far, many different network types have been promoted
for use as those communication infrastructures, including Ethernet, power line car-
rier (PLC), cellular network, telephone/Internet, and short range radio frequency.
However, it is still critical to select a suitable network for a particular application.
On the other hand, wireless communication nowadays is a fast-growing tech-
nology. It has the main advantages of dynamic network formation, low cost, easy
deployment and reduced cable restriction. As the latest mobile communication
network, LTE becomes one promising option for a smart grid.
Therefore, the thesis established a hypothetical smart microgrid in which LTE
is the main communication standard. Firstly, the communication requirements in
this grid were summarized for both monitoring and control purposes. Secondly,
these requirements were compared with the performance of LTE network. Finally,
it was verified that LTE can server better for a smart grid with particular scheduling
scheme.

1.3 Outline
The rest of the thesis is structured as follows. Chapter 2 provides the definition
of problems solved in the thesis report. It discusses the model and experiment
configurations. Chapter 3-4 introduce the communication requirement in a smart
grid. The former chapter briefly summarizes the requirement focusing on latency
and bandwidth in a smart grid. The latter discusses the wireless communication
requirement in a hypothetical smart microgrid. Chapter 5 gives theoretical perfor-
mance analysis in terms of latency and bandwidth for LTE based on the study of the
3GPP Release 8. Chapter 6 presents the experiment results. Chapter 7 introduces
a simple scheduler design based on time-domain PF, and the simulation results
are given. Finally, Chapter 8 gives the conclusion of this thesis and suggestions of
possible future work.
Chapter 2

Problem Formulation

In this chapter, we give the formulation of the problems that we are to addressed
in this master thesis project. There are several main tasks in this project, including
communication requirements summary for a smart grid, performance analysis of
LTE and scheduler design for LTE.

2.1 Problem Definition

For the smart grid communication network, bandwidth and latency are two critical
technical requirements. In this thesis the performance of LTE in bandwidth and
latency was to be investigated in order to evaluate whether it is a suitable net-
work for a smart grid. We were to establish a simple experiment to collect latency
and peak data rates of LTE. After collecting enough data, we would analyse LTE
network performance, compare the performance with smart grid requirements and
design a suitable scheduler for smart objects in grid via LTE. Figure 2.1, Table 2.1
summarizes the required data and results for all problems defined above.

Table 2.1. Problem Formulation

Problem
Index Inputs Output
(Object)
Requirement Summery
1 Literature Latency[ms] and Peak rate[Mbps]
(Smart Grid)
Performance Analysis
2 Experiments Latency[ms] and Peak rate[Mbps]
(LTE)
3 1,2 Results Comparison Is LTE capable?
Scheduler Design
4 Simulation results
(LTE)

6
CHAPTER 2. PROBLEM FORMULATION 7

Figure 2.1. Experiment Structure

2.2 Model for Communication Requirement

There are many literatures describing communication requirements for a smart grid.
In this thesis, we would focus on the requirements of latency and bandwidth for
various length of message delivered in smart grids. To obtain a more accurate
requirement of communication in smart grids, we separated these requirements into
three main parts: a) communication in substations/distributed generation (DG);
b) collecting and dissemination of phasor data; and c) collecting and dissemination
of consumption data. See Chapter 3 for a clear definition of these requirements.
Considering there is relationship between the size of smart grid and specific
communication requirements, we were to establish a smart micro-grid as a more
specific model which consists of several key components, such as a DG, a controller
and different loads.

2.3 Experiment Configurations for LTE Performance

Analysis
Three main hardware devices and one key software “ping” would be used to trans-
mit information data packets in the experiment. Table 2.2 summarizes all devices
used in this experiments.

Table 2.2. Experiment hardware List

Type Connection Description

USB with computer 1. HUAWEI E398
LTE modem
RF with eNodeB 2. SAMSUNG GT-B3730
Computer USB with LTE modem InterCore T5600@1.83GHz RAM4G Windows7
1. Serviced by TELE2 with HUAWEI modem
eNodeB RF with LTE modem
2. Serviced by TELIA with SAMSUNG modem
CHAPTER 2. PROBLEM FORMULATION 8

Here we assume there is an individual cellular network with several users. Each
user has the approximately same priority and property. Thus the information trans-
fer between source and destination can be simplified to sending data packets from
user equipments to servers of network.
In this experiment, user equipment would send Internet Control Message Pro-
tocol (ICMP) echo request packets to the target host using ping command under
Windows 7 via LTE modem, and wait for an ICMP response. In the process, it
would measure the time from transmission to reception (round-trip time, RTT) and
recode any packet loss. The RTT and packet loss data would be used to analyse the
performance of LTE network. On the other hand, it would use the official software
of LTE modems to monitor the peak data rates when it downloads from or uploads
via Internet.
The basic batch files would be used to ping the eNodeBs host in series with
messages in different lengths. Considering that the length of message delivered in
smart grids are almost smaller than 1024 bytes, the length of message set in the
experiment varies from 0 bytes to 1024 bytes. The batch file would ping at least
1000 times for each message, and save the respective latency data into a txt file.
The basic performance of two LTE networks would be investigated and compared
using LTE modems from TELE2 and TELIA respectively.

2.4 Scheduler Design

To optimize the LTE performance for communications in a smart grid, we would cre-
ate a scheduler which gives highest priorities to smart objects. When the scheduler
is working, smart objects can use as many as possible resources of LTE to access net-
work, deliver data messages. The simulation result would indicate how many smart
objects LTE can handle at most under the given communication requirements.
Chapter 3

Communication Requirements for a

Smart Grid

Electrical grid nowadays consists three main subsystem: a) generation subsystem

which produces electricity; b) transmission subsystem which transmits electricity to
load centre; and c) distribution subsystem which continues to transmit electricity
to customers. The existing power grids have severed us with live consumption of
electricity for a long time. However, it has been proven that existing grids cannot
address the expected requirements including optimal generation control, demand
response, energy conservation and reduction of the industry’s overall carbon foot-
print [5]. Simply stated, the smart grid is created and expected to address these
drawbacks of the existing grids. As a result, a smart grid is required to be self-
healing, consumer participation, resist attack, high quality power, accommodate
generation options, enable electricity market, optimize assets, and enable high pen-
etration of intermittent generation sources. The salient features of smart grid in
comparison with the existing grids are shown in Table 3.1.

3.1 Components
To fulfil all the requirements listed in the table, a smart grid needs more components
for monitoring, control and communication. In most cases, a smart grid focuses on
three main areas: a) household devices for consumption automatic meter reading;
b) remote sensing devices for electrical network monitoring and control; and c) dis-
tributed power energy source, such as wind and solar, management. Therefore
there are several key components in a smart grid. They are, Advanced Meter In-
frastructures (AMI) at the houses or buildings, Phasor Measure Units (PMU) for
transmission lines, and remote sensing incorporation of distributed.

9
CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 10

Table 3.1. Comparison of a smart grid with the existing traditional grid [5]

Existing Grid Smart Grid

Electromechanical Digital
One-Way Communication Two-Way Communication
Centralized Generation Distributed Generation
Hierarchical Network
Few Sensors Sensors Throughout
Blind Self-Monitoring
Manual Restoration Self-Healing
Failures and Blackouts Adaptive and Islanding
Manual Check/Test Remote Check/Test
Limited Control Pervasive Control
Few Customer Choices Many Customer Choices

3.1.1 PMU
The Phasor Measurement Unit is considered to be one of the most important mea-
suring devices in the next generation power systems. The distinction comes from
its unique ability of providing synchronized phasor measurements of voltages and
currents in an electrical grid. The ability is achieved by same-time sampling of
voltage and current waveform using a common synchronizing sampling signal from
the Global Positioning Satellite (GPS). The phasor measurements are calculated via
Discrete Fourier Transform (DFT) applied on a moving data window whose width
can vary from fraction of a cycle to multiple of a cycle [6]. In an electrical grid, the
state of system is defined as the voltage magnitude and angle at each bus of the
system. Based on the measurements from PMUs, we can estimate the system state
in real time. Besides enhanced state estimation, PMUs are also used in phase angle
monitoring and control, Wide-Area power system stabilizer and adaptive protection.

Reporting Rate
Generally speaking, PMU measurements are reported at a rate of 20∼60 times a
second, namely 20∼60 Hz. Based on the GPS timing, each utility has its own
Phasor Data Concentrator(PDC) to aggregate and align data from various PMUs.
Measurements from each utility’s PDC are sent to the Central Facility (e.g. TVA’s
SuperPDC) where the data are synchronized across utilities.
CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 11

Message Types
Standard IEEE Std C37.118-2005 defines four message types for PMUs output:
data, configuration, header, and command [7]. The configuration, data and com-
mand messages are binary message; the header message is a human readable mes-
sage. There is an example, shown in Table 3.2, of the data message which carries
the measurements in its frame. The data frame indicates a balanced 3-phase phase-
to-neutral voltage and a constant system frequency. Here the length of the data
message is 52 bytes. In most cases, the lengths are about 100∼200 bytes including
the IP or TCP header (20 bytes).

Table 3.2. Data message example

Size
Field Description Example Value
(Bytes)
Synchronization
SYNC and Frame Formate DataFrame, V1 2 AA 01
Field.
FRAMESIZE Frame Length 52 bytes 2 00 34
PMU/ID number,16-
IDCODE 7734 2 1E 36
bit integer
SOC Second count 9:00 on 6 June 2006 4 44 85 36 00
Time of phasor mea-
16817 µs after the
FRACSEC surement [ms] with 4 00 00 41 B1
second mark
Time Quality.
VA=146356 0◦ 4 39 2B 00 00

Phasor data, 16-bit VB=146356 −120◦ 4 E3 6A CE 7C

PHASORS
integer ◦
VC=146356 120 4 E3 6A 31 83
◦
I1 =10926 0 4 04 44 00 00
16-bit signed integer. +2500 mHz (Nom-
FREQ Nominal frequency in inal 60 Hz with 2 09 C4
millihertz. measured 62.5 Hz).
Rate of change of fre-
DFREQ 2 00 00
quency
ANALOG1=100 4 42 C8 00 00
ANALOG 32 bit floating point ANALOG2=1000 4 44 7A 00 00
ANALOG3=10000 4 46 1C 40 00
Digital data, 16-bit 0011 1100 0001
DIGITAL 2 3C 12
field 0010
CHK CRC-CCITT 2 D4 3F
CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 12

3.1.2 AMI
Since nearly 90% of all power outage and disturbances have their roots in the dis-
tribution subsystem, automatic meter reading systems (AMR) in the distribution
subsystem are indispensable. AMR collects information of consumption records,
alarms and status from customers. Although AMR provides information on the
customers side, it does not address the major issue: demand-side management due
to its one-way communication system. As a result, advanced meter infrastruc-
ture (AMI) which contains two-way communication system is developed. AMI is
capable of both getting instantaneous information from customers and imposing
consumptions of customers. Simply stated, an AMI consists of an AMR, a two-way
communication system, and several specific actuators. Based on its two-way com-
munication system, AMI enable a smart grid to manage consumption demands. Its
main advantages are listed below:
• Realtime pricing. Customers adjust consumption decision based on day-
to-day or hour-to-hour price of electricity. These decisions will affect their
bills.

• Peak shaving. Remove energy consumption from period of high demand,

which helps to reduce need for peak-meeting generation. Given the fact that
20% of today electricity grid’s generation capacity exists to meet peak demand
only[5], peak-shaving helps to reduce and optimize the energy consumption.

• Energy conservation. Reducing usage overall can have substantial environ-

mental and social benefits.

Reporting Rate
Generally speaking, AMI needs to report consumption status at a rate of 4∼6 times
per hour, that is, each consumption measurement is sent every 10∼15 minutes.

Message Types
Referring to the message types for a PMU, we assume there are four different mes-
sage types for an AMI outputs: Data, Configuration, Header, and OutCommand;
one InCommand message for its input [8][9].
• Data message contains the measurements of the consumption from customers
including device ID, meter reading, time stamps, and other identification in-
formation about the customer and the AMI.

• Configuration message contains the configuration information of the AMI.

• OutCommand message informs the customers of real-time price and imposes

the consumption.

• Header: A human readable message.

CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 13

• InCommand message contains commands from supervisor controller.

We use 32 or more bits of floating for representing the meter reading in data
information, 8 bits for each device status, and 20 bytes for basic identification
information (including AMI ID, time date stamp, checksum and so on) in each type
of message. Table 3.3 lists some hypothetical parameters for these message types
[10].

Table 3.3. Four Type messages for AMI

Rate Length
Type Sink Destination Comments
(per hour) (Bytes)

Data AMI Controller 4∼6 <100 4 bytes per meter read-

ing

Configuration AMI Controller 100∼200 Sent to reply a request

only.

Custome 8 bits for on\off switch.

OutCommand AMI <100 Sent when status needs
Devices change.
Header AMI Controller <50
InCommand Controller AMI 100∼200 Sent when necessary.

3.1.3 Distributed Power

Distributed generations (DG) and substations are two of key power suppliers in a
smart grid. Generally, a DG can supply renewable and cheaper electricity such as
wind and solar, or provide power in emergency such as fossil fuel(coal, gas powered).
A substation transfers power from the transmission subsystem to the distribution
subsystem of an area. It transforms high voltage of transmit line to a suitable
low level for local distribution. As a result, it can provide relatively more reliable
electricity. However, in most cases, it can hardly predict the output of a renewable,
clean power generator. Moreover, a fossil fuel generator has heavy air pollution
affecting environment. Furthermore, a substation which relies on the centralized
generations is difficult to be adjusted. In order to produce clean, reliable electricity
power, one promising solution is the hybrid generators which are comprised of two
or more generators. Most hybrid configurations are based on a petroleum-fuelled
engine-driven generator and one or more solar or wind powered generators.

3.1.4 Remote Sensing: Monitoring and Control

Remote sensing of the network has been undertaken for some time in energy dis-
tribution grids using Supervisory Control and Data Acquisition(SCADA) solutions.
CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 14

These systems are deployed to monitor and control other infrastructure utilities
in the high voltage network of the electrical grid. These utilities usually include
high voltage switches, transformers, and transmission lines. When these systems
turns towards to a smart grid, the monitoring and control points are fundamen-
tally extended to the medium and low voltage networks. The utilization of PMUs,
other remote sensors and actuators makes a smart grid capable of better condition
managements.

3.2 Communication Requirement

In practice, there are different communication requirements for various functions
or applications in a smart grid. These can be categorized into requirements for
communication within a substation, communication between substations and con-
trol centres, communication between control centres, collection and dissemination
of electricity data, wide-area control and monitoring and communication for local
protective relaying [11]. As mentioned in Section 2.2, we would focus on communi-
cation in a substation/DG, collecting and dissemination of electricity (phasor and
consumption) data.

3.2.1 Communication in A Substation/DG

Standard IEC 61850-5 [12] introduces the communication requirements for functions
and device models in a substation. Based on IEC 61850-5, there are two independent
groups of performance class:
a) for Control and Protection (P1/P2/P3)
• P1 applies typically to the distribution level of the substation or in cases
where lower performance requirements can be accepted.
• P2 applies typically to the transmission level or if not otherwise specified
by the users.
• P3 applies typically to transmission level applications with high require-
ments, such as bus protection.
b) for Metering and Power quality applications (M1/M2/M3)
• M1 refers to revenue metering with accuracy class 0.5 (IEC 60687) and 0.2
(IEC 60044) and up to the 5th harmonic
• M2 refers to revenue metering with accuracy class 0.2 (IEC 60687) and 0.1
(IEC 60044) and up to the 13th harmonic
• M3 refers to quality metering up to the 40th harmonic.
IEC 61850-5 also defines seven different types of messages in a substation. Table
3.4 lists all these types with their different communication requirements of latency,
rate frequency and resolution.
CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 15

Table 3.4. Messages defined by IEC 61850 for a distribution substation

Performance Latency Rate Resolution

Type Description
Class (ms) (Hz) (bit/bit)
1A Fast Message (type A) P1 10
1B Fast Message (type B) P2,3 3
2 Medium speed Message <100
3 Low speed Message <500
P1 10 480 13/13
P2 3 960 16/16
P3 3 1920 16/18
4 Raw data Message
M1 1500 12/14
M2 4000 14/16
M3 12000 16/18
5 File transfer function >1000
6 Time synchronization
7 Command Message with <500
access control

3.2.2 Collecting and Dissemination of Phasor Data

In reality, measurements from PMUs are initially delivered to devices called Phasor
Data Concentrators (PDCs) where the measurements are time-synchronised, stored
for future reference and forwarded to applications and to Super PDCs. This process-
ing is one function of Wide-Area Measurement System (WAMS) [13]. Furthermore,
phasor measurement is a technology for the WAMS “backbone”, as shown in Figure
3.1. This phasor network consists of PMUs and PDCs. A complete WAMS will
also accommodate measurements of other types, and it will contain many resources
that convert acquired data into useful information.
Here is a practical example. The Western Electricity Coordinating Council
WAMS has reached the following size by the end of 2004:

• 11 PDC units

• 53 integrated PMUs

• 7 stand-alone PMUs

• ∼23 Portable Power System Monitors

• ∼10 monitor units of other kinds

CHAPTER 3. COMMUNICATION REQUIREMENTS FOR A SMART GRID 16

Analysis SCADA Master

PMU
Data Stream to
other Application
PMU Phasor Data Conterator

PMU
Monitoring
or
Recoder

Figure 3.1. General WAMS Structure

WAMS systems are used for both off-line studies and real-time applications.
With real-time WAMS, the continuous measurements feed out as a data stream
applied to on-line applications such as monitoring and control. They are expected
to meet real-time control system requirements with time delay less than 1 second
(typically 100-200ms) [13].

3.2.3 Collecting and Dissemination of Consumption Data

One key feature expected in a smart grid is demand-side management. Thus the
consumption information needs to be collected and analysed by the smart grid
controller. Then suitable commands are sent to customers’ devices for controlling
electricity demand in the following time period. This processing constructs classical
feedback control for demands. The main purpose of this control system is to shift
or remove peak electricity consumption for realizing energy conservation.
The AMI is one key component to fulfil this purpose in a smart grid. An
AMI network should meet certain communication requirements to achieve real-time
feedback control of electricity demand [10].
Reliability: The AMI network must guarantee the arrival of each AMI meter read-
ing and command messages.
Scalability: The designed network for AMI should be able to provide support to
a large number of nodes of AMIs. On the other word, the bandwidth of the
connection network should be large enough to support desired AMI service.
Real time communication: The time of end-to-end delay is required to be short
enough to fulfil the real-time control requirements.
Order: The data packets should be stamped with the time of measurements to
guarantee the order of the data packets in the receiving base.
Chapter 4

A Hypothetical Smart Microgrid

In order to specify the communication requirements, we establish a hypothetical

smart micro-grid which consists of three main subsystem: a) DG, b) WAMS, c) AMI
network. We describe the model in details in next subsection.

4.1 Assumption and Structure of Model

As shown in Figure 4.1, we assume that a 50MW DG works as the only electricity
supplier. Lines 1 and 2 are part of transmission subsystem. Line 3 and 4 connect
the transmission and distribution subsystem. Bus 4 and 6 are distribution bus.
Transformer 2 and 3 are mechanical on-load tap change transformers which used to
control the output voltages remotely. Breakers CB 1∼5 are remote controllable and
used to remove the lines. Load 2 transmits the electricity power to factories, while
Load 3 gives the electricity power to houses or buildings. There is a control centre
outside this DG besides a local controller inside the DG. Engineers can remotely
monitor and control those devices installed in the smart microgrid. In addition,
there are several PMUs deployed on all power buses, while a few AMI infrastruc-
tures are installed in houses, building and factories. Both the phasor measurements
and electricity consumptions are sent to the control centre via LTE network, and
transmitted to the data store system via Ethernet. PMUs and AMIs receive the
commands from control centre via LTE, while DG gets the commands vie Ether-
net. Table 4.1 shows the main parameters for the key components in this smart
microgrid.

4.2 Communication Requirement for LTE Connection

Based on the assumptions above, we conclude the communication requirements for
a smart grid.

17
CHAPTER 4. A HYPOTHETICAL SMART MICROGRID 18

PMUs

LTE
Bus 1 Line 1 Bus 2 Bus 4
50MW
CB 1 CB 2 Bus 5 Bus 6

DG Line 2 Line 4
Load 1
CB 3 CB 4 220V
Transformer 1 Load 3
Transformer 3
Bus 3
CB 5
Line 3

12.47KV
Load 2
Transformer 2

AMIs inside
PMUs on Bus Buildings and
Factories

Control Center

DATA Store System eNB eNB

Figure 4.1. A hypothetical smart microgrid

Table 4.1. Main parameters in the hypothetical smart microgrid

Component Quantity Packet Length (Bytes) Rate Comments

DG 1 50MW
PMU 10 200 Uplink 60Hz
AMI 10,000 100 Uplink every 15min
Others 100 100 Uplink 60Hz IED status
Control Centre 1 100 Downlink depends Control commands
CHAPTER 4. A HYPOTHETICAL SMART MICROGRID 19

1. Latency

• PMU, Control Centre & Others

Recalling to Chapter 3, [13] announces that transmission latency should
be less than 1 second (typically 100∼200 microsecond) for WAMS solu-
tion. Referring to [11], the latency is required to be less than 10 microsec-
ond. For better performance, we choose 10 microsecond as the deadline
for latency in this thesis.
• AMI
Since an AMI reporting rate is less than 1Hz, 1 second latency is short
enough for its real-time operation requirement.

2. Bandwidth
In the worst case, all devices upload their measurements simultaneously. In
other words, all possible data frames need to be sent before the latency dead-
line. Considering the parameters shown in Table 4.1 and latency requirements
listed above, we can easily determine the required bandwidth in both uplink
and downlink for worst case. The results are shown in Table 4.2.

Table 4.2. Communication requirements for worst case

Components Total Length(bits) Latency(ms) Bandwidth(Mbps)

PMU 10 × 2000 10 2
AMI 10, 000 × 1000 1000 10
Uplink
Others 100 × 1000 10 10
SUM 22
(10 + 100) × 1000 10 11
Control Centre
Downlink 10, 000 × 1000 1000 10
SUM 21

4.3 Conclusion
We specified the communication requirements in term of latency and bandwidth
mainly based on IEEE standards and related research. For the hypothetical smart
microgrid, the latency less than 10 ms and the peak rate larger than 22 Mbps are
required. The performances of LTE are determined both theoretically and exper-
imentally in following chapters to evaluate whether LTE fulfils the requirements
obtained above.
Chapter 5

Analysis of Latency and Bandwidth of

LTE

We theoretically estimate the performance of LTE in terms of latency and band-

width based on the standards from 3GPP [14].

5.1 Introduction
5.1.1 Main Techniques in LTE
OFDMA
Orthogonal Frequency-Division Multiple Access (OFDMA) is a multi-user version
of the popular Orthogonal frequency-division multiplexing (OFDM) digital modu-
lation scheme. OFDM realizes information transmission on a radio channel through
variations of a carrier signal’s frequency, phase and magnitude [15]. Rather than as-
signing all transmission information to a single carrier, OFDM cuts that information
into smaller pieces and places each one to a specific subcarrier. The subcarriers are
offset in frequency (δf ) which implements orthogonality to prevent interferences.
A serial stream of binary digits s[n] is to be transmitted as shown in Figure 5.1.
This stream corresponds the data produced by the measurements of a smart grid.
By inverse multiplexing, it is firstly demultiplexed into N parallel streams. Then
each stream is mapped to a complex data sequence (X0 , X1 , ..., XN −1 ) using certain
modulation, such as QAM, 16QAM. An inverse fast Fourier transform (IFFT) is
performed on this sequence, resulting in a set of complex time-domain samples.
These samples are used to generate the transmission signal ν(t).
At the receiver, the signal ν(t) is demodulated using an FFT process to convert
the time-domain samples back to Xn sequence. The original binary digits s[n] are
recovered from Xn using their modulations.
The main advantages of OFMDMA are high spectral efficiency and robustness
against intersymbol interference and fading caused by multipath propagation.

20
CHAPTER 5. ANALYSIS OF LATENCY AND BANDWIDTH OF LTE 21

11
01
11 01 ... 10

...
s[n] 10 Xn IFFT

Binary Modulation Transmitted

stream mapping waveform

Figure 5.1. Construction of a multicarrier OFDM signal

SC-FDMA
Single-carrier FDMA (SC-FDMA) is a frequency-division multiple access scheme.
In essence, as compared to OFDMA, SC-FDMA scheme has just one more FFT
process when creating transmission radio signals as shown in Figure 5.2.

Figure 5.2. Physical layer structure of LTE

Therefore SC-FDMA not only provides the advantages of OFDMA, especially

robust resistance to multipath, but also has lower PAR. However, LTE only uses
SC-FDMA in the uplink because the increased time-domain processing would be a
considerable burden on eNodeBs which have to manage the dynamics of multi-user
transmission [16].

MIMO
In radio, multiple-input and multiple-output, or MIMO (commonly pronounced my-
moh or me-moh), is a technology that makes use of multiple antennas at both the
transmitter and receiver to improve communication performance.
CHAPTER 5. ANALYSIS OF LATENCY AND BANDWIDTH OF LTE 22

5.2 Latency Analysis

The requirement metric by International Telecommunication Union (ITU) for LTE
latency is shown in Table 5.1. Control plane latency is defined as the time required
for the User Equipment (UE) to transit from idle state to active state. In idle
state, the UE does not have a Radio Resource Control (RRC) connection. Once the
RRC is established, the UE turns into the connected state and then into the active
state when it enters the dedicated mode. User plane latency is defined as one-way
transmit time between a packet being available at the IP layer in the UE/E-UTRAN
edge node and the availability of this packet at the IP layer in the E-UTRAN/UE
node. The User plane latency would be focused on in this thesis.

Table 5.1. IMT-A Latency Requirement

Max Latency
Plane
(ms)
Control Plan 100
User Plan 10

User plane latency is calculated for both TDD (Time Division Duplex) and
FDD (Frequency Division Duplex) cases. The components involved in the latency
calculation are:

1. UE processing time

2. Transmission Time Interval (TTI) duration

3. Hybrid Automatic Repeat Request (HARQ) retransmission

4. eNodeB processing time

5.2.1 FDD Mode Analysis

Figure 5.3 shows the FDD components for User plane. In FDD mode HARQ process
is fixed to 8 ms. So HARQ retransmission delay is n × 8 ms, where n is number of
retransmission. Table 5.2 shows the User plane latency calculations for FDD.

5.2.2 TDD Mode Analysis

Figure 5.4 shows the TDD components for User plane. In TDD, HARQ retransmis-
sion time equals the RTT via LTE. Table 5.3 shows the latency calculation for TDD
with 0% and 10% HARQ respectively. Different TDD UL/DL frame allocations lead
to different alignment time. Due to different alignment time, the latency for TDD
varies from 4.1 ms to 5.2 ms even for downlink without HARQ retransmissions.
CHAPTER 5. ANALYSIS OF LATENCY AND BANDWIDTH OF LTE 23

UE TTI eNB

1.5 ms 1 ms 1.5 ms

HARQ

1.5 ms 1 ms 1.5 ms

Figure 5.3. User plane latency components for FDD

Table 5.2. User plane latency analysis for FDD

HARQ 0% HARQ 10%

Components Process
(ms) (ms)
1 UE Processing Time 1.5 1.5
2 TTI Duration(Fixed) 1 1
3 HARQ Retransmission 0 0.8
4 eNodeB Processing Time 1.5 1.5
Total one-way delay 4 4.8

5.3 Bandwidth Analysis

The LTE downlink transmissions from the eNodeB consist of user-plane and control-
plane data from the higher layers in the protocol stack multiplexed together with
physical layer signalling to support the data transmission [17]. The multiplexing
of all these downlink signals is facilitated by the OFDMA described in Section 5.1,
which enables the downlink signal to be subdivided into small units of time and
frequency. The LTE uplink transmissions are similar to the downlink, and it selects
SC-FDMA, as introduced in Section 5.1, as its multiple-access scheme to fulfil some
unique principle characteristics.
Both downlink and uplink transmission resources in LTE possess dimensions of
time, frequency and space. The time-frequency resources are subdivided according
to the following strategies: the largest unit of time is the 10-ms radio frame, which
is further subdivided into ten 1ms subframes, each of which is split into two 0.5-ms
slots. Each slot contains seven OFDM symbols in case of the normal cyclic prefix
(CP) length, or six if the extended CP is configured in the cell. In frequency domain,
CHAPTER 5. ANALYSIS OF LATENCY AND BANDWIDTH OF LTE 24

Downlink
UE TTI eNB
1.5 ms 1 ms 1 ms + tFA

Uplink
UE TTI eNB

1 ms + tFA 1 ms 1.5 ms

Figure 5.4. User plane latency components for TDD

Table 5.3. User plane latency analysis for TDD in uplink

Downlink HARQ Uplink HARQ

Component Process
0% 10% 0% 10%
(ms) (ms) (ms) (ms)
1 UE Processing Time 1.5 1.5 1 1
2 Frame Alignment 0.6∼1.4 0.6∼1.4 1.1∼5 1.1∼5
3 TTI Duration(Fixed) 1 1 1 1
4 HARQ Retransmission 0 0.98∼1.22 0 1∼1.16
5 eNodeB Processing Time 1 1 1.5 1.5
Total one-way delay 4.1∼5.2 5.18∼6.2 4.6∼8.5 5.76∼9.5

resources are grouped in unite of 12 subcarriers, such that one unit of 12 subcarriers
for a duration of one slots is termed a Resource Block (RB) as shown in Figure 5.5.

The smallest unit of resource is the Resource Element (RE) which consists of
one subcarrier for a duration of one OFDM symbol. So a RB is comprised of 84
(with normal CP) or 74 (with extended CP) RE respectively. Within certain RBs,
some REs are reserved for special purpose: a) synchronization signals, b) reference
signals, c) control signalling, and d) critical broadcast system information. The
remaining REs are used for data transmission, and are usually allocated in pairs of
RBs.
In order to evaluate the peak rates in both downlink and uplink, we consider
the best case in both sides of UE and eNodeB. In the downlink, the most efficient
modulation is 64QAM (6 bits) for one OFDMA symbol (RE). If we assume that
there are 100 RB for 20-MHz RF bandwidth in each 0.5 ms with single antenna
CHAPTER 5. ANALYSIS OF LATENCY AND BANDWIDTH OF LTE 25

One radio frame = 10 ms

84 subcarrier

M
ult
One subframe (2 slots)(1 ms)
{
ipl
e
sp
ati
al la
ye
rs
12 subcarrier

1 Resource
element
{

1 Resource block

Figure 5.5. Basic time-frequency resource structure of LTE (normal CP case)

and normal CP, the downlink peak rate could be easily calculated by 100 × 6 ×
84 bits/0.5 ms = 100 M bps. However, the most efficient modulation used in uplink
is 16QAM (4 bits) for one RE. Similarly, the uplink peak rate is obtained by 100 ×
4 × 84 bits/0.5 ms = 67.2 M bps.
Chapter 6

Experiment Results

In this chapter, we estimated the performances of LTE in terms of loss rate, latency
and peak rates in both uplink and downlink experimentally. The experiments are
carried on using two brands of LTE modems from operators TELE2 and TELIA
respectively.

6.1 Loss Rate via LTE

We pinged two IP addresses for more than ten thousand times at different time
interval. Table 6.1 shows the numbers of both the data packets sent by computers
and the data packets lost by communication networks. The results show that the
loss rate of LTE is much lower than 0.01%, which proves LTE is a reliable network.

Table 6.1. Loss rates of LTE communication network

LTE Service Loss Rate

Packets Sent Packets Lost
Operators (%)
TELE2 202,000 5 0.0025
TELIA 80,000 20 0.025

6.2 Latency Analysis via LTE

Because we used ping command to send and receive data packets, the lengths of
data packets only represent the sizes of data packets which are attached after ICMP
headers. Ping command not only sends and receives these ICMP data packets, but
also measures and reports the round-trip time (RTT) for each packet. Figure 6.1
shows the mean values and standard variances of RTT time measured by ping com-

26
CHAPTER 6. EXPERIMENT RESULTS 27

mand for respective small data packets(≤100 bytes), while Figure 6.2 shows those
for larger data packets(≥100 bytes). And Table 6.2 shows several parameters of
RTT values corresponding to the lengths of data packets, including mean, standard
variance, maximum and minimum values.

50

Standard Variance for TELE2

45

Standard Variance of RTT [ms]

Standard Variance for TELIA
Mean Value RTT for TELE2
40 20
Mean Value of RTT [ms]

Mean Value RTT for TELIA

35

30

25

20 10

15

10

0 0
0 10 20 30 40 50 60 70 80 90 100
Length of Data Packet [byte]

Figure 6.1. Mean values and Standard variance of RTT for small data packets.

These table and figures illustrate that when the length of data packets is smaller
than 100 bytes, the RTT is shorter than 20 ms under the service provided by
TELE2, while it is around 20 ms for TELIA. The RTT values increase with the
length of data packets when the length is larger than 100 bytes. But the standard
variances of RTT are approximately the same whatever sizes for the packets for
each service operator. The mean values of RTT under TELE2 are lower than those
under TELIA. However, the standard variances of RTT under TELE2 is around 4∼5
times larger than those under TELIA. This is probably because the measurement
equipment is closer to eNodeB of TELE2 than that of TELIA. The equipment could
use more efficient data modulation schemes and turbo coding to send message via
TELE2 LTE network. As a result, the messages sent by TELE2 modem have shorter
RTT, while there is little probability of being re-transmitted by HARQ in TELIA
LTE network.
If we calculate the latency by dividing RTT by half. The minimum values
for small size packet verifies the theoretical latency which described in Chapter
5. Considering the sizes of data messages used in PMUs, it is 100 bytes packets
that we try to find the distribution of RTT for. Since the ping command uses sort
CHAPTER 6. EXPERIMENT RESULTS 28

50

Mean Value RTT for TELE2

45

Standard Variance of RTT [ms]

Mean Value RTT for TELIA
Standard Variance for TELE2
40 20
Mean Value of RTT [ms]

Standard Variance for TELIA

35

30 15

25

20 10

15

10 5

0 0
100 200 300 400 500 600 700 800 900 1000
Length of Data Packet [byte]

Figure 6.2. Mean values and Standard variance of RTT for large data packets.

Table 6.2. RTT time for different data packets lengths

Length TELE2 (ms) TELIA (ms)

(byte) Mean Std Min Max Mean Std Min Max
50 16.4740 13.1485 10 209 20.3130 2.8826 13 76
100 18.5450 17.8746 10 187 26.4750 2.3721 13 43
200 22.3390 14.5283 12 173 27.2510 3.3976 18 97
300 23.8780 14.6845 11 187 30.3490 3.2186 21 68
400 25.3300 13.0220 19 206 30.3800 4.9175 21 153
500 26.2300 14.6324 19 216 30.3990 5.1371 22 155
... ... ... ... ... ... ... ... ...
1000 29.6577 19.2486 19 222 30.8610 3.5660 21 85
CHAPTER 6. EXPERIMENT RESULTS 29

of quantization filter when calculating the RTT time, we can not use probability
density function directly (pdf) to fit distributions. In this thesis, we integrated pdf
functions for each 1-ms interval, [i − 0.5, i + 0.5](ms) as following:
Z i+0.5
FRTT (i) = fpdf (x)dx for all i = 0, 1, 2, ... (6.1)
i−0.5

Where FRTT (i) is the possibility of measured RTT which equals i, fpdf is the
probability density function of certain distribution, for instance, normal and Poisson
distributions. Before fitting probability distribution, we transfer the data set to
histogram, and then to probability function Pr(X ∈ [i − 0.5, i + 0.5]) as shown in
Figure 6.3. The probability function Pr(X ∈ [i − 0.5, i + 0.5]) was calculated using
Equation 6.2, where n is number of times i occurs inside N samples.
n
Pri = Pr(X ∈ [i − 0.5, i + 0.5]) = for all i = 0, 1, 2, ... (6.2)
N

0.35
200 300

180 0.3
250
160
0.25
Quantity of RTT

140
RTT Value [ms]

200
Probability

120 0.2

100 150
0.15
80
100
60 0.1

40
50 0.05
20

0
0 0 10 15 20 25 30 35 40 45 50
0 100 200 300 400 500 600 700 800 900 1000 10 15 20 25 30 35 40 45 50
Z Z RTT Value [ms]

Figure 6.3. Data Pre-processing. The data set used in this figure is 1000 RTT
values collected with 100 bytes data packets via TELE2 LTE network. The left figure
shows the point of RTT values. The middle one illustrates the histogram, while the
right one shows the probability density function of RTT values.

6.2.1 Model
After the above data preprocessing, both histograms and pdf figures illustrate that
the RTT values seems to fit the mixture distribution because of several obvious
peeks in Figure 6.3 and Figure 6.4. Furthermore, these figures indicate that the
data might appropriately fit a mixture of two or more normal distribution. If we
assume that the possibility of HARQ occurs is p and HARQ happens for j times for a
data packet, the probability for successful sendings can be calculated using Equation
6.3. We assume the mean values and variances of the each normal distribution are
the same. Then we propose to model this mixture of normal distributions as:

PSuccess = pj−1 (1 − p) (6.3)

H (x−µ−jTHARQ )2
X 1 −
fpdf (x) = (1 − p)pj−1 √ e 2σ 2 (6.4)
j=0 σ 2π
CHAPTER 6. EXPERIMENT RESULTS 30

TELE2
TELIA
0.25

0.2
Probability

0.15 Peaks

0.1

0.05

0
10 15 20 25 30 35 40 45 50
RTT Value [ms]

Figure 6.4. Probability density function with 100 bytes data packets via TELE2
and TELIA LTE network.

Where µ and σ are mean and standard variance of normal distribution respectively;
THARQ is the time spent to re-send the data packet; and H indicates the maximum
times of HARQ for sending a packet. Based on Equation 6.1, we determine the
parameters of the distribution by:

for all i = 0, 1, 2, ...

Z i+0.5
FRTT (i) = fpdf (x)dx
i−0.5
H
Z i+0.5 X (x−µ−jTHARQ ) 2
1
= (1 − p)pj−1 √ e− 2σ 2 (6.5)
i−0.5 j=0 σ 2π

6.2.2 Maximum Likelihood & Least Squares Estimations

The methods to estimate the appropriate fitting are the maximum likelihood(MLE)
and least squares estimations(LSE).
CHAPTER 6. EXPERIMENT RESULTS 31

1. Maximum Likelihood Estimation:

Suppose there is a sample x1 , x2 , ..., xn of observations coming from RTT
dataset with FRT T (i). Rewrite FRT T (i) as Pr(i|Θ) which means the probabil-
ity of RTT which equals i given parameter set Θ. Here,

Θ = [µ, σ, p, THARQ ]T (6.6)

Then the joint density function for all observations is obtained as:

Pr(x1 , x2 , ..., xn |Θ) = Pr(x1 |Θ) · Pr(x2 |Θ) · · · Pr(xn |Θ) (6.7)

Now we can get the likelihood function as:

n
Y
L(Θ|x1 , x2 , ..., xn ) = Pr(x1 , x2 , ..., xn |Θ) = Pr(xk |Θ) (6.8)
k=1

In this thesis, we used the logarithm of the likelihood function called log-
likelihood: n X
ln L(Θ|x1 , x2 , ..., xn ) = ln Pr(xk |Θ) (6.9)
k=1

The method of MLE estimates the appropriate parameter set θ by finding a

value of Θ which maximizes its likelihood function ln L(Θ|x):

θMLE = argmax ln L(Θ|x1 , x2 , ..., xn ) (6.10)

θ∈Θ

2. Least Squares Estimation:

We supposed the data set for RTT consists of n points(data pairs)(i, Pri )
obtained by Equation 6.2 as illustrated in Figure 6.4, where i and Pri are
obtained from observations. The model function FRTT (i) is given by Equation
6.5. The goal is to find the parameter set Θ for this model. The least squares
estimation finds its optimum by minimizing the sum of squared residuals:
N
X
ΘLSE = argmin ri2 where ri = Pri − FRTT (i) (6.11)
θ∈Θ i=1

6.2.3 Result
Based on the model given by Equation 6.4, we get the “best” fitting distribution
functions by applying MLE and LSE. Figure 6.5, 6.6 and Table 6.3 show those
parameters and curves estimated by MLE and LSE.
These result shows that it is appropriate to use the mixture normal distribution
model for analysis.
CHAPTER 6. EXPERIMENT RESULTS 32

0.35

Data
0.3
MLE Fit Curve
LSE Fit Curve

0.25
Probability

0.2

0.15

0.1

0.05

0
5 10 15 20 25 30 35 40 45 50 55
RTT Value [ms]

Figure 6.5. RTT values distribution. The data set is collected via TELE2 LTE
network with 100 bytes data packets.

0.25

Data
MLE Fit Curve
0.2 LSE Fit Curve
Probability

0.15

0.1

0.05

0
15 20 25 30 35 40 45
RTT Value [ms]

Figure 6.6. RTT values distribution. The data set is collected via TELIA LTE
network with 100 bytes data packets.
CHAPTER 6. EXPERIMENT RESULTS 33

Table 6.3. Summary fits of RTT values for both TELE2 in Figure 6.5 and TELIA
in Figure 6.6

TELE2 TELIA
MLE LSE MLE LSE
SSE r2 0.8075 0.8075 0.9872 0.9872
Mean Value[ms] µ 13.9111 14.1847 26.3201 26.4449
Standard Variance σ 1.7333 1.7894 2.0160 1.8635
Time for HARQ[ms] THARQ 7.8892 7.5175 8.7179 9.2640
Probability of HARQ p 0.1875 0.2241 0.0175 0.0160
Maximum Repeat times H 3 3 2 2

6.3 Throughput via LTE

We use different softwares and protocols to achieve the peak downlink and uplink
data rate. Table 6.4 shows the peak data rates via both TELE2 and TELIA LTE
networks. It can be seen that these two service operators provides the same peak
rate for both uplink and downlink.

Table 6.4. Peak data rates of LTE communication network

LTE Service Peak Downlink Rate Peak Uplink Rate

Operators (Mbits/s) (Mbits/s)
TELE2 Approaching 30 Approaching 30
TELIA Approaching 30 Approaching 30
Chapter 7

Scheduling

Unlike 3GPP, the scheduling decisions are not made at centralized base station but
at distributed eNodeBs for both downlink and uplink radio transmissions. The
process of LTE scheduling is shown in Figure 7.1.
The main purpose of a scheduler is to allocate suitable physical resources for the
set of users for communication. Because OFDMA and SC-FDMA are used in down-
link and uplink respectively, LTE can schedule resources for users in time domain,
frequency domain and modulation/coding scheme(MCS) domain. In order to make
good scheduling decisions for high performance, a scheduler requires knowledge of
both channel conditions and users’ device conditions. Based on these knowledges,
Adaptive Modulation and Coding(AMC) is used to choose optimal MCS for UEs to
transmit, for example, using QPSK, 16-QAM, or 64-QAM following turbo coding
for certain subcarriers. Ideally, the scheduler needs to know the channel condition
for each subcarrier, each UE in every scheduling time and TTI in LTE. However, due
to limited signalling channel resources, a UE seldom reports all subcarrires channel
conditions, but average condition and several conditions in best subcarriers. The
smallest resource unit allocated by a scheduler to a user is an SB, which consists
two consecutive RBs, spanning TTI of 1 ms and a bandwidth of 180 kHz.

7.1 Scheduler Design

7.1.1 Problem Formulation
As mentioned above, the main purpose of a scheduler is to determine the alloca-
tion of SBs to a subset of UEs in order to maximize some objective function, for
instance, overall system throughput and fairness. Here, due to more specific com-
munication requirements for PMUs to update data measurements, we created a
suitable scheduling scheme in uplink with the following requirements.
• Fulfil communication requirements for PMUs in the smart grid of short latency
and high bandwidth.
• Trade off overall system throughput and fairness for each UE.

34
CHAPTER 7. SCHEDULING 35

UE eNB UE eNB
Data
CQI Buffer Status Report

Scheduling Scheduling

Downlink Signalling Uplink Signalling

Data Transmission Data Transmission

Downlink Scheduling Uplink Scheduling

Figure 7.1. Downlink and uplink scheduling process

Now we can design the scheduler as solving an optimal problem which describes
the allocation performance of the physical resources. In this case, the uplink sched-
uler allocates resources for N users. So the optimal problem can be simplified to
allocate (i, j) resources block in time- and frequency-domain as shown in Figure 7.2,
(c)
which maximizes the sum of utility log Ri,j . The problem is:
N NX
TTI N RB
X X (c)
max log Ri,j (7.1)
c=1 i=1 j=1

where:
(c) (c)
Ri,j = λ(c) xi,j (7.2)
subject to:
P (c)
xi,j ≤ 1 ∀i, j (7.3)
c
P (c)
xi,j ≤ NTTI ∀j, c (7.4)
i
P (c)
xi,j ≤ NRB ∀i, c (7.5)
j
P P (c)
xi,j ≤ L(c) ∀c (7.6)
i j
(c)
xi,j ∈ {0, 1} (7.7)
(c)
where λ(c) is the utility weight function for the c-th user, xi,j is the occupant status:

 1 if (i, j) resource block is occupied by c-th user

(c)
xi,j = (7.8)
 0

otherwise
CHAPTER 7. SCHEDULING 36

NRB
(i,j) resource block

Frequency
1

1 NTTI
Time

Figure 7.2. Resource allocation

There are several constraints on this problem. Equation 7.3 indicates that each
resource block can be allocated to one user at most, and Equation 7.4 and 7.4 limit
the greatest value in time and frequency domain. Equation 7.6 formulates that
resource blocks allocated to UE are limited by each UE transmission demand.

7.1.2 Solutions
We assume that the weight function λ depends on users’ information only. Thus
user devices only report the average channel condition for all available channels in
every TTI. Then, Equation (7.1) becomes:
N
X
max λ(c) l(c) (7.9)
c=1

where:
NX
TTI N RB
X (c)
l(c) = xi,j (7.10)
i=1 j=1

subjects to:
X
l(c) ≤ min(NTTI NRB ,
P
L(c) ) (7.11)
c c
0≤ l(c) ≤ min(NTTI NRB , L(c) ) ∀c (7.12)

Figure 7.3 shows an example of the scheduling allocation, where the value inside
each resource block represents the utility value for each user. The resource blocks
CHAPTER 7. SCHEDULING 37

User H 5 5 5 5 5 5 5 5 5 5 5
User G 7 7 7 7 7 7 7 7 7 7 7
User F 8 8 8 8 8 8 8 8 8 8 8
User E 9 9 9 9 9 9 9 9 9 9 9
User D 8 8 8 8 8 8 8 8 8 8 8
User C 2 2 2 2 2 2 2 2 2 2 2
User B 4 4 4 4 4 4 4 4 4 4 4
User A 3 3 3 3 3 3 3 3 3 3 3
1 k NTTI NRB
Index

Figure 7.3. Allocation example

selected by the scheduler are labelled by red boxes. Obviously the solution for this
optimal problem has to fulfil:
X X
l(c) = min(NTTI NRB , L(c) ) (7.13)
c c

Optimal Solution:
Before solving this problem, we sorted data set λ(c) from largest to smallest to
a value sequence {λi }, where λ1 ≥ λ2 ≥ ... ≥ λN . According to this order, the
allocation solution is also transferred to sequence {li } with up-limit {Li }. Using
discrete dynamic programming, we obtained the solution for each i. The following
expressions show the steps to find the first two optimal solutions.
• To find the optimal solution for l1 :
X X
λ(c) l(c) = λi li
c i
= λ1 l1 + λ2 l2 + ... + λN lN

=⇒ s1 = argmax(λ1 l1 + λ2 l2 + ... + λN lN ) = min(NTTI NRB , L1 ) (7.14)

l1

Proof:
Assume there exist:

{lˆi } = argmax and lˆ1 < s1

X
λi li
{li } i

Denote δ = s1 − lˆ1 > 0 and assume arbitrary ∆li ≥ 0 fulfils:

N
lˆi − ∆li ≥ 0
X
∆li = δ and ∀i = 2, 3, ..., N
i=2
CHAPTER 7. SCHEDULING 38

Then, from:

λ1 ≥ λ2 ≥ ... ≥ λN
N
λ1 δ − λN δ ≥ λ1 δ − λi ∆li ≥ λ1 δ − λ2 δ ≥ 0
P
=⇒
i=2

N
lˆi :
P
IF δ <




i=2



 P ˆ
λi li = λ1 lˆ1 + λ2 lˆ2 + ... + λN lˆN




i



 N
≤ λ1 lˆ1 + λ2 lˆ2 + ... + λN lˆN + λ1 δ −
 P
λi ∆li


=⇒ i=2

= λ1 s1 + λ2 (lˆ2 − ∆l2 ) + ... + λN (lˆN − ∆lN )







ELSE :





N N


λ1 s1 = λ1 lˆ1 + λ1 δ ≥ λ1 lˆ1 + λ1 lˆ2 ≥ λ1 lˆ1 + λi lˆ2 = λi li

 P P P


i=2 i=2 i
=⇒ Conflict with assumption.

• To find the optimal solution for l2 :

s2 = argmax(λ1 s1 + λ2 l2 + ... + λN lN )
l2

 min(NTTI NRB − s1 , L2 )

if NTTI NRB − s1 > 0
= (7.15)
 0

otherwise

Similarly, we can find the solutions set {si } for other li step by step.

7.1.3 Utility Function Design

Since LTE transmits data flows not only for use in a smart grid but also for public
use in a wide area, we would like to determine what data comes from smart grid.
Due to short delay dead-line for PMUs, the scheduler needs to give the highest
priorities to ensure message from PMUs would be transmitted first. As a result, we
created a scheduler which consists of an estimator and a priority calculator. The
estimator determines the possibility for UEs of being in a smart grid based on the
buffer status reports from UEs. The features of a UE used in a smart grid can
be summarized as follows: a) constant data updating rates ṙ = 0, b) equivalent
data packets lengths l˙ = 0 and c) approximately invariant channel qualities q̇ = 0.
Then priority calculator gives the priority value to each UE. With combining all the
components in the scheduler, the whole utility function can be obtained as Equation
7.16 for c-th UE.
λ(c) = WP,c + PPF,c (7.16)
CHAPTER 7. SCHEDULING 39

Here WP,c indicates the weight for the UE in a smart grid and can be obtained by

P
WP,c = (7.17)
0.3ṙ + 0.5l˙ + 0.2q̇

Where PP F,c is traditional scheduling algorithm as given by Equation (7.18) which

is used in variant wireless communication networks. Here (C/I) represents the
channel conditions and R(t) is the throughput within the time interval t.

(C/I)c
PPF,c = (7.18)
Rc (t)
The whole algorithm is concluded as the following Algorithm 1:

Algorithm 1 Scheduling
1: Let M be the set of available RBs at time interval t
2: Let N be the set of schedulable UEs
3: for i = 0 to n do
4: calculate λi based on buffer status reports and channel conditions
5: map data packets of ith UE to required RBs quantity ci using AMC
6: end for
7: Index ← 1;
8: while M 6= ∅ do
9: pick the user k ∈ N with largest value λk
10: if M ≤ ck then
11: assign Index : Index + ck RBs to kth user
12: M ← M − ck
13: Index ← Index + ck
14: else
15: Assign Index : m RBs to kth user
16: M ←∅
17: end if
18: end while

7.2 Simulation
In order to evaluate the performance of our scheduler, uplink system level simula-
tions have been conducted based on 3GPP LTE system model. Table 7.1 summa-
rizes a list of the simulation parameters and assumptions.
We analyzed the performance of the scheduler in terms of latency of PMUs
as well as total throughput. As illustrated in Figure 7.4, it can be seen that the
short latency requirements are fulfilled by using this scheduler which always gives
the highest priority on smart objects. However, in the first TTI, the latency is
much more inconstant than the others. This is because the scheduler is not able to
CHAPTER 7. SCHEDULING 40

Table 7.1. Simulation parameters

Parameter Setting
System bandwidth 5MHz
Subcarriers per RB 12
OFDM symbols per RB 7
RB bandwidth 180Hz
Number of RBs 25
Cell-level user distribution Uniform
Number of PMUs in cell 1,2,3,4,5
Rate for PMUs 60Hz
Data Packet size for PMUs 53bytes
Number of other active users in cell 10
Traffic model Uniform
Transmission time interval (TTI) 1 ms
User move speed Slow
Modulation and coding setting QPSK,16QAM,64QAM
HARQ model None
Simulation Time 50 TTI

allocate smart objects for communication with the highest priority due to limited
information in the first place. Figure 7.5 shows the RBs allocations with 1 PMU
and 10 other UEs.
Based on this simulation, we can easily obtain how many PMUs can be mounted
in a cell fulfilling the 10 microsecond latency requirement. Here we assume that a
PMU data message is 200 bytes and sent at a rate of 60 Hz. Since the scheduler
always gives the highest priorities to PMUs, PMU data messages are always allo-
cated to physical resource first. In a 10-ms interval, LTE can transmit 67 Mbps×10
microsecond= 670 Kbits≈ 67 KB in uplink. Therefore LTE can handle 335 PMUs
in one cell on uplink. For AMIs whose latency deadline is 1 second, data message
length is 100 bytes and sent every 15 minutes, this result is calculated to be 67
Mbps×1 second= 67 Mbits≈ 6.7 MB, which means LTE can accommodate up to
67,000 AMIs communication in uplink even in the worst case.
CHAPTER 7. SCHEDULING 41

6.5

6
Latency [ms]

5.5

4.5

1 PMU
4 2 PMUs
3 PMUs
3.5 4 PMUs
5 PMUs
3
1 2 3 4
i th PMU sent data message

Figure 7.4. Latency simulation of PMU data messages

15

10
CQI

0
50
25 45
40
20 35
30
15
25
10 20
15
5 10

RB 5 TTI

Figure 7.5. Simulation allocation result for 1 PMU and 10 other UEs. Different
colours represent different UEs, besides red colour illustrates PMU data packet.
Chapter 8

Conclusion & Future Work

8.1 Conclusion
This master thesis investigated the performance of LTE network utilized in a smart
grid. The study has proven that LTE network is a promising solution due to its low
latency and large bandwidth.
Firstly, the communication requirements in a smart objects are categorized into
three parts: a) communication in substations/distributed generation (DG); b) col-
lecting and dissemination of phasor data; and c) collecting and dissemination of
consumption data. In each part, we specified the communication requirements in
term of latency and bandwidth mainly based on IEEE standards and related re-
search. The latency less than 10 ms and the peak rate larger than 22 Mbps are
required for a proposed hypothetical smart microgrid in this thesis. Secondly, we
estimated the single antenna LTE performance both theoretically and experimen-
tally. The theoretical analysis indicates that the latency is less than 9.5 ms, the peak
rate is more than 100 Mbps in downlink and 67 Mbps in uplink. The experimental
measurements show that the latency and peak rates of the LTE network provided
by TELE2 fulfil the requirements for the communication in the hypothetical smart
microgrid as summarized in Table 8.1, while the latency of the LTE network pro-
vided by TELIA is a little longer than the required. The latency can be improved
using an appropriate scheduler.
Last, a scheduler was designed to optimize the latency. The simulation results
show the latency is reduced to around 5 ms for smart objects communication via
LTE. In addition, the results show that LTE can handle more than 300 PMUs in a
single cell.
It must be noted that we assume all the physical resources are allocated to data
transmission without considering any resources needed for control signalling in this
thesis.

42
CHAPTER 8. CONCLUSION & FUTURE WORK 43

Table 8.1. Comparison between communication requirements and experiment re-

sults.

Experiment Comments
Type Requirements
TELE2 TELIA TELE2 TELIA
Latency 10 ms 9.2725 ms 13.2375 ms 83.27% ≈0
Peak rate 22 Mbps ≥ 30 Mbps ≥ 30

Here latency results from experiment is obtained by halving mean RTT value of 100 bytes from
Chapter 6. Based on the “best” fitting model, the probability that the latency of each LTE network
will be less than 10 ms is evaluated and filled in Comments column.

8.2 Future Work

Although it has been proven that 3GPP LTE is a promising solution to interconnect
devices in a smart grid, this smart microgrid only includes the components PMU
and AMI. In practice, there are many other potential components which need two-
way communicational infrastructures to help a power grid achieve the goals of a
smart grid. For instance, the transformers, breakers and other devices mounted on
electricity buses need to report their status and receive remote control commands.
More investigation needs to be done before PMU and AMI are deployed in a real
smart grid. In addition, the real time electricity price is still under development.
Furthermore, we only studied the performance of one single antenna LTE net-
work in this thesis. The performance of MIMO LTE needs to be analysed not
only for fulfilling the communication requirements but also for the scheduler de-
sign. Therefore a more complete simulation model for LTE is required. Based
on the model, a scheduler which can allocate physical resources in both time and
frequency domains can be designed.
Reference

[1] 3GPP, LTE Introduction http://www.3gpp.org/LTE

[2] WIKIPEDIA website, 3GPP Long Term Evolution , http://en.wikipedia.

org/wiki/3GPP_Long_Term_Evolution

[3] Takehiro Nakamura, Proposal for Candidate Radio Interface Technologies for
IMT-Advanced Based on LTE Release 10 and Beyond (LTE-Advanced), 2009-
10-15

[4] WIKIPEDIA website,LTE Advanced, http://en.wikipedia.org/wiki/LTE_

Advanced

[5] H.Farhangi, The Path of the Smart Grid, Power and Energy Magazine, IEEE,
2010, ISBN 1540-7977.

[6] A.G.Phadke, J.S.Thorp, A New Measurement Technique for Tracking Voltage

Phasors,Local System Frequency, and Rate of Change of Frequency, Power Ap-
paratus and System,May 1982, ISBN 0018-9510

[7] IEEE Std C37.118, IEEE Standard for Synchropahasors for Power System,22
March 2006

[8] IEC, IEC 61107: Data exchange for meter reading, tariff and load control-
Direct local data exchange, 1996

[9] IEC, IEC 62056-21: Electricity metering Data exchange for meter reading,
tariff and load control: Part 21 Direct local data exchange, 2002

[10] T.Khalifa, K.Naik and A.Nayak, A Survey of Communication Protocols for

Automatic Meter Reading Applications, Communication Survey & Tutorials,
IEEE, Second Quarter 2011, ISSN: 1553-877X .

[11] Carl H.Hauser, David E.Bakken, ..., Security, trust, and QoS in next-generation
control and communication for large power system, International Journal of
Critical Infrastructures, 2007.

[12] IEC, IEC 61850-5: Communication requirements for functions and device mod-
els, 2002

44
REFERENCE 45

[13] J.Y.Cai, Z.Huang, J.Hauer and K.Martin(2005), Current status and experience
of WAMS implementation in North America IEEE/PES Transmission and Dis-
tribution Conference and Exhibition: Asia and Pacific, 2005

[14] 3GPP LTE Encyclopedia, An Introduction to LTE, 2010-12-03

[15] R.Sacchi, P.Lorch, Understanding the use of OFDM in IEEE 802.16(WiMAX)

Agilent Measurement Journal, 2007

[16] M.Rumney, 3GPP LTE: Introducing Single-Carrier FDMA Agilent Measure-

ment Journal, 2008

[17] S.Sesia, I,Toufik, M.Baker, LTE The UMTS Long Term Evolution From Theory
to Practice WILEY, 2009

[18] A.Clark, C.J.Pavlovski, Wireless Networks for the Smart Energy Grid: Appli-
cation Aware Networks, Proceedings of the International Multi-Conference of
Engineers and Computer Scientists, 2010

[19] Motorola, Long Term Evolution(LTE): A Technical Overview, 2010-07-03

[20] Deepti Singhal, Mythili Kunapareddy, and Vijayalakshmi Chetlapalli, White

Paper LTE-Advanced: Latency Analysis for IMT-A Evaluation, Tech Mahindra
Limited, 2010

[21] 3GPP, 3GPP TS 36.211 V8.9.0, 2009-12

Appendix A

Ping Introduction

As mentioned in preview section, the experiment uses ping command to send and
receive data packets. The main advantages of ping are easy implement and under-
stand, relatively accurate latency calculation, Internet Protocol (IP) based applica-
tion, and arbitrary data packets length. Ping operates by sending and waiting for
an ICMP packet, then measures RTT and records the loss rate. Table A.1 shows
the general structure of an ICMP packet. Figure A.1 is one example packet without
additional data captured by Microsoft Network Monitor 3.4 via Ethernet.

Table A.1. ICMP Packet Structure

Bit 0-7 Bit 8-15 Bit 16-23 Bit 24-31

Version Type of service Length
Identification Flags and offset
IP Header
Time to Live
(20 bytes) Protocol Checksum
(TTL)
Source IP address
Destination IP address
Message Type Code Checksum
ICMP Payload
Quench
(8+ bytes)
Data (optional)

It is worthy to mention that the data can be arbitrary length as desired. However
must be less than the maximum transmission unit (MTU) of the network. Here is
an example of ping command which used to collect latency data in this paper.

ping 130.237.32.143 -l 32 -n 4 -f

It means to ping the target host 130.237.32.143 four times with additional 32 bytes

46
APPENDIX A. PING INTRODUCTION 47

Figure A.1. ICMP Packet Example Captured by Microsoft Network Monitor 3.4

data attached and structure the data frame without fragments. The result of ping
looks like following:
Pinging 130.237.32.143 with 32 bytes of data:
Reply from 130.237.32.143: bytes=32 time=19ms TTL=53
Reply from 130.237.32.143: bytes=32 time=18ms TTL=53
Reply from 130.237.32.143: bytes=32 time=17ms TTL=53
Reply from 130.237.32.143: bytes=32 time=18ms TTL=53

Ping statistics for 130.237.32.143:

Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 17ms, Maximum = 19ms, Average = 18ms

The output shows the result of 4 pings to th target,130.237.32.143, with the

results summarized at the end.