Evolution of Smart Grids

SPRINGER BRIEFS IN
ELEC TRIC AL AND COMPUTER ENGINEERING
Zubair Md. Fadlullah

Nei Kato
Evolution of
Smart Grids
123
SpringerBriefs in Electrical and Computer
Engineering
More information about this series at http://www.springer.com/series/10059

Zubair Md. Fadlullah • Nei Kato
Evolution of Smart Grids
123
Zubair Md. Fadlullah Nei Kato
Tohoku University School of Information Sciences
Sendai, Miyagi, Japan Tohoku University
Sendai, Miyagi, Japan
ISSN 2191-8112 ISSN 2191-8120 (electronic)

SpringerBriefs in Electrical and Computer Engineering
ISBN 978-3-319-25389-3 ISBN 978-3-319-25391-6 (eBook)
DOI 10.1007/978-3-319-25391-6
Library of Congress Control Number: 2015954084
Springer Cham Heidelberg New York Dordrecht London

© The Author(s) 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media (www.

springer.com)
Preface
The smart grid is still evolving and its potential landscape is enormous. As a
consequence, the level of complexity of the smart grid is unprecedented. This book
aims to illustrate the highly anticipated opportunities and challenges of the evolving
smart grid. It is essential to identify the scope and challenges of the smart grid in a
comprehensive manner so as to ensure efficient delivery of sustainable, economic,
and secure electricity supplies. The smart grid is the fusion of two core disciplines,
the power grid and the communication networks. Therefore, the motivation behind
this book is to bridge the gap between the research on power grid systems and
on the wireless communication networks through the smart grid concept. In this
vein, this book provides an overview of the smart grid and its key advances in
architecture, distribution management, demand-side response and load balancing,
smart automation, electric storage, power loss minimization, and security. This book
further shows that many of the smart grid challenges can be formulated as trade-off
problems of the utility operator and its customers. This book also demonstrates
how such trade-off problems could be solved using state-of-the-art optimization
techniques such as game theory. Numerical results are also included in the book
to show the effectiveness of the adopted approaches. We believe that the adopted
approaches and the related findings will reveal useful insights for the design of smart
grid and spur a new line of thinking for the performance evaluation of evolving smart
grid systems.
Acknowledgments
The authors would like to acknowledge the support of the Ministry of Education,
Culture, Sports, Science & Technology of Japan, and Tohoku University, Japan.
A very special thanks to Prof. Xuemin (Sherman) Shen, the SpringerBriefs Series
Editor on Wireless Communications. This book would not be possible without his
v
vi Preface
kind support during the process. Thanks also to the Springer Editors and Staff, all
of whom have been extremely helpful throughout the production of this book.
Sendai, Miyagi, Japan Zubair Md. Fadlullah

Nei Kato
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Evolution of Smart Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Challenges with Current Smart Grid Settings . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Aim of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Considered Smart Grid Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Smart Grid Model with Hierarchical Communication Networks . . . . . . 7
2.1.1 Communication Technology Used in HAN . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Communication Technology Used in BAN. . . . . . . . . . . . . . . . . . . . . 10
2.1.3 Communication Technology Used in NAN . . . . . . . . . . . . . . . . . . . . 10
2.2 Smart Grid Communication Packet Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Micro Grids Oriented Smart Grid Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Macro Station and Retailers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Macro Station and Micro Grid Model . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.3 Distributed Micro Grids with Renewable Sources . . . . . . . . . . . . . 13
2.4 Deployment Scenarios and Key Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Practical Implementation and Operating Challenges . . . . . . . . . . . . . . . . . . . 15
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Challenges in Smart Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1 Optimization Challenges from Different Perspectives . . . . . . . . . . . . . . . . . 17
3.1.1 Demand Response (DR): Shaping Electricity
Demand to Match Supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.1.1 Direct Load Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.1.2 Price-Based DR Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.2 Energy Conservation and Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.3 Power Loss Problem in Micro-Grids Based
Smart Grid System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Is Smart Grid Communication Security an Overkill? . . . . . . . . . . . . . . . . . . 20
vii
viii Contents
3.3 Concluding Remarks: Could an All-in-One Optimization

Technique Work in Smart Grid? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Demand Response Challenge in Smart Grid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1 Background into DSM and Load Balancing Problem . . . . . . . . . . . . . . . . . . 25
4.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1 Power System Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1.1 Modeling an Appropriate Energy Cost Function . . . . 27
4.2.1.2 Modeling Load Control on the Customer-Side . . . . . . 27
4.3 Adopted Strategies of Users and Utility Operator
and a Game-Theoretic Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3.1 Strategy of the Users. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3.1.1 Value of Energy Perceived by Users. . . . . . . . . . . . . . . . . . 29
4.3.1.2 Cost of Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3.2 Strategy of the Utility Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4 Playing the Demand Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.5 Comparative Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5 Game-Theoretic Coalition Formulation Strategy
for Reducing Power Loss in Micro Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2.1 Non-cooperative Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.2 Cooperative Coalition Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Coalition Formation Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4 Numerical Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6 On Optimally Reducing Power Loss in Micro-Grids
with Power Storage Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.3 Coalition Formulation Strategy for Micro Grids
with Power Storage Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3.1 Characterization of the Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7 A Distributed Paradigm for Power Loss Reduction in Micro-Grids . . . . 65
7.1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.2 System Model and Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Contents ix
7.3 The Distributed Solution Approach

for the Micro Grids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
8 Security Challenge in the Smart Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
8.1 Security Problem: A Bi-product of Legacy
Communication Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
8.2 Light-Weight Authentication Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
8.3 Security Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
9 Conclusion and Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
9.2 Future Directions and Caveats. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
List of Figures
Fig. 1.1 Evolution of the power grid into a smarter “two-way”

communication-capable one . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Fig. 2.1 Considered smart grid model with hierarchical
communication networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Fig. 2.2 SG communications packet structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Fig. 2.3 Micro-grids oriented architecture of the smart grid . . . . . . . . . . . . . . . . . . . 13
Fig. 4.1 Value of utility for different resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Fig. 4.2 The comparative increase of energy price with the total
energy consumption using the quadratic function and the
extended energy price function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Fig. 4.3 Comparison of the running time between the
conventional methods and the adopted approach . . . . . . . . . . . . . . . . . . . . . . 33
Fig. 4.4 Number of iterations until convergence in case of the
conventional methods and the adopted approach, for
varying numbers of users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Fig. 4.5 Average time needed for each iteration in case of the
conventional methods and the adopted approach . . . . . . . . . . . . . . . . . . . . . . 34
Fig. 4.6 Number of iterations required for the conventional 2
method and the adopted approach for a large number of
users (up to 500). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Fig. 4.7 PAR reduction in the conventional 2 method and the
adopted approach for varying numbers of users . . . . . . . . . . . . . . . . . . . . . . . 36
Fig. 5.1 Example of micro grids oriented system model . . . . . . . . . . . . . . . . . . . . . . . 40
Fig. 5.2 Comparative result of the conventional methods and the
adopted coalition-based power loss minimization approach . . . . . . . . . . 48
Fig. 5.3 The optimal number of MGs in the considered topology . . . . . . . . . . . . . 49
Fig. 6.1 Our considered micro grids system model and the
various power losses occurring in the system . . . . . . . . . . . . . . . . . . . . . . . . . . 54
xi
xii List of Figures
Fig. 6.2 Comparative average power loss per micro grid achieved
in the conventional method and the adopted approach . . . . . . . . . . . . . . . . 62
Fig. 6.3 Comparison of the power load on the macro station in
the conventional method and the adopted approach . . . . . . . . . . . . . . . . . . . 63
Fig. 7.1 The considered system model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Fig. 7.2 A simple example showing how the distributed
algorithm 1 enables power exchange between micro
grids and macro station with minimized power loss. . . . . . . . . . . . . . . . . . . 71
Fig. 7.3 Comparison of the average power loss in the
conventional method and the adopted distributed
algorithmic approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Fig. 7.4 The percentage of cost saving using the adopted
distributed approach compared with the conventional method . . . . . . . 73
Fig. 7.5 Comparison of the power load on the macro station in
the conventional case and the adopted distributed approach . . . . . . . . . . 73
Fig. 7.6 The comparison in centralized and distributed
approaches. (a) Average power loss. (b) The bandwidth cost . . . . . . . . 74
Fig. 8.1 Considered secure and reliable framework for smart grid
communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Fig. 8.2 Adopted light-weight message authentication scheme . . . . . . . . . . . . . . . . 82
Fig. 8.3 Average communication overhead per BAN GW for
varying number of smart meters (HAN GWs) . . . . . . . . . . . . . . . . . . . . . . . . . 86
Fig. 8.4 Average delay at the BAN GW for varying number of
smart meters (i.e., HAN GWs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Fig. 8.5 Memory usage of the proposed and conventional
ECDSA authentication algorithms for different message
volumes received by BAN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Fig. 8.6 Number of HANs supported by the proposed and
conventional authentication schemes for smart grid
communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Chapter 1
Introduction
1.1 Evolution of Smart Grids
This book discusses the smart grid [1–8]. The notion smart grid, linguistically,
implies an electric power delivery grid having some level of “smartness.” Why do
we consider these two elements then? First, consider the power system delivery grid.
It has frequently been referred to as the greatest and most complex machine ever
built in human civilization. In fact, the power grid is almost entirely a mechanical
system, with only modest use of sensors, minimal electronic communication, and
almost a total lack of electronic control. In the last two to three decades, almost all
other industries in the western world have significantly become modernized with
the utilization of sensors, communications, and computational ability leading to far
greater improvement in efficiency, productivity, service quality, and environmental
performance. The power grid system has not evolved, however. There was simply
not much of a need for it to change because of the crude way this massive system
works with a plethora of components including wires, cables, towers, transformers,
circuit breakers, and many more you can possibly think of! When it has needed
expansion, more of these components have just been bolted together without paying
much attention to what may happen in future if the system needs sophistication or
refurbishment. As the power grid started to become bigger in size, its complexity in
terms of monitoring and repairing also increased significantly and manual control of
the system gradually went out of scope. The industry initiated the use of automated
monitoring and power system control with the help of computers began way back
in the sixties. This, coupled with a modest use of sensors, has increased over
time. Still, however, it is far from being ideal as the power companies lack real-
time information on the customer demands and, therefore, are unable to smartly
decide power generation and distribution scheduling tasks. The conventional grids,
therefore, need to become better and smarter; but investors do not see much profit in
© The Author(s) 2015 1

Z.Md. Fadlullah, N. Kato, Evolution of Smart Grids, SpringerBriefs
in Electrical and Computer Engineering, DOI 10.1007/978-3-319-25391-6_1
2 1 Introduction
Fig. 1.1 Evolution of the power grid into a smarter “two-way” communication-capable one
retrofitting the old grid with older metering equipment and limited communication
capability. This is why there is a lot of stress on the conventional power grid systems
to evolve into next generation smart power grid.
So, what contrasting features should the smart grid have? Unlike the traditional
grid, the smart grid aims to exploit sensors, communications, computational ability
and control in some form to enhance the overall functionality of the electric power
delivery system. You may, therefore, define the smart grid as a more naive power
grid system evolving into a smarter one through sensing, communicating, applying
intelligence, and executing control/feedback for adjusting dynamic changes in
primary and renewable electricity supply and demand. The growing trend to
converge toward such a two-way communication-capable smart grid is summarized
in Fig. 1.1. For a power system, this can effectively allow optimization towards
objectives that assure reliability, minimize cost, minimize energy consumption,
manage resources, and alleviate environmental impact [9].
The smart grid concept is synonymous to intelligent grid or future grid, and is
aimed at providing the end-users1 with more stable and reliable power. The common
aspect of the different smart grid proposals consists in the two-way communication
framework between the power source and the consumers, and intelligent sensing
entities and control systems lie along the path of the power source and the end-
users. The sensors have the ability to detect malfunctions or deviations from
normal operational trends which usually require appropriate responses from the
smart grid control center. Furthermore, the responses from the control center need
to be converted into appropriate control messages and transmitted to different
segments of the smart grid. Therefore, the smart grid communication framework
and functionality should be characterized, particularly in terms of its capability of
1
The end-users of the smart grid may also be referred to as simply the users, or as the customers
or consumers in the book.
1.2 Challenges with Current Smart Grid Settings 3
provisioning consumer participation, energy demand and cost estimation, demand

and load balancing, power loss minimization across micro-grids, and resiliency
against malicious threats such as physical and cyber attacks. In a smart grid,
consumers are usually equipped with smart meters, which are able to identify power
consumption in a much more elaborate way in contrast with the conventional energy
meters. While the information gathered by a smart meter can be obtained by the
power company (i.e., the utility provider) for monitoring and billing purpose, the
consumer can also access his or her smart meter to verify the current use and
accordingly adjust the power usage during peak and off-peak hours. In order to
do so, the users may have to periodically send request messages to the utility
provider. By this way, they can actively and dynamically get involved with the power
dissemination process in the smart grid. To implement the two-way communication
between the consumers and the utility provider, the smart meters are often integrated
in the smart grid by using the so-called AMI, which stands for Advanced Metering
Infrastructure. The AMI-based smart meters can be exploited to collect customer
demand profiles and forecast their future demand trends so as to project the
possible demand ahead of time. Furthermore, it should be stressed that a strong
authentication mechanism is a requisite for end-users and smart meters deployed in
the smart grid. Such an authentication mechanism needs to be fast yet secure so that
the smart grid entities may be able to quickly verify each other’s identities and to
allow the users to access smart grid services and resources in a secure manner.
With these multi-directional challenges, even though the smart grid has been
gaining momentum in recent years, researchers are still trying to imagine and
predict how an actual full-scale smart grid communication framework and structure
should be. As a matter of fact, the smart grid vision can only be realized through
the fusion of various disciplines where professionals from power engineering
background will meet the communication engineers. To determine which tech-
nologies from these various disciplines should be exploited to get the best return
has been a daunting task. For example, do we need to enhance the currently
available communication technologies or grasp the upcoming 5G technologies to
suit the smart grid requirements? Can and how the communication technologies
help the smart grid to reduce power loss? How can the smart grid be secure against
well-known easily doable spoofing and man-in-the-middle attacks? Even though
this book cannot give answers to all the challenges and issues related to the smart
grid, we would like to cover some of the most basic yet challenging problems and
address those specific problems to gain some steps towards enabling a reliable and
secure smart grid.
1.2 Challenges with Current Smart Grid Settings
Even though there appeared quite a few researches in the smart grid literature, the
available approaches to address the smart grid communications are still somewhat
limited. This means that a lot of work needs to be re-addressed to provide a
4 1 Introduction
reasonable smart grid models. In other words, the first and foremost agenda is to
agree (or not to agree) on what should be the most realistic power distribution and
communication model for a practical and scalable smart grid. Why is this important?
Because, a smart grid which will cost billions of dollars needs to be designed
in a very careful manner from the beginning. The smart grid will have numerous
electrical appliances connected with each another in a uber-complex fashion so that
they can report back on elements like power consumption and other monitoring
signals. So, fast communication is crucial yet challenging for practical deployment
of the smart grid. In order to realize effective smart grid communication, existing
networking technologies need to be exploited (and even enhanced) to cope with the
multiple services and quality requirements of the residential appliances. The need to
distinguish between the high- and low-priority traffic will be equally important as to
be able to dynamically adapt the network to changing capacity demands in real-time.
Therefore, it is essential that we consider appropriate technologies to implement the
communication networks of smart grid, which may allow the flexible use of existing
capacities without impacting the service quality of the smart grid.
The renewable electricity generation and delivery is another important aspect of
the smart grid. Micro grids (smaller autonomous power generation and dissemina-
tion systems) which could be connected and isolated from the main power grid are
becoming more common. The micro grids are perceived as the means to reduce
unnecessary/fossil-based power generation and distribution, and reduce the total
power loss across the main grid. The echo-friendly or “green” micro grids are,
therefore, gaining lots of attention from the mainstream smart grid community.
But when you deal with a large number of micro grids, substantial reduction in
power loss may be impossible without optimizing the scheduling of local energy
generation and distribution of the micro grids. In addition, the micro grids could
have local storage facilities to locally store energy for selling to their own customers
or even other micro grids at a later time. Should we consider the power generation
and storage losses also in addition to the distribution loss? In addition, how to deal
with such problems? In other words, who will decide the schedule of generation and
storage? A centralized decision maker certainly sounds interesting, but may not be
as impressive in real-time compared to a distributed decision making solution.
Security is another key challenge of the smart grid. The Internet Protocol or
IP-based communication adopted in the smart grid communication means the
malicious threats plaguing the Internet could be very much possible in the smart
grid. The communication links between the smart meters and the control centers,
therefore, need to be secure against various security threats. Given the volume of
messages, delay constraints, and limited hardware capability of the smart meters,
a light-weight message authentication scheme tailored for the smart grid could be
greatly useful.
1.3 Aim of the Book 5
1.3 Aim of the Book
The book mainly focuses on the anticipated opportunities and challenges of the
smart grid. In Chap. 2, we will provide our definition of a smart grid system model
from two perspectives. The first perspective considers an overlay of the power
distribution network and the communication network-based hierarchical smart grid
framework. The second perspective describes a micro grids-based architecture with
a focus on micro grids equipped with renewable energy sources and storage devices.
We will investigate and compare the performance of these two models and discuss
whether a fusion of the two could be possible.
In Chap. 3, the challenges of the smart grid will be more deeply visited. Most
of these challenges can be considered as optimization problems. So, in the chapter,
we will describe how existing works formulated these optimization problems from
different perspectives.
From Chap. 4 onward, the book will transform from a broader narrative to a
more technical discussion. In Chap. 4, we will describe a demand-side management
based on game-theory and its advantages in improving the peak-to-power ratio.
In Chaps. 5–7, we will discuss power loss reduction strategies in micro grids.
These contents of these three chapters are incremental. Chapter 5 will describe
the power loss model of a basic (theoretical) micro grids based system and show
how a centralized game-theoretic approach can effectively improve the power loss.
In Chap. 6, the micro grid model is extended to a more practical one comprising
power storage devices. The chapter shows with this slight extension, the problem
becomes even deeper which needs even more sophisticated algorithm to have a
central decision making. In Chap. 7, we move away from the central decision
making strategy to a distributed paradigm for power loss reduction in micro grids so
as to increase resiliency of the individual micro grids.
In Chap. 8, we shift to a different yet relevant topic where we discuss the need
of a secure framework for smart grid communications. The chapter will introduce
a secure and reliable framework for smart grids, a key part of which is a light-
weight and secure message authentication scheme tailored for the requirements of
smart grid’s AMI. The chapter adopts a Diffie–Hellman key establishment protocol
and hash-based message authentication code, which lets various smart meters at
different segments of the smart grid to establish mutual entity authentication with
low delay and communication overhead.
For interested readers, in the technical chapters (i.e., Chaps. 4–8), technical
analysis is shown to indicate how the adopted approaches can satisfy the desirable
requirements of the respective problems. Some performance evaluation is also
included in the chapters to demonstrate the effectiveness of the adopted approaches.
Finally, the book is concluded in Chap. 9 by summarizing the various challenges
which are critical for a smart grid technology to evolve and mature. Future directions
and caveats are also described in the section.
6 1 Introduction
References
1. M. M. Fouda, Z. M. Fadlullah, N. Kato, R. Lu, and X. Shen, “A lightweight message

authentication scheme for smart grid communications,” IEEE Transactions on Smart Grid,
vol. 2, no. 4, pp. 675–685, Dec. 2011.
2. H. Liang, A. Tamang, W. Zhuang, and X. Shen, “Stochastic information management in smart
grid,” IEEE Communications Surveys Tutorials, vol. 16, no. 3, pp. 1746–1770, Mar. 2014.
3. M. M. Fouda, Z. M. Fadlullah, and N. Kato, “Assessing attack threat against zigbee-based home
area network for smart grid communications,” in 2010 International Conference on Computer
Engineering and Systems (ICCES), Cairo, Egypt, Nov.-Dec. 2010, pp. 245–250.
4. H. Liang, B. J. Choi, W. Zhuang, and X. Shen, “Stability enhancement of decentralized inverter
control through wireless communications in microgrids,” IEEE Transactions on Smart Grid,
vol. 4, no. 1, pp. 321–331, Mar. 2013.
authentication scheme for smart grid communications,” Smart Grid, IEEE Transactions on,
vol. 2, no. 4, pp. 675–685, Dec. 2011.
6. Z. M. Fadlullah, M. M. Fouda, N. Kato, A. Takeuchi, N. Iwasaki, and Y. Nozaki, “Toward intel-
ligent machine-to-machine communications in smart grid,” IEEE Communications Magazine,
vol. 49, no. 4, pp. 60–65, Apr. 2011.
7. A. Abdallah and X. Shen, “Lightweight security and privacy preserving scheme for smart grid
customer-side networks,” IEEE Transactions on Smart Grid, to appear.
8. H. Li, X. Lin, H. Yang, X. Liang, R. Lu, and X. Shen, “EPPDR: an efficient privacy-preserving
demand response scheme with adaptive key evolution in smart grid,” IEEE Transactions on
Parallel and Distributed Systems, vol. 25, no. 8, pp. 2053–2064, Aug. 2014.
9. C. W. Gellings, The smart grid: Enabling energy efficiency and demand response. Fairmont
Press, 2009.
Chapter 2
Considered Smart Grid Model
Before discussing the challenges of the smart grid (and even addressing them), to
have an understanding of how the smart grid is modeled and how it operates would
be useful [1–4]. That is basically the aim of this chapter. To make it interesting to
the readers, two slightly different approaches to the smart grid model are covered
in the chapter. The first one is more like an overlay between the power grid and the
communication system [1]. Hierarchical communication networks are used in this
first approach to allow the bi-directional flow of information between the utility
provider and the customers. Smart meters and AMI are the core technologies
adopted in this case. In the second approach, a micro-grids based smart grid
architecture [5] is considered where the communication networks are somewhat
inline which assume the use of either dedicated or already existing communication
technologies or even the Internet.
2.1 Smart Grid Model with Hierarchical

Communication Networks
Figure 2.1 demonstrates our considered smart grid communication framework

which has been adopted in [1, 2, 6]. The smart grid power transmission and dis-
tribution system is considered in the figure to be separated from the communication
system. First, let us discuss the power distribution network briefly. The power
generated at the power plant(s) is supplied to the customers via the transmission
and distribution substations. The transmission substation is located usually near
the power plant. On the other hand, the distribution substations are located in
different areas and neighborhoods. The transmission substation delivers electric
power over high voltage (>230 kV) transmission lines to the distribution substations

8 2 Considered Smart Grid Model
Fig. 2.1 Considered smart grid model with hierarchical communication networks
that transform the power into medium voltage level and then distribute it to the
building-feeders. The building-feeders convert the medium level voltage into a much
lower one, which becomes usable by user appliances.
From the communication perspective, the smart grid architecture is divided into
a number of hierarchical networks. The transmission substation near the power
plant, and the control centers of the distribution substations are connected with each
another. The mesh network that connects them is typically implemented using high
speed optical fiber to support high volume of smart grid data traffic with the least
possible communication delay.
On the other side of the control centers, the communication framework is
divided into a number of hierarchical networks such as neighborhood area networks
(NANs), building area networks (BANs), and home area networks (HANs). For
simple understanding, consider that each distribution substation covers a single
neighborhood where a NAN is established. Each NAN has a number of BANs.
Each BAN has multiple HANs (basically several apartment-based networks). The
smart meters are the key devices in this communication architecture that act as the
dual-gateway (GW) for energy use monitoring and communication. Depending on
where they are placed, the smart meters are referred to as the NAN GWs, BAN
GWs, and HAN GWs, and they are not necessarily of the same specs and hardware.
For example, a BAN GW, installed alongside a building feeder, can have more
2.1 Smart Grid Model with Hierarchical Communication Networks 9
sophisticated hardware and specs, compared to a HAN GW. On the other hand,
through a cheaper and lower-spec HAN GW, the customer can check energy use and
express his/her energy demand to the grid operator at the control center. A smart
meter having the MSP430F471xx micro-controller can be used as a typical HAN
GW [7] with a random access memory of just 8 kilo bytes and flash memory of
120 kilo bytes. The HAN GW has a 16 MHz CPU, 3/6/7 16-bit analog to digital
converters (ADCs) and programmable gain amplifiers (PGAs), 160-segment liquid
crystal display (LCD), real-time clock (RTC), and 32 32 hardware multiplier.
On the other hand, much more powerful (up to ten times more capability) smart
meters are needed to act as BAN GW with specs like 160 MHz CPU, 128 kilo
bytes random access memory, and 1 mega byte flash memory. As for the NAN GW
configuration, a dedicated computer with Intel Core haswell-i7 CPU, 16 giga bytes
random access memory, and storage scaling up to few terabytes could be considered.
The difference in these smart metering specifications is because the end-users are
likely to deal with significantly lower traffic and usually prefer affordable (i.e., low
cost) smart meters whereas the grid operator at the control center can easily afford
one or more high-spec computers to cope with substantially large volume of data
generated from a large number of customers in the neighborhood.
Based on the existing standards of smart grid, IP-based communication is
preferred to allow seamless inter-connections with HANs, BANs, NANs, control
centers, and the transmission substation. What kind of IP-communication technol-
ogy should be used at each hierarchical network of the smart grid is described below.
2.1.1 Communication Technology Used in HAN
In the considered smart grid, HAN is basically the “last mile.” The HAN allows
the end-users to easily manage their energy use and demand. Let us refer to the
HAN shown on the bottom right of Fig. 2.1. The home network connects the smart
appliances (e.g., television, washing machine, oven, etc. which have their own IP
address) to the HAN GW, which can consistently or periodically communicate with
BAN1. Smart Energy Profile (SEP) Version 1.5 over IEEE 802.15.4 ZigBee radio
communications could be a candidate HAN communication technology because of
its low power requirements (1–100 mW) compared to other home-equipment radio
technologies like the IEEE 802.11 (WiFi) and Bluetooth. Also, ZigBee provides
a reasonable communication range of 10–100 m and incurs moderate cost, and
therefore, it can be considered as the HAN-enabling communication technology.
However, placement of the HAN GW inside the apartment to establish stable
connectivity with all the equipment is still critical.
2.1.2 Communication Technology Used in BAN
Just like a building has several apartments, a BAN has a number of HANs. The
BAN GW at the building feeder can be used to monitor the energy consumption
and project the demand of the building’s residents. Although WiFi could appear
to be a common choice due to its common availability in recent times, a BAN
covering many households in a large apartment building means potential lack of
WiFi coverage. Deploying multiple WiFi access points could be a solution which
may still be subject to serious interference from other access points commonly used
for Internet access and so forth. Therefore, 4G technologies like LTE or WiMax
could be harnessed to cover more areas to facilitate communication between a BAN
and its covered HANs.
2.1.3 Communication Technology Used in NAN
A NAN connects the smart grid’s backhaul with the radio access networks (i.e.,
BANs and HANs). The NAN signifies a specific area like a ward of a city. Through
the NAN GW, the grid operator monitors how much power is being distributed
to a particular neighborhood by the respective distribution substation. Broadband
wireless technologies such as 4G LTE and WiMax can be exploited for connectivity
of the NAN GW and its covered BAN GWs. It is worth noting, however, that the
existing studies indicated smart grid communication network to be separate from
the existing ones for providing Internet services so as to prevent network congestion
and malicious threats.
2.2 Smart Grid Communication Packet Structure
The smart grid data are exchanged between the customers’ smart meters and the
control center through the HAN, BAN, and NAN GWs using IP-based communi-
cation. So, it is important to consider the smart grid data communication packet
structure from a high level as shown in Fig. 2.2. The illustration in the figure
is based on commercial smart meter specifications in [8]. In addition to the raw
message, every packet has the message header, TCP/IP header, and security header.
The message header includes meter ID MAC address, equipment status, and the
Type of Message (ToM). There are nine possible ToMs, which the HAN GW is able
to send to the BAN GW. The function and size of each ToM are also indicated in
the figure.
A nuclear power plant could also be an example of the macro station.
2.3 Micro Grids Oriented Smart Grid Model 11
Fig. 2.2 SG communications packet structure
2.3 Micro Grids Oriented Smart Grid Model
This section of the chapter will describe a different perspective of the smart grid
architecture based on micro grids and its collaboration with the macro station. Let us
first discuss the macro station and its relationship with the retailers before grasping
the concept of micro grids [5, 9–11].
2.3.1 Macro Station and Retailers
The smart grid is a cyber-physical system as it combines the power grid and
communication system (you should understand this by now if you have carefully
followed Sect. 2.1). The detailed information of the smart grid communication
architecture was given in Sect. 2.1, and here, we slightly tilt our perspective to the
power side of the smart grid. The wholesale power market is composed of the
main power plant (i.e., geo-thermal/hydro-electric/nuclear power plants) and the

renewable power generation sources (e.g., solar panels, wind turbines, hybrid car
batteries, etc.), which produce the primary (base) and secondary power, respectively.
The wholesaler does not directly supply the power to the customers. Instead, there
are power “retailers” that usually operate in specific areas to supply power to the
customers living in those areas. Based upon the forecast of customer demands for
the next day, a retailer purchases power from one (or multiple) wholesaler(s), and
sells this power to its own customers. Every retailer has a control center, which you
could possibly relate with Sect. 2.1. The control center carries out authentication,
billing, data aggregation, and direct load control. The smart meter of the customer
and the control center can communicate over wireless technology (Sects. 2.1.1–
2.1.3). Assume that the total power available at a specific retailer is Pmax which
is the sum of the base and secondary power purchased from the wholesale market.
Assume Pdemand to be the demand from all its customers. If Pmax satisfies the demand
requests coming from all its customers, then the control center does not need to
take any action. However, during low-demand usage periods, the retailer could
experience Pmax Pdemand which means that the retailer has purchased excessive
amount of power, which will be wasted leading to financial loss to the retailer.
But if there is a deficit (Pmax < Pdemand ), the retailer has no other option than
instantly purchasing additional power from the wholesale market at a substantially
higher cost. Furthermore, during peak hours, when multiple retailers simultaneously
want to buy additional power, the primary source may have already exhausted its
capacity. Then, the retailers would be forced to buy the extra power from, perhaps
other retailers, which may be even more expensive. Intuitively, in such a scenario, a
retailer’s best option would be to seek secondary sources in its locality and purchase
power from them. These localized secondary sources, which some refer to as the
micro grids, are recently becoming popular.
2.3.2 Macro Station and Micro Grid Model
In this model, the macro station (some refer to it as the MS) represents the whole-
saler of the energy market which usually generates electricity from fossil-based,
non-renewable resources like coal, petroleum, etc. Such a macro station deals with a
high maintenance cost due to somewhat inefficient fuel to electricity generation
and high carbon footprint. Yes, you could possibly imagine a nuclear power plant
as a macro station; but after the 2011-tragedy at Fukushima where the Tokyo
Electric Power Company (TEPCO) had its nuclear power-driven macro station
for the greater north-east Japan, interest toward spatially distributed, smaller yet
independent micro-grids capable of generating and serving their own power from
renewable sources has become paramount. In addition to dealing with emergency
demand in post-disaster scenarios, the micro grids are getting a great deal of interest
from the US military. Compared with the conventional power grids, the smart micro
grids are more flexible since they can quickly adjust their power production in
2.3 Micro Grids Oriented Smart Grid Model 13
Fig. 2.3 Micro-grids oriented architecture of the smart grid
accordance with user-demand. When a micro-grid is not self-sufficient, it can buy

extra power from the wholesaler or even the neighboring micro grids for satisfying
the energy demands of its customers. Thus, the micro grids can act as a viable
complement to the conventional macro station. The micro-grids based smart grid
framework is broadly shown in Fig. 2.3. Readers can check an earlier work of
micro grids based approach for the power distribution system with optimized self-
adequacy in [12].
2.3.3 Distributed Micro Grids with Renewable Sources
The end-users of a micro grid may be typical consumers like residential users,
companies, schools, hospitals, government facilities, and so forth. The end-users
can get power from their subscribed micro grid(s). Because most micro grids these
days have integrated renewable resources like wind, solar, water turbine, etc., in
addition to the purchased power from the macro station. Micro grids exploiting
renewable sources may not always be able to have a stable production and supply of
electricity. This happens because of the complex variation of power demand of the
users depending on the different times of a day [13, 14], the power demand of the
users varies depending on the different times of a day. Particularly during the peak
hours when the electricity consumption of the users is the highest, the micro grids
have to ensure a stable power supply even though instantaneous production is not
possible. So, how can a micro grid cope with the total user-demand approaching or
even exceeding its supply power? The micro grid has two options. First, the micro-
grid can buy additional power from the macro station and/or from other micro grids.
Second, the micro grid can have dedicated power storage devices like batteries, plug-
in hybrid electric vehicles (PHEVs), etc., which may be charged during the off-peak
hours and discharged during the peak hours to meet the user-demand. These two
techniques are not necessarily mutually exclusive.
2.4 Deployment Scenarios and Key Opportunities
The review of available micro grid architectures showed that most of the existing
testbeds are AC (alternating current) micro grids [15] because they are easy to
integrate with the main power grid and most loads (user appliances) which run on
AC. The AC micro grids, however, struggle with maintaining a stable power quality.
This is one of the critical shortcomings of the AC micro grids. In contrast, the DC
(direct current) systems in general are not prone to power quality problems. But
deploying the DC micro grid is not practiced much due to limited use of DC loads by
the customers. HFAC (high frequency AC) micro grids are now gaining popularity
that can help integrating renewable energy sources with the micro grid while
maintaining a reasonable power quality. A HFAC micro grid system has its own
disadvantage, however. It usually needs more control devices, and suffers from a
large voltage drop and long distance power loss. The conventional power generation
source of the micro grids is typically diesel; but renewable sources like solar power
volt, wind, and micro-hydro systems are becoming popular deployment choice for
the micro grids. Because the renewable sources are highly dependent on the ambient
environment, the micro grid usually is deployed with power storage devices to store
power for later use. Most existing test-beds have battery storage while some have
capacitor banks and flywheels as storage devices. Several storage units are deployed
in some micro grids while others do not have any storage unit at all. If the micro
grid system does not have any storage device and only have renewable energy
source, then the main grid connectivity is a very important option for that micro
grid system. There is definitely room for further study on the deployment scenarios
of the micro grids.
To demonstrate that micro grids are not merely research projects, we include
an example of a major aversion of power outage by exploiting a micro grid
using renewable sources in late May 2015 [16]. The Borrego Springs micro grid
supplied energy to 2800 customers for 9 h until the utility operator (San Diego
Gas & Electric) repaired the damage inflicted by a lightning strike. The micro
grid automatically switched between its different power generation sources like
onsite energy, energy storage, and a 26 MW solar generation source to steadily
manage power supply to the customers of the entire community. This shows the
immense potential of micro grids in days to come.
References 15
2.5 Practical Implementation and Operating Challenges
The DC micro grids are not that popular in European countries even though it does
not exhibit much power quality issues. With the advancement of technology, perhaps
more DC-compatible loads and appliances can be designed and implemented to
encourage using the DC micro grids. In case of the HFAC micro grids, the main
barrier to practical implementation is the significantly expensive solar cells and
storage devices. The operating challenge is also to be taken into account since the
clean power generation from renewable sources depend heavily on the environment.
Currently, the initial implementation cost along with operation and maintenance cost
are on the higher side. So, more penetration of renewable sources is expected in the
micro grid systems and further technological improvement is needed to make them
an economically viable option.
At the same time, we have to keep an open eye for the telecommunications needs
for distributed generation in the overall smart grid. The best policy in distribution
planning is to anticipate communications requirements and plan in a comprehensive
way [17]. With the advent of 5G communication technologies by 2020, many
operators would want to take advantage of those and meanwhile just use what is
available to them—LTE, LTE-A, or WiMax broadband technologies in the radio
access part (HAN–BAN and BAN–NAN communication). Therefore, there needs
to be a very comprehensive strategy so as to make a seamless transition to the
newer technologies. Otherwise, there could be adverse effects such as installation
of incompatible or redundant systems.
References

vol. 2, no. 4, pp. 675–685, Dec. 2011.
2. M. M. Fouda, Z. M. Fadlullah, N. Kato, R. Lu, , and X. Shen, “Towards a light-weight message
authentication mechanism tailored for smart grid communications,” in IEEE International
Workshop on Security in Computers, Networking and Communications (SCNC’11), Shanghai,
China, Apr. 2011.
vol. 4, no. 1, pp. 321–331, Mar. 2013.
5. C. Wei, Z. M. Fadlullah, N. Kato, and A. Takeuchi, “GT-CFS: a game theoretic coalition
formulation strategy for reducing power loss in micro grids,” IEEE Transactions on Parallel
and Distributed Systems, vol. 25, no. 9, pp. 2307–2317, Sep. 2014.
vol. 49, no. 4, pp. 60–65, Apr. 2011.
7. Texas instruments - msp430 for utility metering applications. Available online, accessed Aug
2015. “http://focus.ti.com/mcu/docs/mcuorphan.tsp?contentId=31498”.
8. “High-level smart meter data traffic analysis,” Engage Consulting Ltd. for the Energy Networks
Association (ENA), UK, Tech. Rep., May 2010.
9. P. Aristidou, A. Dimeas, and N. Hatziargyriou, “ Microgrid modelling and analysis using game
theory methods,” in Energy-Efficient Computing and Networking, ser. Lecture Notes of the
Institute for Computer Sciences, Social Informatics and Telecommunications Engineering,
N. Hatziargyriou, A. Dimeas, T. Tomtsi, and A. Weidlich, Eds. Springer Berlin Heidel-
berg, 2011, vol. 54, pp. 12–19. [Online]. Available: http://dx.doi.org/10.1007/978-3-642-
19322-4_2
10. C. Wei, Z. M. Fadlullah, N. Kato, and I. Stojmenovic, “On optimally reducing power loss in
micro-grids with power storage devices,” IEEE Journal on Selected Areas in Communications,
vol. 32, no. 7, pp. 1361–1370, Jul. 2014.
11. ——, “A novel distributed algorithm for power loss minimizing in smart grid,” in 2014 IEEE
International Conference on Smart Grid Communications (SmartGridComm), Venice, Italy,
Nov. 2014, pp. 290–295.
12. S. Arefifar, Y. Mohamed, and T. El-Fouly, “Supply-adequacy-based optimal construction of
microgrids in smart distribution systems,” IEEE Transactions on Smart Grid, vol. 3, no. 3,
pp. 1491–1502, Sep. 2012.
13. X. Liang, X. Li, R. Lu, X. Lin, and X. Shen, “UDP: usage-based dynamic pricing with privacy
preservation for smart grid,” IEEE Transactions on Smart Grid, vol. 4, no. 1, pp. 141–150,
Mar. 2013.
14. S. Chen, N. Shroff, and P. Sinha, “Heterogeneous delay tolerant task scheduling and energy
management in the smart grid with renewable energy,” IEEE Journal on Selected Areas in
Communications, vol. 31, no. 7, pp. 1258–1267, Jul. 2013.
15. L. Mariam, M. Basu, and M. F. Conlon, “A review of existing microgrid architectures,” Journal
of Engineering, vol. 2013, no. Article ID 937614, pp. 1–8, 2013.
16. E. Wood, “Borrego Springs Microgrid Averts Long Outage During Lightning Strike Repair,”
(accessed Aug. 2015). [Online]. Available: http://microgridknowledge.com/borrego-springs-
microgrid-averts-long-outage-during-lightning-strike-repair/
17. J. Brandt, “Is telecom a natural transition for utilities?” accessed Aug. 2015.
[Online]. Available: http://www.smartgridnews.com/tech/story/telecom-natural-transition-
utilities/2015-08-11
Chapter 3
Challenges in Smart Grid
Several smart grid pilot projects have been developed in recent year. The rapidly
evolving scenario of smart grids with distributed generation and storage means
dealing with multi-directional challenges involving higher communication capabil-
ities, faster reporting rates, better security, lower power losses, and so forth [1].
Understanding these challenges and then adequately addressing them are a prereq-
uisite for the complex control and management applications of the smart grid. In this
chapter, we describe some of the critical optimization challenges of the smart grid
from different perspectives.
3.1 Optimization Challenges from Different Perspectives
3.1.1 Demand Response (DR): Shaping Electricity

Demand to Match Supply
The smart grid operators need to guarantee that there is sufficient power generation
to service the load in terms of both wattage and volt-amperes reactive. The challenge
can be simply summarized as the classical supply and demand problem. How can
the smart grid keep supply and demand in balance? The traditional way to achieve
this is to continually adjust central generation through demand-side management
(DSM) [2–5], which includes energy efficiency and demand response (DR). Instead
of simply adding more generation to the system, utility operators pay the end-
users to reduce energy consumption since this is easier and cheaper than traditional
generation. To accomplish this, an operator can offer DR programs to encourage
its customers to make short-term reductions in energy demand as a response to a
price signal from the operator to switches on customers’ loads (like air-conditioners,
washing machines, water heaters, lighting arrays, etc.). Usually the DR actions
would be in the range of a few hours.

18 3 Challenges in Smart Grid
3.1.1.1 Direct Load Control
The direct load control (DLC) is an example of DR programs which have helped
throttle back peak demand. DLC examples include shutting down loads, cutting
off industrial process, or cycling it off for a specific time. By this simple yet
effective way, the operator may survive a few critical peak hours instead of having
to provide additional generation. Traditional DLC methods depend on legacy
communications, however, that lack accurate forecast and customer participation
which are necessary to optimize the DR programs. Also, some customers have
privacy concerns regarding DLC simply because they often do not want to give such
control to the utility operator (or anyone else for that matter). To overcome this
privacy issue, additional policies need to be integrated which could differentiate
various types of appliances with different grades of energy usage preferences.
A transparent policy can offer the users a clear choice between utilizing or not
utilizing DLC. In other words, if a user wants to cut energy bill, he/she can give
the utility operator adequate permission to apply DLC.
3.1.1.2 Price-Based DR Programs
Agarwal and Cui [6] argue that charging the customers a flat rate for energy use leads
to “allocative inefficiencies.” To overcome this, smart grid operators can leverage
pricing incentives (such as peak time rebates, time-of-use rates and price response)
to encourage customers to shift usage to off-peak hours. These price-based DR
programs give the customers a clear and up-to-date snapshot of their energy
usage and cost, and deliver more control over their energy bill. Theoretically such
programs sound great because they assume that all the customers will cooperate.
But in reality whether the users will accept the incentive or not is also dictated by
daily habits, psychology, awareness, and other social factors.
3.1.2 Energy Conservation and Efficiency
Energy conservation (load reduction) and efficiency are other DSM techniques
which the smart grid operators can exploit. Energy conservation programs are used
to encourage the users to sacrifice some energy use. In other words, they reduce
the overall demand for electricity by reducing the amount of utility the customer
receives. Example of this includes turning up the thermostat a few degrees in
summer in order to reduce air-conditioning. On the other hand, the energy efficiency
programs aim to reduce the overall demand for electricity while maintaining the
amount of utility the customer receives. For instance, when a user replaces him/her
old air-conditioner with a more energy-efficient model.
How energy demand and supply estimation can be more effectively balanced
has, thus, become a topic of interest. An autonomous and distributed demand-side
energy management system using game theory is an early work [7] in this area.
3.1 Optimization Challenges from Different Perspectives 19
This work shows how game theory has recently emerged as a promising technique
to solve such trade-off of power supply and demand so that both the utility operator
and customers may benefit. These game-theoretic approaches allow the customers
negotiate with the power company or operators until they reach an equilibrium, a
point of “consensus” so to speak, where all the stakeholders feel they have made the
best of the negotiation. The problem is depending on how the optimization problem
or the game is formulated, such a consensus may take a long time to reach, or it may
not at all exist in some cases.
Furthermore, the DSM techniques, particularly the DR programs, usually only
consider either the interest of the utility company or the customer. Consider a simple
example. A power company wants to maximize its revenue while its customers
want to reduce their energy costs. If so, for example, a program that prices peak
electricity up and discourages consumption might actually deprive the utility of
revenue. This case is just one of many in which contrasting demands make it difficult
to design a joint DSM technique that takes into account the interests of the utility
company and its customers simultaneously.
3.1.3 Power Loss Problem in Micro-Grids Based

Smart Grid System
There are different types of power losses associated with power exchange in the
smart grid. For instance, when power is transmitted, heated power distribution lines
result in transmission power loss. Power storage results in storage power loss. Fur-
thermore, different choices might cause different power losses. For example, under
the same physical conditions, transmission power loss caused by obtaining power
from a nearby micro grid is lower than that from a distant micro grid. Therefore,
an efficient algorithm that is able to optimally reduce the total power losses in
smart micro-grid system. Note that whether we consider power transmission in
the traditional power grid or in the smart micro-grid based power delivery system,
power loss is an unavoidable reality.
Even in the state-of-the art HFAC micro grids, large voltage drops and higher
long distance power loss are the key issues that currently limit their practical
implementation. The power loss figure may not appear to be substantial for just one
or two individual micro grids. But when you consider the entire power grid system,
the power loss is just too much. [8] reported the power loss of the 20,279,640 gWh
power consumed by the whole world in 2009, and it indicated that if there was some
way to reduce the power loss even by 1 % of the total power consumption, more
than billions of dollars could have been saved.
Recently, there has been a significant stress on studying the issue of power
loss between the micro-grids and that between an individual micro-grid and the
macro-station recently. In [9], energy management in smart grid powered electrical
appliances investigated with a particular emphasis on energy savings from their
proposed strategy. Arefifar et al. presented systematic and optimized approaches

for the power distribution system into a set of micro-grids with optimized self-
adequacy [10]. In their work in [11], Niyato et al. proposed an algorithm which
optimizes the transmission strategy to minimize the total cost. The problem of
minimizing power losses in distribution networks has traditionally been investigated
by using a single, deterministic demand level. Ochoa et al. presented a novel
algorithm to solve this problem [12]. Kantarci et al. proposed the “cost-aware smart
micro grid network design,” which enables economic power transactions within the
smart grid [13]. The work in [14] proposed operation strategies to be implemented
into the distribution management system for power loss reduction. Some of the other
notable research works, which have addressed power loss can be found in [15, 16].
In one of our earlier works [17], a game-theoretic coalition formulation strategy for
reducing power loss in micro-grids dubbed as GT-CFS was proposed.
3.2 Is Smart Grid Communication Security an Overkill?
In the smart grid, the customers are no longer passive stakeholders. They can
actively get involved to effectively minimize energy consumption by communi-
cating back and forth with the utility operator. Numerous sensing devices, smart
meters, and control units are likely to be between the operator and end-users to
realize the bi-directional communication. IP-based communication technologies are
considered to be the top candidate for setting up HANs, BANs, and NANs. This
means the smart meters and loads are all hooked up to the IP-based communication
network. However, existing IP-based communication networks (e.g., the Internet)
are likely to be susceptible to a wide variety of malicious attacks, such as replay,
traffic analysis, and denial of service (DoS) attacks. The smart grid communication
network is also no exception. Therefore, it is important to properly design smart
grid communication protocols for dealing with all possible security threats [18–20].
However, designing a security protocol from scratch specifically for smart grid
communication may not be practical. Using heavy duty encryption mechanisms
can actually be an overkill since they can significantly add delay and packet
overhead affecting the delay-sensitive communication requirement of the smart grid.
The smart grid needs a light-weight secure protocol suite customized to meet its
communication needs. As a first line of defense against malicious threats to smart
grid communication, a light-weight authentication scheme is required. Just like
the traditional IP-based communication networks, the smart grid communication
framework needs to check whether the parties involved in communication are the
exact entities they appear to be. Hence, the smart grid communication framework
should consider an adequate authentication mechanism [21–28] so that malicious
attackers cannot compromise the secrecy or privacy of the information exchanged
between the utility operator and its customers.
References 21
3.3 Concluding Remarks: Could an All-in-One Optimization

Technique Work in Smart Grid?
How acceptable and widespread the smart grid system will become depends a lot on
how its design can cope with some of the most critical challenges such as balancing
electricity demand and supply, and power loss minimization. Careful design to
combat these challenges is the key. If the utility company directly controls the
residential loads, the customers may perceive this as breach of privacy. If each
customer needs to sign a contract with the utility for DLC, this might not get
beyond large-scale users like malls, warehouses, and factories. These are just some
simple examples to highlight the fact that smart grid challenges are both technical
and social, and therefore, they need cautious design and evaluation. Furthermore,
note that the smart grids are not blackout-proof. It will work fine under normal
operations flattening the load curve, but blackouts can still happen in a smart grid
when the power plants fail to operate during the peak demand hours. Hence, even
in a smart grid, baseload generation, dispatchable generation through DLC, and
storage are required. The smart grid just pushes out the eventuality of a blackout far
enough so that most of the grid system can safely recover from intermittent sources
without increasing the frequency of blackouts. The concluding remark is that an all-
in-one optimization technique in smart grid may not be possible because DSM,
power loss minimization, and security have different and sometimes contrasting
objectives. Instead of an all-in-one solution, an extensive framework with flexible
action-policies can be offered from which the utility operator and customers can
select the best course of action.
References

2. M. M. Fouda, Z. M. Fadlullah, N. Kato, A. Takeuchi, and Y. Nozaki, “A novel demand control
policy for improving quality of power usage in smart grid,” in 2012 IEEE Global Communica-
tions Conference (GLOBECOM), Anaheim, California, USA, Dec. 2012, pp. 5154–5159.
3. C. Gellings, “The concept of demand-side management for electric utilities,” Proceedings of
the IEEE, vol. 73, no. 10, pp. 1468–1470, Oct. 1985.
4. Z. M. Fadlullah, M. Q. Duong, N. Kato, and I. Stojmenovic, “A novel game-based demand side
management scheme for smart grid,” in 2013 IEEE Wireless Communications and Networking
Conference (WCNC), Shanghai, China, Apr. 2013, pp. 4677–4682.
5. Z. M. Fadlullah, D. M. Quan, N. Kato, and I. Stojmenovic, “GTES: an optimized game-
theoretic demand-side management scheme for smart grid,” IEEE Systems Journal, vol. 8,
no. 2, pp. 588–597, Jun. 2014.
6. T. Agarwal and S. Cui, “Noncooperative games for autonomous consumer load balancing
over smart grid,” Computing Research Repository - CoRR, vol. abs/1104.3802, 2011. [Online].
Available: http://arxiv.org/abs/1104.3802
7. A.-H. Mohsenian-Rad, V. Wong, J. Jatskevich, R. Schober, and A. Leon-Garcia, “Autonomous
demand-side management based on game-theoretic energy consumption scheduling for the
future smart grid,” IEEE Transactions on Smart Grid, vol. 1, no. 3, pp. 320–331, Dec. 2010.
8. I. E. A. (IEA), available online, accessed Aug 2015, “http://www.iea.org/statistics/”.

9. Y. K. Tan, T. P. Huynh, and Z. Wang, “Smart personal sensor network control for energy
saving in dc grid powered led lighting system,” IEEE Transactions on Smart Grid, vol. 4,
no. 2, pp. 669–676, Jun. 2013.
10. S. Arefifar, Y. Mohamed, and T. El-Fouly, “Supply-adequacy-based optimal construction of
microgrids in smart distribution systems,” IEEE Transactions on Smart Grid, vol. 3, no. 3,
pp. 1491–1502, Sep. 2012.
11. D. Niyato and P. Wang, “Cooperative transmission for meter data collection in smart grid,”
IEEE Communications Magazine, vol. 50, no. 4, pp. 90–97, Apr. 2012.
12. L. Ochoa and G. Harrison, “Minimizing energy losses: Optimal accommodation and smart
operation of renewable distributed generation,” IEEE Transactions on Power Systems, vol. 26,
no. 1, pp. 198–205, Feb. 2011.
13. M. Erol-Kantarci, B. Kantarci, and H. Mouftah, “Cost-aware smart microgrid network design
for a sustainable smart grid,” in 2011 IEEE GLOBECOM Workshops (GC Wkshps), Dec. 2011,
pp. 1178–1182.
14. I.-K. Song, W.-W. Jung, J.-Y. Kim, S.-Y. Yun, J.-H. Choi, and S.-J. Ahn, “Operation schemes
of smart distribution networks with distributed energy resources for loss reduction and service
restoration,” IEEE Transactions on Smart Grid, vol. 4, no. 1, pp. 367–374, Mar. 2013.
15. J. Aguero, “Improving the efficiency of power distribution systems through technical and
non-technical losses reduction,” in Transmission and Distribution Conference and Exposition
(T&D), 2012 IEEE PES, Dallas, Texas, USA, May 2012, pp. 1–8.
16. T.-I. Choi and Y.-K. Cho, “International business offering related to innovative smart technolo-
gies,” in 2012 IEEE PES Innovative Smart Grid Technologies (ISGT), Washington D.C., USA,
Jan. 2012, pp. 1–8.
vol. 2, no. 4, pp. 675–685, Dec. 2011.
21. H. Zhu, X. Lin, R. Lu, P. Han Ho, and X. Shen, “SLAB: A secure localized authentication and
billing scheme for wireless mesh networks,” IEEE Transactions on Wireless Communications,
vol. 7, no. 10, pp. 3858–3868, Oct. 2008.
22. X. Lin, R. Lu, P. Han Ho, X. Shen, and Z. Cao, “TUA: a novel compromise-resilient
authentication architecture for wireless mesh networks,” IEEE Transactions on Wireless
Communications, vol. 7, no. 4, pp. 1389–1399, Apr. 2008.
23. IEEE P2030 draft guide for smart grid interoperability of energy technology and information
technology operation with the electric power system (eps), and end-use applications and loads.
[Online]. Available: http://grouper.ieee.org/groups/scc21/2030/2030_index.html
24. A. Hamlyn, H. Cheung, T. Mander, L. Wang, C. Yang, and R. Cheung, “Computer network
security management and authentication of smart grids operations,” in 2008 IEEE Power and
Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st
Century, Jul. 2008, pp. 1–7.
25. G. N. Ericsson, “Cyber security and power system communication–essential parts of a smart
grid infrastructure,” IEEE Transactions on Power Delivery, vol. 25, no. 3, pp. 1501–1507,
Apr. 2010.
26. A. R. Metke and R. L. Ekl, “Smart grid security technology,” in Innovative Smart Grid
Technologies (ISGT), Jan. 2010, pp. 1–7.
References 23
27. K. Kursawe, G. Danezis, and M. Kohlweiss, “Privacy-friendly aggregation for the smart-grid,”
in Proceedings of the 11th International Conference on Privacy Enhancing Technologies, ser.
PETS’11. Berlin, Heidelberg: Springer-Verlag, 2011, pp. 175–191. [Online]. Available: http://
dl.acm.org/citation.cfm?id=2032162.2032172
28. M. Kgwadi and T. Kunz, “Securing RDS broadcast messages for smart grid applications,”
in Proceedings of the 6th International Wireless Communications and Mobile Computing
Conference, ser. IWCMC’10. New York, NY, USA: ACM, 2010, pp. 1177–1181.
Chapter 4
Demand Response Challenge in Smart Grid
As indicated in Chap. 3, for the successful deployment of the smart grid,

demand-side management (DSM) or demand response [1–3] is crucial. DSM
means the art of planning and implementation of the electric utility activities that is
aimed to influence the users’ energy consumption by affecting desired changes in
the shape of loads of the utility operator. While DSM aims at producing a change in
the load-shape, it needs to balance the supply of the operator and the demand of the
customers.
The traditional DSM methods either shift or reduce the energy consumption.
Shifting the energy consumption can effectively reduce the aggregate energy load
during the peak hours (which is the main cause of blackouts). To measure the
imbalance in the load-shape of daily energy consumption, the ratio of the highest
peak time energy consumption to the average consumption (PAR) [4] in an entire
day is used. By shifting energy consumption from the peak hours to off-peak hours,
PAR reduction is possible. Another DSM approach is to reduce the energy usage by
encouraging energy-aware consumption patterns [5].
4.1 Background into DSM and Load Balancing Problem
Wang et al. [6] showed case studies of dynamic pricing programs, which were
offered by some US electric companies. The case studies showed improvement of
energy pricing performance through automated devices like smart thermostats to
automatically decrease energy consumption of air-conditioning and central heating
during peak hours. Automatic residential energy consumption scheduling was also
used in [7] with an assumption that the advantages of real-time pricing are currently
limited because currently buildings do not have good automation systems and
customers feel difficult to adjust with dynamic energy pricing. The idea in [7] clearly
shows the potential of using smart pricing in the smart grid. Samadi et al. [5] used

26 4 Demand Response Challenge in Smart Grid
a utility function in an energy schedule optimization in which power generation

capacity over time is fixed. Samadi et al. [5] also used different energy pricing
values for different users. But the customers felt it was unaware that they were
given unequal pricing. To avoid the unfairness issue, the “load uncertainty” was
added to the pricing scheme in [8]. Game theory was used to formulate the energy
consumption scheduling problem in [9]. The objective of this game was to reduce
PAR by shifting energy use. Each player (customer) plays the game with other
players until the game reaches an equilibrium so that everyone becomes happy about
their own PAR reduction. The game does have a distributed flavor, but typically it
needs some central information. On the other hand, a purely centralized scheme
for optimally scheduling energy use [4] was based on a social welfare function.
This social welfare was actually the preference of the users to energy consumption,
which is the difference of the utility and cost of the energy. However, utility of
energy and cost of energy have different units and this was not entirely reasonable.
In literature, we could find other centralized schemes such as [10] which use game-
based interactions between energy companies, retailers, and customers. In the next
section, let us overview an existing system model of smart grid for power, energy
cost, and load control modeling [9].
4.2 System Model
We use the smart grid architecture shown in Fig. 2.1 as the basis of our adopted sys-
tem model. In the adopted model, a single energy supplier (utility operator) having
many customers is considered. The smart meters of the customers are connected
with the operator’s control center, and can schedule the energy consumption of
their residence. Assume that the smart meters can monitor and collect the electricity
consumption data of all electrical loads plugged into the grid. If required, the smart
meters can also turn on or off the loads, and select the appropriate level of energy
consumption for the loads. Furthermore, the smart meters can inform the operator
regarding the energy consumption schedules of the customers. By this way, the
smart meters could be assumed to be capable of informing the power company or
the supplier about users’ energy consumption schedules.
Our considered smart grid model has the following aspects, the power system
modeling, appropriate energy cost function modeling, and load control on
consumer-end modeling [3, 11].
4.2.1 Power System Modeling
We use a popular power system modeling as in [9] which supposes N , a set of users,
who obtain electricity from the utility operator. The number of users is N , jN j.
For each user, n 2 N , lnh denotes the total load during hour h 2 H , f1; : : : ; Hg
4.2 System Model 27
where H D 24. The daily load of the nth user is given by her energy consumption
vector, ln , fln1 ; : : : ; lnH g. Let Lh denote the aggregate load of the users during each
P
hour of a day (i.e., h 2 H). This can be calculated using Lh , h
n2N ln . The
daily peak and average load levels are calculated as Lpeak D maxh2H Lh and Lavg D
1 P Lpeak
h2H Lh , respectively. Now we can calculate the PAR = Lavg D
HP
maxh2H Lh
H Lh
.
h2H
4.2.1.1 Modeling an Appropriate Energy Cost Function
Consider that the hourly energy price changes in the model. The change is
proportional to the system energy consumption in that hour. This is done so as to
give incentive to the customers to prevent them from using much energy during
peak hours. Such a policy can contribute to a lower PAR. The cost function is
an increasing one which means that the energy cost increases with Lh (the total
energy consumption). For each h 2 H, we now have Ch .Lh1 / < Ch .Lh2 /; 8Lh1 < Lh2 .
This assumption makes sure that the higher the energy consumption, the higher the
impact on the energy price hike. Also, the energy cost functions are assumed to be
strictly convex such that for every h 2 H, Ch . Lh1 C .1 /Lh2 / < Ch .Lh1 / C
.1 /Ch .Lh2 /. Here, Lh1 , Lh2 , and are real numbers such that Lh1 ; Lh2 0 and 0 <
< 1.
4.2.1.2 Modeling Load Control on the Customer-Side
Consider An to be the set of loads (including air-conditioners, heaters, refrigera-

tors, etc.) of the nth 2 N customer. For each of her load, an energy consumption
1
scheduling vector xn;a D Œxn;a ; : : : ; xn;a
H
is constructed where xn;a h
means the
corresponding one-hour P energy consumption scheduled for the load a during hour h.
Suppose lnh D x h
; h 2 H denotes her total energy consumption, Lh D
P a2An n;a
n2N nl h
; h 2 H means the aggregate energy consumption of all the users, and
Ln;h D Lh lnh ; h 2 H indicates the aggregate energy consumption of all users
except the nth customer. Then, the smart meter of each customer needs to decide
the optimal schedule of energy consumption vector, xn;a , for all her loads with
some constraints like starting and stopping time of the schedule (˛n;a 2 H and
ˇn;a 2 H, respectively), and the minimum and maximum energy requirement
min max
(En;a and En;a , respectively) to finish the operation. Then, the time interval within
which the loads may be scheduled is given by Hn;a . The power level of each
load, a 2 An , also is limited by the minimum standby power level n;a min
, and the
maximum power level n;a . Then, we have: n;a & xn;a n;a ; 8h 2 Hn;a
max min h max
h
and xn;a D 0; 8h 2 HnHn;a . The feasible energy consumption scheduling set of the
th
n customer is then derived to be:
ˇn;a
X
n D fxn jEn;a
min
h
xn;a En;a
max
;
˛n;a
n;a
min
xn;a
h
n;a
max
; 8h 2 Hn;a ;
h
xn;a D 0; 8h 2 HnHn;a g: (4.1)
In order to improve the assumptions of the existing model, we incorporate a

number of modifications to it by adopting specific strategies of the users and the
utility operator, and formulate a more complete optimization problem.
4.3 Adopted Strategies of Users and Utility Operator

and a Game-Theoretic Solution
In the following, we first discuss our adopted strategies of the users and utility
operators by making some modifications to the earlier described existing system
model described in the earlier section. Based on these strategies, a two-step game is
then presented.
4.3.1 Strategy of the Users
Instead of optimizing the energy consumption schedules of all the users at once
in (4.1), each smart meter needs to optimize the schedule of its user based on
her energy need. Obviously her energy consumption schedule vector belongs to
the feasible set in (4.1). Since the objective is to optimize the users’ pay-off,
suppose that the objective function is the following function of energy consumption
schedule vector.
Wi .x1 ; : : : ; x24 / D Value of Energy Cost of Energy

24
X
D V.Xi / xh Ch .Lh /; (4.2)
hD1
where (x1 ; : : : ; x24 ) is the energy consumption scheduling P24vector with the
energy loads of user i during the hours of the day, Xi D hD1 xh means the
total energy consumption, and V.Xi / represents the value of that amount of energy.
In the remainder of the section, we present an analysis of the value of energy
function and cost of energy function.
4.3 Adopted Strategies of Users and Utility Operator and a Game-Theoretic. . . 29
4.3.1.1 Value of Energy Perceived by Users
Let V.Xi / indicate how much energy value is given by the user i. Because each
user in the power system could have different energy consumption patterns, their
energy consumption schedules are likely to be different also. Even in the case
where some users have the same energy consumption, their attitude and habits may
dictate their perceived energy value. So, it is not so trivial to capture the response
and energy demand of different users toward the same energy price. Hence, we
analytically model the users’ preferences toward energy consumption by adopting
the utility function theory of microeconomics [12].
According to utility theory, a reasonable utility function should satisfy the
characteristics of quadratic and exponential functions. From [13], the exponential
function could present a better model of the user’s preference which motivates us
to adopt the exponential function u.X/ D 1 e!X as its value domain has been
normalized between (0,1). Figure 4.1 shows a plot of utility functions in which !
is a parameter of the users’ tolerance toward energy consumption reduction. The
figure demonstrates that a utility function with a relatively larger value of ! slowly
approaches the upper-bound. This means that the users with large ! values are strict
with the energy curtailment. In other words, their utility values are lower compared
to other users even with the same energy consumption. Because the ! value of a
user is her private information, it should not be revealed to other users. Based on
this utility function, we define the value function of energy as:
V.Xi / D pXmax u.Xi / D pXmax .1 e!Xi /: (4.3)
Fig. 4.1 Value of utility for different resources

In (4.3), Xmax is the maximum amount of energy that the user can use. p is the
average price of a unit of energy.
The adopted utility function has the maximum value of 1. By scaling it up
with the product of average energy price and the maximum amount of energy
consumption of the user, the maximum value of V.Xi / becomes equal to the value
of the maximum energy consumption. What this means is that the user will value
her consumption amount not over the average money she needs to pay to satisfy her
maximum demand.
4.3.1.2 Cost of Energy
Compared to the piece-wise and quadratic linear functions described in [4, 9], the
gap in energy price between peak hours and off-peak hours is large. For example,
PAR D 3 means that the load during peak-hours is three times the average daily
load. This indicates that the difference between the prices of energy is significant.
Due to strict schedules during peak hours, this large difference is inconvenient to the
users. Instead of making such a big jump in the pricing, a proportional increase in the
energy cost in accordance with the total load is important to encourage participation
of users in balancing the PAR. In [6], the energy prices were carefully selected
by considering the reaction of the users and not allowing the differences between
energy prices to exceed three times.
We could intuitively consider that the more drastically the cost function changes,
the better PAR reduction we would get. However, the computation time increases
much when the energy cost function varies dramatically. So, there is actually a trade-
off in selecting the energy price. We adopt the following energy price function with
this trade-off in mind.
Ch .Lh / D ˛Lh log.Lh C 1/; (4.4)
where ˛ is our “price parameter.” The utility operator can tune ˛ to change the daily
energy price to control the energy consumption of the users. The price difference
between the peak and off-peak hours still remains unchanged. The logarithmic
function in (4.4) gives almost a linear shape (Fig. 4.2). The comparison between
our adopted price function and the conventional quadratic price function can also be
noticed in Fig. 4.2.
By substituting (4.3) and (4.4) into (4.2), the user i’s objective function can be
rewritten as:
P24
Wi .x1 ; : : : ; x24 / D pXmax .1 e! hD1 xh /
24
X
˛xh Lh log.Lh C 1/: (4.5)
hD1
4.3 Adopted Strategies of Users and Utility Operator and a Game-Theoretic. . . 31
Fig. 4.2 The comparative increase of energy price with the total energy consumption using the
quadratic function and the extended energy price function
4.3.2 Strategy of the Utility Operator
Now we consider how the utility operator determines the energy pricing scheme.
In our enhanced model, the energy price varies according to the total load during
different hours. For the newly adopted price function, the energy price is now
proportional
to Lh log.Lh C 1/ . This is ensured by setting the energy price function
as Ch .Lh / D ˛Lh log.Lh C 1/ , where Lh is the total load during hour h. Thus, we
have an energy price vector, C.L/ D ˛ L1 log.L1 C1/C: : :CL24 log.L24 C1/ , for
the entire day. Therefore, the supplier can tune the parameter ˛ so as to influence the
price vector, and thus, limit the users’ energy consumption. How can the operator
decide a reasonable ˛ value? Consider a baseline fixed price scheme used in
traditional power grids. In Japan, TEPCO [14] charges 17:87 yen for the first 120 kW
(kilowatts) while 22:86 yen for the next 180 kW for the light residential user with
20 ampere power line. Hence, assume that the operator has the average price for the
energy which it sells to the users. This average price was actually the parameter p
in (4.3). If the operator uses the dynamic price scheme, assume that the total cost
charged for the entire system is equal to the fixed price scheme with the same load.
With this relaxed assumption, we have:
24
X 24
X
p Lh D Lh Ch .Lh /
hD1 hD1
24
X
D˛ Lh2 log.Lh C 1/: (4.6)
hD1
Then, the operator can calculate the energy price parameter ˛ as follows.
P24
p hD1 Lh
˛ D P24 2
: (4.7)
hD1 Lh log.Lh C 1/
Now that we have all the necessary parameters, we are set to show how a demand
game can be played between the operator and its users.
4.4 Playing the Demand Game
The game played between the utility operator and users aims to reduce the system
PAR by optimizing energy schedules of the customers[11]. The optimization
process can be modeled as a two-stage game [15] as shown in Algorithms 1 and 2.
1. The users will try to maximize their pay-offs by optimizing functions (shown
in (4.5)) using interior point method (IPM).
2. The supplier will then adjust the energy price parameter consistent with the user’s
energy consumption schedule according to (4.7).
When the game reaches an equilibrium state, neither users nor supplier will
change their strategies. At the same time, the system PAR and total energy
consumption are reduced.
Algorithm 1 Utility operator’s game.

1: Begin: Collect original schedules from all users
2: All users initialize their schedules from feasible sets
3: End
4: Calculate initial energy price parameter ˛ according to (4.7)
5: Repeat
6: Randomly choose user n 2 N
7: Signal user n to execute algorithm 2
8: Update the new schedule vector Ln from user n
9: Update the energy price parameter ˛ according to (4.7)
10: Until No user wants to change schedule
Algorithm 2 User’s game.

1: Begin: Receive signal from operator
2: Request ˛, vector L
3: User n optimizes schedule by solving the problem in (4.5) by using IPM
4: If xn changes compared to current schedule Then
5: Inform the center of the new schedule vector Ln
6: End If
7: End
4.5 Comparative Performance 33
4.5 Comparative Performance
To demonstrate the effectiveness of our adopted game-theoretic approach, its

performance is compared with two conventional methods (centralized convex opti-
mization and energy consumption game [9] which are referred to as conventional 1
and conventional 2 for easy reference). Using MATLAB [16], a smart grid system
having a single utility operator with multiple users was modeled. The number of
users was varied from 5 to 500. Each user is assumed to be connected to the
operator’s control center via smart meter. Each user is supposed to have 10–15
schedulable loads (washing machines, dish washers, PHEVs, etc.) and 10–15 non-
schedulable loads (lighting banks, heaters, etc.). Table 4.1 lists energy consumption
of some typical residential loads [17]. The settings are randomly generated during
each simulation run. However, the same setting is maintained in a specific situation
for comparing the adopted approach with the conventional methods.
Figure 4.3 shows that the running time of conventional 1 increases rapidly
(almost at an exponential rate) whereas the two game-theoretic methods (the
adopted approach and conventional 2) require low completion time with increasing
numbers of users. Even with a relatively low number of users, the large number
Table 4.1 Energy appliances Appliances Average consumption per day (kW)
and their average
consumption on a daily basis Clothes dryer 2.47
Dishwasher 0.99
Lighting 3.29
Refrigerator 5.89
Washing machine 0.28
Fig. 4.3 Comparison of the running time between the conventional methods and the adopted
approach
Fig. 4.4 Number of iterations until convergence in case of the conventional methods and the
adopted approach, for varying numbers of users
Fig. 4.5 Average time needed for each iteration in case of the conventional methods and the
adopted approach
of parameters seriously influences the running time of convex optimization (i.e.,

conventional 1). Also, it may not lead to convergence for users exceeding 30. So,
conventional 1 seems not to be viable. Figures 4.4 and 4.5 will shed more light into
this issue.
4.5 Comparative Performance 35
In Figs. 4.4 and 4.5, the number of iterations required for each algorithm
to converge and the average time per iteration are demonstrated, respectively.
The running time in the previous analysis is the outcome of these two parameters.
Figure 4.4 demonstrates how the number of iterations needed for conventional 1,
conventional 2, and our adopted approach gradually increases with the number of
users. The average time required for each iteration, however, as demonstrated in
Fig. 4.5, varies slowly for the game-theoretic algorithms while it rises drastically
in case of conventional 1. This corroborates our expectation because the game-
theoretic approaches need to solve just the local optimization problem of each
user. These indicate that conventional 1 is not scalable enough for a large number
of users whereas the game-theoretic approaches perform much better. In the next
comparisons, we only compare conventional 2 and the adopted approach since
conventional 1 was already shown not to be scalable with the growing number
of users.
Figure 4.6 depicts the number of iterations required for the convergences of
conventional 2 and the adopted approach. Their convergence is fast because both
are based on game theory. Also with increasing numbers of users, the ratio between
the number of iterations and the number of users decreases a little. This means
that the bigger the system becomes, the less effect is inflicted by changing a single
user’s schedule. Figure 4.7 demonstrates the comparison of the PAR reduction in
conventional 1 and our adopted approach. The results indicate that the latter achieve
much higher PAR reduction because of its consideration of not only shifting energy
consumption while scheduling, but also its ability to adjust energy consumption
levels during different hours.
Fig. 4.6 Number of iterations required for the conventional 2 method and the adopted approach
for a large number of users (up to 500)
Fig. 4.7 PAR reduction in the conventional 2 method and the adopted approach for varying
numbers of users
4.6 Concluding Remarks
In this chapter, a smart grid infrastructure was considered where the utility operator
and its consumers were demonstrated to play their respective games to optimize
their energy schedules. In the adopted two-stage game in smart grid, the objective
was to achieve a reduction in the system PAR. A pricing scheme was implemented
in which the energy price changes according to the total energy consumption
during each hour so that the customers receive reasonable incentives to follow the
system. Because many parameters need to be optimized, a fully centralized control
using convex optimization did not perform well. The comparative performance also
demonstrated that both the conventional and the adopted approaches which were
based on game theory converged much faster and scaled well with the growing
number of users. In particular, our adopted approach achieved much better PAR
reduction which is desired from the utility operator’s perspective. At the same time,
the users could reduce their energy cost in an effective way.
References
1. M. M. Fouda, Z. M. Fadlullah, N. Kato, A. Takeuchi, and Y. Nozaki, “A novel demand control

policy for improving quality of power usage in smart grid,” in 2012 IEEE Global Communica-
tions Conference (GLOBECOM), Anaheim, California, USA, Dec. 2012, pp. 5154–5159.
2. C. Gellings, “The concept of demand-side management for electric utilities,” Proceedings of
the IEEE, vol. 73, no. 10, pp. 1468–1470, Oct. 1985.
References 37
3. Z. M. Fadlullah, M. Q. Duong, N. Kato, and I. Stojmenovic, “A novel game-based demand side

management scheme for smart grid,” in 2013 IEEE Wireless Communications and Networking
Conference (WCNC), Shanghai, China, Apr. 2013, pp. 4677–4682.
4. P. Samadi, R. Schober, and V. W. S. Wong, “Optimal energy consumption scheduling using
mechanism design for the future smart grid,” in 2011 IEEE International Conference on Smart
Grid Communications (SmartGridComm), Brussels, Belgium, Oct. 2011, pp. 369–374.
5. P. Samadi, A.-H. Mohsenian-Rad, R. Schober, V. Wong, and J. Jatskevich, “Optimal real-
time pricing algorithm based on utility maximization for smart grid,” in 2010 First IEEE
International Conference on Smart Grid Communications (SmartGridComm), Gaithersburg,
Maryland, USA, Oct. 2010, pp. 415–420.
6. J. Wang, M. Biviji, and W. Wang, “Lessons learned from smart grid enabled pricing programs,”
in 2011 IEEE Power and Energy Conference at Illinois (PECI), Urbana, IL, USA, Feb. 2011,
pp. 1–7.
7. A.-H. Mohsenian-Rad and A. Leon-Garcia, “Optimal residential load control with price
prediction in real-time electricity pricing environments,” IEEE Transactions on Smart Grid,
vol. 1, no. 2, pp. 120–133, Sep. 2010.
8. P. Tarasak, “Optimal real-time pricing under load uncertainty based on utility maximization
for smart grid,” in Smart Grid Communications (SmartGridComm), 2011 IEEE International
Conference on, Brussels, Belgium, Oct. 2011, pp. 321–326.
9. A.-H. Mohsenian-Rad, V. Wong, J. Jatskevich, R. Schober, and A. Leon-Garcia, “Autonomous
demand-side management based on game-theoretic energy consumption scheduling for the
future smart grid,” IEEE Transactions on Smart Grid, vol. 1, no. 3, pp. 320–331, Dec. 2010.
10. S. Bu, F. Yu, and P. Liu, “Dynamic pricing for demand-side management in the smart grid,” in
Online Conference on Green Communications (GreenCom), 2011 IEEE, Sep. 2011, pp. 47–51.
11. Z. M. Fadlullah, D. M. Quan, N. Kato, and I. Stojmenovic, “GTES: an optimized game-
theoretic demand-side management scheme for smart grid,” IEEE Systems Journal, vol. 8,
no. 2, pp. 588–597, Jun. 2014.
12. “An introduction to utility theory,” accessed Aug. 2015. [Online]. Available: http://www.
norstad.org/finance/util.pdf
13. G. H. and P. G., “Utility functions: From risk theory to finance,” North American Actuarial
Journal (NAAJ), vol. 2, no. 3, pp. 74–100, 1998.
14. “TEPCO,” accessed Aug. 2015. [Online]. Available: http://www.tepco.co.jp/e-rates/individual/
data/
15. J. Zhang and Q. Zhang, “Stackelberg game for utility-based cooperative cognitive radio
networks,” in Proceedings of the Tenth ACM International Symposium on Mobile Ad Hoc
Networking and Computing, ser. MobiHoc ’09. New York, NY, USA: ACM, 2009, pp. 23–32.
[Online]. Available: http://doi.acm.org/10.1145/1530748.1530753
16. Mathworks - matlab and simulink for technical computing. [Online]. Available: http://www.
mathworks.com/
17. “Typical power consumption,” accessed Aug. 2015. [Online]. Available: http://www.oksolar.
com/technical/consumption.html
Chapter 5
Game-Theoretic Coalition Formulation Strategy
for Reducing Power Loss in Micro Grids
5.1 Background
In this chapter, we move away from the smart-grid demand-side management topic
to another important one of micro grids. Even though a significant progress has
been achieved in the development of the micro grids, the power loss minimization
between the micro grids and also between the macro station and an individual micro
grid is still receiving much attention. Niyato and Wang [1] considered an algorithm
aimed to optimize the transmission strategy in order to minimize the total cost
including the power loss. The power losses minimization in the energy distribution
networks has conventionally been investigated using a single and deterministic
demand level. A novel algorithm to overcome this problem was designed in [2].
The “cost-aware smart micro grid network design” in [3] allows economic power
transactions within the smart grid with manageable power losses. Meliopoulos
et al. [4] also discussed power loss minimization issues by proposing a coordinated
control scheme at real-time with the inclusion of distributed generation resources
(micro grids) with the existing grid. A novel load management solution to coordinate
the charging of multiple PHEVs in a smart grid system was considered in [5]
which also considered the power loss problem. An efficient optimal reconfiguration
algorithm for power loss minimization was studied in [6]. Costabeber et al. [7]
demonstrated that the power loss reduction is viable without central controllers by
exploiting local measurement, communication, and control capability in the micro
grids. Saad et al. [8] used cooperative game theory to formulate novel cooperative
strategies between the micro grids of an energy distribution network aimed to
reduce the power loss. Also, Costabeber et al. [9] and Tenti et al. [10] discussed
cooperative operation of neighboring power processing units to reduce distribution
losses. Corso et al. [11] addressed the daily schedule of distributed generators to
minimize power loss in a micro grid connected to the main grid. A more detailed
methodology to develop an autonomous micro grid for coping with power loss
can be found in [12]. Also, a heuristic-based greedy algorithm was proposed

40 5 Game-Theoretic Coalition Formulation Strategy for Reducing Power Loss: : :
in [13] to reduce the power loss. A review of technologies, methodologies, and

operational approaches for improving the efficiency of power distribution systems
like micro grids can be accessed in [14]. The “autonomous regional active network
management system” project aimed to design a distribution network power loss
management algorithm in [15]. The tie-set graph theory and its application to smart
grid networks was introduced to minimize the power loss in [16]. Power loss of the
energy distribution networks was also considered in [17–21].
Compared to the aforementioned studies addressing the power loss minimization
problem of distribution networks in a scattered manner, we need to have a more
comprehensive insight into this issue in the micro grids context. This chapter
will describe a basic micro grids system model and illustrate how formulation of
effective coalitions between the micro grids can be useful in significantly reducing
the distribution power loss.
5.2 System Model
Figure 5.1 shows our considered system model comprising a number of micro grids.
In this section, our considered system model of the smart grid is presented.
This model considers that the users are supplied electricity by the macro station
and/or a number of autonomous micro grids using distributed, renewable energy
sources like wind farms, solar panels, PHEV batteries, etc. Each micro grid is linked
Fig. 5.1 Example of micro grids oriented system model

5.2 System Model 41
to the macro station through the main power grid. Also, each micro grid is assumed
to have its own customers such as residential users, schools, factories, etc., who
have the capability to notify their respective micro grid(s) regarding energy demands
via AMI. Because the power loss between the macro station and the micro grid is
usually more than that between two neighboring micro grids, the adjacent micro
grids could potentially improve the power loss through forming supportive groups,
which we refer to as coalitions. A coalition can initially form with just one micro
grid (e.g., coalition 2 in Fig. 5.1).
Let the set of micro grids be N . Assume that during each hour, the ith micro grid
(for 8i 2 N ) denoted by MGi produces energy Gi and confronts the total demand
Di from all its users. Let Reqi D .Gi Di / be the surplus of MGi . If Reqi < 0, MGi
needs to buy additional energy to meet the total demand of its users. On the other
hand, MGi has energy surplus and can sell its energy if Reqi > 0. If Reqi D 0, MGi
just meets the demand, and in this case, it is neither an energy seller nor a buyer.
Based on this concept, all the micro grids can be divided into three types—neutral,
seller, or buyer.
The production Gi and the total demand Di are supposed as random numbers in
the real Smart Grid networks [22] since Di depends on many unpredictable factors
like time, user behavior, etc. Furthermore, because many micro grids use renewable
resources, their energy generation is subject to surrounding environment, Gi cannot
be treated as a constant either. So, Reqi is also a random number having a certain
observed distribution.
Additionally, assume that a micro grid acting as seller has enough energy to sell
to the micro grids who assumed the role of buyer(s). On the other hand, assume that
a buyer has sufficient “money” to buy the energy from the seller(s) and/or the macro
station.
Now we can extend the concept of buyers and sellers to the groups or coalitions
formed by a number of micro grids. For instance, Fig. 5.1 shows that there are three
coalitions. In the considered time window (during the current hour), coalition 1 has
energy surplus and can sell energy to other micro grids or the macro station. But,
coalition 2 needs energy and will need to buy energy from another coalition or
multiple other coalitions and/or the macro station. On the other hand, Coalition 3’s
demand and supply are equal, so the micro grids in this coalition do not need to
exchange energy with any other coalition or the macro station.
5.2.1 Non-cooperative Model
Before we can establish the payoff function of coalition, consider a non-cooperative

case as the baseline/comparable model where each micro grid only sells energy to
the macro station or buys energy from the macro station. Let U0 express the medium
level voltage of power transfer between a micro grid and the macro station. Any
energy transfer between the micro grid and the macro station will have power loss.
In this chapter, we are concerned with the basic power losses related to distribution,
namely the power loss over the distribution lines inside the network, and the power
loss due to other factors in the macro station such as friction, corrosion, weather,
dust, etc. If MGi wants to sell Reqi to the macro station when .Reqi > 0/, or buy Reqi
from the macro station when .Reqi < 0/, the power loss Pi0 can be expressed as:
Pi0 D Ri0 I02 C ˛Qi ; (5.1)
where Ri0 denotes the distribution line resistance between the macro station and
MGi , and ˛ is a constant to account for a fraction of power loss due to other factors.
Qi denotes the energy which MGi wants to buy or sell. I0 D UQ0i represents the
current flowing over the distribution line when there is an energy transfer between
the macro station and MGi . Then, we can rewrite (5.1) as:
Ri0 Q2i
Pi0 D C ˛Qi ; (5.2)
U02
where
8
< Reqi W Reqi > 0
Qi D Li W Reqi < 0 (5.3)
:
0 W Reqi D 0:
Here, Li is the total amount of power which needs to be generated (or be made
available to the system) to make sure that MGi is able to get the energy needed to
meet its total demand Reqi . If there is no power loss, Li D jReqi j. Hence, Li >
jReqi j. Li is the solution of following quadratic equation.
Ri0 Li2
Li D Pi0 C jReqi j D C ˛Li Reqi : (5.4)
U02
The number of roots of (5.4) depends on the values of the resistance Ri0 , transfer
voltage U0 , and ˛. For a specific Reqi value, three possible solution-sets of (5.4)
exist (none, one, and two solutions). Because we are interested in minimizing the
value of Reqi , if (5.4) has two roots, the smaller one is to be used. If (5.4) has no
solution, assume that the root is the same as Eq. (5.4) having a single root, which is
.1˛/U 2
Li D 2Ri0 0 .
As expressed by (5.2) and (5.3), the power loss between MGi and the macro
station depends on several factors such as the resistance Ri0 , transfer voltage U0 , ˛,
and Reqi (for either buying or selling). Hence, we can estimate the payoffs of the
micro grids, when the parameters are given. Because in the non-cooperative case
each micro grid can be considered to form a coalition, the payoff of the micro grid
is equal to that of the coalition. So, we can define the non-cooperative payoff (utility)
of each MGi as the total power loss due to the power transfer:
u.fig/ D w2 Pi0 ; (5.5)

5.2 System Model 43
where w2 is the price of a unit power in the macro station. Since we aim to minimize
u.fig/, the negative sign is able to convert the problem into a problem of seeking the
maximum.
5.2.2 Cooperative Coalition Model
Now consider the cooperative coalition model for managing the micro grids acting
as buyers and sellers. In addition to exchanging energy with the macro station,
the micro grids are able to exchange energy with each other. This is particularly
attractive to the micro grids because the power loss during transmission between
the neighboring micro grids is always less than that between the macro station and
a distant micro grid. Furthermore, the micro grids can make collaborative groups or
coalitions as explained earlier to exchange energy with each other so as to minimize
the power loss in the main smart grid and maximize their payoffs in (5.5).
To study the cooperative behavior of the micro grids, consider the framework of
coalition game theory in [22] which defines a coalition game to have a set of players
N , the strategy of players, and the function v: 2N ! R. In this game, v denotes
a function which assigns to every coalition S N a real number representing the
total profits achieved by S. Any coalition S N can be divided into two parts: the
set of sellers (Ss S) and the set of buyers (Sb S) such that Ss [ Sb D S. So, for
a micro grid MGi 2 Ss , Reqi > 0 and it implies that MGi can sell energy to others.
On the other hand, a micro grid MGj 2 Sb with Reqj < 0 implies that it wants to
buy energy from others. Obviously, any coalition S N needs to have at least one
seller and one buyer.
To calculate the payoffs of all the coalitions, let us define the payoff function v(S)
for each S N . For any coalition S D Ss [ Sb , we need to study the local power
transfer between the sellers Ss , the buyers Sb , and the macro station. Consider how
to form the coalition. Note that a micro grid does not join a coalition, if the payoff
of this micro grid in the coalition is below than the payoff out of the coalition. Also,
if a micro grid cannot give payoff to the coalition, it will not join that coalition.
Unlike the non-cooperative model, in the formed coalition in the cooperative
model, there could be multiple micro grids, which can exchange energy with others
or even with the macro station. Let MGi 2 Ss and MGj 2 Sb be a seller and a
buyer, respectively. When MGi and MGj want to exchange power, the power loss
function Pij is:
Rij Q2ij
Pij D ; (5.6)
U12
where Rij denotes the resistance of the distribution line between MGi and MGj . U1 .<
U0 / is the transfer voltage between MGi and MGj . When ˛ D 0, (5.6) becomes the
special case of (5.2). Also, Qij D min.Qi ; Qj / where Qi and Qj are given by (5.3).
It means that if the seller (MGi ) cannot meet the demand of the buyer (MGj ), then
the seller only sells Qi to the buyer. Also, after accounting for the negligible yet
U2
existing power loss between the micro grids, MGj will buy at least 2R1ij amount of
energy (due to the power loss between MGi and MGj ) from MGi .
In any coalition S, the total payoff function consists of three parts: the power loss
between the micro grids which can be obtained from (6.13), the power loss caused
by the micro grid selling energy to the macro station, and the power loss caused by
the micro grid buying energy from the macro station. The second and third parts
are given by (5.2) and (7.7), respectively. Hence, the total payoff function of the
coalition S can be expressed as:
X X X
u.S; / D .w1 Pij C w2 Pi0 C w2 Pj0 /; (5.7)
i2Ss ;j2Sb i2Ss j2Sb
where 2 SS denotes the join order of the micro grids, which decide to join the
coalition S, and SS indicates the set of the micro grids’ order in S. w1 and w2 denote
the price of a unit power in the coalition and that in the macro station, respectively.
Pi0 and Pj0 are given by (5.2) and (7.7), respectively. They represent the power
losses during the energy transfer between the corresponding micro grids and the
macro station. Pij indicates the power loss inside the coalition S, between MGi 2 Ss
and MGj 2 Sb which are also expressed by (6.13). Using (5.7), which represents the
total power loss due to the different power transfers for S, we can define the value
function for the micro grids (N ,v) coalition game:
v.S/ D max u.S; / (5.8)

2SS
5.3 Coalition Formation Game
In a game, the players (participating micro grids in our context) have a number
of options that they can choose from [23]. For example, which coalitions to join,
which coalition/micro grids/macro station to buy the power from, etc. Each player
aims to select the best possible choice. The price of electricity from the micro grid
is assumed to be cheaper than that from the macro station during peak hours and
the power loss between the micro grids is supposed to be less than that between the
micro grid and the macro station. So, an efficient strategy is needed to make sure that
the aggregate energy transfers between the coalitions and the macro station is the
minimum or the aggregate energy transfers within the coalitions is the maximum.
This is to maximize the payoffs of the micro grids. Remember from (5.7) that
u.S; / has three parts. Because the power loss between the macro station and
micro grids is much higher than that between the neighboring micro grids forming a
coalition, the first part is much lower than the second part. Also, similar observation
can be made about the third part. Hence, to maximize (5.8), a strategy is required to
5.3 Coalition Formation Game 45
find the coalition with the appropriate micro grids so as to ascertain the minimum
power loss between the coalition and the macro station, and the maximum power
within the coalition. Our adopted strategy is as follows.
• Initialization: order Ss and Sb in accordance with the requests of the seller/buyer
micro grids, i.e. Sb D fb1 ; : : : ; bk g, and calculate the sum of the sets respectively
and find the least one of them. Assume Sb to be the least one. Then, choose
bl 2 Sb as the objective.
• Step 1: depending on the demand, find the appropriate micro grids in Ss or Sb to
form coalition S with objective. This ensures that the power loss of coalition S is
the minimum in the set of micro grids coalitions and the power exchanged in the
groups is the maximum.
• Step 2: If the remainder of Ss is less than that of Sb , select the largest one in Ss as
the objective. Go to step 1 until there is no availability in the sets or one of the
sets becomes empty.
• Step 3: If the remainder of Ss is more than that of Sb , select the largest one in Sb
as the objective. Go to step 1 until there is no availability in the sets or one of sets
becomes empty.
Revisit (5.8), which shows the maximum total utility produced by any S 2 N
with the minimum power loss over the distribution lines. Therefore, comparing with
the non-cooperative case, the sum of utilities of the micro grids in the cooperative
coalition model increases. In other words, the micro grids receive extra profits by
forming the coalition. After making the coalitions, the micro grids now face a new
problem—how to distribute the extra profits appropriately in the coalition? This
is important for stability because inappropriate allocation of the extra profits will
lead to split of the coalition. The “Shapley” value concept from cooperative game
theory can be exploited for this purpose that assigns a unique distribution (among
the micro grids) of a total surplus produced by the coalition of all the micro grids.
In other words, the Shapley value of a micro grid is simply the contribution of that
micro grid to its coalition. If there exists a coalition game .N ; v/, the Shapley value
can be calculated using:
X jSjŠ.n jSj 1/
i .v/ D .v.S [ fig/ v.S//; (5.9)
nŠ
SN nfig
where n denotes the total number of players (i.e., micro grids) and the sum extends
over all the subsets S of N without the ith player. Imagine that the coalition is
constructed with one player at a time; then each player demands her contribution
v.S [ fig/ v.S/ as an appropriate compensation, and then averages over the
possible different permutations in which the coalition may be formed. Because the
payoffs depend on the micro grids’ order in the coalition, the payoffs are expected
to be different in different orders. Therefore, the contribution of a micro grid to the
coalition is actually independent of the order. So, the fraction in (5.9) calculates the
average of the payoffs in all conditions that is considered to be the contribution of
Table 5.1 The Shapley value Order v.S [ f1g/ v.S/

of player 1
Order 1,2,3 v.f1g/ v.;/=0-0=0
Order 1,3,2 v.f1g/ v.f;g/=0-0=0
Order 2,1,3 v.f2; 1g/ v.f1g/=0-0=0
Order 3,1,2 v.f3; 1g/ v.f1g/=0-0=0
Order 2,3,1 v.f2; 3; 1g/ v.f2; 3g/=1-0=1
Order 3,2,1 v.f3; 2; 1g/ v.f3; 2g/=1-0=1
the player to the coalition. Note that the Shapley value has nothing to do with the
cost of the players. For example, consider three players with costs of 10, 20, and 30
units, respectively, and their payoff function:

1 W S D f1; 2; 3g
v.S/ D
0 W otherwise;
The number of orders in this example is 6. Different orders lead to different

payoffs. Table 5.1 shows the Shapley value calculation for the 1st player which is 13 .
In a similar way, we can compute the Shapely values of the 2nd and 3rd players.
It can be seen that all the players’ Shapely values are 13 which implies that their
contributions for the coalition are the same even though their costs are different.
From (5.9), if two micro grids have equal contributions to the coalition, their
Shapley values are the same, although their individual values could be different.
Also, the Shapely value does not depend on the order of the micro grid in
the coalition. Even though the result may look unfair, it indicates the practical
contribution of the micro grids to the coalition.
Using the adopted strategy for objective selection and the concept of Shapely
value for extra profit allocation in a coalition, an algorithm can now be designed
to formulate coalitions of the micro grids. First, we need to introduce a necessary
definition from [22].
Definition. Assume two collections of disjoint coalitions A D fA1 ; : : : ; Ai g and
B D fB1 ; : : : ; Bj g that are formed out of the same players. For one collection A D
fA1 ; : : : ; Ai g, the payoff of a player k in a coalition Ak 2 A is k .A/ D k .Ak / where
k .Ak / is given for coalition Ak . Collection A is preferred over B by Pareto order,
i.e. A F B, if and only if
A F B , fj .A/ j .B/; 8k 2 A; Bg (5.10)
with at least one strict inequality (>) for a player k. The Pareto order implies that a
group of players prefer to join a collection A rather than B, if at least one player can
improve her payoff when the structure is changed from B to A without curtailing
the payoff of any other player.
5.4 Numerical Results 47
To form the coalition, merge and split [24] rules are defined:
Definition Merge. Merge any set of coalitions fS1 ; : : : ; Sl g where f[liD1 Si g F
fS1 ; : : : ; Sl g, therefore, fS1 ; : : : ; Sl g ! f[liD1 Si g.
Definition Split. Split any coalition f[liD1 Si g where ffS1 ; : : : ; Sl g F [liD1 Si g, there-
fore, f[liD1 Si g ! fS1 ; : : : ; Sl g.
From the above merge and split definitions, notice that some micro grids will
join a new coalition or merge with a bigger coalition, respectively, given that at
least one of them can improve payoff without cutting down the payoffs of any other
micro grid or coalition, respectively. On the other hand, a large coalition could split
into several smaller coalitions (or in worst case disappear) if the micro grids find
that they can leave the coalition or merge with a smaller coalition so as to get more
payoff than that the current coalition. So, a merge or split decision by Pareto order
makes sure that all the involved micro grids agree on it.
By exploiting the merge and split operations, we now describe our algorithm in
the following steps.
1. Each micro grid could obtain information of others (e.g., neighboring micro
grids’ position, buying/selling status, available power, etc.) by using the com-
munication infrastructure or communication technology of smart grid (i.e., smart
meters, BAN/NAN GW, control center, etc).
2. The micro grids will generate the energy, receive the user demands, and estimate
whether to buy or sell energy.
3. The forming coalition stage begins. For a specific partition S D fS1 ; : : : Sk g, each
coalition Si 2 S communicates to its neighbors, uses merge and split operation
rules to find the best cooperative partners to form a bigger coalition or leave the
bigger coalition to form a smaller one so as to get more profits (improve payoff).
The coalitions of micro grids calculate their payoffs by employing (5.7) and (5.8),
find that the payoffs of all of them will increase, and this is the Pareto order in
Eq. (6.20), if they can form a coalition.
4. After the merge and split iterations, the system will compose of disjoint
coalitions, and no coalitions may have any further incentive to perform further
merge or split operation.
Upon such convergence, the micro grids within each formed coalition will start
their energy transfer.
5.4 Numerical Results
In this section, simulation results are presented to evaluate the effectiveness of

our adopted game-theoretic coalition formulation approach for the micro grids.
The considered scenario has a power distribution grid topology of 10 10 km2 .
The macro station is at the center of the grid topology while the micro grids are
Fig. 5.2 Comparative result of the conventional methods and the adopted coalition-based power
loss minimization approach
deployed randomly. Reqi is set to a random value varying from 20 to 20 MW

(negative and positive signs mean whether MGi is a buyer or seller, respectively).
R is set to 0.2 /km, ˛ D 0:02 is set to assumption in [25], and U0 and U1 are
set to 50 and 22 kV, respectively, to reflect practical values in existing smart grid
distribution networks [25]. The price of the each of the unit power loss parameters
is set to w1 D 1 and w2 D 3 (Fig. 5.2).
By varying the numbers of the micro grids from 2 to 50, the average power
loss per micro grid is plotted in Fig. 7.3. Our adopted approach is compared with
the non-cooperative case and also a conventional algorithm called NMS which are
referred to as conventional 1 and conventional 2, respectively. When the number
of micro grids increases, the average power loss changes just a little in the non-
cooperative conventional 1 case. On the other hand, in case of conventional 2 and
the adopted approach, the power losses decrease with the growing numbers of
micro grids. When the number of micro grids is 50, the power loss in the adopted
approach reaches up to significant reduction compared to conventional 2 approach.
This superior performance happens because the power losses within the formed
coalitions in the adopted approach are much lower than those between the macro
station and the micro grids. As a result, when most of the micro grids have joined
coalitions, the overall cost of the users decreases, and thus, the adopted approach
outperforms conventional 2 in terms of improvement of the average power loss.
To reflect the money saving of the users, Fig. 5.3 demonstrates the optimal
number of micro grids in different cases. Remember that the optimal number of
micro grids depends on the area of the considered area and the demands of users
in that area. Therefore, with increasing demands, the users require more electricity
from the micro grids to reduce their cost (compared to procuring electricity from
the macro station). For instance, the optimal number changes from 15 to 50 when
the demands changes from 15 to 55 MW in the considered grid area of 10 10 km2 .
Furthermore, because the resistance is a linear increasing function of distance, a
higher resistance means a higher power loss. So, in a larger area, more micro grids
References 49
Fig. 5.3 The optimal number of MGs in the considered topology
want to minimize the higher power loss. Although not shown in the figure, as
the area changes from 100 to 1000 km2 and the demands are 15 MW, the optimal
number of micro grids to form coalitions becomes from 15 to 17.
In this chapter, we adopted a game-theoretic coalition formulation strategy for

distributed micro grids. The strategy allows the micro grids to form coalitions so that
the power loss is minimized when energy is transmitted from a micro grid to other
micro grids or the macro station. The adopted approach also permits the micro grids
to make decisions on whether to form or break the coalitions while maximizing
their utility functions through alleviating the energy transfer power loss. Through
simulation results, the effectiveness of the adopted approach was shown.
References
1. D. Niyato and P. Wang, “Cooperative transmission for meter data collection in smart grid,”
IEEE Communications Magazine, vol. 50, no. 4, pp. 90–97, Apr. 2012.
2. L. Ochoa and G. Harrison, “Minimizing energy losses: Optimal accommodation and smart
operation of renewable distributed generation,” IEEE Transactions on Power Systems, vol. 26,
no. 1, pp. 198–205, Feb. 2011.
3. M. Erol-Kantarci, B. Kantarci, and H. Mouftah, “Cost-aware smart microgrid network design

for a sustainable smart grid,” in 2011 IEEE GLOBECOM Workshops (GC Wkshps), Dec. 2011,
pp. 1178–1182.
4. S. Meliopoulos, G. Cokkinides, R. Huang, E. Farantatos, S. Choi, Y. Lee, and X. Yu,
“Smart grid infrastructure for distribution systems and applications,” in 2011 44th Hawaii
International Conference on System Sciences (HICSS), Kauai, Hawaii, Jan. 2011, pp. 1–11.
5. S. Deilami, A. Masoum, P. Moses, and M. Masoum, “Real-time coordination of plug-in electric
vehicle charging in smart grids to minimize power losses and improve voltage profile,” IEEE
Transactions on Smart Grid, vol. 2, no. 3, pp. 456–467, Sep. 2011.
6. A. Vargas and M. Samper, “Real-time monitoring and economic dispatch of smart distribution
grids: High performance algorithms for DMS applications,” IEEE Transactions on Smart Grid,
vol. 3, no. 2, pp. 866–877, Jun. 2012.
7. A. Costabeber, T. Erseghe, P. Tenti, S. Tomasin, and P. Mattavelli, “Optimization of micro-grid
operation by dynamic grid mapping and token ring control,” in Proceedings of the 2011-14th
European Conference on Power Electronics and Applications (EPE 2011), Birmingham, UK,
Aug. 2011, pp. 1–10.
8. W. Saad, Z. Han, and H. Poor, “Coalitional game theory for cooperative micro-grid distribution
networks,” in 2011 IEEE International Conference on Communications (ICC) Workshops,
Seattle, Washington, USA, Jun. 2011, pp. 1–5.
9. A. Costabeber, P. Tenti, and P. Mattavelli, “Surround control of distributed energy resources
in micro-grids,” in 2010 IEEE International Conference on Sustainable Energy Technologies
(ICSET), Kandy, Sri Lanka, Dec. 2010, pp. 1–6.
10. P. Tenti, A. Costabeber, D. Trombetti, and P. Mattavelli, “Plug amp; play operation of
distributed energy resources in micro-grids,” in 32nd International Telecommunications Energy
Conference (INTELEC), Orlando, Florida, USA, Jun. 2010, pp. 1–6.
11. G. Corso, M. Di Silvestre, M. Ippolito, E. Sanseverino, and G. Zizzo, “Multi-objective
long term optimal dispatch of distributed energy resources in micro-grids,” in 2010 45th
International Universities Power Engineering Conference (UPEC), Cardiff, United Kingdom,
Aug. 2010.
12. M. Kirthiga, S. Daniel, and S. Gurunathan, “A methodology for transforming an existing distri-
bution network into a sustainable autonomous micro-grid,” IEEE Transactions on Sustainable
Energy, vol. 4, no. 1, pp. 31–41, Jan. 2013.
13. Z. Li, C. Wu, J. Chen, Y. Shi, J. Xiong, and A. Wang, “Power distribution network reconfigu-
ration for bounded transient power loss,” in 2012 IEEE Innovative Smart Grid Technologies -
Asia (ISGT Asia), Tianjin, China, May 2012, pp. 1–5.
14. J. Aguero, “Improving the efficiency of power distribution systems through technical and
non-technical losses reduction,” in Transmission and Distribution Conference and Exposition
(T&D), 2012 IEEE PES, Dallas, Texas, USA, May 2012, pp. 1–8.
15. L. McDonald, R. Storry, A. Kane, F. McNicol, G. Ault, I. Kockar, S. McArthur, E. Davidson,
and M. Dolan, “Minimisation of distribution network real power losses using a smart grid active
network management system,” in 2010 45th International Universities Power Engineering
Conference (UPEC), Cardiff, United Kingdom, Aug. 2010.
16. K. Nakayama and N. Shinomiya, “Distributed control based on tie-set graph theory for
smart grid networks,” in 2010 International Congress on Ultra Modern Telecommunications
(ICUMT) and Control Systems and Workshops, Moscow, Russia, Oct. 2010, pp. 957–964.
17. T.-I. Choi and Y.-K. Cho, “International business offering related to innovative smart technolo-
gies,” in 2012 IEEE PES Innovative Smart Grid Technologies (ISGT), Washington D.C., USA,
Jan. 2012, pp. 1–8.
18. M. Ismail, M.-Y. Chen, and X. Li, “Optimal planning for power distribution network with
distributed generation in Zanzibar island,” in 2011 International Conference on Electrical and
Control Engineering (ICECE), Yichang, China, Sep. 2011, pp. 266–269.
19. N. Katie, V. Marijanovic, and I. Stefani, “Profitability of smart grid solution application in
distribution network,” in 7th Mediterranean Conference and Exhibition on Power Generation,
Transmission, Distribution and Energy Conversion (MedPower 2010), Agia Napa, Cyprus,
Nov. 2010, pp. 1–6.
References 51
20. Z. M. Fadlullah, Y. Nozaki, A. Takeuchi, and N. Kato, “A survey of game theoretic approaches
in smart grid,” in 2011 International Conference on Wireless Communications and Signal
Processing (WCSP), Nanjing, Jiangsu, China, Nov. 2011, pp. 1–4.
21. C. Zhang, W. Wu, H. Huang, and H. Yu, “Fair energy resource allocation by minority game
algorithm for smart buildings,” in Design, Automation Test in Europe Conference Exhibition
(DATE), 2012, Dresden, Germany, Mar. 2012, pp. 63–68.
22. H. Li and W. Zhang, “QoS routing in smart grid,” in 2010 IEEE Global Communications
Conference (GLOBECOM 2010), Miami, Florida, USA, Dec. 2010.
24. K. R. Apt and A. Witzel, “A generic approach to coalition formation,” Computing Research
Repository - CoRR, vol. abs/0709.0435, 2007. [Online]. Available: http://arxiv.org/abs/0709.
0435
25. J. Machowski, J. W. Bialek, and J. R. Bumby, Power system dynamics : stability and
control. Chichester, U.K. Wiley, 2008, rev. ed. of: Power system dynamics and stability / Jan
Machowski, Janusz W. Bialek, James R. Bumby. 1997. [Online]. Available: http://opac.inria.
fr/record=b1135564
Chapter 6
On Optimally Reducing Power Loss
in Micro-Grids with Power Storage Devices
6.1 Background
Chapter 5 described a strategy to allow the micro grids to make coalitions and
exchange power with other micro grids and/or the macro station. In this chapter,
we are going to carry on from that strategy and enhance the model further by
considering micro grids to use power storage devices. Recently, the micro grid
developers and operators are exhibiting a great deal of interest in using lithium-
ion batteries and flow batteries. The battery technologies have become quite mature
and they are currently capable enough to provide exceptional renewable power
integration in the micro grids based energy systems [1, 2]. In [3], it was demon-
strated how the storage devices can supplement energy generation to consumption
to achieve a balance between energy demand and supply within the micro grid.
Necessity of optimal control of the power storage devices of the micro grid was also
indicated in [3]. A new photovoltaic power generation and load power consumption
prediction algorithm was designed in [4] which was specifically designed for a
residential storage controller. However, these existing works usually focused on a
single micro grid and did not consider the power losses impacting the entire system
including the macro station and multiple micro grids. Instead, they had a localized
approach such as how to reduce the power loss within a given micro grid, how to
charge and discharge the storage device periodically, and so forth. In this chapter, we
adopt a different approach by focusing on a scalable total power loss minimization
approach across the entire smart grid.
6.2 System Model
Our considered power storage equipped micro grids model is shown in Fig. 6.1.
As shown in the figure, at the center of the model, there is the macro station, which is
the utility operator of a given region. The macro station could be one of the retailers

54 6 On Optimally Reducing Power Loss in Micro-Grids with Power Storage Devices
Fig. 6.1 Our considered micro grids system model and the various power losses occurring in the
system
or even the power wholesaler. The macro station has its own customers such as
residential users, factories, etc. Additionally, it is connected to the micro grids which
are located in its area of operation. There are power distribution lines between the
macro station and these distributed micro grids. Thus, power can be exchanged
between the macro station and the micro grids. In normal mode of operation, the
macro station sells power to the micro grids. However, if needed (particularly in
peak hours), the macro station may also buy surplus power of the micro grids. The
micro-grids are the secondary power generation sources in the considered area.
They can use diesel and also integrate with simple or hybrid renewable energy
sources like solar, wind power, small hydro, geothermal, waste-to-energy, hydraulic,
and combined heat and power systems. The micro grids have their own customers
(which may overlap with the customer base of the macro station) in their locality.
The power produced in a micro grid is transmitted to its users according to their
demand. The micro grids are assumed to have power storage devices like lithium-
ion batteries, rechargeable car batteries, flywheel energy storage, etc. The storage
devices get charged up during the off-peak time and discharge during the peak hours
to meet the energy demand of the customers. In addition to their initial deployment
cost, the power storage devices suffer from storage power loss during charging and
discharging. Therefore, it is important to consider the storage power loss of the
micro grid along with power generation and transmission losses.
Assume N is the set of micro-grids, and N , jN j. Also assume that during tth
hour, the total user-demand of the ith micro grid MGi (i 2 N / is Di .t/, and MGi can
supply power Ui .t/ by discharging its charged power storage devices. Our model
takes into account four types of power losses—generation power loss (see [5]),
6.2 System Model 55
storage power loss [6], transmission power loss, and conversion power loss [7]. Even
though other types of power losses are possible, they are usually negligible and our
model does not include them. Assume for power m, the associated power losses
are storage power loss PLS.m/, generation power loss PLG.m/, and transmission
power loss PLT.m/. The general observation is that PLS.m/ PLG.m/ PLT.m/.
Let the generation power be denoted by Gi .t/. Consider Gi .t/ and Di .t/ as random
numbers in the real smart grid networks from the work in [8]. To account for the
generation power loss, the supply power Ui .t/ can be expressed as:
Ui .t/ D .1 /Gi .t/; (6.1)
where is a fraction of the generation power loss. Based on (6.1), the generation
power loss during tth hour denoted by PLGi .t/ is given by:
PLGi .t/ D Gi .t/ (6.2)
The generation power of MGi has a maximum and a minimum [8]. This means
that its power generation Gi .t/ has a range.
Gmin Gi .t/ Gmax ; (6.3)
where Gmin and Gmax are the minimum and maximum of the generated power
by MGi .
Many factors influence the quantity of power in the power storage devices like
power self-discharge, the cost of power charging, etc. To quantify the storage power
loss PLSi .t/ to account for those factors, a function of charged power Si .t 1/ in the
time interval .t 1/ is adopted:
PLSi .t/ D ˇSi .t 1/; (6.4)
where Si .t 1/ denotes the charged power of the power storage devices in time
(t 1), and ˇ is a fraction of the storage power loss. Also, the power storage device
has a maximum capacity. To avoid over-discharge, the storage power Si .t/ needs to
satisfy the following condition:
Smin Si .t/ Smax ; (6.5)
where Smin and Smax are the minimum and maximum of the stored power, respec-
tively. When the supply power is more than the users’ power demand, some power
will be charged in the power storage devices. Otherwise, the power storage devices
will discharge to satisfy the demand. The storage power is calculated as:
Si .t/ D .1 ˇ/Si .t 1/ C Si .t/; (6.6)

where Si .t/ denotes the operation of charging or discharging during tth hour. Si .t/
needs to satisfy the following.

0 W discharge
Si .t/ D (6.7)
>0 W charge;
When the storage power and supply power fall below the user-demand, or when
the micro grid has surplus power (to sell), the micro grid needs to exchange power
with the macro station and/or with its neighboring micro grids. Let Wi .t/ be the
power which MGi wants to exchange with other micro grids or the macro station:
Wi .t/ D Si .t/ C Ui .t/ Di .t/; (6.8)
where Ui .t/ and Si .t/ are given by (6.1) and (6.7), respectively. If .Wi < 0/, the
supply power and storage power of MGi is below its users’ aggregate demand.
So, MGi needs to buy power from other micro grids or the macro station.
If .Wi > 0/, MGi has surplus power which it can sell other micro grids in need
of energy, or even to the macro station.
Now let us consider the other types of power losses. When power transmission
happens between the macro station and micro grids, transmission power loss and
conversion power loss happen [7]. In addition, generation power loss needs to be
considered. Because to comply MGi ’s request for additional energy, other micro
grids or the macro station need to generate the power which will be associated with
generation power loss. Thus, the total power loss between MGi and the macro station
can be expressed as:
PLi0 .t/ D PLTi0 .t/ C PLCi0 .t/ C PLGi0 .t/ (6.9)
Bi .t/2 Ri0
According to Chap. 5, PLTi0 .t/ D U02
and PLCi0 .t/ D ˛Bi .t/. There-
fore, (6.9) can be re-written as:
Bi .t/2 Ri0
PLi0 .t/ D C .˛ C /Bi .t/; (6.10)
U02
where Ri0 is the power line resistance between MGi and the macro-station, U0
denotes the transmission voltage between MGi and the macro-station, ˛ is a fraction
of the conversion power loss, and Bi .t/ indicates the power that MGi wants to
exchange with the macro station. Bi .t/ could be any one of the following:
8
< Wi .t/ W Wi .t/ > 0
Bi .t/ D Li .t/ W Wi .t/ < 0 (6.11)
:
0 W Reqi D 0;
6.2 System Model 57
where Li .t/ is the total amount of power which needs to be generated to ensure that
MGi can obtain the power required to meet its demand Wi . Li is the solution of
following quadratic equation:
Ri0 Li2 .t/

Li .t/ D PLi0 .t/ C jWi .t/j D C .˛ C /Li .t/ Wi .t/: (6.12)
U02
For a given Li .t/, three possible solution sets of (6.12) exist (zero, one, and two
solutions). To minimize the value of Li .t/, we choose the smaller one when (6.12)
has two roots. If (6.12) has no solution, assume that the root is the same as (6.12)
.1˛/U02
with a single root, which is Li .t/ D 2Ri0
.
Similar with (6.10), the power loss between the micro grids can be estimated.
Since the voltage among the micro grids are medium level voltage, the conversion
power loss can be somewhat neglected. Therefore, if MGi and MGj want to exchange
power, the power loss function PLij can be expressed as follows.
Rij B2ij .t/

PLij .t/ D PLTij .t/ C PLGij .t/ D C Bij .t/; (6.13)
U12
where Rij is the resistance of the distribution line between MGi and MGj . U1 is the
transmission voltage between MGi and MGj . Then, Bij can be expressed as:

Bi .t/ W jBi .t/j jBj .t/j
Bij .t/ D (6.14)
Bj .t/ W otherwise;
where Bi and Bj are given by (6.11). It implies that if the seller MGi cannot meet
the demand of the buyer MGj , then MGi only sells Bi to MGj . Also, MGj will buy at
U2
least 2R1ij amount of power from MGi due to the power loss between MGi and MGj .
Finally, the total power losses PTAi .t/ of MGi can be given by:
X
PLAi .t/ D PLGi .t/ C PLSi .t/ C PLi0 .t/ C PLij .t/ (6.15)
j
Therefore, the total payoff function of the coalition C is as follows.

X
u.C; …; t/ D PLAi .t/ (6.16)
i
where … 2 CC denotes the join order of the micro grids, which decide to join
the coalition C, and CC is the set of the micro-grids’ order in C. If a micro grid
wants to maximize its payoff, it has to minimize the total power losses. This means
that the micro grid wants the maximum of (6.16). Hence, by using (6.16), which
indicates the total power losses incurred by the different power transmissions for C,
it is possible to define the value function for the micro grids (N ,v) coalition game:
v.C; t/ D max u.C; …; t/: (6.17)

…2CC
Consider the micro grids as “players” as part of a cooperative game. When the
strategies are given, these players can adopt the best choice. The strategy of the
players and the coalition game which they play are described next.
6.3 Coalition Formulation Strategy for Micro Grids

with Power Storage Devices
Based on the earlier described system model, the micro grids could have different
strategies or choices to make coalitions. Each micro grid, however, confronts the
challenge of how to find the best strategy from a set of all possible strategies. For
example, if a micro grid wants to make a coalition with other micro grids, it needs to
know which micro grids are the most appropriate ones. A game-theoretic approach
is adopted to overcome this issue [9].
The total power losses inside a micro grid comprise generation power loss (from
renewable sources of the micro grid) and storage power loss, respectively. Consider
this as the “inner power loss.” On the other hand, the “external power loss” happens
when power is exchanged among the micro grids or between a micro grid and the
macro station. Hence, the external power loss comprises generation, transmission,
and conversion power losses. From the system model, we understand that the inner
power loss is less than the external power loss in the same condition. Additionally,
the storage power loss is lower than the generation power loss inside the micro
grids. Hence, when a micro grid’s storage power, its aggregate user-demand, and the
generated power are given, the minimum of the inner power loss, the operation of
charging and discharging Si , and the required power Wi can be estimated. The next
challenge is how to form the coalition to optimally reduce the total power losses.
If MGi wants to join the coalition C, it will increase the payoff of the coali-
tion. The “marginal payoff function” per unit power of MGi for the coalition is
expressed as:
u.C C fig; … ; t/ u.C; …; t/

MPF.i; C/ D ; (6.18)
BiC
where C C fig denotes the new set including set C and MGi . … and … are the join
orders for the new and old sets, respectively. BiC is the power which MGi wants to
exchange with the coalition C. This means extra profits per unit power when MGi
joins the coalition C. Because different micro grids will lead to different MPFs,
based on (7.10), the micro grids can make the best decisions to form coalition.
When the coalitions are formulated, because power loss among the micro grids is
less than that between the macro station and the micro grids, there are extra payoffs.
In order to distribute these extra payoffs, the “Shapley value” [10] is adopted (similar
6.3 Coalition Formulation Strategy for Micro Grids with Power Storage Devices 59
to its use in Chap. 5). When there is a coalition game .N ; v/, the Shapley value of
player i (i.e., MGi ) can be calculated using:
X jCjŠ.N jCj 1/
i .v/ D .v.C [ fig/ v.C//; (6.19)
NŠ
CN nfig
where N is the number of players and the sum extends over all the subsets C of
N without the ith player. v.C/ is given by (6.17). For the micro grids, it is the
contribution of forming the coalition. Therefore, the extra payoffs will be distributed
based on (6.19).
The set of micro grids can be divided into two disjoint subsets (set of buyers
Cb and set of sellers Cs ) depending on the value of Wi . Cb consists of the micro
grids MGp .p 2 jCb j/ with (Wp < 0), and Cs is composed of the micro grids
MGq .q 2 jCs j/ with (Wq > 0). If (Wi D 0), then MGi can be considered as either
a buyer or a seller. To maximize the payoffs of coalitions, each micro grid finds its
appropriate partners so as to maximize (6.17) using the following steps.
• Initialization: At first, Si .t/; Di .t/ and Gi .t/ are given for the tth hour. The oper-
ation Si .t/ and the demand Wi .t/ are evaluated. The set N is divided into Cs
and Cb . Cs and Cb are sorted in descending order according to the requests of the
selling or buying micro grids, i.e., Cs D fs1 ; : : : ; sk g. The sum of sets is calculated
and the lesser one of the two sets is determined.
For ease of explanation, assume that Cs is the lesser one. Then, select sl 2 Cs
as the objective.
• Step 1: Based on (7.10), find the appropriate micro grids in Cs or Cb to form
coalition C with an objective that the profit of coalition C is the maximum. This
step indicates that the power loss of micro grids in coalition C is less than that
between the macro station and the micro grids belonging to C.
• Step 2: If the remainder of Cs is less than that of Cb , choose the biggest one in
Cs . Go to step 1 until there is no availability in the sets or one of the sets becomes
empty.
• Step 3: If the remainder of Cs is more than that of Cb , choose the biggest one
in Cb . Go to step 1, until there is no availability in the sets or one of the sets
becomes empty.
• Step 4: Calculate the “Shapley value” of the micro grids, and distribute the extra
payoffs.
Based upon the aforementioned steps for objective selection and the concept of
Shapely value for extra profit distribution in a coalition, we adopt an algorithm to
formulate distributed coalitions of micro grids in the remainder of this section.
Before the algorithm can be formally described, let us introduce an important
definition from [8].
Definition. Consider two collections of disjoint coalitions A D fA1 ; : : : ; Ai g and
B D fB1 ; : : : ; Bj g that are formed out of the same players. For one collection A D
fA1 ; : : : ; Ai g, the payoff of a player k in a coalition Ak 2 A is k .A/ D k .Ak / where
k .Ak / is given by (6.17) for coalition Ak . Collection A is preferred to B by Pareto

order, i.e. A F B, if and only if
A F B , fk .A/ k .B/; 8k 2 A; Bg; (6.20)
with at least one strict inequality (>) for a player k.

The Pareto order signifies that a group of players prefer to join a collection A
instead of B, if at least one player is able to improve its payoff when the structure is
changed from B to A without reducing the payoff of any other.
To form the coalition, two distributed rules are needed: merge and split [11]
defined as follows:
Definition Merge. Merge any set of coalitions fS1 ; : : : ; Sl g where f[liD1 Si g F
fS1 ; : : : ; Sl g, hence,fS1 ; : : : ; Sl g ! f[liD1 Si g.
Definition Split. Split any coalition f[liD1 Si g where ffS1 ; : : : ; Sl g F [liD1 Si g,
hence,f[liD1 Si g ! fS1 ; : : : ; Sl g.
The merge and split definitions help micro grids to maximize their payoffs, and
find the proper micro grids to form coalitions. Since the micro grids act as players
of a cooperative game, we adopt a coalition formation algorithmic approach for
micro grids with power storage devices by exploiting the merge and split operations.
First, in our adopted approach, each micro grid is able to share information (e.g., its
position, neighboring micro-grids, etc.) with the others using the communication
infrastructure or communication technology. Second, the micro grids generate
the power, get the demands of the users, charge/discharge power into/from their
storage devices, and take a decision whether to buy or sell the power. Third, the
coalition formulation stage begins in which the merge or split processes occur as
follows. Given a partition C D fC1 ; : : : Ck g, each micro grid in each coalition
Ci 2 C communicates with its neighboring micro grids. Using these negotiations,
the coalitions exchange the information with others that include the numbers of
coalitions, their capabilities, etc. The information helps the coalitions to find the
best cooperative partners to make a larger coalition or leave the larger coalition to
create some smaller ones to obtain more profits (payoffs). The coalitions or micro-
grids calculate their payoffs using (6.16) and (6.17). When they find that all their
payoffs will increase by merging into a particular coalition, they reach the Pareto
order as shown in Eq. (6.20).
6.3.1 Characterization of the Game
Characterization of the adopted game involves proving its stability, convergence,

and optimality.
Definition. A coalition C: = fC1 ; : : : ; Ck g is Dhp -stable if the following two

conditions are satisfied.
i 2 f1; : : : ; kg and for each partition fP1 ; : : : ; Pl g of the coalition Ci :
(a) for each P
v.Ci / ljD1 v.Pj /. P
(b) for each set T f1; : : : ; kg: i2T v.Ci / v.[i2T Ci / [12].
Lemma 6.3.1. The coalition formed by the proposed algorithm is Dhp -stable [7].
Lemma 6.3.2. The considered (N ,v) micro grids coalition game converges to the
Pareto optimal Dhp -stable partition, if such a partition exists. Otherwise, the final
partition is merge-and-split proof [7].
Lemma 6.3.3. The total power loss incurred in the proposed scheme is minimum.
Proof (Reduction to Absurdity). Assume that the total power loss is not minimized.
If the adopted approach can help the micro grids to obtain demand W and
corresponding total power loss PLA is not minimum, there exists a demand W
which is different from W and the corresponding total power loss PLA is minimum.
Then, PLA < PLA , if W > W . It is because that the inner power loss is less
than the external power loss under the same condition. It means that the use of
storage power or generation power is better than purchased power from other micro
grid(s)/the macro station. Therefore, PLA is not minimum. If W < W , it means
that there is power surplus while the demand is not satisfied. This condition is not
reasonable since it arises contradiction. Hence, the total power loss in the adopted
approach is minimum.
Theorem 6.3.4. There exists a global optimal result for each micro grid using the
adopted approach.
Proof. From Lemmas 6.3.1 to 6.3.3, we know that the model results in a global
optimal result.
From Theorem 6.3.4, the Pareto optimal is the only stable situation. So, the merge
and split operations help the micro grids to maximize their utilities (i.e., minimize
the total power loss) until the Dhp -stable situation occurs. In such a situation, no
micro grid can decrease its total power losses without increasing the total power
losses of the other micro grids.
Some comparative results are included to show the effectiveness of our adopted
approach compared with a conventional non-cooperative method [4]. In the con-
ventional method, the micro grids only exchange power with the macro station and
they cannot exchange power with the other micro grids. Our considered scenario
considers a power distribution grid topology, area of which is 10 10 km2 . The
macro station is placed at the center of the grid while the micro grids are arbitrarily
deployed in the topology. The resistance between the micro grids is the same as that
between the macro station and any micro grid, and its value is set to R D 0:2 /km.
The fraction of power transmission ˛, the fraction of power generation , and
the fraction of power storage ˇ are set to 0.02, 0.05, 0.01, respectively based
on [5, 6, 13]. The power demand Di of MGi is derived from a uniform distribution
between 10 and 742 MW based on [8]. The power generation Gi is obtained using a
uniform distribution between 10 and 326 MW. Suppose that the capacity of power
storage device is 200 MW, and the minimum storage power is 10 MW. U0 and U1
are set to 50 and 22 kV, respectively, based on [13]. The prices of a unit power
loss are set as w1 D 1 and w2 D 3 [7]. In the adopted approach, the users
send the information to the corresponding micro grids, and the micro grids also
exchange the information to other micro grids or the macro station if necessary.
Assume that the micro grids can communicate with the macro station through an
optical backbone network, capacity of which is 100 Mbps. For simplicity, each
micro grid is assigned to satisfy the demand of 100 users. The packet size for user
to micro grid communication is set to 102 bytes [14] while that for inter-micro grid
communication is considered to be 112 bytes.
Figure 6.2 shows the average power loss per micro grid for different numbers of
the micro-grids ranging from 5 to 50. In the conventional method, the power loss
per micro grid does not improve since the micro grids only obtain power from the
macro station. On the other hand, in case of the adopted approach, the average power
loss is significantly improved with the increasing numbers of the micro grids. This
happens because the coalitions are formed by the micro grids with the objective of
optimally alleviating the power loss. When the micro-grids could successfully make
the coalitions in the adopted approach, they could exchange power with other micro
grids instead of the macro station leading to the reduction of the average power loss.
Figure 6.3 demonstrates that the micro grids want to purchase the power from the
macro station for N D 20 micro grids in the conventional method as well as in the
adopted approach. Suppose that the peak period of the day is from 12 p.m. to 9 p.m.
Fig. 6.2 Comparative average power loss per micro grid achieved in the conventional method and
the adopted approach
6.5 Concluding Remarks 63
Fig. 6.3 Comparison of the power load on the macro station in the conventional method and the
adopted approach
Also suppose that the situations of the micro grids are remain fixed since their initial
random deployment in the grid. Even though in both schemes, the micro grids have
the power storage devices that they could charge in off-peak time and discharge in
peak time, notice that compared with the conventional method, the result achieved
(i.e., the burden in terms of the power load inflicted upon the macro station) by the
adopted game-theoretic approach is lower. This happens because the micro grids
in the adopted approach can buy power from their neighboring micro grids instead
of the distant macro station. Thus, both the users and the macro station can obtain
benefits from forming coalitions through the adopted approach.
In this chapter, a game-theoretic coalition formulation method for micro grids with
power storage devices was discussed. Our adopted approach allows the micro grids
to take a decision on whether to charge or discharge their power storage devices, and
to find other micro grids (i.e., appropriate neighbors) to build coalitions to optimally
minimize the total power losses. The adopted approach also allows the micro grids
to make decisions on whether to form or split the coalitions while alleviating the
power losses due to power generation, transmission, and storage. In addition, it was
shown that the adopted approach offers a stable, convergent, and optimal solution.
Furthermore, the effectiveness of the adopted approach was demonstrated compared
to a conventional method.
References
1. S. Minnihan, “Microgrids focus on security, creating needs for UPS storage,” accessed Aug.
2015. [Online]. Available: http://www.smartgridnews.com/story/microgrids-focus-security-
creating-needs-ups-storage/2013-05-29
2. C. Hill, M. Such, D. Chen, J. Gonzalez, and W. Grady, “Battery energy storage for enabling
integration of distributed solar power generation,” IEEE Transactions on Smart Grid, vol. 3,
no. 2, pp. 850–857, Jun. 2012.
3. C. Ahn and H. Peng, “Decentralized voltage control to minimize distribution power loss of
microgrids,” IEEE Transactions on Smart Grid, vol. 4, no. 3, pp. 1297–1304, Sep. 2013.
4. Y. Wang, X. Lin, and M. Pedram, “Accurate component model based optimal control for energy
storage systems in households with photovoltaic modules,” in 2013 IEEEGreen Technologies
Conference, Denver, CO, USA, Apr. 2013, pp. 28–34.
5. E. Bendict, T. Collins, D. Gotham, S. Hoffman, and D. Karipdes, “Losses in Electric
Power Systems,” accessed Aug. 2015. [Online]. Available: http://docs.lib.purdue.edu/cgi/
viewcontent.cgi?article=1270&context=ecetr
6. IEA-ETSAP and IRENA Technology Policy Brief E18, “Electricity storage,” accessed Aug.
2015. [Online]. Available: https://www.irena.org/DocumentDownloads/Publications/IRENA-
ETSAP%20Tech%20Brief%20E18%20Electricity-Storage.pdf
8. Z. Li, C. Wu, J. Chen, Y. Shi, J. Xiong, and A. Wang, “Power distribution network reconfigu-
ration for bounded transient power loss,” in 2012 IEEE Innovative Smart Grid Technologies -
Asia (ISGT Asia), Tianjin, China, May 2012, pp. 1–5.
vol. 32, no. 7, pp. 1361–1370, Jul. 2014.
10. L. S. Shapley, “A value for n-person games,” Contributions to the theory of games, vol. 2,
pp. 307–317, 1953.
11. K. R. Apt and A. Witzel, “A generic approach to coalition formation,” Computing Research
Repository - CoRR, vol. abs/0709.0435, 2007. [Online]. Available: http://arxiv.org/abs/0709.
0435
12. K. R. Apt and T. Radzik, “Stable partitions in coalitional games,” Computing Research
Repository - CoRR, vol. abs/cs/0605132, 2006. [Online]. Available: http://arxiv.org/abs/cs/
0605132
13. J. Machowski, J. W. Bialek, and J. R. Bumby, Power system dynamics : stability and
control. Chichester, U.K. Wiley, 2008, rev. ed. of: Power system dynamics and stability / Jan
Machowski, Janusz W. Bialek, James R. Bumby. 1997. [Online]. Available: http://opac.inria.
fr/record=b1135564
vol. 2, no. 4, pp. 675–685, Dec. 2011.
Chapter 7
A Distributed Paradigm for Power Loss
Reduction in Micro-Grids
7.1 Background
In the earlier two chapters, we showed that in order to enhance the efficiency of
the smart grid in terms of power loss reduction, the micro grids can exchange
power with other micro grids instead of the macro station. If the micro grids
have more than one neighbor, which neighbor can be chosen as the most appropriate
partner for building the power loss reducing coalition is a key challenge [1, 2].
The coalition formation approaches in the earlier chapters are actually subject to
central decision making. In other words, the micro grids have to take their problem
up to the control center of the macro station which evaluates all the parameters
and then makes optimal decisions pertaining to micro grids’ coalition formation.
Then, the control center has to send the decision back to the micro grids. This
means that the micro grids have to go back and forth with the control center since
it is the only entity which has the global information (e.g., which micro grids
have energy shortage and by how much, which micro grids have surplus power
and how much they can sell in the next hour, etc.). In this chapter, we discuss a
distributed decision making approach to suit the physically distributed micro grids
to allow them to make autonomous decisions regarding energy transfer and coalition
formation. The distributed decision making may not be optimal as the centralized
version. But it may save precious time for the micro grids to take prompt decisions
rather than relying on the centralized entity to make decisions on their behalf.
The adopted approach is essentially different from other existing decentralized
algorithms like [3, 4] in terms of its objective to minimize the power losses.

66 7 A Distributed Paradigm for Power Loss Reduction in Micro-Grids
Fig. 7.1 The considered system model
7.2 System Model and Problem Formulation
Our proposed system model is described with the aid of Fig. 7.1. Consider that
there are three layers in this model. The top layer consists of the macro station,
which is typically the regional utility operator. The macro station can also be the
energy wholesaler or the regional retailers. A single macro station is shown in the
figure, but multiple macro stations are also possible on the top layer. The macro
station is able to exchange power with the secondary power sources on the second
layer. These secondary power sources are the micro grids with renewable/hybrid
power generation sources. For simplicity, suppose that the macro station has enough
power to satisfy the demands of the micro grids and receive the surplus power
from the micro grids. Following Chap. 6, each micro grid is physically linked to
the macro station. However, compared with the macro station, the micro grids could
be deployed much nearer to the customers, who are in the third and bottom layer of
the considered model. Therefore, the customers could be linked to the micro-grids
directly. The micro grids supply power to the customers to satisfy their demands.
Additionally, the micro grids can exchange power with their neighbors or even with
the macro-station, when supply and demand are unequal. Because they just know
the location of their neighbors, a distributed algorithm could be adopted. A purely
distributed algorithm could help the micro grids to find proper partners so as to
minimize the total power loss of the smart grid. The smart meters are installed at
the customer-premise through which the customers can send the demands to the
micro-grids.
Let N indicate the set of the micro-grids and N D jN j. In the given time period
(e.g., one second), for the ith micro grid denoted by MGi , we define a real function
Di .t/ as the current remaining power of MGi and it can be expressed as:
Di .t/ D Gi .t/ Wi .t/: (7.1)

7.2 System Model and Problem Formulation 67
Here, Gi .t/ and Wi .t/ are the generation power and the aggregate user-demand
of MGi , respectively. It means that MGi wants to obtain power to satisfy its demand
.Di .t/ < 0/. MGi has a power surplus to sell .Di .t/ > 0/, or its supply equals
its demand .Di .t/ D 0/. The micro grids can be divided into two types, namely
“exporters” and “importers” (synonymous to sellers and buyers, respectively, as
described in Chaps. 5 and 6). The exporters have surplus to sell while the importers
need additional amount of power to satisfy the customer-demands. If the current
remaining power of MGi is zero, MGi is considered to be either an exporter or an
importer, and it cannot have any impact. Additionally, the demand of customers
Wi .t/ and production power Gi .t/ are always considered as random numbers in the
real smart grid networks [5]. Hence, the value of Di .t/ is considered as a random
number with a certain observed distribution.
When Di .t/ ¤ 0, MGi will exchange power with other micro grids or the macro
station. It will then lead to power loss. For simplicity, two types of power losses are
considered in our model even though other types of losses can be accommodated
as mentioned in Chap. 6. The first one is power loss due to transmission (PLT) [5].
To alleviate the power loss of long-distance transmission, the voltage between the
primary power station to the macro station is high (e.g., 50 kV [5]) while that
between the micro grid and the macro station is medium (e.g., 22 kV [5]). Therefore,
high voltage needs to be converted into medium level. This process will cause the
power loss due to conversion (PLC).
First, the power loss between two micro-grids are considered. Based on [5], if
MGi transmits power to MGj , the power loss function Pij .t/ can be expressed as:
Rij Q2ij .t/

Pij .t/ D ; (7.2)
U12
where Rij is the resistance of the distribution line between MGi and MGj . U1 is the
transfer voltage between MGi and MGj , and it is less than U0 . In this model, we do
not consider the power loss of transforming voltage levels between MGi and MGj .
Also, Qij .t/ is defined as:
( Q2ij .t/Rij
Dj .t/WjDi .t/j > jDj .t/j
Qij .t/ D U12 (7.3)
Di .t/ Wotherwise:
If MGi sells power to MGj , the current remaining power Di .t/ will be updated as:
Di .t/ D Di .t/ Qij .t/: (7.4)
If MGi buys power from MGj , Di .t/ is updated as follows:
Di .t/ D minfDi .t/ C Qij .t/ PLij .t/; 0g: (7.5)

After exchanging power with the other micro grids, if Di ¤ 0, MGi will exchange
power with the macro station. In this process, consider PLT and PLC ([5]). If MGi
wants to sell Di .t/ to the macro station .Di .t/ > 0/ or buy Di .t/ from the macro-
station .Di .t/ < 0/, we are able to express the power loss PL0i .t/ as follows:
R0i Q20i .t/

PL0i .t/ D C ˛Q0i .t/; (7.6)
U02
where R0i is the distribution line resistance between the macro-station and MGi ,
U0 is the voltage of power transfer between MGi and the macro station, and ˛ is a
fraction of power loss caused by voltage conversion. For simplicity, ˛ is treated as
a constant [5]. Q0i .t/ is the power that MGi wants to buy or sell.
The value of Q0i .t/ is any of the following.
(
Q20i .t/Rij
U02
C ˛Q0i .t/ Di .t/ W Di .t/ < 0
Q0i .t/ D (7.7)
Di .t/ W otherwise:
Based on (7.2) and (7.6), in a given time-slot t, the total power loss PLAi .t/ of
MGi is given by:
X PLij .t/
PLAi .t/ D PL0i .t/ C : (7.8)
j
2
If MGi exchanges power with MGj , the power loss PLij .t/ should not be
calculated twice. Therefore, PLAi .t/ includes half of PLij .t/.
Our target in this chapter is to allow the micro grids to make distributed decisions
so as to minimize the total power loss. Hence, the objective function is
X
Minimize PLAi .t/
i
s:t: Di .t/ Gi .t/ C i .t/ 8i 2 N ; (7.9)

P
where i .t/ D sign.Di .t//Q0i .t/ PL0i .t/ C j .sign.Di .t//Qij .t/ PLij .t//,
sign.Di .t// D 1 if Di .t/ < 0, and sign.Di .t// D 1 otherwise. Hence, our condition
is that the demand at each micro grid does not exceed the sum of the amount of the
remaining produced power and the power it exchanged with other micro grids and
the macro station.
7.3 The Distributed Solution Approach for the Micro Grids 69
7.3 The Distributed Solution Approach

for the Micro Grids
Based on the system model in Sect. 7.2, the total power loss of the smart grid could
be calculated. However, unlike the centralized algorithms considered in Chaps. 5
and 6, the micro grids do not need to acquire the total information (i.e., the global
snapshot of all other micro grids and all the parameters). Assume that the micro
grids only have knowledge of the locations of one-hop neighbor(s), and are able to
exchange power with it/them.
At the beginning of time-slot t, MGi estimates Wi .t/ from the aggregate demand
of its customers. To satisfy Wi .t/, MGi generates power Gi .t/. If the current
remaining power of MGi is Di .t/ ¤ 0, MGi will exchange power with its
neighbor(s). MGi will exchange information of Di .t/ with its neighbors. Based on
the remaining power of neighbors, MGi generates a set of Potential Exchange power
Neighbors (PEN). This set implies that if MGj 2 PEN, MGi has the opportunity to
exchange power with MGj . If PEN of MGi has more than one element, it needs to
select a proper neighbor to exchange power so as to minimize the total power loss.
The “Reducing power loss per Unit exchanged Power” (RUP) of MGi and MGj for
the micro grid pair can deal with this problem. If MGi exchanges power with MGj ,
the function can be expressed as:
PL0i .t/ C PL0j .t/ PLij .t/

RUP.Qij .t// D : (7.10)
jQij .t/j
The above function represents potential extra payoffs (reducing power loss) per
unit exchange power, if MGi joins the coalition. PL0i .t/ and PL0j .t/ indicate power
losses if the same power Qij .t/ was exchanged with the macro station by both micro
grids in the current coalition. Merging them could replace these two by power
exchange between them, with power loss PLij .t/. Higher values of RUP imply
saving power per unit power. Therefore, based on (7.10), the micro grids can make
the best decisions to merge their coalitions.
MGi calculates RUP (Eq. (7.10)) of PEN and sorts PEN in a descending order
according to RUP. Then, MGi considers the first element j from PEN, if PEN is not
empty. MGi sends Di .t/ to MGj , waits for the response from MGj unless time-out
occurs. If MGi receives “accept” response from MGj , it will exchange power with
MGi , based on Di .t/ and Dj .t/, delete j from PEN, and update Di .t/. MGj will be
deleted when MGi is waiting for the response from MGj and time-out occurs, or the
response is “reject.” At the same time, MGi receives offers from its neighbors as
well. If MGi is waiting for the response from the first element of PEN j and receives
the offer from other neighbor k.k ¤ j/, the status of MGk will be set as “hold,”
and the hold message is returned. The neighbor k will not be deleted until time-out
occurs. The above action will not repeated until Di .t/ D 0 or PEN is empty. After
exchanging power with one-hop neighbors, Di .t/ is updated by the quantities of
exchanged powers. If Di .t/ ¤ 0, MGi will exchange power with the macro station.
Algorithm 1 Distributed algorithm for micro grids’ power exchange. Input: Wi .t/,
Gi .t/, Output: Qij .t/, Q0i .t/, and PLAi .t/
1: BEGIN
2: for each MGi
3: Loop
4: Estimate Di .t/, based on Wi .t/, Gi .t/ (Eq. 7.1)
5: While (Di .t/ ¤ 0)
6: Share Di .t/ with neighbors of MGi .
7: Calculate PEN, RUP of PEN, and sort PEN
8: While (Di .t/ ¤ 0 and PEN is not empty)
9: Get first element of PEN j
10: If (j:status=hold and time-out occurs)
11: Remove j from PEN
12: Else
13: Send offer to MGj and wait for response
14: while (time-out does not occur)
15: If (the response from j =“accept”)
16: exchange power with j, calculate Qij .t/, PLij .t/ based on Eqs. 7.2, 7.3, update Di .t/
and remove j from PEN.
17: Endif
18: If (receive offer from neighbor k and k ¤ j)
19: k:status=hold and send “hold” message to k
20: Endif
21: Endwhile
22: Delete j from PEN
23: Endif
24: Endwhile
25: If .Di .t/ ¤ 0/
26: Compute Q0i .t/, PL0i .t/ based on 7.6 and 7.7 and Di .t/=0
27: Else
28: If received offer from neighbor l
29: Send “reject” to neighbor l
30: Endif
31: Endif
32: Endwhile
33: Endloop
34: END
These steps are summarized in Algorithm 1 [6]. For example, suppose that there
are one macro station and five micro grids (denoted by MG1 to MG5 as shown in
Fig. 7.2). In this example, RUP12 D 1:2, RUP23 D 2, RUP34 D 3:5, and RUP45 D 4.
Based on these RUPs, MG1 will send offer to MG2, MG2 will send offer to MG3,
MG3 will send offer to MG4, and MG5 will send offer to MG4. Because MG4
sends offer to MG5 and waits for the response, MG4 will send “hold” to MG3. In
the same way, MG3 and MG2 send “hold” to MG2 and MG1, respectively. When
MG4 receives response from MG5, they will exchange power. Because PL45 D 0:2,
Q45 D 1:3. After that D4 D D5 D 0 and MG4 sends “reject” to MG3. When MG3
receives “reject,” it will activate the offer of MG2 and exchange power with MG2
Fig. 7.2 A simple example showing how the distributed algorithm 1 enables power exchange
between micro grids and macro station with minimized power loss
(Q23 D 3). Hence, MG1 receives “reject” from MG2, after power transmission
between MG2 and MG3 (D2 D D3 D 0). Eventually, MG1 will exchange power
with the macro station (Q01 D 3:5).
Theorem 7.3.1. The solution of Algorithm 1 is Pareto Optimal.
Proof. Assume that the solution .a1 ; a2 ; : : : ; aN / is not Pareto Optimal. Therefore,
there exists at least a micro grid l 2 N, which can adjust its action al to al so
as to augment its utility while utilities of others will not be diminished. In other
words, u.al ; al / < u.al ; al /. Because the algorithm could help micro grids to find
the most appropriate neighbors and exchange power to maximize their payoff, the
micro grids cannot augment their utilities by changing the solution of the algorithm.
Therefore, u.al ; al / u.al ; al / 8al 2 Al . This result contradicts the previous
assumption that the solution is not Pareto Optimal.
In this section, numerical results are demonstrated to show the effectiveness of

the adopted distributed algorithm compared to (1) the distributed algorithm in [5]
whereby the micro grids choose the nearest neighboring micro-grid to exchange
power with, and (2) the centralized approach in Chap. 6. Our considered scenario
comprises a power distribution grid topology, area of which 10 10 km2 . The
macro station is deployed at the center of the grid, and the micro grids are deployed
randomly in the topology. Each micro grid is linked with its one hop neighboring
micro grid and the macro station. Similar to the assumption made by Saad et al. [5],
the power demands of the customers Wi .t/ of MGi is derived from a Gaussian
Fig. 7.3 Comparison of the average power loss in the conventional method and the adopted
distributed algorithmic approach
distribution between 10 and 316 MW. The power generation Gi .t/ is obtained from
a Gaussian distribution between 10 and 316 MW. The resistance between the micro
grids is the same as that between the macro station and any micro grid. R is set
to 0.2 /km. The fraction of power conversion parameter ˛ is set to 0.02 based
on [7]. U0 and U1 are set to 50 and 22 kV, respectively based on the practical values
considered in [7]. The prices of the each of the unit power are set as w1 D 1 and
w2 D 3 [1].
Figure 7.3 demonstrates the average power loss per micro grid for different
numbers of micro grids ranging from 5 to 50 in case of the conventional method
and our adopted distributed algorithmic approach. When the number of micro grids
increases, the power losses decrease in both schemes. However, the distributed
approach shows better performance because all the relevant power loss components
are considered to achieve near-optimal minimization of the total power losses while
finding the micro grids their suitable one-hop neighbors to exchange power.
Figure 7.4 shows the percentage of cost saving using the adopted distributed
approach in contrast with the conventional method. When the number of micro grids
increases, the percentage becomes larger. This is because the adopted technique
helps the micro grids to discover proper neighbors so as to minimize the total power
losses, and thereby, saves money. Thus, it allows the entire power grid to save a
significant amount of money in contrast with the conventional method.
Figure 7.5 demonstrates that the macro station in the conventional method needs
to supply more power for the micro girds to satisfy their user-demands than that
in the adopted distributed approach. The reason is that in the conventional case,
the micro grids only exchange power with the nearest one-hop neighboring micro
grids without considering the total power loss while by using the adopted distributed
algorithm, the micro grids can exchange power with others so as to reduce the total
Fig. 7.4 The percentage of cost saving using the adopted distributed approach compared with the
conventional method
Fig. 7.5 Comparison of the power load on the macro station in the conventional case and the
adopted distributed approach
power loss. Higher power loss will cause higher power load from the macro station.
Therefore, the adopted approach helps the macro station to decrease the peak of
power generation and improve its efficiency.
Furthermore, consider the comparison of the performance of the centralized
method in Chap. 6 and the adopted distributed approach. In the centralized case,
the micro grids send demands to the control center of the macro station. The macro
station will help all the micro grids to find proper neighbors as it knows all the
information of the micro grids. However, in the distributed approach, the micro grids
Fig. 7.6 The comparison in centralized and distributed approaches. (a) Average power loss.
(b) The bandwidth cost
only know the demands of one hop neighbors. By using the distributed approach,
the micro grids will send the demands to one-hop neighbors so as to minimize the
total power loss. If time-out occurs, the micro grids will give up this neighbor and
try to exchange power with other neighbors. Admittedly, this process could add to
the total power loss. Therefore, the average power loss in the distributed approach
is slightly higher than that in the centralized approach as shown in Fig. 7.6a. On the
other hand, because the micro grids do not send information to the control center,
the communication bandwidth cost in the distributed approach is lower than that in
the centralized approach as demonstrated in Fig. 7.6b. It means that by using the
distributed approach, precious communication bandwidth could be saved for more
important tasks such as accommodating AMI-oriented communication of many
more end-users in the micro grids.
References 75
In this chapter, we discussed a novel cooperative power exchange algorithm for

micro grids from a purely distributed perspective. This allows the micro grids to
form coalitions so as to minimize the total power loss. Through numerical results,
the effectiveness of the approach was verified. Comparative results showed its
superior performance compared to a conventional distributed method. Also, some
trade-off in the performance between the centralized approach and this distributed
one could be seen.
References

vol. 32, no. 7, pp. 1361–1370, Jul. 2014.
3. M. E. Elkhatib, R. E. Shatshat, and M. M. A. Salama, “Decentralized reactive power control
for advanced distribution automation systems,” IEEE Transactions on Smart Grid, vol. 3, no. 3,
pp. 1482–1490, Sep. 2012.
vol. 4, no. 1, pp. 321–331, Mar. 2013.
5. W. Saad, Z. Han, and H. Poor, “Coalitional game theory for cooperative micro-grid distribution
networks,” in 2011 IEEE International Conference on Communications (ICC) Workshops,
Seattle, Washington, USA, Jun. 2011, pp. 1–5.
6. C. Wei, Z. M. Fadlullah, N. Kato, and I. Stojmenovic, “A novel distributed algorithm for
power loss minimizing in smart grid,” in 2014 IEEE International Conference on Smart Grid
Communications (SmartGridComm), Venice, Italy, Nov. 2014, pp. 290–295.
7. J. Machowski, J. W. Bialek, and J. R. Bumby, Power system dynamics : stability and control.
Chichester, U.K. Wiley, 2008, rev. ed. of: Power system dynamics and stability / Jan Machowski,
Janusz W. Bialek, James R. Bumby. 1997. [Online]. Available: http://opac.inria.fr/record=
b1135564
Chapter 8
Security Challenge in the Smart Grid
8.1 Security Problem: A Bi-product of Legacy

Communication Networks
Whether we consider the overall smart grid or the micro grids paradigm (as dis-
cussed in Chap. 2), security is an important issue which cannot be overlooked.
The available smart metering technologies including the AMI often arise privacy
concerns since they rely on centralizing personal energy consumption information
of the users at their smart meters. In the Netherlands, there was a legal ruling in
2009 to make it mandatory to consider privacy issues of using smart meters [1].
Also, in the US policy adopted by NIST stated that there should be privacy for
design approach for smart grid communications [2]. These privacy concerns can
be addressed by appropriately authenticating the smart meters. Such a solution,
however, needs to take into consideration the limited resources (like low memory
and computational capacity) of the smart meters. So, any authentication scheme
for smart grid communication requires careful design so as to enforce adequate
security while placing minimal burden on the already limited resources of the smart
meters. In this chapter, we discuss a light-weight message authentication method
for securing communication among various smart meters at different points of the
smart grid. Our adopted method is based on the Diffie–Hellman key establishment
protocol and hash-based message authentication code that allows smart meters to
make mutual authentication and achieve message authentication in a light-weight
fashion. That is, it does not result in high delay and exchanges few signal messages
in the message authentication phase.
Securing smart grid communication relies on two important factors [3]—
communication delay and large volume of messages exchanged in the smart grid.
For example, consider the case that the control center accidentally misses inputs
from some HAN GWs. This may influence the control center’s decision in upcoming
energy distribution scheduling. To avoid any potential emergency situation which
could occur at any time, the smart grid communication system needs to have

78 8 Security Challenge in the Smart Grid
Table 8.1 Example of power Load Power requirement (kW/h)

requirements of different
loads in a typical HAN Air conditioner 1
Fridge 0.2
Microwave oven 0.1
Lighting arrays 0.05
Desktop computer 0.2
the ability to handle the message delivery to the control center via the BAN and
NAN GWs with the minimum possible delay. The power requirements of the HAN
devices/loads which are listed in Table 8.1 are sent to the respective BAN by meter
periodic data read (i.e., ToM#2 (revisit Fig. 2.3)). Each raw periodic request message
has a size of 32 bytes. If we consider the mandatory headers, the packet size can,
therefore, be roughly .50 C 32 D/ 82 bytes. Additionally, we need to consider
TCP/IP headers and optional security headers if any security protocol is used. If the
BAN GW experiences congestion, the packet may be delayed to be sent to the NAN
GW and control center. Furthermore, it may also get dropped if the RAM and the
on-chip flash of the BAN GW are full. The BAN GW memory can be full due to
several reasons like multiple messages arriving from different HANs at the same
time, limited processing capability of the BAN GWs means the buffer remains full
until the processing is complete, etc. If that is the case, then the BAN GW may
ask the HAN GW to retransmit the required packets. This could lead to increased
communication delay. Practically speaking, the smart grid communication delay
should be in the order of a few milliseconds according to [3, 4]. But it is tough to
achieve this in a large scale smart grid. So, how to minimize the communication
delay becomes one key challenge.
Hauser et al. [3] further report that the smart grid communication network needs
to be able to accommodate more messages simultaneously without having any major
impact on the communication delay. The large volume of messages in the smart grid
communication is likely to influence the available bandwidth. Consider a model
where a control center, connected with 10,000 feeders (and BAN GWs), serves
100,000 customers. Assume that each HAN GW generates a message every second
to the BAN GW [5] during peak hours. Then the total number of messages generated
per second is 100,000. Additionally, the BAN GWs generate messages to each other
and also to the control center through the NAN GW. If the average packet size is
100 bytes, the required transmission line bandwidth in the radio access network is
estimated to be 800 Mbps.
As demonstrated in the above example, any secure smart grid communication
framework requires to have light-weight operations to (1) avoid possibly high
communication delay, and (2) reduce communication overhead by cutting down
unnecessary signal messages. Also, remember that the security headers contribute
to the increased packet size as well (as illustrated in Fig. 2.3). This shows the impor-
tance of a light-weight authentication mechanism for HAN/BAN/NAN GWs in the
context of delay-sensitive and bandwidth-intensive smart grid communication.
8.2 Light-Weight Authentication Mechanism 79
However, the existing methods for facilitating smart grid communications

security are often inherited directly from the other communication networks
which do not necessarily take into account the context of specific communication
requirement of the smart grid. Also, most security systems do not consider secure
framework to reliably authenticate the smart meters operating on the different
levels in the smart grid. For example, the BAN GW should authenticate the
requesting HAN GWs while the NAN GW should be able to authenticate its
BAN GWs. The cryptographic overheads may add a substantial portion to the
packet size. Cryptographic operations also contribute to significant computation
cost, particularly in the receiver-end which needs to verify the message. Thus,
in the earlier example comprising 100,000 users, the number of messages to be
verified per second by the NAN GW becomes significantly high. Also, there is
processing delay at the respective smart meters to decrypt the arriving encrypted
messages. This increases the communication delay. Because the conventional Public
Key Infrastructure (PKI) methods are not adequate for the strict delay requirement
of smart grid communications, a light-weight verification algorithm tailored for
smart grid communications is needed so as to allow faster processing of incoming
messages.
In addition, the smart meters are susceptible to various attacks found in literature.
The use of IP-communication technologies in smart grid means that the power
system can become vulnerable to cyber-attacks listed in Table 8.2. To mitigate this
problem, a security framework is required, which can take into consideration various
design objectives [6–10].
8.2 Light-Weight Authentication Mechanism
To address the earlier listed threats, we adopt a framework with security and relia-
bility guarantees. The secure and reliable framework for smart grid communications
have to achieve the following objectives.
1. Source authentication and message integrity: The smart meters need to be able
to check the origin of a received packet and deliver the packet without any
alteration. For instance, if a BAN GW is to receive a packet from a HAN GW, the
BAN GW needs to authenticate the HAN GW. After successful authentication, it
needs to check whether the packet is unaltered or not.
2. Small communication overhead and fast verification: The security scheme
needs to be efficient in terms of low communication overhead and acceptable
processing delay. In other words, a large number of message signatures from
numerous smart meters need to be verified in a short time.
3. Conditional privacy preservation: The actual identity of a smart meter (which
includes the owner’s name, apartment number, etc.) needs to be concealed by
appropriate encryption technology.
Table 8.2 Security threats against smart grid communications and the security requirements to
address these problems
Attack Impact on SG Security requirement
Sniffing on smart meters Same problem as Encrypted packets:
conventional network tougher for decoding traffic
Traffic analysis Difficult to detect Change encryption keys
periodically
Denial of service (DoS), Can extract keys from second Authenticated sharing of
wireless jamming and generation Zigbee chips [11] resources and/or channels
interference
DoS buffer overflow attack May delete the content Debug programs and
of smart meters protocol thoroughly
Reconfigure attack Install unstable firmware on Only permit secure firmware
smart meter(s) and upgrade from authenticated CCs
electronic appliances
Spoofing – Impersonate smart meters Authenticate smart meter
– Increase victim’s bill over Internet
– Lower attacker’s own bill Protocol Security (IPSec)
Man-in-the-Middle (MitM) May impersonate Secure communication
smart meters [12] over IPSec
Replay attack – Store current data – Use time-stamp and
(during low power usage) time-synchronization
of smart meter at smart meters and CC
– Then send the stored data to – Use time-variant nonce
the utility company at a later
time (during high
power usage) [13]
4. Prevention of internal attack: A HAN GW owner should hold only her own
cryptographic key, and she should not be able to access any neighboring
HAN GWs’ keying materials. Thus, even if a smart meter is compromised, an
adversary cannot use the compromised smart meter to access other smart meters’
sensitive information.
5. Maintaining forward secrecy: It should be assured that a session key derived from
a set of long-term public and private keys will not be compromised if one of the
(long-term) private keys is compromised in future.
Figure 8.1 shows our considered security framework for establishing a secure
communication environment in the smart grid [8]. The framework has three parts—
authentication, communication management, and network analysis, monitoring and
protection. The smart meters are required to be authenticated before they can
participate in the communication with other BAN/NAN GWs. The authentica-
tion scheme can be based on protocols such as Diffie–Hellman, SIGn-and-MAc
8.2 Light-Weight Authentication Mechanism 81
Fig. 8.1 Considered secure and reliable framework for smart grid communications
(SIGMA), and Internet Key Exchange (IKEv2). The communication management

module has two parts—message encryption/decryption and end-to-end protection.
Existing cryptographic algorithms like Data Encryption Standard (DES), Advanced
Encryption Standard (AES), Rivest, Shamir, and Adleman (RSA) public key
encryption, etc., can be used to encrypt the communication. For the end-to-end
protection, Internet Protocol Security (IPSec) or virtual tunnel could be used. In the
network analysis, monitoring, and protection module, smart meters are supposed
to act as monitoring stubs which are equipped with anomaly and/or signature-based
intrusion detection algorithms in order to detect malicious threats listed in Table 8.2.
If the system finds any attack and detects a security update, it contacts a secure
server to download appropriate patches or firmware updates. The monitoring stubs

may also provide appropriate responses to the detected attacks. The focus of this
chapter is, however, on the first step of the framework. In other words, designing
an appropriate authentication method, which is light-weight and suited for delay-
sensitive and bandwidth-intensive smart grid communications, is stressed upon in
the chapter while other features of the smart grid security framework can be treated
as open problems which require more attention in future.
Suppose that HAN GW i and BAN GW j have their private and public key pairs.
HAN GW i has public and private keys denoted by PubHAN_GWi and PrivHAN_GWi ,
respectively. On the other hand, BAN GW j has public and private keys denoted by
PubBAN_GWj and PrivBAN_GWj , respectively. The Diffie–Hellman key establishment
protocol [14] is adopted for the initial handshake between the HAN and BAN GWs.
Suppose G D hgi is a group of large prime order q such that the Computational
Diffie–Hellman (CDH) assumption holds. In other words, given ga ; gb , for unknown
a; b 2 Zq , it is difficult to estimate gab 2 G. Based on the CDH assumption, our
adopted light-weight message authentication method is shown in Fig. 8.2, and its
detailed steps are described below.
1. HAN GW i selects a random number a 2 Zq , computes ga , and sends ga in an
encrypted request packet to BAN GW j.
HAN GWi ! BAN GWj W fijjjjjga gfencrgPubBAN_GWj
2. BAN GW j decrypts it and sends an encrypted response consisting of gb . Here, b

is a random number.
BAN GWj ! HAN GWi W fijjjjjga kgb gfencrgPubHAN_GWi
Fig. 8.2 Adopted light-weight message authentication scheme

8.3 Security Analysis 83
3. After receiving BAN GW j’s response packet, HAN GW i recovers ga ; gb with its
private key. If the recovered ga is correct, BAN GW j is authenticated by HAN
GW i. Then, with gb and a, HAN GW i can calculate the shared session key
Ki D H.ijjjjj.gb /a /, where H W f0; 1g ! Zq is a secure cryptographic hash
function, and sends gb to BAN GW j in the plaintext form.
4. When the correct gb is received by the BAN GW j, BAN GW j authenticates
HAN GW i, and calculates the same shared session key Ki D H.ijjjjj.ga /b /.
5. To ensure data integrity in the late transmission, a Hash-based Message Authen-
tication Code (MAC) generation algorithm using the shared session key Ki is
utilized. The generated MAC denoted by HMACKi is based on the message Mi
and recorded time instance of sending the message Ti , where Ti is used to prevent
possible replay attacks. Then, HAN GW i transmits the following to the BAN
GW j:
HAN GWi ! BAN GWj W fMi kTi kHMACKi gfencrgKi
Since Ki is shared between BAN GW j and HAN GW i itself, BAN GW j

can verify the authenticity of the sender and integrity of Mi . By this way, it can
provide the NAN GW with the authenticated messages.
8.3 Security Analysis
Now let us analyze the security of the adopted light-weight message authentication
method to verify whether the required security properties are satisfied.
• The adopted method can provide mutual authentication. In the adopted method,
because ga is encrypted with BAN GW j’s public key, only if the adopted public
key encryption technique is secure, then BAN GW j is the only entity who can
recover ga with the corresponding private key. So, when HAN GW i receives the
correct ga in Step 3, HAN GW i can realize that its counterpart is indeed BAN
GW j. With the same reasoning, because gb is encrypted with HAN GW i’s public
key, BAN GW j can also authenticate HAN GW i if it can receive the correct gb
in Step 4. Thus, the adopted method can provide mutual authentication between
HAN GW i and BAN GW j.
• The adopted method can establish a semantic-secure shared key in the mutual
authentication environment. The semantic security of the shared key under the
chosen-plaintext attack means that an adversary A cannot distinguish the actual
shared key Ki from the ones randomly drawn from the session key space, when A
is given ga ; gb and Z 2 G, where Z is either the actual shared key Ki or a random
value R drawn from the session key space, according to a random bit ˇ 2 f0; 1g,
i.e., Z D Ki when ˇ D 0, and Z D R is returned when ˇ D 1. Let ˇ 0 2 f0; 1g
be A’s guess on ˇ. Then, the semantic security indicates PrŒˇ D ˇ 0 D 12 . Now
assume that there exists an adversary A that can break the semantic security of
the shared key with a non-negligible advantage " D 2 PrŒˇ D ˇ 0 1 within

the polynomial time. We can use the adversary A’s capability to solve the CDH
problem, i.e., give .g; ga ; gb / for unknown a; b 2 Zq , to estimate gab 2 G.
First, the adversary A is given .g; ga ; gb /, and also permitted to make qH distinct
queries on the random oracle H in the random oracle model [15]. To account for
these random oracle queries, let us maintain a H-list. When a new query Ci 2 G is
asked for the session key shared between HAN GW i and BAN GW j, we choose
a fresh random number Zi 2 G, set H.ijjjjjCi / D Zi , put .ijjjjjCi ; Zi / in H-list, and
return Zi to A. When the adversary A makes a query on the session key, we flip a
coin ˇ 2 f0; 1g, and return a random value Z ? 2 G.
Let E be the event that C D gab has been queried by A to the random oracle H. If
the event E does not take place, A has no idea on the session key Ki D H.ijjjjjgab /,
then we have:
N D 1
PrŒˇ D ˇ 0 jE
2
and
N PrŒE
PrŒˇ D ˇ 0 D PrŒˇ D ˇ 0 jE PrŒE C PrŒˇ D ˇ 0 jE N
1 N
D PrŒˇ D ˇ 0 jE PrŒE C PrŒE
2
1 1 PrŒE
PrŒE C .1 PrŒE/ D C
2 2 2
Additionally, because
1 "
" D 2 PrŒˇ D ˇ 0 1 ) PrŒˇ D ˇ 0 D C
2 2
we have PrŒE ". Since H-list contains qH entries, we may pick up the correct
Ci D gab and solve the CDH problem with the success probability 1=qH given that
the event E does occur. Combining the above probabilities, we have:
"
SuccCDH D 1=qH PrŒE :
qH
However, this result is in contradiction with the CDH assumption. Hence, the
adopted method can also establish a semantic-secure shared key. It is worth noting
that if either HAN GW i or BAN GW j is compromised, the mutual authentication
environment cannot be achieved. However, the compromise of either HAN GW i
or BAN GW j’s private key will not have impact on the security of the previous
session keys. Therefore, the adopted method can also achieve perfect forward
secrecy [14].
• The adopted method can provide an authenticated and encrypted channel for the
late successive transmission. Since both HAN GW i and BAN GW j hold their
shared session key Ki , the late transmission fMi kTi kHMACKi gfencrgKi can achieve
integrity in addition to confidentiality. Additionally, the embedded timestamp
Ti prevents possible replay attacks. Thus, the adopted method can provide an
authenticated and encrypted channel for the late successive transmissions.
In summary, our adopted method is secure and suitable for the two-party
communication in a smart grid environment.
The effectiveness of the adopted message authentication method is evaluated using

MATLAB [16]. For the smart grid topology in this evaluation, 10 NANs, each
having 50 BANs, are considered. The number of HANs in each BAN is varied from
a range of 10–140. Table 8.3 lists the other relevant parameters. The performance
of our adopted method is compared with conventional ECDSA, which was used as
a secure authentication protocol for smart grid demand response communications
in [17]. In simulations, AES-128 algorithm was used to encrypt the packets to
be transmitted using the shared session key, Ki , generated during the proposed
authentication mechanism. In order to compare with this, we used ECDSA-256
authentication and encryption in simulations because its security level is comparable
to that of 128-bits cryptography [18]. Note that only the messages exchanged
between HANs and their corresponding BAN are considered for authentication.
Furthermore, the session key is considered to be generated at the beginning of every
fresh session.
The size of the HAN packet bound for the BAN is considered to be 102
bytes, which is sufficient to contain the users’ power requirements and request
to the control center. The size of the generated MAC is set to 16 bytes based
on the RACE Integrity Primitives Evaluation Message Digest (RIPEMD-128)
Table 8.3 Simulation Parameter Value

parameters used for
evaluation. BAN GW CPU clock 160 MHz
Number of HANs 10–140
HAN message generation interval (ı) 10 s
TCP header 20 bytes
Message header 50 bytes
Raw message 32 bytes
Hash header in proposed authentication scheme 16 bytes
ECDSA certificate size 125 bytes
ECDSA signature size 64 bytes
Simulation time 800 s
Fig. 8.3 Average communication overhead per BAN GW for varying number of smart meters
(HAN GWs)
Fig. 8.4 Average delay at the BAN GW for varying number of smart meters (i.e., HAN GWs)
algorithm. This is done to utilize its resiliency against collision and pre-image
attacks. The HAN message generation interval (ı) is considered to be 10 s to
correspond with highly frequent need for demand-response communications in the
smart grid. Communication overhead and message decryption/verification delay
are compared in Figs. 8.3 and 8.4 for the conventional method and the adopted
approach. Figure 8.3 shows the average communication overhead (in kilo bytes)
experienced by the BAN GWs for varying numbers of smart meters. It should be
noted that only one session per HAN GW with the BAN GW is supposed. When
the number of smart meters is low, both the adopted approach and conventional
method result in small overheads (below 5 kilo bytes). But the communication
overheads start to gradually rise with the increase in smart meters. In case of the
conventional method, the rise in the average communication overhead is much more
significant than the adopted approach. The conventional method suffers from higher
communication overheads mainly because of the certificate and signature included
in each packet.
Next, Fig. 8.4 demonstrates the comparison between the adopted approach
and the conventional method in terms of decryption/verification delay per BAN
GW. OpenSSL package is used to measure the delays for both the adopted and
conventional schemes [19] on a computer running 3.0 GHz Intel Xeon Processor
(E5450) and Linux distribution of Debian 4.0. To simulate the BAN GW, the
experimental values (e.g., decryption time) were scaled by 19.2 times to fit the
160 MHz of the BAN GW. As shown by the plot in Fig. 8.4, the decryption delay
increases linearly for both the schemes. However, the conventional method shows
higher decryption delay compared to that of the adopted approach since the latter
provides a secure authentication process followed by AES encryption, which is
faster than the conventional one relying on heavier signature verification plus
decryption at the BAN for every message arriving from each HAN.
The memory usage of the adopted and conventional schemes over time for
changing message volumes received by a given BAN GW is demonstrated in
Fig. 8.5. The memory usage consists of two upper bounds, namely the RAM
boundary and the RAM plus flash memory boundary that comprise 128 KB and
1 MB, respectively. Different message rates are used. The legends in the graph use
the format of px and cx which mean the adopted approach and the conventional
method with the message rate of x per ı. When the message rate is 50 per ı, the
conventional method consumes about 50 kilo bytes of memory of the BAN GW.
The adopted approach shows similar memory use. But notice when the number of
Fig. 8.5 Memory usage of the proposed and conventional ECDSA authentication algorithms for
different message volumes received by BAN
Fig. 8.6 Number of HANs supported by the proposed and conventional authentication schemes
for smart grid communications
messages per ı arriving at the BAN GW increases to 100, the conventional method
gets overwhelmed with the high number of messages coming from the high number
of HANs, and it exceeds the RAM and flash memory bound after 570 s. Compared
to this, the adopted approach uses much lower memory usage (approximately 100
kilo bytes) and continues to sustain this until end of simulation. However, when the
number of apartments in a given building is further increased resulting in a higher
message reception rate of 130 messages per ı at the BAN GW, the results change
dramatically. In this case, the conventional method uses up all the available memory
at the BAN GW very fast (within just 220 s of the beginning of the simulation). On
the other hand, the adopted approach remains below 270 kilo bytes of the overall
available memory throughout the conducted simulation. This superior performance
of the adopted approach is because of its less processing in decrypting the packets
that result in less queuing time in the RAM and the flash memory.
It may also be interesting to evaluate how may HANs are supported by the two
compared schemes in terms of the available RAM and flash memory at the BAN
GW over time. Figure 8.6 demonstrates this. In the conventional method, when
the number of HANs per BAN exceeds 81, the memory usage starts to increase
with time. This implies that after a while the memory usage will overflow the
memory space of the BAN GW (i.e., 1152 kilo bytes consisting of 1 mega byte flash
memory and 128 kilo bytes of RAM). At that point, the messages arriving from
the HANs are dropped and not served within the BAN GW queue. For example,
for 95 HANs supported by a particular BAN, the conventional method requires
around 1260 kilo bytes of memory space to avoid any drop of messages. On the
other hand, the adopted approach can accommodate 127 HANs within the BAN
which is significantly higher than that achieved by the conventional method. This is
possible because the adopted approach can process the messages arriving from the
HANs in the BAN memory space much quicker than the conventional method.
References 89
In this chapter, we described a light-weight message authentication scheme tailored

for the requirements of smart grid communications based on Diffie–Hellman
key establishment protocol and hash-based authentication code. A detailed security
analysis of the adopted approach verifies its ability to satisfy the desirable security
requirements within a secure and reliable smart grid communications framework.
Also, comparative results demonstrate its superior performance over a conventional
method.
References
1. C. Cuijpers and B.-J. Koops, “Het wetsvoorstel ‘slimme meters’: een privacytoets op basis van
art. 8 evrm.” Tilburg University, the Netherlands, Tech. Rep., Oct. 2008, accessed Aug. 2015.
[Online]. Available: http://www.consumentenbond.nl/morello-bestanden/pdf-algemeen-2008/
onderzoek_UvT_slimme_energi1.pdf
2. US National Institute for Standards and Technology (NIST), “The Smart Grid Interoperability
Panel Cyber Security Working Group: Smart Grid Cybersecurity Strategy and Requirements,
revision 1,” accessed Aug. 2015. [Online]. Available: http://csrc.nist.gov/publications/nistir/
ir7628/nistir-7628_vol2.pdf
3. C. H. Hauser, D. E. Bakken, I. Dionysiou, K. H. Gjermundrod, V. S. Irava, J. Halkey, and
A. Bose, “Security, trust, and QoS in next generation control and communication for large
power systems,” International Journal of Critical Infrastructures, vol. 4, no. 1/2, 2008.
4. R. Vaswani and E. Dresselhuys, “Implementing the right network for the smart grid:
Critical infrastructure determines long-term strategy,” Silver Spring Networks, White Paper,
accessed Aug. 2015. [Online]. Available: http://www.silverspringnet.com/pdfs/whitepapers/
SilverSpring-Whitepaper-UtilityProject.pdf
5. A. Aggarwal, S. Kunta, and P. K. Verma, “A proposed communications infrastructure for the
smart grid,” in Proc. Innovative Smart Grid Technologies (ISGT), Gaithersburg, Maryland,
USA, Jan. 2010.
vol. 2, no. 4, pp. 675–685, Dec. 2011.
9. M. M. Fouda, Z. M. Fadlullah, and N. Kato, “Assessing attack threat against zigbee-based home
area network for smart grid communications,” in 2010 International Conference on Computer
Engineering and Systems (ICCES), Cairo, Egypt, Nov.-Dec. 2010, pp. 245–250.
10. M. M. Fouda, Z. M. Fadlullah, N. Kato, R. Lu, , and X. Shen, “Towards a light-weight message
authentication mechanism tailored for smart grid communications,” in IEEE International
Workshop on Security in Computers, Networking and Communications (SCNC’11), Shanghai,
China, Apr. 2011.
11. T. Goodspeed, “Extracting keys from second generation zigbee chips,” LAS VEGAS, NV,
USA, Jul. 2009.
12. S. Blake-Wilson, “Securing the smart grid,” AuthenTec Embedded Security Solutions, White
Paper, accessed Aug. 2015. [Online]. Available: http://www.authentec.com/white-paper.cfm
13. M. Carpenter, T. Goodspeed, B. Singletary, E. Skoudis, and J. Wright, “Advanced metering
infrastructure attack methodology,” InGuardians, White Paper, Jan. 2009, accessed Aug. 2015.
[Online]. Available: http://inguardians.com/pubs/AMI_Attack_Methodology.pdf
14. D. R. Stingson, Cryptography Theory and Practice, 3rd ed. CRC Press, Nov. 2005.
15. M. Bellare and P. Rogaway, “Random oracles are practical: a paradigm for designing efficient
protocols,” in Proceedings of the 1st ACM conference on Computer and communications
security, ser. CCS ’93. New York, NY, USA: ACM, 1993, pp. 62–73. [Online]. Available:
http://doi.acm.org/10.1145/168588.168596
16. Mathworks - matlab and simulink for technical computing. [Online]. Available: http://www.
mathworks.com/
17. M. Kgwadi and T. Kunz, “Securing RDS broadcast messages for smart grid applications,”
in Proceedings of the 6th International Wireless Communications and Mobile Computing
Conference, ser. IWCMC’10. New York, NY, USA: ACM, 2010, pp. 1177–1181.
18. G. Calandriello, P. Papadimitratos, J.-P. Hubaux, and A. Lioy, “Efficient and robust pseudony-
mous authentication in vanet,” in Proceedings of the 4th ACM international workshop on
Vehicular ad hoc networks, ser. VANET ’07. New York, NY, USA: ACM, 2007, pp. 19–28.
[Online]. Available: http://doi.acm.org/10.1145/1287748.1287752
19. “OpenSSL: Cryptography and SSL/TLS Toolkit,” accessed Aug. 2015. [Online]. Available:
http://www.openssl.org/
Chapter 9
Conclusion and Future Directions
In recent years, research attention on the smart grid comprising distributed power
generators such as micro-grids has significantly increased [1–7]. The smart grid con-
tinues to evolve to date by integrating electric power engineering technologies with
communication networks, through a plethora of power instrumentation sensors and
smart meters deployed between the electricity generator and consumers. Through
these assorted technologies, the smart grid is expected to offer the consumers
the long-cherished opportunity to communicate with their utility provider. This
empowers the consumers to express their energy needs in an interactive manner with
the provider, and also allows them to monitor and regulate their own energy use.
However, the smart grid communication and power distribution systems are subject
to many complex issues such as architecture with legacy support, varying demand
response and load management, varying price of power, power loss in smaller micro-
grids, electricity storage, security, and so forth. These issues lead to various decision
making challenges.
The aforementioned challenges are also interesting from whichever perspective
we look at them. For example, we can formulate some problems considering that
the consumers should receive the incentives, while other problems may focus on the
benefit of the provider. In other words, what is an opportunity for the consumers
could very well be a de-motivating factor, i.e., a trade-off challenge for the utility
provider, and vice versa. Such a trade-off can be depicted as an optimization
problem, and could be solved with state-of-the-art optimization techniques such
as game theory. Therefore, it is essential to identify the scope and challenges of
the smart grid in a comprehensive fashion so that efficient delivery of sustainable,
economic, and secure electricity distribution may become possible in the smart grid.
In this book, we also described some relevant techniques to address these challenges
and include numeric results to show how these techniques can be effective. The
content of the book can be useful for readers of a broad spectrum. The initial

92 9 Conclusion and Future Directions
chapters could be useful for interested readers with basic knowledge in electric grids
and communication networks while the later chapters might offer more insight to the
advanced readers who have been active in the smart grid research area.
Frankly speaking, smart grid adoption is taking its time, and is somewhat
different in each country. But the fact of the matter is—smart grids are happening.
With the promises smart grid offers, many challenges follow. Adequate demand
response, power loss minimization, and security are three core challenges (that
this book highlighted on) to be effectively addressed in smart grid design and
implementation. The smart grid does not only hold the potential to improve the
energy sector but also it is being considered as an important IoT (Internet of Things)
application which is currently being implemented around the world. Just like the
smart grid, the IoT is about inter-connecting devices at a large scale. The AMI of
the smart grid continuously monitors every electricity point within a house using
which the way electricity is consumed can actually be modified. This information
at a wider scale (e.g., in the city level) can be used for maintaining the load balance
within the energy grid assuring a high quality of service. Scaling further, with the
global focus on energy management, the IoT is expected to extend the connected
benefits of the smart grid beyond the distribution, automation and monitoring being
done by utility operators.
9.2 Future Directions and Caveats
The rule of engaging the energy market is swiftly changing with the smart grid. For
many years, energy operators considered that the demand was a statistically random
process. The operators traditionally measure only the total consumption of each
customer instead of its distribution over time. With state-of-the-art smart meters
able to report energy use profiles over time, the operators can come up with time-
dependent tariffs to give customers an incentive to lower their peak consumption.
If these tariffs are transparent to the users at real-time, and more importantly if they
are reasonably stable over time, the customers can schedule their consumption to
flatten their usage pattern to effectively reduce their average energy bill. In practice,
this will largely depend on the good will of the operators also, whether they want to
give their customers more control in determining their own usage or not still remains
a concern.
The micro grids based power generation from renewable sources creates random
and unpredictable changes in the total power generation. This means even more
dynamic adaptation of demand at real-time is needed. The change in tariffs will be
much more frequent in such cases and this may be bothersome the customers to
make decisions in real-time. The customers could input their preference into their
smart meter beforehand (e.g., accept the tariffs within certain ranges and hours, etc.),
and they could review and change the preference when they have the time. This
means that it comes down to the utility operators to actively manage the regulation
of energy use based on pre-decided inputs from the customers using bi-directional
machine-to-machine technology.
References 93
Smart grid is an exciting technology to many, and out of excitement, the

pro-smart grid camp often forgets that the communication network and home
controllers of the smart grid do come with a cost. Developing and deploying
the communication network, controllers, and smart grid applications at will is far
from an optimal model. They need to support general-purpose machine-to-machine
communications and applications to decrease the marginal cost of the smart grid
use cases. A viable vision is to consider these as IoT [8, 9] applications. With
such a bigger vision, the smart grid concept can be further extended to other utility
services. The improvement of gas, water, and waste-water use can be achieved if
the corresponding utility providers also adopt similar means. IoT can be exploited
effectively to seamlessly integrate all these. The connectivity and accessibility
which the IoT offers could enhance the user experience and efficiencies permitting
greater interaction and control for the customers. Also, the IoT can deliver more data
for manufacturers and utility providers to reduce costs through diagnostics and city-
wide meter reading capabilities. Thus, IoT will become instrumental in realizing
more connected, cost-effective, and smarter utility grids.
The smart grid standards are still evolving. Some countries are keener in having
and advancing the technology than others. This means that political will to invest
time and money into the smart grid is going to be a critical factor in determining
its success. Also, the governments have to reach out to the masses by spreading
consumer education on the impact and benefits of the smart grid technology, and
empower them to take the right steps to protect their privacy. Besides government-
backed initiatives, private sector investment into smart grid is increasing remarkably.
The tech-giant Google is also showing a great deal of interest into the smart grid
market by supporting renewable energy projects, data-center efficiency, and home
energy management software and hardware tools [10]. Google itself is a major
investor in the intermittent green energy market such as wind and solar projects
around the globe. By holding a 37.5 % stake in the Atlantic Wind Connection,
Google is emerging as a stakeholder in the next-generation grid projects. Its
competitors like Honeywell and Apple are also following suit, particularly in
building energy management technologies. This gives us a sense that smart grid
is going to further evolve in future, and thus, will reshape our life for the better, and
revolutionize the business models in the energy sector dramatically.
References

vol. 2, no. 4, pp. 675–685, Dec. 2011.
vol. 49, no. 4, pp. 60–65, Apr. 2011.
vol. 4, no. 1, pp. 321–331, Mar. 2013.
94 9 Conclusion and Future Directions

vol. 2, no. 4, pp. 675–685, Dec. 2011.
8. Y. Kawamoto, H. Nishiyama, N. Kato, N. Yoshimura, and S. Yamamoto, “Internet of things
(IoT): Present state and future prospects,” IEICE Transactions on Information and Systems,
vol. E97-D, no. 10, pp. 2568–2575, Oct. 2014.
9. Y. Kawamoto, H. Nishiyama, Z. M. Fadlullah, and N. Kato, “Effective data collection via
satellite-routed sensor system (SRSS) to realize global-scaled Internet of Things,” IEEE
Sensors Journal, vol. 13, no. 10, pp. 3645–3654, Oct. 2013.
10. J. S. John, “What Is Google Plotting for the Smart Grid?” accessed Aug. 2015.
[Online]. Available: http://www.greentechmedia.com/articles/read/what-is-google-plotting-
for-the-smart-grid

Evolution of Smart Grids

Uploaded by

Copyright:

Available Formats

Evolution of Smart Grids

Uploaded by

Document Information

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

Evolution of Smart Grids

Uploaded by

Copyright:

Available Formats

SPRINGER BRIEFS IN

ELEC TRIC AL AND COMPUTER ENGINEERING

Zubair Md. Fadlullah

More information about this series at http://www.springer.com/series/10059

Evolution of Smart Grids

ISSN 2191-8112 ISSN 2191-8120 (electronic)

Library of Congress Control Number: 2015954084

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media (www.

Sendai, Miyagi, Japan Zubair Md. Fadlullah

3.3 Concluding Remarks: Could an All-in-One Optimization

7.3 The Distributed Solution Approach

Fig. 1.1 Evolution of the power grid into a smarter “two-way”

1.1 Evolution of Smart Grids

© The Author(s) 2015 1

provisioning consumer participation, energy demand and cost estimation, demand

1.2 Challenges with Current Smart Grid Settings

1.3 Aim of the Book

1. M. M. Fouda, Z. M. Fadlullah, N. Kato, R. Lu, and X. Shen, “A lightweight message

2.1 Smart Grid Model with Hierarchical

Figure 2.1 demonstrates our considered smart grid communication framework

© The Author(s) 2015 7

2.1.1 Communication Technology Used in HAN

2.1.2 Communication Technology Used in BAN

2.1.3 Communication Technology Used in NAN

2.2 Smart Grid Communication Packet Structure

Fig. 2.2 SG communications packet structure

2.3 Micro Grids Oriented Smart Grid Model

2.3.1 Macro Station and Retailers

main power plant (i.e., geo-thermal/hydro-electric/nuclear power plants) and the

2.3.2 Macro Station and Micro Grid Model

Fig. 2.3 Micro-grids oriented architecture of the smart grid

accordance with user-demand. When a micro-grid is not self-sufficient, it can buy

2.3.3 Distributed Micro Grids with Renewable Sources

2.4 Deployment Scenarios and Key Opportunities

2.5 Practical Implementation and Operating Challenges

1. M. M. Fouda, Z. M. Fadlullah, N. Kato, R. Lu, and X. Shen, “A lightweight message

3.1 Optimization Challenges from Different Perspectives

3.1.1 Demand Response (DR): Shaping Electricity

© The Author(s) 2015 17

3.1.1.1 Direct Load Control

3.1.1.2 Price-Based DR Programs

3.1.2 Energy Conservation and Efficiency

3.1.3 Power Loss Problem in Micro-Grids Based

proposed strategy. Arefifar et al. presented systematic and optimized approaches

3.2 Is Smart Grid Communication Security an Overkill?

3.3 Concluding Remarks: Could an All-in-One Optimization

1. H. Liang, A. Tamang, W. Zhuang, and X. Shen, “Stochastic information management in smart

8. I. E. A. (IEA), available online, accessed Aug 2015, “http://www.iea.org/statistics/”.

As indicated in Chap. 3, for the successful deployment of the smart grid,

4.1 Background into DSM and Load Balancing Problem

© The Author(s) 2015 25

a utility function in an energy schedule optimization in which power generation

4.2 System Model

4.2.1 Power System Modeling

4.2.1.1 Modeling an Appropriate Energy Cost Function

4.2.1.2 Modeling Load Control on the Customer-Side

Consider An to be the set of loads (including air-conditioners, heaters, refrigera-

In order to improve the assumptions of the existing model, we incorporate a

v.S/ D max u.S; / (5.8)

A F B , fj .A/ j .B/; 8k 2 A; Bg (5.10)

Gmin Gi .t/ Gmax ; (6.3)

Smin Si .t/ Smax ; (6.5)

k .Ak / is given by (6.17) for coalition Ak . Collection A is preferred to B by Pareto

A F B , fk .A/ k .B/; 8k 2 A; Bg; (6.20)