E 1613501382

Pandian Vasant
Petronas University of Technology, Malaysia

Nader Barsoum
Curtin University, Malaysia
Jeffrey Webb
Swinburne University of Technology, Malaysia
Innovation in Power,
Control, and Optimization:
Emerging Energy Technologies
Innovation in power, control, and optimization: emerging energy technologies / Pandian Vasant, Nader Barsoum, and Jeffrey
Webb, editors.
p. cm.
Summary: This book unites research on the development of techniques and methodologies to improve the performance of
power systems, energy planning and environments, controllers and robotics, operation research, and modern artificial
computational intelligent techniques-- Provided by publisher.
Includes index.
ISBN 978-1-61350-138-2 (hardcover) -- ISBN 978-1-61350-139-9 (ebook) -- ISBN 978-1-61350-140-5 (print & perpetual
access) 1. Electric power system stability. 2. Power resources--Research. 3. Power resources--Economic aspects. I.
Vasant, Pandian. II. Barsoum, Nader N. III. Webb, Jeffrey, 1963-
TK1010.I45 2012
621.31--dc23
2011026251
British Cataloguing in Publication Data
A Cataloguing in Publication record for this book is available from the British Library.
All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Senior Editorial Director: Kristin Klinger
Director of Book Publications: Julia Mosemann
Editorial Director: Lindsay Johnston
Acquisitions Editor: Erika Carter
Development Editor: Myla Harty
Production Editor: Sean Woznicki
Typesetters: Jennifer Romanchak, Chris Shearer
Print Coordinator: Jamie Snavely
Cover Design: Nick Newcomer
Published in the United States of America by
Engineering Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: cust@igi-global.com
Web site: http://www.igi-global.com
Copyright 2012 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.
Library of Congress Cataloging-in-Publication Data
Editorial Advisory Board
Milorad Bojic, University of Kragujevac, Serbia
Gianfranco Rizzo, University of Salerno, Italy
Janos Sebestyen Janosy, KFKI Atomic Energy Research Institute, Hungary
Ivan Zelinka, Thomas Bata University in Zlin, Czech Republic
Davor Skrlec, University of Zagreb, Croatia
Nikolai I. Voropai, Energy Systems Institute, Russia
Monica Chis, Siemens Program and System Engineering, Romania
Cengiz Kahraman, stanbul Technical University, Turkey
Valentina E. Balas, Aurel Vlaicu University of Arad, Romania
Arturo Suman Bretas, Universidade Federal do Rio Grande do Sul, Brazil
Etienne Kerre, Ghent University, Belgium
Radu Emil Precup, Politehnica University of Timisoara, Romania
Gerardo Maximiliano Mendez, Instituto Tecnologico de Nuevo Leon, Mexico
Table of Contents
Foreword by Ivan Zelinka .................................................................................................................. vii
Foreword by Igor Tyukhov ................................................................................................................. ix
Preface .................................................................................................................................................... x
Acknowledgment ................................................................................................................................xiii
Chapter 1
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems ........................ 1
Nicolay Voropai, Energy Systems Institute, Russia
Irina Kolosok, Energy Systems Institute, Russia
Elena Korkina, Energy Systems Institute, Russia
Alexey Paltsev, Energy Systems Institute, Russia
Anna Glazunova, Energy Systems Institute, Russia
Victor Kurbatsky, Energy Systems Institute, Russia
Nikita Tomin, Energy Systems Institute, Russia
Alexander Gamm, Energy Systems Institute, Russia
Irina Golub, Energy Systems Institute, Russia
Roman Bershansky, Energy Systems Institute, Russia
Daniil Panasetsky, Energy Systems Institute, Russia
Dmitry Efmov, Energy Systems Institute, Russia
Dmitry Popov, Energy Systems Institute, Russia
Christian Rehtanz, University of Dortmund, Germany
Ulf Hger, University of Dortmund, Germany
Chapter 2
Hopfeld Lagrange Network for Economic Load Dispatch .................................................................. 57
Vo Ngoc Dieu, Asian Institute of Technology, Thailand
Weerakorn Ongsakul, Asian Institute of Technology, Thailand
Chapter 3
Renewable Energy and Sustainable Development................................................................................ 95
Abdeen Mustafa Omer, Energy Research Institute, UK
Chapter 4
Demand-Side Response Smart Grid Technique for Optimized Energy Use ....................................... 137
Fouad Kamel, University of Southern Queensland, Australia
Marwan Marwan, Queensland University of Technology, Australia
Chapter 5
Soft Computing and Computational Intelligent Techniques in the Evaluation of Emerging Energy
Technologies ....................................................................................................................................... 164
Selcuk Cebi, Karadeniz Technical University, Turkey
Cengiz Kahraman, Istanbul Technical University, Turkey
hsan Kaya, Yldz Technical University, Turkey
Chapter 6
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms:
Kurdistan Electric Network Case Study ............................................................................................. 198
Mohammad Saleh, University of Kurdistan, Iran
Hassan Bevrani, University of Kurdistan, Iran
Chapter 7
Many-to-Many Assignment Problems: Lagrangian Bounds and Heuristic ........................................ 220
Igor Litvinchev, Nuevo Leon State University, Mexico
Socorro Rangel, So Paulo State University, Brazil
Chapter 8
Power Systems Investments: A Real Options Analysis ...................................................................... 248
Joo Zambujal-Oliveira, Instituto Superior Tcnico & Technical University of Lisbon, Portugal
Chapter 9
Optimal Confguration and Reconfguration of Electric Distribution Networks ................................ 268
Armin Ebrahimi Milani, Islamic Azad University, Iran
Mahmood Reza Haghifam, Tarbiat Modares University, Iran
Chapter 10
A Descriptive Approach for Power System Stability and Security Assessment ................................. 293
A. G. Tikdari, University of Kurdistan, Iran
H. Bevrani, University of Kurdistan, Iran
G. Ledwich, Queensland University of Technology, Australia
Chapter 11
Analyses and Monitoring of Power Grid ............................................................................................ 315
Rana A. Jabbar, Rachna College of Engineering and Technology, Pakistan
Muhammad Junaid, Rachna College of Engineering and Technology, Pakistan
M. A. Masood, Rachna College of Engineering and Technology, Pakistan
A. Bashir, Rachna College of Engineering and Technology, Pakistan
M. Mansoor, Rachna College of Engineering and Technology, Pakistan
Chapter 12
Solving Fuzzy Optimization Problems of Uncertain Technological Coeffcients with Genetic
Algorithms and Hybrid Genetic Algorithms Pattern Search Approaches .......................................... 344
Pandian Vasant, University Technology Petronas, Malaysia
About the Contributors .................................................................................................................... 369
Index ................................................................................................................................................... 377
vii
Foreword
Since the beginning of our civilization, the human race has had to confront numerous technological
challenges such as finding the optimal solution of various problems including control technologies,
power sources construction, and energy distribution, amongst others. These examples encompass both
ancient as well as modern technologies like automatic theatre controlled by special programmes in
ancient Greece, the first electrical energy distribution network in USA, mechanical, electronical, as well
as computational controllers, et cetera. Technology development of those and related areas has had and
continues to have a profound impact on our civilization and lifestyle.
The topics discussed in this book are of these mentioned areas and mutually joined into a compre-
hensive text, which while discussing the specific selected topics, give a deeper insight to the interdis-
ciplinary fusion of those modern and promising areas of emerging technologies. This book discusses
the mutual intersection of interesting fields of research, as hybrid renewable energy and energy saving,
solar and fuel cells, power systems, chaos and power quality, soft computing, simulators, and software
engineering, amongst others. Novel techniques are also discussed in this book, which are able to handle
tasks such as control of various technological and energetical systems, optimization by means standard,
as well as novel methods. Together with many interesting emerging technologies, a reader will also find
in the book various mathematical and algorithmical methods used for proposed technologies including
models like fuzzy logic, neural network, evolutionary algorithms, or Hybrid System Optimization.
Therefore, this book titled Innovation in Power, Control and Optimization: Emerging Energy
Technologies, edited by Pandian Vasant, Nader Barsoum, and Jeffrey Webb, is a timely volume to be
welcomed by the community focused on power control and optimization as well as computational intel-
ligence community and beyond. This book is devoted to the studies of common and related subjects in
intensive research fields of emerging technologies. For these reasons, I enthusiastically recommend this
book to our scientists and engineers working in the above mentioned fields of research and applications.
Ivan Zelinka
Czech Republic January 2011
Ivan Zelinka (born 1965) is currently working at the Technical University of Ostrava (VSB-TU), Faculty of Electrical Engi-
neering and Computer Science. He graduated consequently at Technical University in Brno (1995 MSc.), UTB in Zlin (2001
Ph.D.) and again at Technical University in Brno (2004 assoc. prof.) and VSB-TU (2010 - professor). Before academic
career he was an employed as TELECOM technician, computer specialist (HW+SW) and Commercial Bank (computer and
LAN supervisor). During his career at UTB he proposed and opened 7 different lectures. He also has been invited for lectures
at 7 universities in different EU countries plus role of the keynote speaker at the Global Conference on Power, Control and
viii
Optimization in Bali, Indonesia (2009) and Interdisciplinary Symposium on Complex Systems (2011), Halkidiki, Greece.
He is responsible supervisor of grant research of Czech grant agency GAR named Softcomputing methods in control, co-
supervisor of grant FRV - Laboratory of parallel computing. He was also working on numerous grants and two EU project
like member of team (FP5 - RESTORM) and supervisor (FP7 - PROMOEVO) of the Czech team. Currently he is a head of
Department of Applied Informatics and in total he has been supervisor of more than 27 MSc. and 19 Bc. diploma theses. Ivan
Zelinka is also supervisor of doctoral students including students from the abroad. He was awarded by Siemens Award for his
Ph.D. thesis, as well as by journal Software news for his book about artificial intelligence. Ivan Zelinka is a member of British
Computer Socciety, Machine Intelligence Research Labs (MIR Labs - http://www.mirlabs.org/czech.php), member of expert
team of company DaySpring Global Multinational Inc., division Knowledge Management & Mining division (see also http://
www.dsgm.ca/consulting.asp), IEEE (committee of Czech section of Computational Intelligence), a few international program
committees of various conferences and three international journals (Associate Editor of MSE, Hindawi, http://www.hindawi.
com/journals/mse/editors.html, Editorial Council of Security Revue, http://www.securityrevue.com/editorial-council/, Edito-
rial board - Journal of Computer Science, Riga, Latvia). He is author of journal articles as well as of 5 books in Czech and 8
chapters in 6 books in English language.
ix
Foreword
This book is a challenge!
The challenge to a reader, who is interested in new power technologies, but does not have a solid
technical background.
Possibly, reader has to have the technical skills.
This book does not look like a fiction, but it is a scientific fiction which is becoming a reality!
To read fiction or scientific fiction is much easier than to read scientific books.
It reminded me my story about reading books on hard sciences, such as quantum mechanics.
When I was the first year student at the Moscow Power Engineering Institute, the Russian transla-
tion of famous American Physicists Richard Feynman Physics Course became available in our student
bookstore. There were many complicated concepts and formulas over there but I enjoyed this book
because between the formulas I could find very exciting belletristic.
I hope the reader will find here the exciting belletristic pieces with combination of strict technical
approaches and formulas describing various hot topics.
The reader will find answers to the question: what is happening in new emerging power technologies?
This book is a challenge because it covers a wide spectrum of problems from optimal configuration
of electric distribution networks to smart grids, and from monitoring power grid to renewable energy
technologies.
The energy market, as you know, is on the verge of a vast transformation.
Just take a look at this book and see what kind of energy innovation is appearing!
Igor Tyukhov
Moscow State University of Environmental Engineering, Russia
Igor I. Tyukhov is Executive Director of UNESCO Chair Ecologically clean engineering at the Moscow State University of
Environmental Engineering (MSUEE) and Deputy Chair Holder of the UNESCO Chair Renewable Energy and Rural Elec-
trification (part time) at the All-Russian Research Institute for Electrification of Agriculture (VIESH), member of the Interna-
tional Solar Energy Society, and Associate Editor of Solar Energy Journal. He graduated from the Moscow Power Engineering
Institute, V. A. Fabrikant Physics Department, 1972. He got degree kandidat technicheskikh nauk (Ph.D.) at 1979. He has
been with the Moscow Power Engineering Institute for more than 30 years teaching various physics disciplines and conduct-
ing research work on solar energy, solar concentrators, optical metrology, and semiconductor technology . Dr. Tyukhov was
visiting scholar at George Mason University (GFDP, 1999/200 academic year), at University of Oregon (Fulbright Program,
2002/2003), and at the Oregon Institute of Technology (2003). Dr. Tyukhov is expert in the field of photovoltaics and renewable
energy. He is author more than 200 papers, more than 20 patents, several chapters in the books and coauthor of monograph.
x
Preface
Many engineering systems and science problems suffer from the issue of developing a system that can
cope with variations of system or control parameters, measurements uncertainty, and complex multi-
objective optimization criteria. The need for a priori knowledge and the inability to learn from past
experience make the design of robust, adaptive, and stable systems a difficult task. Currently, research
on energy resources is of great importance for future oil replacements, particularly in vehicles and
other transportation. Computational intelligence has been proven to provide successful solution of com-
plex optimization problems by fuzzy logic, neural networks, evolutionary algorithms, and genetic algo-
rithms. They include system identification, parameter estimation, multi-objective optimization, robust
solutions, adaptive systems, self-organization, and failure analysis.
This book aims to provide relevant theoretical frameworks and the latest empirical research findings
in these areas. It is written for professionals who want to improve their understanding of the strategic
role in the area of power, control, and optimization. Each book chapter is written by experts in their
particular field of expertise.
Chapter 1 of this text describes coordinated intelligent operation and emergency control of electric
power systems.
In Chapter 2 a Hopfield Lagrange network (HLN) is proposed for solving economic load dispatch
(ELD) problem. HLN is a combination of Lagrangian function and continuous Hopfield neural network
where the Lagrangian function is directly used as the energy function for the continuous Hopfield neural
network.
The increased availability of reliable and efficient energy services that stimulates new develop-
ment alternatives such as solar, wind, et cetera is discussed in Chapter 3. This chapter elaborates on
the potential for such integrated systems in the stationary and portable power market in response to the
critical need for a cleaner energy technology. Anticipated patterns of future energy use and consequent
environmental impacts (acid precipitation, ozone depletion, and the greenhouse effect or global warm-
ing) are comprehensively discussed in this chapter.
Chapter 4 describes a dynamic smart Grid concept which enables electricity end-users to be acting
on controlling, shifting, or curtailing own demand to avoid peak-demand conditions according to infor-
mation received about electricity market conditions over the Internet.
The global warming and energy need requires developing emerging energy technologies for the
electricity, heat, and transport markets are subject of discussion in Chapter 5. In this chapter are also
discussed in great detail the emerging energy technologies that aim at increasing efficiency of energy
utilization processes from energy sources and diminish CO
2
exhalation. The main aim of the chapter
is to exhaustively present soft computing and computational intelligent techniques in the evaluation of
emerging energy technologies.
xi
Chapter 6 presents an overview of key issues and technical challenges in a regional electric net-
work, following the integration of a considerable amount of wind power. A brief survey on wind power
system, the present status of wind energy worldwide, common dynamic models, and control loops for
wind turbines is given.
Modified Lagrangian bounds and a greedy heuristic are proposed and discussed in Chapter 7 for
many-to-many assignment problems taking into account capacity limits for tasks and agents. A feasible
solution recovered by the heuristic shown to speed up the subgradient technique to solve the modified
Lagrangian dual. A numerical study is presented to compare the quality of the bounds and to demonstrate
the efficiency of the overall approach.
Energy projects with extended life cycles and initial investments can be unprofitable under discount
cash flow methods. Therefore, real options analysis has become relevant as a pricing technique for these
types of projects, with private risks and high investment levels. Following this question, the work pre-
sented in Chapter 8 analyses different real options approaches to select the most acceptable for investing
decisions in the energy sector.
Power loss reduction is considered as one of the main purposes for a distribution systems design-
ers and operators especially for recent non-governmental networks. Moreover, the nature of power
loss challenges different methods to solve this problem, while various studies indicate effectiveness
of reconfiguration and its high portion for this case. Thus, reconfiguration can be introduced as an
optimization procedure to obtain economical high quality operation by changing the status of sectional-
izing switches in these networks. Some major points, such as using different switch types, considering
number of switching, and time varying loads which are almost neglected or not applied simultaneously
in most pervious essays are discussed in Chapter 9.
In Chapter 10, the power system is considered as a continuum, and the propagated electromechani-
cal waves initiated by faults and other random events are studied to provide a new scheme for stability
investigation of a large dimensional system. For this purpose, the measured electrical indices (such as
rotor angle and bus voltage) following a fault in different points among the network are used, and the
behavior of the propagated waves through the lines, nodes, and buses is analyzed. The impact of weak
transmission links on a progressive electromechanical wave using energy function concept is addressed.
In Chapter 11 analyses and monitoring of the power grid in Pakistan is presented.
Finally, in Chapter 12, a solution is proposed to a certain nonlinear programming difficulties related to
the presence of uncertain technological coefficients represented by vague numbers. Only vague numbers
with modified s-curve membership functions are considered. The proposed methodology consists of novel
genetic algorithms and a hybrid genetic algorithm pattern search (Vasant, 2008) for nonlinear program-
ming for solving problems that arise in industrial production planning in uncertain environments. Real
life application examples in production planning and their numerical solutions are analyzed in detail.
The new method suggested has produced good results in finding globally near-optimal solutions for the
objective function under consideration.
xii
The editors of this text want to thank all the contributors to this text for their time, energy and invalu-
able expertise that we believe will make this book a success and extremely valuable resource in the area
of power management, control and optimization of engineering problems.
Pandian Vasant
Petronas University of Technology, Malaysia
Nader Barsoum
Curtin University, Malaysia
Jeffrey Webb
Swinburne University of Technology, Malaysia
xiii
Acknowledgment
We would like to this golden opportunity to sincerely thank the following friends and colleagues of us
for their valuable help and strong support of book chapters of the manuscript. Their marvelous feedback,
opinion, constructive comments, and suggestions for the improvement of the overall outstanding qual-
ity of the book chapters are gratefully acknowledged.
Nikolai Voropai, Energy Systems Institute, Russia
Dragica Vasileska, Arizona State University Tempe, USA
Igor Litvinchev, Nuevo Leon State University, Mexico
Hassan Bevrani, Kumamoto University, Japan
Cengiz Kahraman, Istanbul Technical University, Turkey
Rainer Burkard Graz, University Of Technology, Austria
Milorad Bojic, University of Kragujevac, Serbia
Gerardo M. Mendez, Instituto Technologico de Nuevo Leon, Mexico
Blanca Prez Gladish, Universidad de Oviedo, Spain
Furthermore, we sincerely thank the group of IGI Global at Hershey PA, USA, for their great help
and excellent support on this book project. In particular, special thanks go to Ms. Jan Travers, Ms. Myla
Harty, and Mr. Dave De Ricco of IGI Global for their great help.
Last but not least, we sincerely express our sincere thanks and appreciation to members of PCO
Global for their great support.
Pandian Vasant, Nader Barsoum, Jeffrey Webb
January 2011
1
Copyright 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Chapter 1
DOI: 10.4018/978-1-61350-138-2.ch001
Nicolay Voropai
Energy Systems Institute, Russia
Irina Kolosok
Elena Korkina
Alexey Paltsev
Anna Glazunova
Victor Kurbatsky
Nikita Tomin
Alexander Gamm
Irina Golub
Roman Bershansky
Daniil Panasetsky
Dmitry Efmov
Dmitry Popov
Christian Rehtanz
University of Dortmund, Germany
Ulf Hger
University of Dortmund, Germany
Coordinated Intelligent
Operation and Emergency
Control of Electric
Power Systems
ABSTRACT
This chapter presents the following approaches and developments: (1) the approach to power system
state estimation based on structural and functional decomposition. PMU measurements are used to
coordinate the solutions obtained in individual areas; (2) a non-iterative method to calculate voltage
2
Coordinated Intelligent Operation and Emergency Control of Electric Power Systems
INTRODUCTION
Last decades electric power industry is being lib-
eralized and restructured in many countries. This
process forces governments and science to learn
what market structures are the most efficient and
how regulation methods influence the industry and
welfare of the people. Consumers are interested
in optimizing their activity in a new environment
and investors tend to accurately forecast the pros-
pects of companies in electric power industry and
related industries. The growing interest in this
field generates necessity to exchange ideas and
research results.
The trends in expansion of electric power
systems and changes in the conditions of their
operation have led to complicate power system
operation, increased its changeability and unpre-
dictability that call for prompter and more adequate
response of controls systems.
Operational management deals with emer-
gency state prediction concept (probability of
emergency state occurrence). Prediction functions
can be realized by means of different advice-giver
software. But anyway, the final decision (control
action) is realized by a system operator who for
a variety of reasons is not able to realize rapid
and economically ineffective control actions that
would prevent an accident development. In such
a manner, emergency control schemes do not
predict the possible development of the normal,
emergency or postemergency states but operate
only when the disturbance has occurred.
It is possible to suppose that there are two main
disadvantages of the existing control ideology.
First disadvantage is the absence of fast control
actions realization in operational management.
And the second one is the absence of prediction
procedures in emergency control.
According to the authors opinion, these dis-
advantages may have been one of the causes of
the blackouts that took place all over the world
over the last several decades. Describing the
disadvantages authors suggest a possible ways of
developing and improving of the existent control
systems (Panasetsky, 2009).
The main idea is that the new methods that deal
with voltage instability and cascade line tripping
must complement and do not contradict to the
existing ideology. The new control system can be
built by using distributed intelligence principles.
The distributed intelligence is taken to mean the
multi-agent systems.
Development of systems and devices for
monitoring the state of energy and electrical
equipment (devices and systems of diagnostics)
and also monitoring the Electric Power System
(EPS) operation conditions seems to be highly
important because of radically changed develop-
ment trends and complicated operating conditions
of large-scale Interconnected Power Systems
(IPSs) (Kurbatsky, 2009; Voropai, 2010).
Modern systems for measurement of power
system state variables and their control, new
communication and information processing
systems, etc. allow creation on a new basis and
with essentially higher efficiency one of the most
important stages of power system control their
operation and emergency control.
magnitude and phase at all the buses, allowing the state parameters of EPS to be obtained fast enough;
(3) the approach to the super short-term forecasting of state variables on the basis of neural network
technologies and algorithms of nonlinear optimization that is realized in the ANAPRO software; (4)
an analysis of the possibility of determining weak ties and cut sets in EPS; (5) a control system based
on the multi-agent technique; (6) the development of selective automatic systems intended to prevent
and eliminate out-of-step operation on the basis of synchronized voltage phase measurements obtained
from Phasor Measurement Units.
3
The chapter includes: new conditions and
requirements for power system monitoring and
control; wide area monitoring of power system
state using new technologies and tools; compre-
hensive emergency control systems and defense
plans for large power systems; coordinated con-
cepts and control systems for improving stability
and security of power systems; coordinated wide
area monitoring and control of large power sys-
tems; development of computer technologies and
modern tools towards smart power grids
The Problem of Monitoring,
Forecasting and Control in
Electric Power System
The trends in expansion of EPSs and changes
in the conditions of their operation have led to
considerable transformations in their structure and
operation. These transformations are conditioned
by the following factors:
Increase in scale, expansion of territories
to be serviced, interconnection of dif-
ferent power systems for joint operation
which results in creation of interregion-
al, interstate and intercontinental power
interconnections;
Decentralization of power supply due to a
wider use of distributed generation sources
that are connected to the distribution net-
work nodes;
Restructuring of power industry, which of-
ten makes its structure drastically different
from the technological structure of EPS as
a technically single unit and from its con-
trol structure;
Liberalization of relationships in the elec-
tric power industry which leads to many
participants of relationships with different,
often opposing, interests, in the course of
expansion, operation and control of power
system.
All these factors essentially complicate power
system operation, increase its changeability and
unpredictability, raise danger of severe emergen-
cies with undesirable development and massive
consequences for a system and consumers and,
therefore, call for prompter and more adequate
response of control systems. This generates the
need to improve and develop principles and sys-
tems of power system operation control which
can be based on:
New systems for measurement of operat-
ing parameters (PMU) and their control
(FACTS, energy storage devices etc.) that
signifcantly improve EPS observability
and controllability;
Modern communication systems, new in-
formation technologies and artifcial intel-
ligence methods, highly effcient comput-
ers, which totally change the processes
of acquisition, transmission, presentation
(visualization) and use of information on
power systems;
Effcient mathematical control theory
methods in multicriteria non-coincident
conditions.
Based on the above circumstances the so
called Smart Grid concept was developed (Of-
fice, 2003 and European, 2006). This problem
is very often associated with the integration of
renewable energy sources in EPS as well as with
distribution electric network based on the informa-
tion technologies and artificial intelligence in the
distributed control systems of power supply and
power consumption (Chuang et al. 2004; Amin et
al. 2005). The most comprehensive understanding
of the Smart Grid concept to date has been given
by Shahidehpour (2009). In general the Smart
Grid concept can be represented as a set of the
following components:
Generation (enhancement of reliability
and economic effciency of electric ener-
4
gy production through the use of modern
highly intelligent systems for monitor-
ing and control, integration of renewable
sources, distributed generation and energy
storage devices on the basis of the Internet
technologies);
Electric transmission network (wide scale
monitoring of operating conditions and
their control with the help of new devices
and technologies (FACTS, PMU, artifcial
intelligence, etc.) in order to provide pow-
er supply reliability and electric network
controllability);
Substations (automation of substations that
are based on the advanced electric equip-
ment by using modern systems for diag-
nosing, monitoring and control on the basis
of information and computer technologies
in order to provide reliability and control-
lability of substations);
Distribution electric network (a radical
increase in its controllability and reliabil-
ity through introduction of distributed mi-
croprocessor-based control and protection
systems with the use of new information,
computer and Internet technologies);
Consumers (equipping them with highly
intelligent systems intended for electricity
control and metering, demand-side man-
agement and load control in emergency
situations).
In Russia power industry reforms and new
methods and technical tools of control have also
encouraged the development of principles and
methods of EPS dispatching and emergency con-
trol (Ayuev et al, 2008; Voropai, 2008).
Let us consider the problems of improving the
principles and enhancing the efficiency of systems
intended for EPS operation and emergency control.
The most important directions here include an
increase in control adaptability and expansion of
coordination among control stages, devices and
systems. For this purpose it is necessary to develop
an efficient system of wide scale monitoring and
forecasting of operating conditions and control of
EPS. Figure 1 shows power system states and the
blocks of problems on monitoring, forecasting and
control that correspond to these states.
The blocks of monitoring and forecasting of
the EPS normal, pre-emergency and post-emer-
Figure 1. Time diagram of events in the system for monitoring, forecasting and control in EPS
5
gency operating conditions include the following
problems:
System state estimation;
Forecasting the parameters of expected
operating conditions. Forecasting is nec-
essary because during the state estimation
procedure current state is estimated with
some delay, while monitoring and control
problems require some advance estimation
of system state (to control is to foresee);
let us note that for these two blocks of
problems the advance time can vary;
Detection of weak points in the system in
the expected operating conditions;
Determination of margins for transfer ca-
pabilities of ties in the expected condi-
tions; This is necessary to effciently use
the margins in the operating conditions and
automatic control through appropriate con-
trol actions;
Visualization of the expected conditions;
Determination of indices and criteria for
transition from normal to pre-emergency
conditions and, vice versa, from post-
emergency to normal conditions.
A special explanation should be given concern-
ing the last problem. It relates to the following
basic principles of power system operation control
in a market environment:
under normal operating conditions ef-
fcient contract relationships among the
participants of the wholesale markets for
electricity, capacity and ancillary services,
that are coordinated on a market basis;
under a threat of emergency conditions a
transition from market criteria of operation
control to the centralized principles of dis-
patching control;
under emergency and post-emergency con-
ditions the use of strictly centralized dis-
patching and automatic emergency control.
The enumerated problems have certain
specificity when applied to normal conditions,
on the one hand, and to pre-emergency and
post-emergency conditions, on the other hand.
Pre-emergency and post-emergency conditions
(at system restoration) require higher speed of
algorithm operation and smaller time intervals
between individual states for which the problems
of operation monitoring and control are solved.
A separate problem is monitoring of emergency
conditions. Dynamic nature of emergency condi-
tions requires that the results of monitoring be
promptly transferred to dispatcher. They should
have an integrated form and indicate dangerous
points in the system in terms of undesirable de-
velopment of emergency. The primary informa-
tion, however, obtained as a result of emergency
conditions monitoring, should arrive at the auto-
matic control devices and systems with the view
of their adaptation to the current power system
emergency state.
In addition to monitoring and forecasting of
the operating conditions the efficient adaptive
methods and algorithms for choosing the points
of control action application and adjustments on
the basis of advanced achievements of the control
theory and artificial intelligence play an important
part in increasing the adaptability of control. The
methods and algorithms are implemented by dis-
patcher and automatic control systems and provide
adaptation of controls to the current system state
and possible disturbances.
It is necessary to considerably expand and
improve coordination of control. This can be done
in the following directions:
1. Expansion of EPS operation control
coordination:
A. In terms of time - from coordinated
design of control systems to their
implementation through dispatching
and automatic devices;
6
B. In terms of situation coordination of
on-line dispatching, continuous auto-
matic and discrete emergency control.
2. Extension of a range of devices intended
for coordinated control of EPS operating
conditions:
A. Development of traditional systems,
including automatic voltage regulators
(AVR) and speed governors of synchro-
nous machines, automatic emergency
control devices, etc.;
B. Use of new devices for measurement
and control PMU, FACTS, energy
storage systems, etc.
3. Development and extension of principles and
systems for coordinated online dispatching
and emergency control to distribution electric
networks with distributed generation.
4. Harmonization of commercial interests of
participants in the markets for electricity,
capacity and ancillary services and the need
to provide EPS reliability and survivability.
5. Development of new criteria and new
methods for monitoring, forecasting and
control of operating conditions in order to
provide effective coordinated control for
all participants of the wholesale market,
power system reliability and survivability.
Implementation of the foregoing positions
will allow one to considerably enhance the
operation efficiency, controllability, reli-
ability and survivability of modern power
systems.
Decomposition of Power System
State Estimation Problem
with the Use of PMU Data for
Large Dimension Schemes
State estimation (SE) of electric power systems
is an important procedure that allows on-line cal-
culation of state variables for a current scheme of
electric network on the basis of Tele-Information.
The obtained calculated model of power system is
then used to solve various technological problems
to effectively control electric power system.
Currently the System Operator - Central
Dispatching Office of Russias UES is creating
an integrated computational model that reflects
most completely the topology and operation of
UES, to solve a set of online dispatching control
problems instead of previous models that varied
in degree of detail and were applied to solve
individual problems.
The single computational model of UES/RPS
covers the entire backbone network of 220 kV
and higher; the lines of lower voltage classes,
that are important for market participants in terms
of correct description of power supply volumes,
boundaries of federal network company, interstate
power flows, electricity outputs of power plants;
and power plants with an installed capacity above
5 MW and large consumption nodes (Ayuyev,
2005). Currently the single computational model
includes about 7000 nodes, 10000 branches and
800 generators.
Similar situation is observed in the dispatching
practices in other countries. Creation of Western
European Union for the Coordination of Transmis-
sion of Electricity (UCTE), North American Elec-
tric Reliability Corporation (NERC) that embraces
most of the North-American power systems, etc.
has lead to the necessity to make calculations for
very large and sophisticated systems.
The calculations for a large system encounter
the problems related to the un-homogeneity of
calculated schemes, large volumes of various
data to be processed and the requirement for high
speed software.
Besides, the need for online state estimation of
such systems increases the burden on the available
computing resources in the EPS Control Center.
The distributed data processing at decomposi-
tion of the state estimation problem is an effective
method of solving these problems. Decomposi-
tion of power system state estimation problem is
addressed in a great number of scientific papers
in Russia (Gamm, 1983; Gamm, 1995) and other
7
countries (Clements, 1972; El-Kleib, 1992; Iwa-
moto, 1989 and others).
Until recently state estimation in EPS was
mainly based on the SCADA measurements: volt-
age magnitudes, branch power flow, nodal power
injections and, occasionally, current magnitudes.
The advent of WAMS (Wide-Area Measurement
System) that contains phasor measurement units
(PMU) as the main measurement equipment
(Phadke, 2002) makes it possible to synchro-
nously and accurately control the EPS state and
essentially improve the results of state estimation
(Gamm, 1997). The use of PMU measurements
offers new possibilities in decomposition of the
state estimation problem.
The paper considers the algorithm of state esti-
mation by the test equation technique (Clements,
1972) that employs structural decomposition of
state estimation problem, i.e. division of the cal-
culated scheme into subsystems, and functional
decomposition of the SE problem (detection of
bad data, state estimation on the basis of quadratic
and robust criteria). The two-level algorithm is
proposed to divide the calculated scheme into
subsystems for state estimation by the test equa-
tion technique. Application of the test equation
technique that allows the values of measured
variables to be fixed by setting zero variances
for them, and placement of Phasor Measurements
Units at boundary nodes (El-Kleib, 1992) make it
possible to essentially simplify the procedure of
coordinating the solutions obtained for separate
subsystems. In addition the paper presents the
algorithm of PMU placement at boundary nodes.
The example of calculation for a fragment of real
power system is given.
Decomposition of State
Estimation Problem
Decomposition of the state estimation problem
is based on structural (by subsystems) and func-
tional (by the problems solved) decomposition.
The structural decomposition is made by dividing
the calculated scheme into subsystems by one or
another method (Gamm, 2007). The functional
decomposition is made in accordance with the
problems solved within the SE procedure. The
main of them are: analysis of network topology
(formation of current calculated scheme); analysis
of observability; analysis of bad data; calculation
of estimates and calculation of steady state with
regard to the estimates obtained.
Methods of Structural Decomposition
The calculated scheme can be divided into subsys-
tems by the following techniques: decomposition
utilizing geographical characteristics (Gamm,
1995; Falcao, 1995), decomposition by boundary
nodes (Gamm, 1983; Clements, 1972), by tie-lines
(Gamm, 1983; Gamm, 1995; Abdel-Rahman,
2001), based on the structure of gain matrix
(Wallach, 1981), by Danzig-Wolf decomposition
algorithm (El-Kleib, 1992), and others.
The main algorithms of SE problem decompo-
sition suggest dividing the calculated scheme into
subsystems whose boundaries are either nodes or
branches. In this case the SE problem is solved
iteratively unless the boundary conditions are
met. The method of decomposition with bound-
ary nodes has been chosen for realization in the
offered algorithm.
In this case the equality of voltage magnitudes
and phases at the boundary nodes should be met [2]:
U U U
i j k
= = = ... ; (1)

i j k
= = = ... ; (2)
Besides the boundary balance relationships
should be met. For example for boundary node l,
common for the i, j, , k-th subsystems
P P U U
l lm l l m m
m s i j k
s
+ =
=

( , , , )
, ,...,

0 ; (3)
8
Q Q U U
l lm l l m m
m s i j k
s
+ =
=

( , , , )
, ,...,

0 ,
(4)
where
s
- a set of nodes of the s-th subsystem,
that are adjacent to the l-th node.
The Use of PMU Data in Structural
Decomposition
Development and improvement of software for
monitoring and control of power systems at a
qualitatively new level have become possible ow-
ing to WAMS (Wide-Area Measurement System)
that allows the EPS state to be controlled synchro-
nously and accurately. The devices for measuring
phasors (Phasor Measurement Units) are the basic
measurement equipment in this system.
The results of solving the state estimation
problem can be essentially improved by using
the PMU data. As compared to the standard set
of measurements received from SCADA system
PMUs placed at a node provide accurate (the error
is 0.2-0.5%) measurements of voltage magnitude
and phase at this node as well as the magnitudes
and phases of currents in the branches adjacent
to this node.
The possibilities of using synchronized pha-
sor measurements for distributed state estimation
were discussed in (Zhao, 2005; Jiang, 2007; Jang,
2008 and others). In (Zhao, 2005) the method is
suggested to decompose the calculated scheme
into the areas with PMU to be installed in each
area. The data from these PMU are then used to
solve a coordination problem. In (Jiang, 2007)
the authors suggest placing PMU at a basic node
of each area. The PMU measurements coordinate
the SE problem solution of each area. (Jang,
2008) presents a diakoptic-based distributed SE
algorithm, in which PMUs are used to coordinate
voltage angles of each area SE solution.
The chapter suggests the use of PMU measure-
ments for distributed state estimation to coordinate
the solutions obtained for individual areas. For
PMU placement under SE problem decomposi-
tion the algorithm based on the annealing method
was developed. They will be presented in Section
4. Compared to the PMU-based decomposition
methods proposed in (Zhao, 2005; Jiang, 2007;
Jang, 2008) the number of PMUs in our case study
did not increase, yet made it possible to perform
parallel state estimation by subsystems, to solve
coordination problem without iteration and obtain
an optimal but not pseudo-optimal, as in (Zhao,
2005), solution that coincides with the solution
for the entire network.
Placement of PMUs at boundary nodes makes
it possible to register boundary variables U and
measured highly accurately. In this case the
operating conditions of some subsystems can
be calculated independently of one another and
solution of the coordinated problem consists in
calculating nodal injections by (3), (4) using the
estimates of power flows.
Algorithm of Structural Decomposition Using
Test Equations
The idea of decomposing the state estimation
problem with PMU placement at boundary nodes
is rather attractive. In reality, however, due to
high cost of PMUs they can only be used when
the number of boundary nodes is small.
To calculate large inhomogeneous schemes
the authors propose a method of dividing the
calculated scheme with respect to voltage levels
(Wallach, 1981). This method decreases essen-
tially a negative impact of un-homogeneity of
calculated scheme and telemetric information in
calculation of subsystems of one voltage class but
for the complex scheme inevitably leads to a large
number of boundary nodes. Therefore, the paper
proposes a two-stage algorithm to decompose the
calculated scheme into subsystems that combines
the positive features of both approaches.
At the first stage the scheme is divided into
rather large areas with minimum number of inter-
system ties and boundary nodes. This decomposi-
tion can be made on the basis of administrative
9
division, for example, the entire scheme of Russias
UES is decomposed into regional power subsys-
tems of large regions in the country that operate
in parallel or it can be decomposed artificially
into separate areas by special algorithms (Gamm,
1995). PMUs are placed at the boundary nodes of
the areas. Highly accurate measurements obtained
from PMU make it possible to register the values
of magnitudes and phases of nodal voltages at the
boundary nodes and make calculations for the
areas in parallel.
At the second stage the calculated scheme of
each area in turn is divided into subsystems that
correspond to the levels of nodal voltages. The
calculations start with the subsystem of the highest
voltage level (750-500 kV). Normally this part of
the scheme is well provided with highly accurate
telemetry and contains a basic node. Then the
calculations are made successively for the rest
of the subsystems. The subsystems are ranked
by voltage levels (220 kV, 110 kV, etc.). Every
time the node bordering the subsystem of higher
voltage level is chosen as a basic one.
After the calculation of the low level sub-
systems a coordination problem is solved for
all areas. In this case boundary conditions (1),
(2) are met automatically, and the coordination
problem implies calculation of nodal injections
at the boundary nodes on the basis of power flow
estimates obtained for each area (Equations (3)
and (4)).
Functional Decomposition
The functional decomposition of the SE problem
is performed in accordance with the problems
solved within the state estimation procedure. The
main of them are: analysis of network topology;
analysis of observability; analysis of bad data;
calculation of estimates and steady state by the
estimates obtained.
The current calculated scheme is formed for
the entire scheme. Bad data analysis and calcula-
tion of estimates and steady state are performed
by the test equation technique for each subsystem
of a certain voltage class before solving the state
estimation problem (Gamm, 2007).
State estimation is made according to two
criteria: the method of weighted least squares
and the robust criterion that allows the estimates
to be obtained and bad data to be suppressed
simultaneously.
Control is transferred to one or another state
estimation program depending on operation of
the bad data detection program. In case of bad
data detection or their absence the program for
calculation of estimates operates on the basis of
the least squares method. However, if it is im-
possible to detect erroneous measurements and,
hence, identify bad data the program operates
according to the robust criterion (Gamm, 2005).
State estimation is made starting at the upper level
of the structural decomposition.
Full Algorithm
The full algorithm for solving the state estima-
tion problem based on structural and functional
decomposition is as follows.
1. The complete calculated scheme of EPS is
decomposed into rather large areas. Phasor
measurement units are placed at the boundary
nodes of subsystems. In the subsystems that
have no basic node of the complete scheme
one of the boundary nodes with PMU of the
highest voltage class is chosen as a basic
one. Measurements of nodal injections at
boundary nodes are excluded from the vector
of measurements.
2. At the second stage of decomposition the
calculated scheme of each area is divided into
subsystems that correspond to the levels of
nodal voltages. The boundaries of the sub-
systems are the nodes adjacent to the nodes
of the voltage class of this subsystem. For
example for the 750-500 kV voltage class
subsystem the nodes with the voltage of
220 kV are boundary nodes and vice-versa.
10
3. The calculation starts with the subsystem of
the highest voltage level (750-500 kV) for
each subsystem. Normally this part of the
scheme is well provided with highly accurate
measurements and contains a basic node. The
state estimation algorithm for subsystems
with boundary nodes is as follows:
A. For each subsystem that contains
boundary nodes the problem of bad
data detection is solved by the test
equation technique.
B. In case of bad data detection or their
absence the state estimation is made
according to least squares method.
C. In the event that erroneous measure-
ments cannot be detected and hence it is
impossible to detect bad data, the state
estimation is made according to the
robust criterion (bad data suppression).
4. The rest of the subsystems in the scheme are
successively calculated. They are ranked by
voltage level (220 kV, 110 kV, etc.). Every
time the node bordering the subsystem of
higher voltage level is chosen as a basic node.
The estimates of the boundary variables of
the state vector that are obtained at the upper
level of decomposition are registered.
5. The injections at boundary nodes between
the subsystems of different voltage class are
calculated.
6. After all subsystems of the first level of
decomposition have been calculated similar
problem is solved for the boundary nodes
with PMU.
Placement of PMU at Decomposition
of the State Estimation Problem
For decomposition of power system state estima-
tion problem it is necessary to maintain accurate
values of voltage magnitudes and phases at
boundary nodes of subsystems for iteration-free
solution of coordination problem. A simple but
not an optimal solution is placement of PMUs at
all boundary nodes. Based on the measurements
to be received from the placed PMUs the voltage
magnitude and phase at a neighboring node can be
calculated using the electrical circuit equations. In
the paper the vector voltage measurement obtained
by the equations is called the calculated PMU.
The study shows that the accuracy of parameters
of the calculated PMU practically equals the ac-
curacy of measurements of the physical PMU
(Kolosok, 2009). With an optimal combination of
physical and calculated PMU at all boundary nodes
of subsystems it is possible to determine voltage
magnitudes and phases required to coordinate
the solutions obtained for individual subsystems.
In order to minimize the number of PMUs we
analyze not only the list of boundary nodes but
the list of internal lines within subsystems that
are incident to these nodes as well. The bound-
ary nodes may happen to belong to one and the
same subsystem and bound one and the same
line. Then it is enough to place a physical PMU
at one end of the line and a calculated PMU at
the other. To choose the optimal PMU placement
the algorithm based on the simulated annealing
method was developed.
A fragment of a real scheme divided into 3
subsystems (Figure 2 and Figure 3) will be con-
sidered for illustration of the algorithm operation.
Figures 2 and 3 show a variant of system divi-
sion into subsystems with boundary lines: between
subsystems I and II - lines 2-5 and 4-7, between
subsystems I and III lines 3-9 and 4-9, and
between subsystems II and III lines 7-10.
The objective function of the annealing method
in this problem has the form:
min
var
E
K K
K K
PMU iants
calculated subsystems
=
+
+
1 2
3 4
, (5)
where
K
PMU
1 the number of placed PMUs;
11
K
iants
2
var
the number of calculated PMU
variants, their existence implies that there
are redundant PMUs in the scheme.
K
calculated
3 the maximum number of cal-
culated PMUs enabling one to obtain ac-
curate,U ;
K
subsystems
4 the number of subsystems
with PMUs. It is desirable that every sub-
system has at least one PMU.
The problem starts with the choice of random
nodes, preferably those with maximum connectiv-
ity, and with the assignment of PMUs with accurate
measurements of
pmu pmu ij ij
U I
pmu pmu
, , , to these
nodes. Further the possibility of obtaining the
calculated PMU at the ends of incident branches
is determined. At each step the objective function
is calculated. If the result E
new
, obtained by (5)
after some iteration is less than the previously as-
sumed optimal result ( E E
new opt
< ), a new variant
of assignment is taken E E
opt new
= , if the result is
not less - whether or not the return to the previous
step occurs depends on the value of wrong decision
probability P E e
E k T
b
( )
/( )

=

, where
E
k T
b

the Metropolis criterion, an analog of Boltzmann
factor.
Figure 2. Placement of PMUs in intersystem lines.
PMUs at nodes 4,6,12
Figure 3. Placement of PMUs at boundary nodes. PMU at node 7
12
The criterion of operation completion is
min E .
The practical result of searching and obtain-
ing the optimal placement of PMU at division of
system into subsystems with boundary lines is
given in Table 1, line 1. Figures 2 and 3 show that
with an account taken of the calculated scheme
topology the optimal result is given by PMUs
(black squares) placed at boundary node 4 and
nodes 6 and 12 which are adjacent to the bound-
ary ones. The circles denote the nodes for which
the calculated PMUs are obtained.
Figure 3 presents the variant of dividing system
into subsystems by boundary nodes: between
subsystems I and II - nodes 5 and 7, between
subsystems I and III nodes 4 and 7, between
subsystems II and III boundary node 7. Table
1, line 2, shows the optimal PMU placement
under the following division into subsystems: the
optimal result E
opt
= 0 14 . is given by the PMU
(black square), installed at boundary node 7 which
is common for all subsystems, the calculated
PMUs provide pseudo-measurements at nodes
4,5,8,10, which are adjacent to the boundary ones
(denoted by circles).
The calculations show that the number of PMUs
placed at decomposition of the scheme is consider-
ably lower than the number of boundary nodes.
Calculation Example
In order to test the efficiency of the suggested
decomposition algorithm of state estimation
the calculations were made for a real scheme
consisting of 107 nodes and 175 branches. The
calculations were based on real measurements.
The efficiency of the algorithm was assessed by
comparing the results of calculations made for
subsystems to the results of the calculation made
for the entire scheme.
At the first stage the Genetic Algorithm (Ko-
losok, 2003) was used to divide the entire scheme
into two subsystems containing 55 and 52 nodes
including 6 boundary nodes in which the PMU
data (measurements of magnitudes and phases of
nodal voltages) were modeled. The calculations
of these subsystems were carried out in parallel
which reduced the time of solving the SE problem
almost twice: from 0.49 s to 0.27 s.
At the second stage of decomposition each
of the subsystems in turn was decomposed into
three subsystems corresponding to the voltage
levels of 500 kV, 220 kV, 110 kV and lower. The
calculation of these subsystems according to the
above algorithm was made successively; therefore
the full time of solution could increase. However,
owing to the improved convergence of the iteration
processes in the calculation of subsystems of the
same voltage class the total time of the calcula-
tion for all the three subsystems practically did
not change.
More efficient operation of bad data detection
algorithm and application of the robust criterion
of SE (Kolosok, 2003) for two of six subsystems
improved considerably the results of state estima-
tion: the value of the SE objective function at the
point of solution decreased almost by a factor
Table 1. The practical result of searching and obtaining the optimal placement of PMU at division of
system into subsystems with boundary lines
Division into
subsist.
PMU placement at
nodes, K
PMU
1
Variants of calculated
PMU, K
iants
2
var
Calculated nodes,
K
calculated
3
The number of
subsist. K4
Function
By lines 4,6 and 12 0 2,3,5,7,8,9,10,11 3 0,27
By nodes 7 0 4,5,8,10 3 0,14
13
of 6 and the estimates at boundary nodes were
noticeably improved.
Structural and functional decomposition of
state estimation problem is an effective method
to solve the problems arising during calculation
of large schemes.
The proposed two-level algorithm for structural
decomposition of the SE problem allows one to
simultaneously process the data for local sub-
systems of considerably smaller dimensionality;
decrease the adverse impact of un-homogeneity of
the calculated scheme and telemetric information
when calculating one-voltage-class subsystems;
essentially simplify solution of the coordination
problem which, in this case, does not require itera-
tive calculations by subsystems; and reduce the
time for SE problem solving for the entire scheme.
Functional decomposition of the SE problem
allows one to coordinate interaction between the
problems solved at different levels, organize a
flexible choice of the method to solve one or an-
other state estimation problem for each subsystem,
integrate the methods of artificial intelligence
and numerical methods, accelerate the process
of measurement processing, and, thus, reduce the
time of system state estimation.
The use of measurements from PMUs placed
at boundary nodes of subsystems allows state esti-
mation to be performed for individual subsystems
independent of one another. Solution of coordi-
nated problem in this case implies calculation of
nodal injections at boundary nodes and does not
require iterative calculations.
The developed algorithm of placing PMU by
the annealing method and the use of calculated
PMU make it possible to reduce the number
of PMUs required for solving the coordination
problem which in this case is essentially lower
than the number of boundary nodes.
Simulation calculations as well as the calcula-
tion of a real scheme demonstrate the efficiency of
the suggested approach to electric power system
state estimation.
PMU for Fast Calculation of Steady
State in Electric Power Systems
The current state of an electric power system
(EPS) can be properly determined by a minimum
set of state variables that will make it possible to
uniquely determine all the rest of state variables.
Such a set of variables is called a state vector. As
a rule it is voltage magnitudes and phases
x = { , } U .
The speed of determining the current system
state depends on the speed of calculating all state
vector components. As a rule the state vector
components are calculated in the process of state
estimation problem solution by the iterative
methods. In emergency situations the speed of
achieving the result can prove to be insufficient.
Measurement of these components by using PMU
is the most attractive method. PMU installed at
the bus provides measurements of voltage mag-
nitude U
i
and phase
i
at this bus, current mag-
nitude and angle value between voltage and
current
ij
in the entire branches incident to this
bus or in some of them subject to transmission
capacity of communication channels (Phadke,
2002). In practice, however, installation of PMU
at each bus is impossible.
The paper presents a non-iterative method to
calculate voltage magnitude and phase at all the
buses, allowing the state parameters of EPS to be
obtained fast enough.
Problem Statement
To calculate load flow in EPS fast, the complex
network configurations are reduced to a radial
form and the voltage magnitude and phase at all
buses are calculated directly. For this purpose an
optimal set of measurements from the SCADA
system and PMU and the known relations between
the state variables in EPS are applied.
The calculated PMU is installed at bus j. Instal-
lation of the calculated PMU means calculation of
14
voltage magnitude and phase at the neighboring
buses by the formulas:
U U I R X
j i ij ij ij
= + ( cos sin ) , (6)

j i
ij
i
arctg
I X R
U R X
=

( cos sin )
cos sin
.
(7)
The phases
k
at bus k are calculated through
measurements of ( U P Q
k k j k j
, ,

). To do this, the
active and reactive voltage drops ( , )
' ''
U U
are calculated by the expressions
U
P R Q X
U
k j k j
k
'
=
+

(8)
U
P X Q R
U
k j k j
k
''
=

(9)
then
k
is determined

k j
j
arctg
U
U U
=
''
'
. (10)
With the measurements shown in Figure 3 the
voltage magnitude at the bus k is calculated by
the formula:
U U U U
k k
= + ( ) ( )
' ''

2 2
. (11)
k
is calculated by expression (5), where
( , )
' ''
U U are determined through
P Q U
j k j k k
,
,
.
Calculation of , U at all EPS Buses
Calculation of , U starts with the reference bus
that is determined in accordance with (Glazu-
nova, 2009). PMU is installed at the reference
bus. The process of searching for the reference
bus reveals buses and branches with insufficient
measurements for determination of , U at the
next bus. Their calculation can be continued by
installation of additional PMUs. These buses are
also called reference buses.
Algorithm for Calculation of , U
1. Reduction of the scheme to a radial form
(removal from the network graph of buses
forming loops).
2. Determination of reference buses, from
which the voltages are calculated in each
radial scheme. Installation of PMUs at the
reference buses.
3. Installation of the calculated PMUs at the
buses adjacent to the reference ones.
4. Calculation of voltage magnitude and phase
at all buses of each radial scheme.
5. Calculation of voltages at the removed
buses. To do this, the state estimation of
each scheme is performed, with the number
of buses equal to m+1, where m number
of branches adjacent to the removed one.
6. Comparison of values of voltage phasors
that are obtained at point 4 with the values
calculated at point 5. If the difference is
less than the threshold d, then go to point
7, otherwise to point 11. The threshold is
determined by the formula:
d = 3 ,
where - variance of measurements. For voltage
u
2
2 = ; d Bm
u
= = 3 2 4 24 * . ;
15
d
o
= 0 2 . .
7. Calculation of the values of power flows
in the branches, limited by the buses where
voltage is corrected.
8. Adjustment of the injection values at the
buses, where voltage is corrected and at the
adjacent buses.
9. Calculation of the values of power flows in
the boundary branches, i.e. those connect
two radial schemes.
10. Analysis of the obtained results. Comparison
of the calculated values of power flows with
measurements. If the difference is less than
the threshold d, then go to point 11.
Where d MW
p
= = 3 25 15 * ,
d MVAr
q
= = 3 100 30 * .
11. Generation of the signal on the error in
measurements.
12. End of the algorithm work.
Practical Results
Examples of the 13-bus and 14-bus schemes il-
lustrate the possibility for non-iterative calcula-
tion of load flow by the measurements from the
SCADA system and data of PMU.
13-Bus Scheme
The states of the 13-bus scheme (Figure 4) were
calculated on the basis of real data.
Figure 4. Fragment of a real scheme
16
The given network scheme is reduced to the
radial form by removal from the graph of bus 3
as the most connected (Figure 4). The dashed lines
show branches adjacent to the removed bus. Bus
13 is a reference bus. For the non-iterative calcu-
lation of load flow to be made on the basis of
measurements presented in Figure 4 an addi-
tional PMU should be installed at bus 10. Branch
7-8 becomes a boundary one.
As a result the transformed scheme has two
radial schemes with the vertex at bus 13 (13-1-
11; 13-2; 13-12-5-4-6-7), and the vertex at bus
10 (10-8; 10-9), one boundary branch (7-8), four
branches adjacent to the removed bus.
PMUs are installed at buses 13, 10. Voltage
magnitudes and phases are calculated at the
neighboring buses 1, 2, 12, 8, 9 by using the data
of PMU from Equations (6), (7).
Voltage magnitudes (if it is needed) and phases
for other buses are calculated by the following
measurements from the SCADA system:
At buses 4, 6 by the active and reactive power
flows at the beginning of the branch (10), (11).
At buses 11, 5, 7 by the active and reactive
power flows at the end of branch and voltage in
the following succession:
1. Calculation of voltage drop from (8), (9);
2. Calculation of
k
by Equations (10), where
k=5, 7, 11.
Voltage at the bus removed from the scheme
shown in Figure 4 is calculated by separation of
the 5-bus scheme (Figure 5).
For the obtained scheme the state estimation
is performed based on available measurements
by the algorithm described in (Gamm, 2007).
Figure 5. Fragment of a real scheme
17
Voltages calculated by Equations (6)-(11) are
applied as initial data. Here it is taken into con-
sideration that the calculated PMUs are installed
at buses 2, 9, 12, i.e. the voltage magnitude and
phase at these buses are measured with a high
degree of accuracy. Then the following condition
is analyzed:
U U d
SE cal
< , (12)
where U
SE
estimated voltage values, U
cal

voltages calculated in the radial scheme. If condi-
tion (7) is true, it means that there are no errors
in measurements and the obtained state is adequate.
This condition holds true for bus 4:
509 508 4 25
SE cal
< . . (13)
The values of active and reactive power flows
in the branch 34 are calculated.
Balances at buses 3, 4 are maintained by correc-
tion of the injection values there. The calculation
results are given in Table 2.
The second column presents the values of
measurements. The results of SE of the 13-bus
scheme are given in column 3. The fourth column
shows results of SE for the 5-bus scheme. The
fifth column presents voltage values calculated
by the suggested algorithm. The Table 2 indicates
that the voltage phasor obtained by the non-iter-
ative method coincides with SE results.
The values of active and reactive power flows
in the boundary branch 7-8 are calculated based on
the known voltages. The values of obtained flows
and the values of flows from SE, measurements
of flows are shown in Table 3.
Table 2. Calculation results with real data
Measurement Estimate Estimate of 5-bus scheme Non-iterative method
2 3 4 5
U

U
U
cal

cal
753 753 0 753 0
746 747 0.103 747 0.10 747 0.103
517 503 0.118 503 0.11
502 505 0.159 506 0.16 505 0.160
505 505 0.171 505 0.173
498 500 0.153 501 0.154
497 504 0.114 504 0.117
512 507 0.104 508 0.107
515 500 0.096 500 0.09 501 0.099
515 503 0.102 503 0.102
741 743 0.193 743 0.191
512 502 0.113 502 0.11 502 0.113
785 740 0.083 740 0.083
18
Analysis of the obtained values reveals that
the power flows in the boundary branch are
within the normal range:
P P d
meas ss
< . (14)
Hence, the steady state of the 13-bus scheme
is calculated.
14-Bus Scheme
Figure 6 presents a 14-bus test IEEE scheme.
SS of the considered scheme is calculated by
using the described algorithm.
1. For reduction of scheme to the radial form
3 buses (2, 4, and 6) are removed from the
network graph. Two radial schemes with the
vertices at buses 9 and 5 are obtained.
2. Installation of PMUs at buses 9 and 5.
3. Installation of calculated PMUs at buses 1,
2, 4, 6, 7, 10, 14.
4. Calculation of voltages at the buses of
each radial scheme: 14-13-12; 10-11; 7-8.
Calculation starts with the buses, where the
calculated PMUs are installed.
5. State estimation for the 3-bus scheme (2-
3-4). For the schemes with removed buses
the problem of SE is not solved because of
installation of the calculated PMUs there.
Voltage magnitude and phase at bus 3 are
calculated through SCADA measurements
and the calculated PMU data.
6. Adjustment of power flows in the boundary
branches 6-12; 6-11, 6-13, 4-3. The calcula-
tion results are shown in Table 4.
7. Analysis of the obtained results.
Table 4 shows that condition (14) for the
boundary branches is true (the difference between
the calculated flow and the measurement is less
than the threshold equal to 15 MW and 30 MVAr).
Hence, the steady state for the 14-bus scheme is
calculated.
The results of SE and the results obtained on
the basis of the suggested technique are pre-
sented in Table 5.
Data of PMUs are shown by bold type and
data of the calculated PMUs by italic type.
From Table 5 it is seen that the voltage values
obtained by different techniques coincide.
Table 3. Values of power flows in the boundary branch
Ranch P Q
Meas SE SS Meas SE SS
78 -266 -264 -258 83 80 79
Table 4. Power flow values in boundary branches
Branch P Q
SE SS SE SS
6 11 6.89 7.4 11.4 9.2
612 8.18 8.01 9.8 2.12
613 17.9 21.7 22.39 10.8
34 -23 -27 -5 -8
19
Determination of the Quality of Results
The quality of results is determined by applying
the objective function value of state estimation:
( )
( ())
x
y y x
i i
i
i
m
=

=
2
2
1

y vector of measurements, x estimates (or
calculated values in the radial schemes) of the
state vector components, m number of measure-
ments.
The values of criteria for different calculations
are given in Table 6.
The Table reveals that the values of criteria
are almost the same for all calculation methods.
It means that the EPS steady state can be calcu-
lated by the non-iterative method approximately
with the same accuracy as by SE.
1. The algorithm of non-iterative calculation of
load flow that is based on reduction of the
calculated scheme of EPS to a radial form is
suggested. It applies a minimum number of
PMUs and an optimal set of measurements
from the SCADA system.
2. The steady states of the test 14-bus scheme
and a fragment of the real scheme are cal-
culated. The EPS state calculated by the
suggested method is shown to coincide with
the results of state estimation.
Figure 6. A 14-bus IEEE scheme
20
Artificial Intelligence
Technologies for Monitoring
Large Power Interconnections
Implementation of market principles in planning
and control of operating conditions, expansion of
the area of coordinating operation control of EPSs
in terms of time (from design of control systems
to their realization by dispatching and automatic
devices) and situation (coordination of dispatch-
ing, continuous automatic and discrete emergency
control) all cause fast dynamics of change in EPS
operating conditions. As a result the problem on
working out principles of interconnection of EPSs
of different geographical length and with different
structure of an electric network is considerably
complicated. The principles developed apply dif-
ferent norms, standards, control algorithms, etc.
Development of systems and devices for moni-
toring the state of energy and electrical equipment
Table 5. Calculation results of steady state
N
Estimate Non-iterative method
3 5
U

U
cal

cal
1 380 0 380 0
2 397 -0.0604 397 -0.0604
3 383 -0.12 381 -0.12
4 387 -0.1027 387 -0.1027
5 387 -0.0874 387 -0.0874
6 406 -0.1363 406 -0.1363
7 403 -0.1277 403 -0.1277
8 414 -0.1310 414 -0.1323
9 401 -0.1404 401 -0.1404
10 399 -0.1402 399 -0.1402
11 401 -0.1375 402 -0.1374
12 400 -0.1401 401 -0.1356
13 399 -0.14 402 -0.1407
14 393 -0.1423 393 -0.1423
Table 6. The values of criteria
Scheme Real data
SE Non-iterative method
13 20.35 22.70
14 20.4 25
21
(devices and systems of diagnostics) and also
monitoring the EPS operation conditions seems to
be highly important because of radically changed
development trends and complicated operating
conditions of large-scale IPSs.
It should be underlined that in these conditions
a system of operation monitoring is a powerful tool
to study dynamic characteristics of EPSs in real
time for different system states. The system en-
hances control efficiency of normal and emergency
conditions in the current market environment.
Artificial Intelligence Methods
in Problems of Operating
Condition Monitoring
Effective organization of the system of IPS
operating condition monitoring is possible by
an extensive involvement of new tools for the
analysis and calculations of operating conditions,
primarily technologies of artificial intelligence.
It should be underlined that the methods of
artificial intelligence have nothing to do with algo-
rithmic calculations and therefore, do not require
complex computational mathematical models to be
constructed for an object. The artificial intelligence
methods reproduce (copy) individual functions of
the creative activity of human brain, which makes
it possible to find optimal decisions in a large set
of possible states in the shortest time.
The following are primarily the main technolo-
gies of artificial intelligence:
Models applying an apparatus of artifcial
neural networks (ANN) (Haykin, 2006,
Ossovsky, 2004)
Genetic Algorithms (Goldberg, 1989)
Hybrid and fuzzy systems (Pospelov,
1986).
The Kohonen maps applied for visualization
and analysis of schemes and operating conditions
in EPS may be given as an example of practical
use of the artificial intelligence technologies. Use
a self-organizing Kohonen network in the struc-
ture of the main system of data processing of the
WAMS system is offered in (Handshin, 2006).
The Kohonen network in this case is capable
of analyzing schemes and operating conditions
on-line and off-line. For the given ANN
the problem is solved by division of the studied
situations into clusters with close typical change
of processes in EPS. Hence, the whole variety of
schemes and operating conditions can be visual-
ized effectively and dangerous contingencies in
the studied EPSs can be detected timely.
The experimental studies have shown (Kur-
batsky, 2009) that the Kohonen maps applied allow
an adequate detection of inadmissible combina-
tions of the scheme and operating parameters
and a correct choice of necessary control actions.
In addition, a key monitoring problem is the
problem of forecasting state variables of EPS.
In this case within the system state estimation
(Glazunova, 2009; Gamm, 2007) it is supposed
to forecast all the state variables for a very short
time span. The lead time can be from several
seconds (an interval of obtaining information on
EPS state variables in a unit time) to one minute.
Information on the system state is obtained by us-
ing the SCADA system and from PMU recorders.
Normally modern monitoring systems employ
for these purposes traditional forecasting models
such as ARIMA and Kalman filter. These make
it possible to obtain rather good forecasts pro-
vided the dynamics of random variations of state
variables is represented as a stationary Gaussian
process.
In the cases where state variables are rather
variable and nonstationary it is more sensible to use
neural network models. In the course of numerous
studies and calculations the traditional forecasting
methods have proved to be insufficiently accurate
in modeling state variables of power systems as
compared to the neural network methods.
In spite of this fact it is necessary to emphasize
that the advantage of the neural network forecast
as compared, for example to ARIMA and Kalman
22
filter is most pronounced in the cases where the
trend of analyzed time series differs from the linear
one and the initial data contain a considerable value
of irregular component (Borovikov, 2006). In the
cases, however, where time series of the process
contains a linear trend the traditional approaches
practically do not differ from the neural network
ones in terms of forecasting accuracy.
Super Short-Term Forecasting
of State Variables on the Basis
of Artificial Neural Networks
The authors propose the approach to super short-
term forecasting of state variables on the basis
of neural network technologies and nonlinear
optimization algorithms. The approach is imple-
mented in the ANAPRO software (Kurbatsky,
2008, Kurbatsky 2009).
The use of nonlinear optimization algorithms
in the approach, namely, the methods of simulated
annealing (SA) and neuro-genetic input selection
(NGIS) (Haykin, 2006, Ossovsky, 2004), provides
the procedure of choosing the best forecasting
model for each individual sampling. For example
in the process of learning sampling analysis based
on the NGIS algorithm individual input data can
be rejected as less informative. This method
represents optimization on the basis of random
search techniques and combines the capabilities
of GA and PNN/GRNN networks to automatically
determine optimal combinations of input variables.
The PNN/GRNN networks allow the best
results to be remembered, which improves
the previous results. Owing to the radial layers
with Gaussian function in the structure of PNN-
algorithm, bad data in the input sampling can be
reduced to minimum.
The SA technique makes it possible to ana-
lyze the properties of the initial sampling and
organize a competition-based system between
different neural network forecasting models when
in the process of nonlinear optimization the best
forecasting model is selected. This competition
procedure is based on the criterion of minimiz-
ing the total risk of ANN error (Ossovsky, 2004).
This allows the balance between the reliability of
learning and the quality of model to be reached,
thus making the neural network forecasting model
generalize better.
R(w)=E
w
(W)+Ec
(
W) (15)
E
w
(W) - standard performance measure that
depends on both the network itself (model) and
the input data; E
c
(W) - complexity penalty depends
solely on the ANN itself and is determined on the
basis of preliminary data on the model structure;
regularization parameter.
The ANN types considered in the studies
were: radial basis function (RBF) and general-
ized regression neural network (GRNN) (Haykin,
2006, Ossovsky, 2004). They make it possible to
obtain rather accurate results in super short-term
forecasting.
In the case of complex computational forecast-
ing problem (for example when learning sample
contains many additional factors) the principle
of committee machine (CM) is used [Haykin,
2006] as a basis or the proposed approach. The
CM represents a neural network system consisting
of combination of neural networks-experts which
allow one to find a general solution, y (n), which
has priority over each solution of an individual
expert, y
k
(n).
The calculation practice shows that for differ-
ent combinations of learning and test samples the
SA procedure should be started 3-6 times to form
the neural networks - experts.
Experimental Calculations
The calculations show (Kurbatsky, 2009) that the
suggested neural network approach within the
ANAPRO software makes it possible to much
more efficiently solve this problem as compared
to the other traditional forecasting methods. This
is related to the use of nonlinear optimization
23
algorithms as a basis for the approach. The al-
gorithms are used to analyze the time series and
networks on the basis of Gaussian radial functions,
in particular, the GRNN structure. The GRNN
networks are learned practically in a flash, which
is extremely important for on-line forecasts, and
are robust to the presence of bad data.
The experimental calculation results obtained
while minutely forecasting active power flow in
the 500 kV intersystem transmission lines that
connects two large-scale interconnected power
systems are presented in Figure 7 and Table 7.
The calculation data show that the neural
network model allows one to decrease the mi-
nutely forecast error almost twice as compared
to the forecast made by the ARIMA model.
As it is seen from Figure 7 the time series of
change in the active power flow on this section
of ESR is extremely variable and non-stationary
which does not allow the traditional forecasting
models to be used.
As is seen the neural network forecast on the
basis of the proposed approach allows one to
decrease the forecasting error by a factor of 2 as
compared to the ARIMA model.
The need for wider application of monitoring
systems is growing increasingly urgent in the light
of the future possibility to interconnect the main
power grids of Europe and Russia. The techno-
logical progress in the last decades has shown that
the use of perspective information technologies,
first of all ANN, can provide reliable operation
of future interconnected power grids of EU and
Russia, optimal use of energy resources on vast
territories and mutually beneficial electricity
trade between different regions in terms of market
requirements.
Study of the Properties of a
Large Electric Power System
by Using Singular Analysis
Transfer capability of a tie determines the maxi-
mum power which can be transmitted over this tie
without deteriorating system reliability. Electric
power system has so called weak ties. The transfer
capabilities of these ties will be the first to achieve
their limit if the changes occur in the system
conditions and this limit will vary depending on
the system state.
Transition to market leads to the power flows
close to the maximum admissible ones. Therefore
it is important to find the methods to detect weak
ties in electric cutsets in which stability losses are
most probable and to organize monitoring of their
transfer capabilities. It is also very important to
study the factors that determine the weakeness
of ties and their transfer capabilities and find the
methods to reinforce the ties. Accurate knowledge
of weak ties and their transfer capabilities is very
important to plan maintenance and management
of wholesale electricity markets.
With gradual increase in the power transmitted
along the tie line the maximum power will be
reached in the end at which there is no solution
to the load flow problem. This is the result of
degeneration of the Jacobean matrix J. The matrix
relates the changes of active P and reactive
Q powers with changes in phases and
Table 7. Telemetry forecasting errors on the basis of ARIMA and ANN
Model
MAPE for interval,%
20:01 20:02 20:03
ANN of GRNN type
(60-83-60)
12,5 18,1 17.1
ARIMA (2,2,0)
AR(2)
32,1 57,1 49,2
24
magnitudes U of nodal voltages in the system
of linearized equations of power balances
P
Q
J
U
P P U
Q
Q U U

(16)
Degeneration of the Jacobi matrix in a gen-
eral case, as is shown in [Venikov, 1975], is a
local index of the global phenomenon related to
the loss of static stability. The sign of degeneration
can be closeness to zero of the determinant
det( ) J , equality of the conditionality number
cond J ( )
max min
= to infinity, where
max
and
min
are maximum and minimum singular values
of the Jacobi matrix.
The minimum singular value of the Jacobi
matrix is another index, since the closer
min
to
zero the closer the current conditions to the limit
ones in terms of static stability.
In (Voitov, 2000) the authors analyze the pos-
sibility of applying
min
for estimation of static
stability with respect to voltage. In the paper the
singular analysis technique is used to detect the
ties with the parameters and/or high active
power flows that lead to a considerable rise in
voltage phases and deterioration of the Jacobi
matrix conditionality. The study on interrelation
between the Jacobi matrix conditionality and
transfer capabilities of ties and cut sets employs
the method of contribution factor (Bialek, 2000).
Singular Decomposition of
the Jacobean Matrix
Singular decomposition of the n n asymmetri-
cal Jacobi matrix can be presented in the form:
J W V w v
T
i i
i
n
i
T
= =
=

1
, (17)
Figure 7. Use different models for short-term forecast 60 second ahead of active power plow
25
where W w w w
n
=( , , ..., )
1 2
and V v v v
n
=( , , ..., )
1 2

- orthogonal matrices of left and right singular
vectors,
1 2 3
....
n
- singular val-
ues given in an ascending order.
If
1
=
min
is considerably lower than the
rest of the singular values then, all other condi-
tions being equal, the least contribution to the
change in phases and magnitudes of nodal volt-
ages is made by the first term of the decomposi-
tion (17). Taking into account the inversion of
matrix J this can be written as follows:
U
J
P
Q
w
P
Q
T
1
1 1
(v
1
/ )
.
(18)
I n t r o d u c t i o n o f s c a l a r v a l u e
S w
P
Q
T 1
1 1
=
( / ) , which is called the first

generalized disturbance shows that the maximum
changes in phases and magnitudes of voltage will
occur at the nodes that are called sensors in
(Golub, 1995) and correspond to the maximum
components of the first right singular vector.
Similar to the first generalized disturbance
taking into account the expression (1) and decom-
position (2) we can write down the scalar value
of t he f i r s t ge ne r a l i z e d r e s pons e
F v
U
T 1
1 1
=
( )

. It shows that the maximum
contribution to the change in phases and magni-
tudes of voltages are made by the changes in
injections at the nodes that correspond to the
maximum components of the left singular vector.
The elements of the network scheme whose
parameters change having the greatest impact on
the minimum singular value
1
, are called weak
places in (Golub, 1995). The change in the active
power f l ow al ong t he t i e k - l
P P
kl kl kl kl
= ( ) is largely determined
by the value of
kl
. This value can be estimated
by the difference of the components of the right
first singular vector ( ) v v
k l 1 1
that correspond
to the phases of nodal voltages at the k-th and l-th
nodes and can be applied as an index of the tie
weakness. The larger the change in
ij
with
increase in the flow along the tie the faster the
maximum of the transmitted power will be reached
in it and the Jacobi matrix will degenerate.
Another index of tie weakness that allows
one to determine the network parameters or state
variables f, that make the greatest impact on the
conditionality of the Jacobi matrix is the derivative
of the minimum singular value with respect to f
=
1 1 1
f w J f v
T
( ) .
For example the tie k l can be called weak
if the change in its admittance y
kl
leads to the
maximum change in
1
= =
( )
1
1 1 1 1
2 2
2
y
w J y v w w
P
y
P
U y
Q
kl
T
kl U
kl kl

( )

y
Q
U y
v
v
kl kl
U
2
1
1

(19)
It follows from expression (19) that the mini-
mum singular value
1
is affected not only by the
values of branch admittances but also by the
parameters of current conditions which may lead
to a change in ranking the ties in terms of weak-
ness under varying conditions. Another possibil-
ity to determine the ties whose weakness does not
change or changes little under varying conditions
is to study the derivative of the minimum singu-
lar value of the symmetrical nodal admittance
matrix with respect to admittances of ties (Golub,
1995).
By changing the admittance of weak ties it
is possible to increase or decrease the minimum
singular values and thus to improve or deteriorate
26
the conditionality of the Jacobi matrix and the
nodal admittance matrix.
Contribution Factor of
Power Transmission
The contribution factor method (Bialek, 2000)
makes it possible to determine for a known load
flow between what ties in the electric network the
power of each generator node is distributed and
in what proportion. If to assume that all tie lines,
along which the power is transmitted from gen-
erator node j, have equal transfer capabilities then
with an increase of power at the generator node
power flow in the tie that has the maximum coef-
ficient a
kl
j
will reach its limit value earlier than
the flows in the other ties. The algorithm (Golub,
1995; Gamm, 2003) can be used to determine the
coefficients a
kl
j
. It suggests searching for the paths
from each generator on the directed graph with
the orientation of branches that coincides with
the direction of flows along these branches.
If power of the generator node increases by
the value p , the power flow in the tie k - l will
increase by the value p a p
kl kl
j
= , where
a
kl
j
is a contribution factor that determines the
contribution of generator node j to the flow along
the tie k - l from the node k to the node l.
In the subsystem transmitting power of a
specific generator, the ties ranked as weak on the
basis of singular analysis may have low contribu-
tion factors and vice versa the ties that are not
weak may have high contribution factors. To find
the compromise solution the ties of each subsys-
tem can be assigned the weight equal to the
product of the tie weakness index and the contri-
bution factor, for example the weighting coeffi-
cient for the tie k - l can be represented by
( ) v v a
k l kl
j
1 1
.
The information on the maximum weighting
coefficients makes it possible to detect weak cut
sets in every subsystem. This procedure is rather
simple and implies that the most loaded weak
ties are excluded unless the studied subsystem is
divided into independent subsystems.
Case Study
Let us consider a 14-node network scheme (see
Figure 8). Nodes 1, 3, 101, 201 and 203 are gen-
erating, node 101 is slack, nodes 4, 6, 100, 202 are
load ones. Directions of active power flows for the
base network state that is far from the steady-state
stability limit are shown in the scheme by arrows.
The plan for detection of weak ties and cut
sets of the test scheme is the following:
1. Determination, for the base conditions by
using the singular analysis, of the nodes
with the maximum response to generalized
disturbances, the nodes that cause maximum
generalized disturbances and also the weak
cut sets, the changing conditions of which
have the greatest effect on the Jacobi matrix
conditionality.
2. Analysis of factors influencing the transfer
capability of weak ties.
3. Confirmation of the fact that with the increase
in power transmitted on the path along weak
ties the limit of transmitted power is reached
in the most loaded weak tie.
Components of the right singular vector that
correspond to phases of the nodal voltages are
presented in Figure 9a, curve . They determine
nodes 200-203 as nodes with sensor phases and
separate two groups of coherent generators 201,
203 and 1, 3, whose voltage phases have a similar
response to external disturbances.
Analysis of the components of the left singu-
lar vector that correspond to nodal voltage
phases, Figure 9b, curve , reveals that generation
change at nodes 201 and 203, should lead to the
maximum response of voltage phases.
Curves in Figures 9c and 9d illustrate the
values of difference of the right singular vector
27
components that correspond to node numbers at
the tie ends and derivatives of the minimum sin-
gular vector of the Jacobi matrix with respect to
tie admittances. Analysis of the curves reveals ties
200-8, 8-5 and 202-100 as weak. The ranging of
weak ties using the first and second criteria differs.
Therefore, it is possible to suppose that weakness
of the 202-100 depends to a greater extent on
the conditions and weakness of ties 200-8, 8-5
on their admittances, the latter is proved by the
studies carried out in (Golub, 1995).
We will show that with the increasing active
power flows on the path with loaded weak ties the
limit of transmitted power is found in such ties.
Let us determine by the method of contribution
factor the subsystems in Figure 10 transmitting
power from:
Generator nodes 1 and 3 to load node 4 -
subsystem );
Slack node 101 to load nodes 4 and 6 - sub-
system b);
Generator node 201 to load nodes 4, 6 and
202 - subsystem c) that includes weak ties
200-8, 8-5 and 202-100;
Generator node 203 to load nodes 4, 6,
202 - subsystem d) that includes weak tie
202-100.
The effect of the increase in power transmitted
over weak ties on the Jacobi matrix conditional-
ity will be illustrated by considering the simulta-
neous equal increase (decrease) of active power
of generation at nodes 201 and 203, whose con-
tribution to the generalized disturbance is maximal,
by the value p . This will lead to decrease (in-
crease) of generation at base node 101 by the
value 2 p plus the changing losses.
Figure 8. Scheme of the 14-node test network
28
Figure 9. The values of components the singular vector and derivative
29
Figure 11a shows the change of minimum
singular value of the Jacobi matrix with increase
and decrease in the active power of generation at
nodes 201 and 203. When generation at node 201
or 203 increases by the value larger than
16 75 . p , steady-state stability is violated with
sharp decrease of conditionality number. The best
conditionality of the Jacobi matrix is determined
with the decreasing generation by the value
35 p with respect to the base state denoted
by null. In Figure 11b curves b are constructed
for the limit conditions and characterize the in-
Figure 10. Subsystems transmitting active power from certain generator to load nodes
Figure 11. Change of the minimum singular value of the Jacobean matrix and the voltage phase difference
30
creasing sensitivity due to deterioration of the
Jacobi matrix conditionality.
Change in the difference of voltage phases in
the ties on the path of power transmission from
generator nodes 201 and 203 in subsystems and
d, Figure 11b, shows that the Jacobi matrix is
degenerated, because the phase difference in the
202-100 becomes high.
The weakness of the 202-100 is also confirmed
by the impact of increasing admittances of weak
ties on the minimum singular value of the Jacobi
matrix and the limit value of generation at nodes
201 and 203 that is shown in Table 8. With practi-
cally the same impact on the conditionality number
the increasing admittance of tie 202100 leads
to much higher transfer capability of the network
in comparison with ties 200-8, 8-5.
Table 9 illustrates the change in contribution
factors of ties in subsystems and d with the
increasing generation at nodes 201 and 203 and
also the values of weighting coefficients in the
initial and limit conditions.
Weak tie 202-100 in subsystem d is the most
heavily loaded. With the increase of generator
node power the contribution factor of this tie also
increases. Tie 202-100 in subsystem d has maxi-
mum weighting coefficients in both initial and
heavy conditions.
In the initial and heavy conditions in subsystem
ties 200-8, 8-5 and 202-100 are the most loaded
weak ties. With the change of power at node 201
the contribution factors of these ties vary insig-
nificantly. When the power of node 201 changes
by the value 35 p , at which the Jacobi
matrix conditionality is the best, tie 202-100 is
no longer weak and even its tripping does not lead
to instability. The weak cutest is observed in the
most loaded weak ties 202-100 and 8-5 of sub-
system .
On the basis of the singular analysis of the
Jacobi matrix and the method of contribution
factors the index for determination of the most
loaded weak ties is suggested. In these ties with
the increase in transmitted power the limit of
transfer capability is reached earlier than in other
ties, which leads to steady-state stability loss.
Further studies on weak ties should provide for
consideration of the constraints imposed on op-
erating parameters.
Table 8. Impact of increasing admittances of weak ties in per cent (Y %) on the limit power p of
generation at nodes 201 and 203 and the minimum singular value of the Jacobi matrix
Increase of
Y%
p

1
=
min
202-100 200-8 8-5 200-8 8-5
0 16.75 16.75 16.75 4.59 4.59 4.59
10% 19.88 18.25 18.25 4.59 4.88 4.85
20% 23.75 19.81 19.94 5.13 5.18 5.13
30% 28.63 21.56 21.69 5.43 5.53 5.43
40% 34.94 23.44 23.69 5.77 5.93 5.77
50% 43.25 25.56 25.81 6.13 6.4 6.13
31
T
a
b
l
e

9
.

C
h
a
n
g
e

o
f

t
h
e

c
o
n
t
r
i
b
u
t
i
o
n

f
a
c
t
o
r
s

f
o
r

t
i
e
s

i
n

s
u
b
s
y
s
t
e
m
s

a
n
d

d

w
i
t
h

i
n
c
r
e
a
s
e

o
f

g
e
n
e
r
a
t
i
o
n

a
t

n
o
d
e
s

2
0
1

a
n
d

2
0
3

b
y

t
h
e

v
a
l
u
e

p

a
n
d

t
h
e

w
e
i
g
h
t
i
n
g

c
o
e
f
f
i
c
i
e
n
t
s

(
)
v
v
a
k
l
k
l j
1
1

s
h
o
w
n

i
n

b
o
l
d

t
y
p
e

f
o
r

t
h
e

i
n
i
t
i
a
l

a
n
d

l
i
m
i
t

c
o
n
d
i
t
i
o
n
s

p
T
i
e
s

o
f

s
u
b
s
y
s
t
e
m
s

d
2
0
0
-
8
8
-
5
5
-
6
5
-
2
2
-
4
1
0
0
-
6
2
0
0
-
2
0
2
2
0
2
-
1
0
0
1
0
0
-
4
2
0
2
-
1
0
0
1
0
0
-
4
1
0
0
-
6
-
4
5
.
0
0
.
9
9
9
0
.
9
8
5
0
.
8
9
6
0
.
0
7
6
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
1
0
.
0
0
1
0
.
0
0
1
-
2
5
.
0
0
.
9
5
0
0
.
9
4
0
0
.
8
1
0
0
.
1
1
1
0
.
1
1
1
0
.
0
0
2
0
.
0
4
1
0
.
0
1
1
0
.
0
0
1
0
.
2
6
0
0
.
0
2
4
0
.
0
5
3
-
1
5
.
0
0
.
9
2
9
0
.
9
0
8
0
.
7
5
0
0
.
1
3
0
0
.
1
3
0
0
.
0
0
5
0
.
0
7
1
0
.
0
3
0
0
.
0
0
2
0
.
4
2
8
0
.
0
3
5
0
.
0
5
3
-
5
.
0
0
0
.
9
0
0
0
.
8
8
4
0
.
7
1
0
0
.
1
4
2
0
.
1
4
1
0
.
0
0
7
0
.
0
4
8
0
.
0
4
8
0
.
0
0
3
0
.
5
3
0
0
.
0
3
7
0
.
0
7
5
0
.
0
0
0
.
9
0
1
0
.
8
7
3
0
.
7
0
0
0
.
1
4
6
0
.
1
4
5
0
.
0
0
7
0
.
0
9
8
0
.
0
5
7
0
.
0
0
4
0
.
5
7
0
0
.
0
3
7
0
.
0
7
3
0
.
0
0
0
.
3
5
3
0
.
4
2
5
0
.
0
0
8
0
.
0
1
4
0
.
0
0
3
0
.
0
0
2
0
.
0
0
0
0
.
0
3
6
0
.
0
0
1
0
.
3
6
5
0
.
0
1
3
0
.
0
1
8
4
.
0
0
0
.
8
9
5
0
.
8
6
5
0
.
6
8
7
0
.
1
4
8
0
.
1
4
7
0
.
0
0
7
0
.
1
0
4
0
.
0
6
3
0
.
0
0
4
0
.
6
0
0
0
.
0
3
6
0
.
0
6
9
8
.
0
0
0
.
8
8
9
0
.
8
5
6
0
.
6
7
3
0
.
1
4
9
0
.
1
4
8
0
.
0
7
2
0
.
1
1
0
0
.
0
6
9
0
.
0
0
4
0
.
6
2
0
0
.
0
3
5
0
.
0
6
5
1
2
.
0
0
0
.
8
8
2
0
.
8
4
5
0
.
6
5
0
0
.
1
4
9
0
.
1
4
8
0
.
0
0
7
0
.
1
1
7
0
.
0
7
6
0
.
0
0
4
0
.
6
4
6
0
.
0
3
4
0
.
0
6
1
1
6
.
0
0
0
.
8
6
9
0
.
8
2
3
0
.
6
3
0
0
.
1
4
5
0
.
1
4
3
0
.
0
0
8
0
.
1
3
0
0
.
0
8
7
0
.
0
0
4
0
.
6
6
8
0
.
0
3
3
0
.
0
5
8
1
6
.
7
5
0
.
8
5
0
0
.
8
0
5
0
.
6
1
3
0
.
1
3
9
0
.
1
3
6
0
.
0
0
8
0
.
1
4
1
0
.
0
9
5
0
.
0
0
5
0
.
6
7
0
0
.
0
3
3
0
.
0
6
0
1
6
.
7
5
0
.
3
3
6
0
.
3
8
7
0
.
0
0
3
0
.
0
0
2
0
.
0
0
0
0
.
0
1
4
0
.
0
0
1
0
.
0
7
3
0
.
0
0
1
0
.
5
1
3
0
.
0
0
4
0
.
0
0
6
32
A Multi-Agent Approach to
Coordination of Different
Emergency Control Devices
Against Voltage Collapse
Power Industry spends a lot of money to protect
a power system against different severe distur-
bances. Nevertheless, large interconnected Power
Systems throughout the world are frequently
subjected to widespread blackouts which interrupt
millions of consumers and cost billions of dollars.
Analysis of the recent blackouts showed, that
the most severe interruptions occurred in highly
loaded interconnected power systems due to EHV
line disruption followed by multiple contingencies
(CIGRE, 2007). These accidents highlighted the
deficiency of the existing protection systems that
cannot maintain the integrity of the transmission
grid during multiple contingencies (Lachs, 2002).
Power system behavior in an emergency state is
characterized by complex interaction between dis-
crete and continuous control devices. Continuous
control devices are automatic voltage regulators,
turbine governors, FACTS devices, etc.
Discrete control devices are different pro-
tection relays, under load tap changers, etc.
Currently both continuous and discrete control
devices substantially use local signals only and
do not coordinate their actions with each other.
Absence of coordination between discrete and
continuous control devices is the shortcoming of
the existing protection system and it may lead to
blackout. This section presents a control system
based on the multi agent approach. The control
system provides coordination between discrete
and continuous control devices to prevent volt-
age instability.
Voltage Instability Mechanism
To understand the importance of the discrete and
continuous control devices coordination, one
should understand the mechanism of voltage
instability that may occur any time after the first
severe contingency and lead to blackout.
Existing practice shows that if protection
system works correctly, most power systems have
sufficient stability to withstand the first heavy
disturbance in EHV transmission system. The
post-disturbance phase represents a deceptively
calm period that lasts several minutes with a normal
level of frequency and then voltage collapse that
lasts seconds (Lachs, 1992).
The first heavy disturbance leads to increase in
the reactive power losses and reactive power out-
put of rotating units in the vicinity of the affected
region. So, the first disturbance effects influence
only the affected region, being initially a local
problem. But some time after, the lack of reac-
tive power in the affected region might increase
considerably, leading to voltage collapse in the
neighboring regions and even in the whole system.
This happens because if the disturbance is not dealt
with timely, the after-effects spread out through the
EHV transmission network and actuate different
control devices such as automatic voltage regula-
tors, automatic transformer tap changers, current
protection relays, etc. These control devices act
at the different speed, respond to changes in the
immediate vicinity and act without coordination
with one another. Their actions in response to the
post-disturbance conditions are actually the main
cause of power system breakdown; consequently,
the timely control of the discrete and continuous
control devices under the post-disturbance condi-
tions is the only means to prevent voltage collapse
of the whole system (Lachs, 2002). Undoubtedly,
the absence of different control actions coordina-
tion during the post-disturbance period can cause
different types of instability. But first of all, one
should cope with voltage instability because it
was the main cause of the recent blackouts. New
system protection system philosophy has to be
proposed to prevent voltage instability during the
post-disturbance period.
33
A Multi-Agent Approach Outlines
There are a number of definitions for what an
agent is. This fact testifies to the difficulty in de-
fining the notion of multiagent systems (MAS).
General definition says that MAS is a distributed
and coupled network of intelligent hardware and
software agents that work together to achieve the
global goal. Agents are autonomous structures
and they operate with each other through differ-
ent mechanisms.
MAS could have different architectures. Reac-
tive architecture is one of them. It is based on a
simple stimulus-response mechanism triggered
by sensor data. Its advantage is a faster but not
reason better response in dynamic environments.
Agents in reactive architecture are also simpler
in design than agents that are more intelligent.
Power systems are already using many reactive
agents such as protective relays, automatic volt-
age regulators, etc. However, the fact that these
simple reactive agents have extremely narrow
knowledge about one another, results in some
disadvantages, for instance, lack of coordination.
Another type of the MAS architecture is layered
(hybrid) architecture that allows both reactive and
deliberative agent behavior.
Another key component of the MAS is a com-
munication principle. If agents need to cooperate
and be coordinated, they have to communicate
with one another by using some communication
language. Currently, the most used communica-
tion language is the FIPA (The Foundation for
Intelligent, Physical Agents). FIPA standards can
be found in [The Current]. Coordination among
agents can be provided by using different ap-
proaches including organizational structuring and
distributed multiagent planning.
Organizational structuring provides coordina-
tion through the definition of roles, communication
paths and authority relationship. Organizational
structuring is the easiest way to resolve conflicts
among agents and provide their coherent behavior.
Power system control centralization is an example
of organizational structuring: there is an agent
(control center) which has some knowledge of
the current and the prospective system states and
establishes rules for other agents according to
hierarchical structure of the MAS. However, such
an approach sometimes is impractical, because it is
hard to design such a central controller, especially
when the latter has a little time for collecting a lot
of information to provide control actions.
Another approach to agent coordination is a
distributed multi-agent planning. In order to avoid
inconsistent or conflicting actions, agents can
build a multi-agent plan that details all the future
agent actions and interactions required to achieve
their global goal. In the process of working agents
communicate in order to build and correct their
individual plans until all conflicts are removed.
We believe that MAS that is likely to be used for
protection against voltage collapse should have
layered architecture and use distributed multi-
agent planning approach as a perspective way to
provide coordination between different control
devices during the post-disturbance period. For
better understanding of the multi-agent approach
principles see (Bellifemine, 2007; McArthur,
2007; Taylor, 1991).
System Protection Philosophy
A new system protection philosophy is needed to
control the post-disturbance phenomenon. A new
protection system must detect the critical situa-
tion and coordinate the work of control devices to
exclude any possibility of voltage instability. So,
how can the new protection system identify the
critical situation and what kind of control actions
should the system use to control the capacity of
available reactive power resources?
Parameters-Indicators
The main symptoms that precede the voltage col-
lapse are considerable reduction of transmission
34
voltage levels and increase of reactive power
outputs on rotating units (Lachs, 2003). Reduction
of voltages and increase of rotating unit excitation
were proposed in different papers to indicate the
proximity to voltage collapse. Thus, these two
criteria may be used to detect the critical situation
appearance and activate protection system.
Control Actions
Power industry has already used the philosophy
of load shedding by selecting non-essential load
to prevent frequency reduction. The analysis
of recent blackouts showed that the rapid load
shedding is usually the only way to prevent the
collapse of the whole system (CIGRE, 2007). On
the one hand, load shedding should be as fast as
possible, on the other hand, it should be optimal.
The optimal load shedding scheme can be realized
by using different optimization procedures, but
it is hard to solve optimization problem for any
possible situation in advance, because the number
of situations is too big. This means that some op-
timization computations should be made during
the post-disturbance period. In spite of the fact
that there is a number of optimization techniques
that can be used to calculate emergency control
actions quickly, the amount of input data required
to solve the problem is usually too big. The state
estimation alone can take from tens of seconds to
minutes. However, load shedding under the post-
disturbance conditions has to work faster. Hence,
load shedding procedure has to use less complex
methods to control post-disturbance phenomenon.
The following simple countermeasures to control
post-disturbance phenomenon were proposed in
(Lachs, 1992):
Countermeasure 1. Fast tap changing on
transmission substation transformers.
Countermeasure 2. Raising terminal volt-
age on selected synchronous condensers
and hydro generators.
Countermeasure 3. Fast tap changing on
selected generator transformers.
Countermeasure 4. Strategic load shed-
ding at selected transmission substations
only if voltage levels and reactive outputs
do not meet the requirements, or some
transmission lines are overloaded.
Countermeasure 5. Rearranging genera-
tor MW outputs. Connecting part of the
disconnected load.
Countermeasures 13 have approximately the
same execution time and their main purposes are
to impede the sharp increase of series reactive
power losses, to increase transmission line charg-
ing and to inhibit tap changing on subtransmis-
sion and distribution transformers. Load is shed
(Countermeasure 4) only after countermeasures
1 3. This will decrease the amount of the load
to be shed. Countermeasure 5 considers an opti-
mization procedure. The optimization procedure
takes much more time in comparison with coun-
termeasures 14 and provides post-emergency
operation optimization.
Thereby, countermeasures 14 provide fast
control of the post-disturbance phenomenon to
avoid voltage collapse and countermeasure 5 pro-
vides long-time-period post-emergency operation
optimization. The proposed control principles can
be applied to various parts of the grid that work
independently. Briefly, the control actions aim to
control the capacity of the available reactive power
resources and do not let reactive power demand of
the affected region increase beyond their sustain-
able capacity to exclude the possibility of voltage
instability (Bellifemine, 2000).
The proposed control system can be built by
using distributed intelligence principles. The
distributed intelligence is taken to mean the multi-
agent system.
Multi-Agent Control System Structure
The proposed multi-agent control system provides
reactive power control to prevent generator trip-
35
ping and preserve load bus voltages within the
normal range.
A power system is used to illustrate the main
principles of the proposed multi-agent approach.
This power system is a part of the modified 24 bus
IEEE One Area RTS-96 system (see Figure 15).
It is divided into two subsystems - Subsystem A
and Subsystem B that correspond to transmission
and subtransmission plus distribution systems
respectively. The proposed MAS consist of two
types of agents: Load Agents and Generator
Agents. Any agent at any time has the following
set of local data:
Local state variables (primary and second-
ary voltages, power fows, etc.).
Operating characteristics of the local
equipment (generator terminal voltage, tap
range of the tap changer, excitation current
of the generator, etc.).
Any agent has two goals:
Local goal. It consists in maintaining lo-
cal state variables and equipment operating
characteristics within the normal range.
Global goal. It consists in voltage collapse
prevention.
To make different parts of the proposed MAS
system work independently, each agent must
know only about the limited number of agents,
which influence his activity most. For instance,
Load Agents, installed at Bus101 Bus103 in
Subsystem A must know much about the agents
in Subsystem B, because all these agents can
influence them. On the other hand, in spite of
the fact that agents in Subsystem B could know
much about one another, they must know only
about three agents in Subsystem A: Load Agents,
installed at Bus101 Bus103, because these three
agents can only influence them. In this case, sub-
transmission system produces minimal influence
on transmission system.
MAS Ontology
Agents communicate with each other, by us-
ing some communication language. According
to FIPA standards, messages exchanged by
agents have a number of fields and in particular:
sender, receiver, communicative intention (also
calledperformative), content, language, ontol-
ogy and some fields used for control. Ontology
is the vocabulary of symbols and their meanings.
For the effective communication, both the sender
and the receiver must ascribe the same meaning to
symbols. Ontology can include different elements
such as agent actions, terms, concepts, etc. Agent
actions indicate actions that can be performed by
some agents. Terms are expressions identifying
entities (abstract or concrete) thatexist in the
world. For voltage control purposes, following
the simplest Voltage Control Ontology can be
proposed:
Agent actions of the Voltage Control Ontology:
Increase Reactive Power.
Stop Reactive Power Increase.
Start Load Shedding.
Terms of the Voltage Control Ontology:
Owner.
Voltage Rate.
The Voltage Control Ontology Usage
Principles will be Given further
Generator Agent
Generator Agent obtains local information about
excitation current of the generator, primary and
secondary voltages at the generating substation,
active power flows and transformer tap ranges.
If excitation current goes beyond of its normal
range, Generator Agent tries to decrease it to
exclude the possibility of the generator tripping.
Generator Agent sends messages to other agents
that can decrease the shortage of the reactive power
in the affected region. The sent messages apply
36
FIPA Request Interaction Protocol and include
Increase Reactive Power action of the Voltage
Control Ontology. The sequence diagram for the
Request Interaction protocol used by the Generator
Agent is depicted in Figure 12.
Before sending a message, Generator Agent
could use a rule set to identify whether receiver
is able to help him. In our research, we used the
following simple rule: Generator Agent do not
send Request message to another agent if electric
coupling between them has become too weak.
For instance, if Bus202 Bus203 active power
flow is equal to zero, Generator Agent at Bus 203
does not send Request message to Generator Agent
at Bus 202.
In response to his request, Generator Agent can
receive either Refuse or Agree message. Agree
message means that Request Interaction proto-
col participant starts to increase reactive power.
Sometime later, Generator Agent will receive
Inform-Done message with Stop Reactive Power
Increase action, which means that the participant
stopped increasing reactive power. Thus, Gen-
erator Agent always knows when reactive power
increase in his subsystem is stopped. If reactive
power increasing is stopped, but Generator Agent
is still overexcited, it starts Load Shedding pro-
cedure (see Figure 13).
FIPA Contract Net Interaction Protocol is used
in Load Shedding procedure. In this protocol, the
initiator wishes to optimize some function that
characterizes the Load Shedding Procedure. We
use minimal voltage rate function, but of course,
it could be function, which includes some eco-
nomic aspects. Generator Agent sends n Call for
Proposal messages to Load Agents and solicits
from them m proposals and k refuses. The propos-
als contain voltage rates at primary buses of the
Load Agents. After that, Generator Agent accepts
j proposals and sends j Accept-Proposal mes-
sages to those Load Agents which have the low-
est voltage rates at their primary buses. When
Load Agent receives Accept-Proposal message it
Figure 12. FIPA Request Interaction Protocol
37
starts to shed the load until its primary voltage
will not increase up to the specified value.
Now consider situation when Generator Agent
receives Request message. First, it analyzes oper-
ating characteristics of the generator and if they
are within the normal range it starts to increase
reactive power output according to the algorithm,
presented in Figure 14.
Load Agent
Load Agent obtains local information about pri-
mary and secondary voltages at the substation,
transformer tap ranges and active power flows.
Load Bus agent takes part in Load Shedding pro-
cedure. It also can shed the load independently
in case of critical voltage drop. If it is installed at
transmission system substation, Load Agent can
take part in reactive power regulation. In this case,
Load Agent changes transmission transformer tap
ratio until primary voltage will not decrease or
secondary voltage will not increase up to specified
values. Changing transmission transformer tap
ratio, Load Bus agent must coordinate its actions
with generators in transmission system.
Multi-Agent Control System
Implementation
The success of multi-agent system mainly depends
on the availability of appropriate technology
(development tools, programming languages)
that allows its implementation. Any kind of
programming language could be used for MAS
realization, but object-oriented languages are more
suitable, because the concept of agent is close to
the concept of object.
The computer model of the proposed MAS
for power system voltage stability control was
implemented in JADE. JADE has become a firm
favorite with researchers in power engineering in
recent years. JADE implements a famous object-
oriented language Java. Agents, developed for
Figure 13. FIPA Contact Net Interaction Protocol
38
the JADE platform consist of three basic layers:
a message handling layer; a behavioral layer; a
functional layer. Message handling layer is respon-
sible for the sending and receiving of messages
from other agents. The behavioral layer provides
control of when an agent has to implement some
task. The functional layer embodies the action
the agent can perform. JADE provides program-
mers with the following ready-to-use functions:
full compliance with the FIPA specifications;
efficient transport of asynchronous messages; a
simple agent life-cycle management; a library of
interaction protocols, etc. For further information
about JADE platform, see (Bellifemine, 2007),
(Taylor, 1997), (Milano, 2005).
Necessary power flows and time domain
simulations were carried out in Matlab/PSAT
environment (Milano, 2008). Java capabilities of
the JADE environment were used to implement
communication between Matlab/PSAT and JADE.
To provide communication between Matlab
and JADE, Box Agents are used. Box Agents are
Java objects that contain different data structures.
During Time Domain Simulation, information
about power system operating conditions at each
integration step passes from Matlab environment
to JADE by means of Box Agents. After that,
agents inside JADE environment process this
information, produce control actions if needed,
put information about control actions inside Box
Agents and pass Box Agents back to Matlab en-
vironment. Thus, there is no need to use computer
hard disc during the simulation, all computations
are performed inside the main memory and simu-
lation process is faster.
The proposed MAS software realization allows
one to use complex Matlab/PSAT routines and to
model complex behavior of the agents.
Figure 14. Reactive power output increasing algorithm of Generator Agent
39
Case Study
The Test System
Modified IEEE One Area RTS-96 system is used
as a case study (see Figure 15). Initially this
test power system contained 24 buses and had
no dynamic elements. During modification, the
following changes in the test system structure
were made:
To explore the infuence of the ULTCs
actions during low voltage conditions,
transformers equipped with ULTCs were
installed between subtransmission system
and distribution system loads.
Each load was modeled as 50% constant
impedance and 50% constant current for
both active and reactive components.
Each generator was modeled by six order
dynamic model and was equipped with
Type I Turbine Governor (TG) and Type II
Automatic Voltage Regulator (AVR) (see
PSAT documentation).
Three machines connected to Bus201
Bus203 in subtransmission system were
equipped with over excitation limiters
(OXLs) (see PSAT documentation)
After modification, IEEE One Area RTS-96
system contains 42 buses. Parameters of the
unmodified 24-bus test system can be found in
PSAT test folder (Milano, 2008). Parameters of
the modified 42-bus test system can be found
in [Modified]. For better understanding of the
transient process, agents were installed only at
the buses depicted in Figure 15.
Figure 15. A part of the modified IEEE One Area RTS-96 system
40
Disturbance
To test the proposed MAS for an extreme con-
tingency, the following sequence of disturbances
is examined:
2 seconds after the simulation starts. Loss
of the generator connected to the Bus 201.
40 seconds after the simulation starts. Loss
of Bus208 Bus207 line.
Preliminary Remarks to the Simulation Process
During the simulation process, two types of au-
tomatic systems are considered:
Automatic system based on conventional
principles
Automatic system based on multi-agent
principles.
Both automatic systems do not provide for
decentralized Under Voltage Load Shedding
(UVLS) scheme. Undoubtedly, decentral-
ized ULVS scheme is an effective means of
preventing voltage collapse and it should be
provided for both conventional and multi-
agent automatic systems. However, the
main purpose of the simulation is to dem-
onstrate the MAS advantages in relation to
reactive power sources coordination for the
purpose of generator tripping prevention. It
should also be mentioned, that the proposed
centralized multi-agent ULVS scheme dif-
fers from conventional centralized ULVS
scheme, because it is actuated without time
delay in case when there is no available re-
active power in a subsystem.
Dynamic Simulation for Automatic System
Based on Conventional Principles
Conventional automatic system includes the fol-
lowing set of the decentralized devices:
TG and AVR at each generator.
OXLs at the generators, connected to
Bus201 Bus203.
OXLs maximum field currents for generators
connected to Bus202 and Bus203 are 3 and 2.5
respectively. OXLs maximum voltage output
signal is 0.1.
ULTCs are installed at the subtransmission
substations Bus204 Bus210. ULTC time
delay for the frst tap movement is 20 sec-
onds. ULTC time delay for subsequent tap
movements is 5 seconds. ULTC tap range
is 12 steps.
Voltage reductions at load substations during
the simulation are shown in Figure 16a. The change
of rotor currents during simulation is represented
in Figure 16b.
After the first disturbance, rotor current of the
generator, connected to Bus203, reaches its ther-
mal limit, and AVR reference voltage of the
generator starts to decrease. 20 seconds after the
first disturbance, ULTCs on all transformers at
the affected subtransmission substations starts to
work. This leads to further decrease of generator
203 AVR reference voltages. Compensating reac-
tive power shortage, generator 202 increases its
excitation current. After the second disturbance,
rotor current of generator 202 reaches its thermal
limit and rotor current of generator 203 exceeds
its thermal limit. AVR reference voltages of both
generators continue to decrease and after a while,
this will lead to generator 203 tripping and to the
voltage collapse.
Dynamic Simulation for Automatic System
Based on Multi-agent Principles
In addition to the set of local devices, represented
for conventional automatic system, multi-agent
automatic system also includes ULTCs for trans-
mission transformers at Bus101 Bus103. Trying
to exclude generator tripping, multiagent automatic
system coordinates the work of local devices.
Voltage reductions at load substations during the
simulation are shown in Figure 17a. The change
41
Figure 16. Changes of rotor current and in HV substation voltage level
42
Figure 17. Changes of rotor current and in HV substation voltage level
43
of rotor currents during simulation is presented in
Figure 17b.
After the first disturbance, rotor current of the
generator, connected to Bus203, reaches its thermal
limit and the generator sends request message to
generator 202 and to the transmission transformers,
connected to Bus101 Bus103. Transmission
transformers at Bus101 Bus103 as well as gen-
erator 202 are trying to decrease reactive power
shortage.
Their joint actions decrease generator 203
excitation current. Excitation current becomes
lower than its thermal limit, and generator 203
AVR reference voltages starts increase. After the
second disturbance, rotor currents of both genera-
tors reach their thermal limits and generators send
request messages to each other and to transmis-
sion transformers at Bus101 Bus103, but in this
case, the generators receive refuse messages and
immediately start load shedding procedure. Thus,
during the transient process, rotor currents of the
generators remain within the normal range. This fact
excludes the possibility of the generator tripping.
The absence of the control devices coordina-
tion during the post-disturbance period is one of
the main causes of the voltage instability, which
permanently occurs in power systems all over the
world. The proposed multi-agent control system
provides reactive power control by coordinating the
work of different discrete and continuous control
devices in a post-disturbance period. The reactive
power control in a post-disturbance period prevents
generator tripping and maintains load bus voltages
within the normal range. The efficiency of this ap-
proach has been proved by numerical simulations.
Advanced Emergency Control
System for Prevention and
Elimination of Power System Out-
Of-Step Operation Using PMU
Out of-step operation in power system intercon-
nections is one of the most severe emergency
conditions. It is related to the loss of stability in
power system interconnections which may cause
damage to equipment, interruption of power supply
to consumers and unwanted development of emer-
gency processes with severe consequences for the
entire interconnection and its parts (Sovalov, 1988;
Pourbeik, 2006; Mkrov, 2005 et al). With the
future possibility of an interconnection between
UCTE and IPS/UPS power systems by use of AC
tie lines the occurrence of out-of-step operation at
the interface between these power systems can be
dangerous for both of them and result in undesir-
able consequences for the systems and consumers.
Measures are therefore required to detect, prevent
and eliminate out-of-step conditions.
Special automatic out-of-step protection sys-
tems (OSPS) have been used in electric power
systems for reliable, timely and selective detec-
tion and elimination of out-of-step conditions
(Sovalov, 1988; Gonik, 1988; Brinkis, 1975). The
most effective system is the so called selective
OSPS which is based on the angle measurement
(Brinkis, 1975; Quintana, 1991). Previously the
difficulties of comparing angles at different po-
sitions within the network made us use indirect
angle calculation. Most popular methods of this
calculation are based on determination of current
amplitude or complex impedance at a connection
point of automatic system (Brinkis, 1975). In this
case the system is represented by a two-machine
equivalent with regard to cutset of the ties in
which the OSPS is installed. The parameters of
the two-machine equivalent are determined on
the assumption that the motion of generators in
the initial system along both sides of the cutset at
issue is coherent (Brinkis, 1975; Narovlyansky,
2005 et al). This assumption is based on the fact
that kinetic energy of generators mutual oscilla-
tions in the transient process under disturbance,
in the case of out-of-step conditions, passes to
the kinetic energy of the out-of-step motion of
two groups of generators along both sides of the
cutset at which the out-of-step conditions occur,
while inter-machine oscillations within these two
groups of generators decrease essentially.
44
The easiest way to implement an indirect
method for calculating the transmission angle is the
use of angle dependence of transmission current.
The disadvantage of this method is a wide scatter
of operating angle values of the OSPS under the
assumed current pickup settings due to various
possible compositions and structure of ties in the
cutset in different schemes and conditions of power
system. Besides, this dependence is nonlinear.
The lesser error is obtained by the use of the so
called phantom scheme, i.e. by modeling of the
voltage phasor of a point located at some distance
from the site of automatic system placement, for
example a receiving end of transmission line.
Such an approach is used in the OSPS installed in
power systems in Russia. Particularly complicated
conditions for selective operation of OSPS occur in
the multi-frequency out-of-step conditions along
several cutsets (Brinkis, 1975).
Further development of selective OSPS has
resulted in the creation of a multifunctional
device. The device makes it possible not only to
eliminate the out-of-step conditions if they have
occurred but also to prevent their occurrence. It
has two stages of control actions (Brinkis, 1975):
the control actions of the first stage are intended
to prevent the loss of stability and for this purpose
generation is disconnected in the surplus part of
the system and fast reserve is used (or secondary
load is shed) in the deficient part. If these control
actions are insufficient and fail to prevent out-of-
step operation the control actions of the second
stage are triggered and split power interconnection.
The use of synchronized voltage phase mea-
surements obtained from PMU offers principally
new capabilities of implementing the selective
OSPS and selective out-of-step protection system
(Phadke, 2008). Some OSPSs have been lately
suggested on the basis of PMU. In (Centeno,
1997, Bozchalui, 2006) in order to reveal tran-
sient instability the equal-area criterion is used
when representing the system by a two-machine
equivalent. Its parameters are determined by the
complex values of power system state variables.
In (Padiyar, 2006) the measurements of voltage
phases and differences of their first derivatives are
used to forecast power system stability losses ac-
cording to the criterion based on energy function. In
(Yutian, 2008) an integrated criterion is suggested
to reveal the center of oscillations with the use of
estimates of the voltage magnitude projection at
some point of the tie line between two parts of the
system and current along this tie line when using
the two-machine equivalent of power system on
the basis of the generators motion coherence in
these two parts of the system which is estimated
on the basis of currently measured angles.
The authors of this chapter suggest the prin-
ciples of creating a modified selective out-of-step
protection and prevention system (SOSPPS) with
the use of PMU measurements. Its efficiency is
demonstrated on the test power system.
Principles of Designing a
Modified SOSPPS
The Scheme of Interrelation Between States
and Control Actions
Loss of synchronism in power system operation
at a cutset can be caused by two main reasons:
The maximum admissible transfer capabil-
ity of the cutset is exceeded and, thus, the
a periodic static stability of the system is
lost;
The transient stability is lost as a result of
disturbance on one of the ties or near the
considered cutset.
In both cases an indicator for loss of synchro-
nism and the beginning of out-of-step conditions
is the difference in voltage phases on the ends of
the most critical tie line of the considered cutset.
In other words there is some maximum value of
the voltage phase difference
lim
as
, whose excess
indicates the beginning of the out-of-step condi-
tions.
45
In order to avoid this critical situation it is
necessary to maintain the cutset load at the level
not exceeding some admissible level correspond-
ing to
lim lim
<
as
. The difference between
lim

and
lim
as
should take into account irregular varia-
tions of flows along the tie lines and the need to
ensure transient stability of power systems under
standard disturbances. In Russia dispatching
centers of power systems use the recommended
values of transmission loading margins under
normal and post emergency states. In the power
systems of UCTE there are no similar explicit
recommendations. Nevertheless, setting the value

lim lim
<
as
can be expedient.
Thus, power systems can have four states
(Figure 18): secure, dangerous, emergency (out-
of-step conditions) and post-emergency. The se-
cure state of power systems is determined by the
condition
ij
<
lim
. PMU measurements are used
to trace the current value of
ij
t
( )
. The dangerous
state of power system occurs at
ij
>
lim
. For
the power system to return to a secure state it is
necessary to perform control actions to reduce
the loading of the cutset by decreasing the gen-
eration of power plants on the transmitting side
and by using fast reserve (or disconnecting sec-
ondary consumers) on the receiving side. If these
control actions are sufficient the power system
returns to the normal state. However, if the control
actions are insufficient the system passes to the
emergency state (out-of-step conditions) which
is eliminated by disconnecting the cutset (by di-
viding the power system). Should the splitting of
the power system be unsuccessfully, then an
emergency situation can develop and the post-
emergency state may turn out to be severe and
even turn into a blackout. In the event of a suc-
cessful splitting the generation and load in both
subsystems are balanced by generation discon-
nection in the surplus subsystem and by auto-
matic frequency load shedding in the deficient
subsystem. Post-emergency state in this case will
be less severe as compared to the previous one.
Figure 18. A scheme of interrelation between power system states and control actions in SOSPPS
46
It should be noted that traditionally in Russias
power system dispatching practice automatic load
shedding and OSPS are considered separately due
to the fact that the former is adjusted and operates
using the power flow values as an indicator of
overload while the latter is intended for the trans-
mission angle action, with transmission angles
being determined by indirect methods (Sovalov,
1988). The use of one and the same indicator
which is the difference between PMU-measured
voltage phases on both ends of power transmission
allows one to consider both types of automatic
systems as a single integrated emergency control
system.
Criteria for Actions of SOSPPSs Stages
As noted above the conditions for transition from
normal (safe) state to the emergency state is for-
mulated as
ij
>
lim
. Hence the criterion for
action of the cutset unloading stage of SOSPPS
will look as
C t
act
ul
ij
=
( )
>
( )

lim
. (20)
In the event that the action of the cutset unload-
ing stage is insufficient or inefficient, then the
difference in the voltage phases along the critical
tie line of the cutset at its overload continues to
increase and reaches the value
lim
as
. This indicates
the loss of aperiodic static stability of the power
system along the considered cutsets and the need
to split the system. The criterion for action of the
SOSPPSs division stage will have the form
C t
act
as
ij
as
=
( )
>
( )

lim
. (21)
To formulate the criterion for action of SOS-
PPSs division stage according to the conditions
of transient instability of power system under
large disturbances it is necessary to use the second
derivative of the difference between the voltage
phases of a critical time line in the cutset. Decrease
in the second derivative of voltage phase differ-
ence indicates conservation of transient stability
of EPS. System transition to an emergency state
(out-of-step condition) is revealed provided that
at least for three cycles of measurements by using
PMU, the second derivative value of voltage phase
difference for the critical tie line at the cutset does
not go down below some small value d
min
. In
this case each cycle may account for several scores
of milliseconds. Theoretically, d
min
= 0 , how-
ever practically this value is not equal to zero
because of errors and noise in measurements and
also inaccurate determination of the second de-
rivative due to discrete measurements. Determina-
tion of the acceptable value d
min
is an indepen-
dent problem.
Hence, for the direct power flow through the
cutset from node if to node j the criterion for action
of SOSPPSs division stage subject to transient
stability will have the form
C t
d t
dt
d
ij
ij
+
=
( )
>
( )
( )

0
2
2 min
...

( )
d t T
dt
d
ij S
2
2
2
min
, (22)
where T
S
cycle length between the PMU mea-
surements.
The corresponding criterion C
for the reverse

power flow through the cutset (from node j to
node i) is determined in a similar way.
The general criterion of transition to an
emergency state (out-of-step condition) and the
action of SOSPPSs division stage is written in
the following way:
C C C C
as act
as
=
( ) ( )
+
. (23)
47
As a result EPS is split into two isolated
subsystems.
Related Problems
The formulated approach for improving SOSPPS
based on PMU measurements includes only its
basic principles illustrated by the test example
later in Chapter III. Actually consideration should
also be given to the related problems to be solved
in the course of approach application. Below are
the main problems.
As indicated above, it is reasonable to measure
voltage phases on the basis of PMU on the ends
of the critical tie line at the cutset. The problem
is that the increase of transmitted power along the
cutset results in different loading of individual tie
lines at the cutset. Such a situation is caused by the
parameters of the tie lines and also the structure
and parameters of adjacent electric networks. This
un-homogeneity of an electric network is revealed
in different disturbance-sensitivity of nodes and tie
lines at the loaded cutset (Voitov, 1999). In other
words those elements affect operation parameters
(voltage, power flows etc.) changing to variable
extents. Hence, conditions for transition to the
out-of-step operation are formed first of all in the
most sensitive tie line. The tie line turns out to be
critical at the cutset and it is expedient to place
PMUs on its ends.
It should be noted that significance of a critical
time line at the cutset requires additional studies.
It is explained by the fact that with the start of
out-of-step condition change in the voltage angles
will be observed in all tie lines of the cutset. It
is important to establish the extent to which the
change in tie line loading at the beginning of
out-of-step condition is significant in terms of
the efficiency of SOSPPS operation.
Another problem is the necessity for verifica-
tion of criterion (23). It has two components. The
first is associated with available errors and noise
in the measurements by using PMU, delays in in-
formation transmission, measurement frequency,
etc. These technical properties should be studied
thoroughly and individually. And the results of
studies will determine specific features of the
designs of automatic systems.
The other component of the problem is the ac-
curate determination of the second derivatives of
variation of voltage phase differences. It depends
on the monotone change of phase differences and
the length of intervals between measurements, as
well as on measurement errors. The problem is
that the second derivatives must be calculated by
numerical differentiation of measured parameters.
This question also requires thorough additional
studies.
The results of studies on both components
determine efficiency of using criterion (22) and
in particular, certainty in setting the value d
min
The issue about selectivity of action of the
modified SOSPPS at the multi-frequency out-of-
step condition for the case of stability loss at
several cutsets remains to some extent open. It
seems that selectivity of work of automatic systems
should be sufficiently high and acceptable, since
the change of voltage phases on the ends of tie
lines at the cutset is a quite definite indicator of
the beginning out-of-step condition. Here the
value
lim
as
is close to 90 and the value
lim
is
uniquely determined by the state variables of the
critical tie line of the cutset and the required
margins of its transfer capability.
Test Studies
Let the test electric power system operates in
post-emergency conditions when the tie 8-5 is
loaded at 90% of its maximal transmission capa-
bility, and voltage mutual angles equal
8 5
= 36,6
and
202 100
= 50,5.
Let us consider as a disturbance the unsched-
uled disconnecting the one of two lines of the tie
8-5. The behaviors of voltage mutual angles and
their time-derivatives without any control actions
are shown in Figure 19.
48
Suppose the system is equipped with SOSPPS,
and a starting value for action of the cutset unload-
ing stage (see criterion (20)) is set as
min( ) 8 5
=
52,2. At the time of t = 0,4 s the control actions
will be realized as partial disconnections (of
generation in power surplus part of the system
and of load in its deficient part). Suppose the
larger disconnections are to be highly undesirable
through technical and/or economical limitations.
The behaviors of voltage mutual angles and
their time-derivatives with above mentioned
control actions are shown in Figure 20.
Figure 20 demonstrates the inefficiency of
undertaken control actions for providing the sys-
tem stability. The non-periodic growth of mutual
angles in the cutset, which appeared virtually
straight after the disturbance, lasts also after these
actions (although not so fast as without them). If
we suppose (only with a view to exemplify the
study) the inadmissibility of further mutual asyn-
chronous motion for the power system, then the
starting value for action of the cutset division
stage (see criterion (21)) in accordance with Fig-
ure 20 is to be set as
lim( ) 8 5
as
= 57,3. At that case
the system is being divided at the time of t = 0,5
s (when the first derivative of the angle reaches
its maximum
d
dt
8 5
= 44 grad/s).
After splitting the system into two separate
subsystems (one of them with the surplus and other
with the lack of active power) each subsystem
faces the challenge of bringing the frequency to
admissible level. The solution is in the further
reducing the power (of the generation and load
respectively). This reducing may ensue less than
that would be required under saving the parallel
Figure 19. Time behavior of ( ), d/dt ( ) and d
2
/dt
2
( ),) in the absence of control actions:
a) for the tie 8-5, b) for the tie 202-100
49
operation with unified frequency. In general the
system splitting involves the more beneficial
effect the higher power of the subsystem which
separates with the power lack.
CONCLUSION
In conclusion, we note the following:
1. Essential sophistication of the operat-
ing conditions of current EPSs enhances
danger of heavy system emergencies and
requires improvement and development of
the principles and control systems of EPS
operating conditions. For this purpose it is
necessary to apply new methods and tools
for measuring operating parameters, their
transfer, processing and application of
operating conditions control of EPS. The
suggested basic principles of the system of
monitoring and forecasting the operating
conditions and control of EPS substantially
enhance efficiency and adaptability of
the coordinated operation and emergency
control in EPS. The results described in the
paper illustrate efficiency of the approach
and applied methods and information
technologies.
2. Structural and functional decomposition
of state estimation problem is an effective
method to solve the problems arising during
calculation of large schemes. The proposed
two-level algorithm for structural decom-
Figure 20. Time behavior of ( ), d/dt ( ) and d
2
/dt
2
( ) when maximum permissible unload-
ing the cutset: a) for the tie 8-5, b) for the tie 202-100
50
position of the SE problem allows one to
simultaneously process the data for local
subsystems of considerably smaller dimen-
sionality; decrease the adverse impact of in
homogeneity of the calculated scheme and
telemetric information when calculating
one-voltage-class subsystems; essentially
simplify solution of the coordination prob-
lem which, in this case, does not require
iterative calculations by subsystems; and
reduce the time for SE problem solving for
the entire scheme.
3. Dynamic state estimation with application
of the Kalman filter can be used to forecast
all the EPS state variables for a short period
of time. The regular filter adjustment im-
proves the forecast quality. Measurements
from PMU that are also applied as the state
vector components and the precise measure-
ments improve the forecast results.
4. The need for wider application of monitor-
ing systems is growing increasingly urgent
in the light of the future possibility to in-
terconnect the main power grids of Europe
and Russia. The technological progress in
the last decades has shown that the use of
perspective information technologies, first
of all ANN, can provide reliable operation
of future interconnected power grids of EU
and Russia, optimal use of energy resources
on vast territories and mutually beneficial
electricity trade between different regions
in terms of market requirements.
5. Specific features of using the singular
analysis have been studied to determine
sensor and weak points in EPS consisting
of a great number of subsystems. Separation
of sensor and weak points for the intercon-
nected power system is shown to be possible
for each subsystem independently.
6. The absence of the control devices coordi-
nation during the post-disturbance period
is one of the main causes of the voltage
instability, which permanently occurs in
power systems all over the world. The pro-
posed multi-agent control system provides
reactive power control by coordinating the
work of different discrete and continu-
ous control devices in a post-disturbance
period. The reactive power control in a
post-disturbance period prevents generator
tripping and maintains load bus voltages
within the normal range. The efficiency of
this approach has been proved by numerical
simulations.
7. After splitting the system into two separate
subsystems (one of them with the surplus
and other with the lack of active power)
each subsystem faces the challenge of
bringing the frequency to admissible level.
The solution is in the further reducing the
power (of the generation and load respec-
tively). This reducing may ensue less than
that would be required under saving the
parallel operation with unified frequency.
In general the system splitting involves the
more beneficial effect the higher power of
the subsystem which separates with the
power lack.
ACKNOWLEDGMENT
The study was supported by the Grants of Lead-
ing Scientific School of RF#1857.2008.8 and of
Russian Foundation of Basic Researches #09-08-
91330 and by Federal Agency for Science and
Innovations within Federal Program R&D in Pri-
ority Areas of Russias Science and Technological
Complex Development for 2007-2012. The study
was supported by the call FP7-ENERGY-2008-
RUSSIA, FP7 Cooperation Work Programme:
Theme 5 Energy
51
REFERENCES
Abdel-Rahman, K., Mill, L., Phadke, A., De La
Ree, J., & Liu, Y. (2001). Internet based wide area
information sharing and its roles in power system
state estimation. In Proc. IEEE Power Eng. Soc.
Winter Meeting, vol. 2, (pp. 470 475). 28 Jan.
1 Feb. 2001.
Amin, S. M., & Wollenberg, B. F. (2005). Toward
a smart Grid: Power delivery for the 21st century.
IEEE Power and Energy Magazine, 3(5), 3441.
doi:10.1109/MPAE.2005.1507024
Ayuev, B. I., Shulginov, N. G., & Koshcheev, L.
A. (2008). Development of principles, algorithms
and tasks of emergency control in Russias UPS.
Proceedings of the 3rd International Scientific and
Technical Conference Energy System: Control,
Competition, Education, (p. 6). Ekaterinburg, Rus-
sia: Ural State Technical University, (in Russian)
Ayuyev, B. I., Erohin, P. M., Neuymin, V. G.,
Mashalov, E. V., & Shubin, N. G. (2005). The
software complex of optimal power flow solu-
tion for United Power System of Russia in a
competitive market. In Proceedings IEEE Power
Tech Conference, St. Petersburg.
Barzam, A. B. (1964). System automatic devices.
Moscow, Russia: Energiya. (in Russian)
Bellifemine, F., Caire, G., & Greenwood, D. (2007).
Developing multi-agent systems with JADE. Lon-
don, UK: Wiley. doi:10.1002/9780470058411
Bellifemine, F., Caire, G., Trucco, T., & Rimassa,
G. (2000). JADE programmers guide. CSELT &
University of Parma. Retrieved from http://www.
jade.tilab.com/doc
Bialek, W. (2000). Tracing-based unifying frame-
work for transmission pricing of cross-border
trades in Europe. In Proc 2000 International
Conference on Electric Utility Deregulation and
Restructuring, and Power Technologies, London.
Borovikov, V. P. (2006). Forecasting in STATIS-
TICA system in Windows environment: Theory
and intensive practice on computer: In V. P. Bo-
rovikov & G. I. Ivchenko (Eds.), Tutorial, 2
nd
ed.,
revised and enlarged. Moscow, Russia: Finansy i
statistika (in Russian)
Bozchalui, M. C., & Sanaye-Pasand, M. (2006).
Out-of-step relaying using phasor measurement
unit and equal area criterion. Proceedings of IEEE
Power India Conference.
Centeno, V. E. A. (1997). An adaptive out-of-step
relay. IEEE Transactions on Power Delivery,
13(1), 131138.
Chuang, A., & McGranaghan, M. (2004). Func-
tions of a local controller to coordinate distributed
resources in a smart grid. Proceedings of 2004
IEEE PES General Meeting, IEEE Press, Pitts-
burgh, USA, (p. 7).
CIGRE Task Force. (2007, April). C2.02.24: CI-
GRE defense plan against extreme contingencies.
Clements, K. A., Denison, O. J., & Ringle, R. J.
(1972). A multy-area approach to state estimation
in power system networks. In Proc. IEEE Power
Eng. Soc. Meeting, San Francisco, CA. (Paper
C72 465-3).
El-Kleib, A. A., Nieplocha, J., Singh, H., &
Maratukulam, D. J. (1992). A decomposed state
estimation technique suitable for parallel proces-
sor implementation. IEEE Transactions on Power
Systems, 7(3), 10881097. doi:10.1109/59.207322
ENAS. (2004). Methodological guidelines on
power system stability. (in Russian)
Etingov, P. V. (2006). Application of the systems of
transients monitoring to control FACTS devices.
Proceedings of the Workshop Methodological
Problems in Reliability Studies of Bulk Energy
Systems, Kharkov National Technical University,
Kharkov, Ukraine, (p. 6). (in Russian)
52
European Commission. (2006). European smart
Grids technology platform: Vision and strategy
for Europes electricity networks of the future (p.
45). Luxemburg.
Falcao, D. M., Wu, & Murphy L. (1995). Parallel
and distributed state estimation. IEEE Trans-
actions on Power Systems, 10(2), 724730.
doi:10.1109/59.387909
FIPA. (2010). The current set of standard FIPA
specifications. Retrieved from http://www.fipa.
org/repository/standardspecs.html
Gamm, A., Grishin, Y., Glazunova, A., Kolosok,
I., & Korkina, E. (2007). New EPS state estima-
tion algorithms based on the technique of test
equations and PMU measurements. In Proc of
the International Conference PowerTech2007,
Lausanne.
Gamm, A. Z. (1983). Decomposition algorithms
for solution of power system state estimation
problem. Electronnoye Modelirovanie, 3, 6368.
Gamm, A. Z., & Golub, I. I. (1995). The prob-
lem of weak places in electric power systems. In
Proc. 1995 International Conference on Electri-
cal Power Engineering Power Tech, Stockholm,
(pp. 542-546).
Gamm, A. Z., & Golub, I. I. (2003). Contribution
factor of active and reactive power transmission
in electric power system. Elektrichestvo, 3, 916.
Gamm, A. Z., & Grishin, Yu. A. (1995). Distrib-
uted information processing in automated power
system control systems. In Proc. V Int. Workshop
Distributed information processing. Novosibirsk,
Russia, (pp. 243-247). (in Russian)
Gamm, A. Z., Kolosok, I. N., & Paltsev, A. S.
(2007). Methods for decomposition of EPS state
estimation problem when solving it on the basis of
multiagent technologies. In Scientific Proceedings
of Riga Technical University Power and Electrical
Engineering, Riga, (pp. 205-214).
Gamm, A. Z., Zaika, R. A., & Kolosok, I. N.
(2005). Robust methods of EPS state estimation
based on the test equations and their software
support using genetic algorithms. [in Russian].
Electrichestvo, 10, 28.
Glazunova, A., Kolosok, I., & Korkina, E. (2008,
April). Test equation method for state estimation
using PMU measurements. Proc. of Conf. Moni-
toring of Power System Dynamics Performance.
Saint Petersburg.
Glazunova, A., Kolosok, I., & Korkina, E. (2009).
PMU placement on the basis of SCADA measure-
ments for fast load flow calculation in electric
power systems, Proceedings of the International
Conference PowerTech2009, Bucharest.
Goldberg, D. (1989). Genetic algorithms in search,
optimization and machine learning.
Gonik Ya, E., & Iglitsky, E. S. (1988). Automatic
elimination of out-of-step condition. Moscow,
Russia: Energoatomizdat. (in Russian)
Handshin, E. (2006). Asset management and sur-
vivability of electric enrgy systems during large
disturbances. In E. Handshin, E. Hauptmeter, &
D. Hause (Eds.), Liberalization and modernization
of power systems: risk assessment and optimiza-
tion for asset management: The 3
rd
International
workshop, Irkutsk Energy Systems Institute (pp.
156-168).
Haykin, S. (2006). Neural networks: A compre-
hensive foundation (2nd ed.). Williams Publishing
House.
Iwamoto, S., Kusano, M., & Quantana, V. H.
(1989). Hierarchical state estimation using
a fast rectangular-coordinate method. IEEE
Transactions on Power Systems, 4(3), 870879.
doi:10.1109/59.32574
53
Jang, W., Vittal, V., & Heydt, G. T. (2008, No-
vember). Diakoptic state estimation using Pha-
sor Measurements Units. IEEE Transactions on
Power Systems, 23(4), 15801589. doi:10.1109/
TPWRS.2008.2002285
Jiang, W., Vittal, V., & Heydt, G. T. (2007, May).
A distributed state estimator utilizing synchro-
nized Phasor measurements. IEEE Transactions
on Power Systems, 22(2), 563571. doi:10.1109/
TPWRS.2007.894859
Kolosok, I. N., Korkina, E. S., & Paltzev, A. S.
(2009). Decomposition of the state estimation
problem at calculation for the schemes of large
dimensionality. Proceedings of International
Conference Liberalization and Modernization
of Power Systems: Coordinated Monitoring and
Control towards Smart Grids, Energy Systems
Institute, Irkutsk, Russia, (p. 7). (in Russian)
Kolosok, I. N., & Zaika, R. A. (2003). A genetic
algorithm as a means of enhancing the efficiency
of information validation methods for the power
system control. In Proc. 30 International Conf.
Information Technologies in Science, Education,
Business and Protection of Natural Resources.
Gurzuf, Ukraine, (pp. 145147).
Kurbatsky, V. G., & Tomin, N. V. (2008). Ap-
plication of the ANAPRO software for analysis
and forecasting of state parameters and process
characteristics in electric power systems. Proceed-
ings of the 8
th
Baikal All-Russian Conf. Information
and Mathematical Technologies in Science and
Management, part 1 (pp. 91-99). Irkutsk, Russia:
SEI SB RAS.
Kurbatsky, V. G., & Tomin, N. V. (2009). Use of the
ANAPRO software to analyze and forecast operat-
ing parameters and technological characteristics
on the basis of macro applications. Proceedings
of IEEE Bucharest Power Tech (p. 7). Bucharest,
Romania: IEEE Press.
Lachs, W. R. (1992, May). Voltage instability in
interconnected power systems: A simulation ap-
proach. IEEE Transactions on Power Systems,
7(2), 753761. doi:10.1109/59.141782
Lachs, W. R. (2002, May). Controlling Grid
integrity after power system emergencies. IEEE
doi:10.1109/TPWRS.2002.1007916
Lachs, W. R. (2003, February). A new horizon
for system protection scheme. IEEE Transactions
TPWRS.2002.807065
Li, L., Liu, Y., Mu, H., & Yu, Z. (2008). Out-of-step
splitting scheme based on PMUs. Proceedings of
DRPT2008 Int. Conf. Nanjing, China.
Makarov Yu, V., Reshetov, V. I., Strojev, V. A., &
Voropai, N. I. (2005). Blackout prevention in the
United States, Europe and Russia. Proceedings
of the IEEE, 93(11), 19421955. doi:10.1109/
JPROC.2005.857486
McArthur, S. D. J., Davidson, E. M., Catterson, V.
M., Dimeas, A. L., Hatziargyriou, N. D., Ponci, F.,
& Funabashi, T. (2007, November). Multi-agent
systems for power engineering applicationz - Part
I. IEEE Transactions on Power Systems, 22(4),
17431752. doi:10.1109/TPWRS.2007.908471
Milano, F. (2005). An open source power system
analysis toolbox. IEEE Transactions on Power
Systems, 20(3), 11991206. doi:10.1109/TP-
WRS.2005.851911
Milano, F. (2008, February). Documentation for
PSAT, version 2.0.0.
Mogilko, R. (2008, April). SMART-WAMS re-
corder of transient variables: Experience in design
and implementation experience - Prospects for
development. Proc. of Conf. Monitoring of Power
System Dynamics Performance, Saint Petersburg.
54
Narovlyansky, V. G., & Nalevin, A. A. (2005). A
method for determination of equivalent param-
eters of power system network during out-of-step
condition. [in Russian]. Elektrichestvo, 8, 1521.
Office of Electric Transmission and Distribution,
United State Department of Energy. (2003). Grid
2030: A national version for electricitys second
100 years (p. 52). Washington, DC.
Ohura, Y., E. A. (1990). A predictive out-of-step
protection system based on observation of the
phase difference between substations. IEEE
Transactions on Power Delivery, 5(4), 661667.
doi:10.1109/61.103664
Ossovsky, S. (2004). Neural networks for informa-
tion processing (trans. I. D. Rudinsky). Moscow,
Russia: Finansy i statistika. (in Russian)
Padiyar, K. R., & Krishna, S. (2006). Online
detection of loss of synchronism using energy
function criterion. IEEE Transactions on Power
Delivery, 21(1), 163171. doi:10.1109/TP-
WRD.2005.848652
Panasetsky, D. A. (2009). Multi-agent approach
to coordination of different emergency control
Devices against voltage collapse. Proceedings
of International Conference Liberalization and
Modernization of Power Systems: Coordinated
Monitoring and Control towards Smart Grids,
Energy Systems Institute, Irkutsk, Russia, (pp.
171-176). (in Russian)
Panasetsky. (n.d.). Modified 24 bus IEEE one
area RTS-96 test system. Retrieved from http://
panasetsky.net/wpcontent/uploads/FILES/d 024
my.mdl
Phadke, A. G., & Thorp, J. S. (2008). Synchronized
phasor measurements and their applications.
Berlin, Germany: Springer. doi:10.1007/978-0-
387-76537-2
Phadke, G. (2002). Synchronized Phasor measure-
ments. A historical overview. In Proc. IEEE/PES
Transmission and Distribution Conference, N.1,
(pp. 476-479).
Pospelov, D. A. (Ed.). (1986). Fuzzy sets in the
control and artificial intelligence models. Mos-
cow, Russia: Nauka. (in Russian)
Pourbeik, P., Kundur, P. S., & Taylor, C. W.
(2006). The anatomy of a power grid blackout.
IEEE Power and Energy Magazine, 4(5), 2229.
doi:10.1109/MPAE.2006.1687814
Quintana, V. H., & Mueller, H. (1991). Partitioning
of power networks and application to security con-
trol. IEE Proceedings. Generation, Transmission
and Distribution, 138(6), 535545. doi:10.1049/
ip-c.1991.0067
Shahidehpour, M. (2009). Smart Grid: A new
paradigm for power delivery. Proceedings of
IEEE Bucharest Power Tech (p. 7). Bucharest,
Romania: IEEE Press.
Sovalov, S. A., & Semenov, V. A. (1988). Emer-
gency control in power systems. Moscow, Russia:
Energoatomizdat. (in Russian)
Taylor, C. W., & Erickson, D. C. (1997, January).
Recording and analyzing the July 2 cascading
outage. Computer Applications in Power Systems,
10(1), 2630. doi:10.1109/67.560830
Venikov, V. A., Stroev, V. A., Idelchik, V. I.,
& Tarasov, V. I. (1975, August). Estimation of
electrical power system steady-state stability in
load flow calculations. IEEE Transactions on
Power Apparatus and Systems, 94(3), 152159.
doi:10.1109/T-PAS.1975.31937
Voitov, O. N., Voropai, N. I., Gamm, A. Z., Golub,
I. L., & Efimov, D. N. (Eds.). (1999). Analysis
of Inhomogenities in Electric Power Systems,
Novosibirsk, (p. 302).
55
Voropai, N. I. (2008). Problems and directions
of control system development in electric power
systems. Proceedings of the 3rd International
Scientific and Technical Conference Energy
System: Control, Competition, Formation (p.
7). Ekaterinburg, Russia: Ural State Technical
University. (in Russian)
Voropai, N. I., & Etingov, P. V. (2006). Develop-
ment of adaptation methods of fuzzy logic power
system stabilizer. Proceedings of 2006 IEEE PES
General Meeting, IEEE Press, Montreal, Canada,
(p. 7).
Voropai, N. I., Etingov, P. V., Oudalov, A. S.,
Germond, A., & Sherkaoui, R. (2005). Congestion
management using coordinated control of FACTS
devices and load management. Proceedings of 15th
PSCC, University of Liege, Liege, Belgium, (p. 7).
Voropai, N. I., Kolosok, I. N., Kurbatsky, V. G.,
Etingov, P. V., Tomin, N. V., Korkina, E. S., & Palt-
sev, A. S. (2010). Intelligent coordinated operation
and emergency control in electric power systems.
Proc. Conference on Control Methodologies and
Technology for Energy Efficiency (CMTEE-2010),
Vilamoura, Portugal.
Wallach, Y., & Handschin, E. (1981). An ef-
ficient parallel processing method for power
system state estimation. IEEE Transactions on
TPAS.1981.316852
Wang, X. M., & Vittal, V. (2004). System island-
ing using minimal cut sets with minimum net flow.
IEEE PES General Meeting, Denver, USA.
Zhao, L., & Abur, A. (2005). Multiarea state esti-
mation using synchronized phasor measurements.
IEEE Transactions on Power Systems, 20(2),
611617. doi:10.1109/TPWRS.2005.846209
56
APPENDIX: LIST OF AUTHORS
The problem of monitoring, forecasting and control in electric power system: Voropai N.I.
Decomposition of power system state estimation problem with the use of PMU data for large di-
mension schemes: Kolosok I.N., Korkina E.S., Paltsev A.S.
PMU for fast calculation of steady state in electric power systems: Glazunova A.M.
Artificial intelligence technologies for monitoring large power interconnections: Kurbatsky V.G.,
Tomin N.V.
Study of the properties of a large electric power system by using singular analysis: Gamm A.Z.,
Golub I., Bershansky R.G
A Multi-Agent Approach to Coordination of Different Emergency Control Devices Against Volt-
age Collapse: Panasetsky D. A
Advanced Emergency Control System for Prevention and Elimination of Power System out-of-Step
Operation Using PMU: Voropai N.I, Rehtanz C., Efimov D.N., Popov D.B., Hger U.
57
Chapter 2
DOI: 10.4018/978-1-61350-138-2.ch002
Vo Ngoc Dieu
Asian Institute of Technology, Thailand
Weerakorn Ongsakul
Asian Institute of Technology, Thailand
Hopfeld Lagrange Network
for Economic Load Dispatch
ABSTRACT
In this chapter, a Hopfeld Lagrange network (HLN) is proposed for solving economic load dispatch
(ELD) problems. HLN is a combination of Lagrangian function and continuous Hopfeld neural network
where the Lagrangian function is directly used as the energy function for the continuous Hopfeld neu-
ral network. In the HLN method, its energy function augmented by Hopfeld terms from the continuous
Hopfeld network could damp out oscillation of the conventional Hopfeld network during the conver-
gence process. Consequently, the proposed HLN can overcome the disadvantages of the conventional
Hopfeld network in solving optimization problems for its simpler implementation, better global solution,
faster convergence time, and larger scale applications. The proposed method has been tested on differ-
ent ELD problems including all thermal units, thermal units with fuel constraint, and both thermal and
hydro units. The obtained results from the test cases have shown that the proposed method is effective
and effcient for solving the ELD problems. Therefore, the HLN method is the new contribution to the
development of new methods for solving optimization problems in power systems.
58
Hopfeld Lagrange Network for Economic Load Dispatch
INTRODUCTION
Power system engineering has the longest his-
tory of development among the various areas
within electrical engineering. Since the practical
numerical optimization methods applied to power
system engineering and operation, they have been
playing a very important role in economic opera-
tion of power systems. The value contributed by
the power system optimization is considerable
in economics for large utilities from fuel cost,
operational reliability, and security.
In power system operation, the problem allocat-
ing among the available thermal power generating
units to the customers load demands in an eco-
nomic, secure and reliable way has been identified
and received great attention since the beginning
of the 20
th
century (Happ, 1977; Chowdhury &
Rahman, 1990). The problem has been formulated
as an optimization problem which is to minimize
fuel cost of overall online generating units while
satisfying load demand and other constraints over
a considered schedule time. This problem has been
frequently known as the economic load dispatch
(ELD) problem. The simplest form of the ELD
problem can be considered as to minimize the
total fuel cost through determining the allocation
of power generation of each units among a set of
committed units in a thermal power plant subject
to a constraint that total power generation equals
to load demand or among a set of committed units
in different thermal power plants subject to a con-
straint that total power generation equals to load
demand plus transmission power loss. This type
of ELD problem is referred to the conventional
problem applied in vertically integrated power
systems. The more complicated ELD problems
can be arranged in ascending order as more con-
straints added:
The generator capacity limits are
considered.
There are conficts among the consid-
erations such as economy, security, and
emission.
Fuel, labor and maintenances constraints
are added.
The optimal power solution considers both
active and reactive control variables.
The problem is considered in deregulated
environment.
In fact, the ELD problem involves the solu-
tion of two different problems. The first one is
pre-dispatch problem or unit commitment (UC)
problem which requires selecting optimal units
among the available ones to meet the expected
load demand with a specified reserve of operation
over a scheduled time horizon. The last one is the
on-line ELD which requires satisfying the load
demand among the actually committed units so
as to minimize total cost of supplying the require-
ment of the system.
Due to the great values contributed to the
economic operation of power systems, the ELD
problems have been attracted the attention many
researchers from 1920s with several solution meth-
ods proposed to deal with more complicated and
larger scale problems. One of the earliest methods
to find economic results for power generation
schedule was known as equal incremental method,
where the optimum solution is obtained when all
the marginal cost of the committed generating
units are equal (Stahl, 1930, 1931). Since digital
computers were used in 1950s the numerical
methods have been continuously developed for
solving more complicated and larger problems
with more efficient and better quality solution.
Several solution methods have been proposed
for solving the problems including conventional,
artificial intelligent, and hybrid methods. Many
conventional methods have been widely used for
solving the ELD problems such as dynamic pro-
gramming (DP) (Liang, Glover & Glover, 1992),
linear programming (LP) (Wells, 1968), quadratic
59
programming (QP) (Irving & Sterling, 1985), inte-
rior point method (IPM) (Ponnambalam, Quintana
& Vanelli, 1992), and Lagrangian relaxation (LR)
(Fisher, 1973). Conventional methods require the
models of optimal generation scheduling problems
to be represented as piecewise linear or polyno-
mial approximations of monotonically increasing
nature. However, such an approximation may lead
to sub-optimal solution, resulting in huge loss of
revenue over the time. Hence, the new trend in
the recent time is to use more realistic models of
hydro and thermal plants.
In recent years, meta-heuristic optimization
techniques have attracted much attention of re-
searchers due to their ability to seek for global
optimal solution for problems with complicated
constraints. These methods have proved to be
very efficient since they do not place any restric-
tion on the shape of the cost curves and other
non-linearity in model representation. Although
these heuristic methods do not always guarantee
to find the globally optimal solution, they can
provide a reasonable solution (sub-optimal near
globally optimal) in a sufficient computational
time. Among the many artificial intelligence
based methods, some popular ones have been
implemented for solving the ELD problems
consisting of Hopfield neural network (HNN)
(Park, Kim, Eom, & Lee, 1993), simulated
annealing (SA) (Wong & Fung, 1993), tabu
search (TS) (Ongsakul, Dechanupaprittha, &
Ngamroo, 2004), ant colony search algorithm
(ACSA) (Song, Chou, & Stonham, 1999), genetic
algorithm (GA) (Bakirtzis, Petridis & Kazarlis,
1994), evolutionary programming (EP) (Wong
& Yuryevich, 1998), differential evolution (DE)
(Nomana & Iba, 2008), and particle swarm
optimization (PSO) (Jeyakumar, Jayabarathi &
Raghunathan, 2006). Most of the conventional
methods can offer good solution in a short com-
putational time but they can only deal with simple
and small or medium-scale problems meanwhile
most of the meta-heuristic methods can deal with
more complicated and larger scale problems.
However, the meta-heuristic methods may suffer
from slow convergence and local optimum for
large-scale problems. A clear consensus is pres-
ently heading toward the hybrid models, which
are the combinations of both conventional and
non-conventional methods and can handle the
present day complicated problems commonly
seen within developed countries. Moreover,
parallel techniques are also designed for solution
methods to reduce computational time in large-
scale problems. However, the development of
these techniques is still limited due to hardware
dependence. On the other hand, developing new
methods which can deal with complicated prob-
lems and obtain fast solution is being continued.
In the recently new trends, hybrid systems are
considered as promising methods and widely
used for optimal generation scheduling.
Many hybrid systems have been proposed for
solving the ELD problems such as fuzzy logic
and GA (Song, Wang, Wang & Johns, 1997),
hybrid EP and sequential QP (Attaviriyanupap,
Kita, Tanaka & Hasegawa, 2002), combined
GA and TS (Ruangpayoongsak, Ongsakul &
Runggeratigul, 2002), hybrid PSO and sequential
QP (Victoire & Jeyakumar, 2004), hybrid EP and
LP (Somasundaram, Lakshmiramanan & Kup-
pusamy, 2005), combined DE and QP (Coelho
& Mariani, 2006), etc. Hybrid systems can be
usually the combinations whether among the
conventional methods or between the conven-
tional methods and the meta-heuristic methods
or among the meta-heuristic methods to utilize
the advantages of element methods. Therefore,
the hybrid systems can deal with more compli-
cated problems and obtain better solution with
shorter computational time than many single
meta-heuristic search methods.
In this chapter, a new method based on a com-
bination of continuous Hopfield neural network
and Lagrangian function is proposed for solving
the ELD problems including:
Basic ELD problem: This is a very simple form
of ELD problem. The problem is to minimize an
60
objective of total cost for thermal generating units
subject to power balance constraint and power
generation limits and ramp rate constraints for
each unit during a schedule time horizon, usually
1 hour. The objective function considered for
this problem here is a sum of quadratic fuel cost
functions of all online units and the power loss
in transmission system is computed via Krons
loss formula where the power loss is a function
of power outputs of units.
Fuel constrained ELD problem: This problem
is more complex than the basic ELD problem with
more objective and constraints added. The objec-
tive of the problem is to minimize both total fuel
cost and emission level of thermal generating units
while satisfying different constraints including
power balance, fuel delivery, fuel storage, genera-
tor capacity limits, fuel delivery limits, and fuel
storage limits for a certain period of time. Both
the objectives of fuel cost and emission considered
in this problem are modeled as quadratic func-
tions which are the function of power outputs of
generating units.
Hydrothermal ELD problem: This problem
includes both thermal and hydro generating units
but only fuel cost of thermal units is considered
while neglecting the total cost of hydro units since
it is very small and negligible. The hydraulic
constraints for hydro units are also included in
addition to the constraints for the thermal units.
The purpose of this problem is to minimize total
fuel cost of thermal generating units subject to
power balance, continuity of reservoir head,
and generator capacity limit constraints. The
functions considered in this problem including
fuel cost of thermal units and water discharge
and reservoir head variation for hydro units are
quadratic functions.
BACKGROUND
Continuous Hopfield neural networks (Hopfield,
1982, 1984) have been widely used for solving
optimization problems in different fields. The
continuous Hopfield neural networks are a recur-
rent network type that operates in an unsupervised
manner. The action of a Hopfield network is based
on the minimization of its energy function which
is mapped from an optimization problem to the
network will converge to a solution of the problem.
One of the advantages of the Hopfield network is
that it can efficiently handle variable limits by its
sigmoid function. However, the applications of the
Hopfield network to the optimization problems
are limited to simple problems with linear con-
straints due to the complex process of mapping
from the problem to the neural network. Moreover,
large number of iterations and oscillation during
convergence process are also the major concerns
that the Hopfield network can be suffered while
solving optimization problems.
In the implementation of the conventional
continuous Hopfield neural network (HNN) to
the ELD problems (Park et al., 1993; Su & Chiou,
1997), an energy function for the problem is pre-
defined including objective and constraints associ-
ated with weighting factors, and then mapped into
the HNN to determine the weighting connections
for the neurons. There are many drawbacks for
such an application. Firstly, the solution by the
HNN for the problem is sensitive to the selected
weighting factors associated with the defined
energy function which may lead to local optimal
solution if these weighting factors are not carefully
tuned. Secondly, HNN is very difficult to deal
with the complicated problems with nonlinear
constraints since the problem constraints have to
be linearized before implementing in HNN. Lastly,
the computational effort of HNN for solving the
ELD problem is high since it needs large number of
iterations to obtain optimal solution, thus it is dif-
ficult for HNN to deal with large-scale problems.
Some improvements for the conventional HNN
have been proposed to overcome the mentioned
drawbacks. To speed up the computational time for
HNN, the sigmoid function of continuous neurons
has been linearized (Su & Chiou, 1997, 2000).
61
By doing this way, the solution for the problem
is directly found with a very short computational
time similar to analytical approach. However, this
improvement can be applied to simple problems
with non-binding inequality constraints. Another
improvement of HNN has been proposed in (Yal-
cinoz & Short, 1997) where the energy function
of HNN is formulated in quadratic programming
form for identifying with quadratic programming
problem to determine the weight connections for
the neurons. For this improvement, the improved
HNN can deal with large-scale problems and its
speed of calculation has been also considerably
improved. Nevertheless, this formulation can only
apply to quadratic programming problems with
linear constraints. An adaptive HNN has been
proposed in (Lee, Sode-Yome & Park, 1998;
Lee, Nuroglu & Sode-Yome, 2000) to speed up
the convergence of HNN by adaptively adjusting
slope, bias, and learning rates of neurons. With
the new adjustment techniques, the adaptive
HNN can deal with non-convex problems and
its computational speed has been also improved.
However, the implementation of this neural net-
work is similar to the conventional HNN; that is
the weight connections for neurons have to be
pre-determined and the problem constraints have
to be linearized. Recently, a modification for HNN
has been proposed in (da Silva, Nepomuceno &
Bastos, 2004) by representing the energy of HNN
in two terms, a confinement term that groups the
constraints and an optimization term that conducts
the network output to the equilibrium points. The
minimization for energy function of this modi-
fied HNN is conducted in two stages for the two
corresponding terms of the energy function. This
modified HNN can deal with more complicated
problem compared to the conventional HNN and
its ability to find global optimal solution has been
also improved. However, for implementation this
modified neural network, the problem constraints
also need to be linearized.
In this research, a new improvement of con-
tinuous Hopfield neural network, called Hopfield
Lagrange network (HLN), is proposed to over-
come to the difficulties of the Hopfield network.
In the proposed HLN method, the Lagrangian
function is directly used as the energy function
of the Hopfield network. The advantages of the
proposed neural network over to the conventional
Hopfield network are as follows:
The proposed neural network is not nec-
essary to predefne an energy function as-
sociated with penalty factors and map the
problem into the Hopfeld network for
determining the synaptic interconnections
among the neurons.
Since the proposed neural network uses
Lagrangian function as the energy function
for the Hopfeld network, it can effciently
handle constraints of the problems without
causing constraint mismatch. Moreover,
the proposed neural network is not limited
to the simple problems with linearized con-
straints as the Hopfeld network, especially
for the time-coupling constraints.
The proposed neural network can give a
very fast convergence to the optimal solu-
tion compared to the conventional Hopfeld
network.
The proposed neural network can eas-
ily deal with large-scale and complicated
optimization problems via Lagrangian
relaxation.
The proposed HLN model is solved using
sub-gradient technique with updating step sizes
which will be easily tuned for each problem while
the slope of sigmoid function for continuous
neurons can be fixed. With the new improve-
ments, the proposed HLN method could solve any
optimization problems that the Hopfield network
can. Moreover, the solution quality obtained by
the proposed neural network is also higher than
that from the conventional HNN. Therefore, the
newly proposed neural network could be one of
62
the new options for solving optimization problems
in power systems.
HOPFIELD LAGRANGE NETWORK
AND APPLICATIONS
Hopfield Lagrange Network
for Optimization Problem
Optimization Problem Formulation
The constrained optimization problem is formu-
lated as follows:
Min f x
k
( ) (1)
subject to
g x
i k
( ) = 0i = 1, , M (2)
x x x
k k k ,min ,max
k = 1, , N (3)
where f(x
k
) objective function to be minimized;
g
i
(x
k
) equality constraint;
x
k
independent variable;
x
k,min
, x
k,max
lower and upper bounds of variable
x
k
.
Hopfield Lagrange Network Algorithm
The Lagrangian function for the problem is for-
mulated as follows:
L f x g x
k i i k
i
M
= +
=
( ) ( )
1
(4)
where
i
is Lagrangian multiplier associated with
constraint i.
To apply the Lagrangian function in HLN, the
continuous and multiplier neurons corresponding
to independent variables and Lagrangian multipli-
ers are respectively needed. In HLN, the neurons
representing continuous variables are called
continuous neurons, and the neurons representing
Lagrange multiplier are called multiplier neurons.
The energy function for HLN is formulated
based on the Lagrangian function as follows:
E f V V g V g V dV
k x i i k x
i
M
c
V
k
N
k x
= + +
=
( ) ( ) ( )
, , ,
,
1
1
0
1

(5)
where V
k,x
output of continuous neuron k corre-
sponding to x
k
; V
i,
output of multiplier neuron i
corresponding to
i
; g
c
-1
inversed sigmoid function
of continuous neurons.
The last term in (5) is the Hopfield term of
continuous neurons where its global effect is
displacement of solutions toward the interior of
the state space (van den Berg & Bioch, 1993).
The dynamics of the neural network are defined
such that the energy function (5) should be mini-
mized with respect to the continuous neurons and
maximized with respect to the multiplier neurons.
The network dynamics are defined as follows:
dU
dt
E
V
f V
V
V
g V
V
U
k x
k x
k x
k x
i
i k x
k x
k x
,
,
,
,
,
,
,
,
( ) ( )
=

(6)
dU
dt
E
V
g V
i
i
i k x
,
,
,
( )
= (7)
where U
k,x
total inputs of continuous neuron k cor-
responding to the output V
k,x
; U
i,
total inputs of
multiplier neuron i corresponding to the output V
i
.
The inputs of neurons at iteration n are updated
based on the dynamics from (6) and (7) as follows:
63
U U
E
V
k x
n
k x
n
x
k x
,
( )
,
( )
,
=

1
(8)
U U
E
V
i
n
i
n
i
,
( )
,
( )
,

= +

1
(9)
where
x
and
are positive updating step sizes for

the inputs of continuous and multiplier neurons,
respectively.
The sigmoid function of continuous neurons
for determining the relationship between the
inputs and outputs is defined by a monotonically
increasing function as follows:
V g U
x x
U x
k x c k x
k k
k x k , ,
,max ,min
, ,min
( ) tanh = =

+ ( )
+
2
1

(10)
where is a positive scaling factor known as
slope which determines the shape of the sigmoid
function. The shape of the sigmoid function is
given in Figure 1.
The transfer function to determine the outputs
of multiplier neurons from their inputs is defined
by a linear function as follows:
V g U U
i m i i , , ,
( )

= = (11)
The diagram for the proposed HLN is given
in Figure 2.
Selection of Parameters
The proper parameter selection will guarantee
rapid convergence for the neural network. So far,
there is no method to find optimal parameters for
the neural network. Therefore, the parameters are
Figure 1. Sigmoid function of continuous neurons with different slopes where x
k,max
= 1 and x
k,min
= 0
64
tuned based on experiments. Based on experiments
some observations are draw as follows.
If the slope < 1, the neural network converges
very fast but the obtained solution may be local
optimum. In contrast, if the slope > 1, the solu-
tion from the neural network is global optimum
but slightly slower convergence, the larger value
of used the better solution obtained. Therefore,
the preferable values for is ranging from 10 to
100, since higher values of lead to longer to
converge but the improvement in the obtained
solution is inconsiderable.
The updating step sizes
x
and
for neurons
usually depend on the problem being considered.
It is observed that the larger the values of the
updating step sizes, the closer the discrete system
behavior of the neural network, producing values
at the upper and lower limits of each neuron. On
the contrary, the smaller the values of updating
step sizes, the slower convergence of the neural
network. To determine these parameters, a small
value will be chosen first and then gradually in-
crease until the network behaves like a discrete
system. The proper values will be obtained. Note
the values of the updating step sizes are usually
smaller than 1.
Initialization
The neural network requires initialization for each
neuron. It is observed that a good initialization for
multiplier neurons can speed up the convergence
process. However, this initialization does not af-
fect on the final solution by the neural network.
In this research, the initial outputs of continu-
ous neurons are initialized based on medium
start, e.g. the outputs of continuous neurons are
initialized at middle point of independent variable
limits as follows:
V x x
k x k k ,
( )
,max ,min
0
1
2
= +
( )
(12)
where V
k,x
(0)
is the initialization of V
k,x
of continu-
ous neuron k.
The multiplier neurons can be initialized at
zero. However, to speed up the convergence pro-
cess of the neural network, the multiplier neurons
is initiated by the solution of the equation in (6),
Figure 2. Discrete-time implementation of Hopfield Lagrange network
65
in which total inputs of neurons are neglected.
The obtained solution is the initial value for the
multiplier neurons:
V
f V
V
g V
V
i
k x
k x
i k x
k x
,
( ) ,
,
,
,
( ) ( )
0
=
(13)
where V
i,
(0)
is the initial value of V
i,
of multiplier
neuron i.
The initial inputs of neurons are calculated
based on their inversed sigmoid and transfer
functions in (10) and (11), respectively as follows:
U
V x
x V
k x
k x k
k k x
,
( ) ,
( )
,min
,max ,
( )
ln
0
0
0
1
2
=

(14)
U V
i i ,
( )
,
( )

0 0
= (15)
where U
k,x
(0)
and U
i,
(0)
are the initial values of
continuous and multiplier neurons, respectively.
Stopping Criteria
The algorithm of the neural network will be termi-
nated when ever maximum error from the neural
network is lower than a pre-specified threshold
or the maximum allowable number of iterations
is reached.
The maximum error at iteration n from the
neural network is defined as follows
Err g V V V V V
n
i k x
n
k x
n
k x
n
i
n
i
n
max
( )
,
( )
,
( )
,
( )
,
( )
,
(
max ( ) , , =
1

11)
{ }

(16)
where Err
max
(n)
maximum error from the neural
network;
pre-specified threshold;
N
max
maximum allowable number of iterations.
Overall Procedure
The overall procedure of the proposed HLN for
solving optimization problem is described as
follows:
Step 1: Select parameters for the neural
network.
Step 2: Initialize all neurons using (12)
- (15).
Step 3: Choose a threshold and the maxi-
mum number of iterations N
max
. Set n = 1.
Step 4: Calculate dynamics of neurons us-
ing (6) - (7).
Step 5: Update total inputs of neurons us-
ing (8) - (9).
Step 6: Calculate outputs of neurons using
(10) - (11).
Step 7: Calculate maximum error Err
max
(n)

using (16).
Step 8: If n < N
max
and Err
max
(n)
> , n = n
+ 1 and return to Step 4. Otherwise, stop.
The algorithm is also represented in a flow
chart as shown in Figure 3.
Proof of Convergence for HLN
To illustrate how the dynamics of neural network
from (6) and (7) cause the energy function in (5) to
be minimized with respect to continuous neurons
and maximized with respect to multiplier neurons,
the effects on energy function due to the status
changes in the continuous neurons and multiplier
neurons are investigated.
Consider the effect of the status change in the
continuous neurons on the energy function:
dE
dt
E
V
dV
dt
k x
k x
=

,
,
(17)
Substituting V
x,k
in (10) into (17):
66
dE
dt
E
V
dg U
dU
dU
dt
k x
c k x
k x
k x
=

,
,
,
,
( )
(18)
Substituting (6) into (18):
dE
dt
dg U
dU
dU
dt
c k x
i
k x
=

( )
, ,
2
(19)
Since g
c
(U
k,x
) is a monotonically increasing
function as shown in Figure 1, the value of the
derivative dg
c
(U
k,x
)/dU
k,x
is always positive. Con-
sequently, the right hand side of Equation (19) is
always negative. Therefore, the energy function
(5) is always minimized when there is a change
in the status of the continuous neurons.
On the other hand, the effect of a change in
the status of the multiplier neurons associated
with equality constraints on the energy function
is considered as follows:
dE
dt
E
V
dV
dt
i
i
=

,
,
(20)
Substituting V
i,
in (11) into (20):
dE
dt
E
V
dg U
dU
dU
dt
i
m i
i
i
=

,
,
,
,
( )
(21)
Substituting (7) into (21):
dE
dt
dU
dt
i
=

,
2
(22)
Figure 3. Algorithm for HLN
67
It is obvious that the right hand side of equa-
tion (22) is always positive. Therefore, the energy
function always seeks for maximum value when
there is a status change on the multiplier neurons
associated with equality constraints.
HLN for Solving Basic ELD Problem
Problem Formulation
The objective of the basic ELD (BELD) problem
here is to minimize total cost of thermal generating
units of a system over some appropriate period
(one hour typically) while satisfying various
constraints including power balance, generator
power limits, and ramp rate constraints.
Mathematically, the BELD problem is formu-
lated as follows:
Min F a b P c P
i i i i i
i
N
= + +
( )
=
2
1
(23)
subject to
Power balance constraint
P P P
i
i
N
L D
=
=
1
0 (24)
P PB P B P B
L i ij j
j
N
i
N
i i
i
N
= + +
= = =

1 1
0
1
00
(25)
Generator operating limits
P P P
i i i ,min ,max
; i = 1, 2, , N (26)
Ramp rate constraints
P P UR
i i i

0
, if generation increases (27)
P P DR
i i i
0
, if generation decreases
(28)
where a
i
, b
i
, c
i
fuel cost coefficients for unit i;
B
ij
, B
0i
, B
00
transmission loss formula coeffi-
cients;
DR
i
ramp down rate limit of unit i (MW/h);
N total number of online units;
P
D
total load demand of the system (MW);
P
i
output power of unit i (MW);
P
i
0
initial output power of unit i (MW);
P
i,min
, P
i,max
lower and upper generation limits of
unit i (MW);
P
L
total network loss of the system (MW);
UR
i
ramp up rate limit of unit i (MW/h).
HLN Implemented to the BELD Problem
The Lagrangian function L of the problem is
formulated as follows:
L a b P c P P P P
i i i i i
i
N
D L i
i
N
= + +
( )
+ +
= =

2
1 1

(29)
To represent in HLN, N continuous neurons and
one multiplier neurons are required. The energy
function E of the problem is formulated based on
the Lagrangian function as follows:
68
E a bV cV
V P P V g
i i pi i pi
i
N
D L pi
i
N
c
= + +
( )
+ +
+
=
=
2
1
1
1
(( ) V dV
V
i
N
pi
0
1
=

(30)
where V
output of multiplier neuron representing

Lagrangian multiplier ;
V
pi
output of continuous neuron i representing
for output power P
i
.
The dynamics of neurons inputs are derived
as follows:
dU
dt
E
V
b cV V
P
V
U
pi
pi
i i pi
L
pi
pi
=

= ( )

\
)
'
!
2 1
11
11
+
1
1
1
'
!
1
11
+
1
1
1

(31)
dU
dt
E
V
P P V
D L pi
i
N
= +

= +
=
1
(32)
where
= +
=
P
V
B V B
L
pi
ij pj
j
N
i
2
1
0
(33)
U
input of multiplier neuron corresponding to

the output V
.;
U
pi
input of continuous neurons corresponding
to the outputs V
pi
.
The algorithm for updating inputs of neurons
is defined:
U U
E
V
pi
n
pi
n
i
pi
( ) ( )
=

1
(34)
U U
E
V
n n

( ) ( )
= +

1
(35)
The outputs of continuous neurons represent-
ing for output power of units are calculated via
the sigmoid function:
V g U
P P
U
pi c pi
i high i low
pi
= =

\
)
( )
l
l
l
( ) tanh
, ,
2
1 PP
i low ,

(36)
where the new generator limits are redefined as
follows:
P P P UR
i high i i i , ,max
min , = +
{ }
0
(37)
P P P DR
i low i i i , ,min
max , =
{ }
0
(38)
P
i,high
maximal possible power output of unit i;
P
i,low
minimal possible power output of unit i.
The outputs of multiplier neurons are defined
by a transfer function as follows:
V

= = g U U
m
( ) (39)
The outputs of neurons are initialized by:
V
P P
pi
i i ( ) ,max ,min 0
2
=
+
(40)
V
N
b cV
P V
i i pi
L pi i
N
( )
( )
0
0
1
1
2
1
=
+

=
(41)
The maximum error for the network at iteration
n is calculates as follows:
Err P V V
P P V V
n n
pi
n n
D L pi
n
i
N
max
( ) ( ) ( ) ( )
( )
max , ,
max ,
=
{ }
= +
=

1
ppi
n
pi
n n n
V V V
( ) ( ) ( ) ( )
,
1 1

(42)
69
Numerical Results
A test system consists of three online thermal
generating units supplying a load demand of 850
MW with unit data given in Table 1. Ramp rate
constraints are neglected.
The transmission loss coefficients B are given
by:
B
ij
=

0 3 0 0
0 0 9 0
0 0 1 2
10
4
.
.
.

For implementation of HLN to the problem,
the slope of sigmoid function is fixed at = 100
and the updating step sizes for neurons are tuned
for each case. The maximum number of iterations
and the maximum error for the neural network are
set to 2,500 and 10
-4
, respectively. The proposed
HLN method is coded in Matlab and run on a
2.1 GHz PC.
When power loss is neglected, the selected
updating step sizes are
i
= 0.0225 and
= 0.0005,
and the initial output values of the neurons are set
based on (40) and (41) as follows: V
pi
(0)
= [375
250 125]
T
, V
(0)
= 9.0283. The corresponding
inputs for the neurons are calculated based on
the inverse functions of sigmoid function (36)
for continuous neuron and transfer function (39)
for multiplier neurons as follows: U
pi
(0)
= [0 0 0]
T

and U
(0)
= 9.0283. The proposed HLN produces
a total cost of 8,194.05 ($/h) with the obtained
solution as follows:
= V
= 9.1502 ($/MWh)
P
1
= V
p1
= 394.037 (MW)
P
2
= V
p2
= 333.496 (MW)
P
3
= V
p3
= 122.467 (MW)
In this case, the proposed HLN method finds
the optimal solution with 27 iterations in 0.005
seconds. The maximum error of computation Er-
r
max
, energy function of Hopfield neural network E,
energy production cost , and unit power outputs
during the convergence process of HLN are given
in Figures 4, 5, 6, and 7, respectively. In Figure 4,
the maximum error is high at the beginning and
when it is lower than the pre-specified threshold
of 10
-4
, the algorithm will stop due to the stop-
ping criteria satisfied. The energy function of the
HLN as shown in Figure 5 also varies the same
way with maximum error with high value at the
beginning and getting lower during the iterative
process and reaches the minimum point as the
stopping criteria satisfied. In contrast, the energy
production cost in Figure 6 is lower at the begin-
ning this is because the total power generation
from the units is less than the load demand which
causes the power balance constraint unsatisfied.
When the total power generation from the units
increases to satisfy the power demand the energy
production cost also increases. The power outputs
of generating units in Figure 7 oscillate at the
beginning and then they reach the stable state
as iterations increased. As observed, the optimal
solution is obtained when all maximum error,
energy function, and energy production reach the
stable region; that is the difference between two
consecutive iterations is inconsiderable.
Table 1. Data for the three-generating unit system
Unit a
i
($/h) b
i
($/MWh) c
i
($/MW
2
h) P
i,max
(MW) P
i,min
(MW)
1 561 7.92 0.00156 600 150
2 310 7.85 0.00194 400 100
3 78 7.97 0.00482 200 50
70
Figure 5. Energy function of HLN for three-unit system neglecting power loss
Figure 4. Maximum error of HLN for three-unit system neglecting power loss
71
Figure 6. Energy production cost for three-unit system neglecting power loss
Figure 7. Power generation for three units neglecting power loss
72
When power loss is included, the selected
updating step sizes are
i
= 0.02 and
= 0.001,
and the outputs of neurons are initialized at V
pi
(0)

= [375 250 125]
T
, V
(0)
= 9.3312. Their correspond-
ing inputs of neurons determined by inverse
functions of sigmoid and transfer functions are
U
pi
(0)
= [0 0 0]
T
and U
(0)
= 9.3312, respectively.
The total cost produced by the HLN method is
8,344.21 ($/h) and the solution is found as follows:
= V
= 9.5302 ($/MWh)
P
1
= V
p1
= 435.409 (MW)
P
2
= V
p2
= 299.666 (MW)
P
3
= V
p3
= 130.746 (MW)
P
L
= 15.821 (MW)
In this case, the HLN method finds the optimal
solution with the same number of iterations as in
the case neglecting power loss in 0.010 seconds.
In the obtained solutions, all constraints are
met; that is, generator outputs are within there
lower and upper limits and total power generation
from the generators totally meets the load demand
plus power loss (if included) requirement. Obvi-
ously, in the case with power loss neglected, the
total operation cost and energy production cost
are lower than those for the case with power loss.
HLN for Solving Fuel
Constrained ELD Problem
Problem Formulation
Fuel constrained ELD (FELD) problem or fuel
scheduling is an important part of utility for opera-
tion and planning since it is a complex problem
of very large dimensions with a wide range of
time periods and a large set of constraints and
variables. The fuel used by a generating unit may
be obtained from different contracts at different
prices. Fuel contracts are generally under a take-
or-pay agreement including both maximum and
minimum limits on delivery of fuel to generating
units over life of the contract. The fuel storage is
usually within a specified limit to allow for inac-
curate load forecasts and the inability to deliver
on time of suppliers (Asgarpoor, 1994).
Assuming that the entire schedule time hori-
zon is divided into M subintervals each having
a constant load demand and that all generating
units are available and remain on-line for M
subintervals. The objective is to simultaneously
minimize generation cost and emission level of
generating units over the M subintervals such that
the constraints for power balance, fuel delivery
and fuel storage for any given subinterval as
well as maximum-minimum fuel delivery, fuel
storage, and generator operating constraints for
each generating unit are satisfied.
The problem formulation for a system hav-
ing N thermal generating units scheduled in M
subintervals is as follows (Basu, 2002).
Min {F
fc
+ F
em
} (43)
F t a b P c P
fc k fi fi ik fi ik
i
N
k
M
= + +
( )
= =

2
1 1
(44)
F t a b P c P
em k ei ei ik ei ik
i
N
k
M
= + +
( )
= =

2
1 1
(45)
subject to
Power balance constraints
P P P
ik
i
N
Lk Dk
=
=
1
0 ; k = 1,, M (46)
P P B P B P B
Lk ik ij jk
j
N
i
N
i ik
i
N
= + +
= = =

1 1
0
1
00

(47)
Fuel delivery constraint
73
F F
ik
i
N
Dk
=
=
1
0 ; k = 1, , M (48)
Fuel storage constraint
X X F t Q
ik ik ik k ik
= +
1
; i = 1, , N (49)
Q d e P f P
ik i i ik i ik
= + +
2
; k = 1, , M (50)
P P P
i ik i ,min ,max
; i = 1, , N
;
k = 1, , M
(51)
Fuel delivery limits
F F F
i ik i ,min ,min
; i = 1, , N; k = 1, , M
(52)
Fuel storage limits
X X X
i ik i ,min ,max
; i = 1, , N; k = 1, , M
(53)
The fuel storage at subinterval k in (49) can be
rewritten in terms of initial fuel storage as follows:
X X F t Q
ik i il l il
l
k
= + ( )
=
0
1
(54)
where a
ei
, b
ei
, c
ei
emission coefficients for thermal
unit i;
a
fi
, b
fi
, c
fi
fuel cost coefficients for thermal unit
i;
B
ij
, B
0i
, B
00
transmission loss formula coeffi-
cients;
d
i
, e
i
, f
i
fuel consumption coefficients for ther-
mal unit s;
F
Dk
fuel demand of the units during subinterval
k, in tons;
F
em
emission function of generating units;
F
fc
fuel cost function of generating units;
F
ik
fuel delivery for thermal unit i during subin-
terval k, in tons;
F
i,min
, F
i,max
lower and upper fuel delivery limits
for thermal unit i, in tons;
M number of subintervals of scheduled period;
N total number of thermal units;
P
Dk
load demand of the system during subinter-
val k, in MW;
P
Lk
transmission loss of the system during sub-
interval k, in MW;
P
ik
output power of thermal unit i during subin-
terval k, in MW;
P
i,min
, P
i,max
thermal unit i, in MW;
Q
ik
fuel consumption function of thermal unit i
in subinterval k, in tons/h;
t
k
duration of subinterval k, in hours;
X
ik
fuel storage for unit i during subinterval k,
in tons;
X
i,min
, X
i,max
lower and upper fuel storage limits
for thermal unit i, in tons.
HLN Implemented to the FELD Problem
The Lagrange function L of the problem is for-
mulated as follows:
74
L t a b P c P a b P c P
k fi fi ik fi ik ei ei ik ei ik
i
N
k
=
( )

( )
l
l
l
= =
2 2
1 11
1 1 1
M
k Lk Dk ik
i
N
k
M
k ik
i
N
Dk
P P P F F
\
)
= = =

\
)
( )
\
)
=
= =
k
M
ik ik i il l il
l
k
i
X X F t Q
1
0
1 1
NN
k
M

=1

(55)
To represent in HLN, 3NM continuous
neurons and (N+2)M multiplier neurons are
required. The energy function E of the problem
is formulated based on the Lagrangian function
in terms of neurons as follows.
E t a b V c V a b V c V
k fi fi pik fi pik ei ei pik ei pik
i
= ( ) ( )
l
l
l
=
2 2
1
NN
k
M
k Lk Dk pik
i
N
k
M
k fik
i
N
V P P V V V

=
= = =

\
)

1
1 1 1

\
)
( )
=
=
F
V V X V t Q
Dk
k
M
ik xik i fil l il
l
k
1
0
1
\
)

= =

i
N
k
M
c
V
c
V
c
g V dV g V dV g V
pik fik
1 1
1
0
1
0
1
( ) ( ) ( ))dV
V
i
N
k
M xik
0
1 1

\
)
= =

(56)
where V
pik
output of continuous neuron represent-
ing for output power P
ik
;
V
fik
output of continuous neuron representing
for fuel delivery F
ik
;
V
xik
output of continuous neuron representing
for fuel storage X
ik
;
V
k
output of multiplier neuron associated with
power balance constraint;
V
k
fuel delivery constraint;
V
ik
fuel storage constraint.
The dynamics of HLN for updating neuron
inputs are defined as follows:
dU
dt
E
V
t b c V t b c V
V
P
pik
pik
k fi fi pik k ei ei pik
k
L
=

( )

( )

2 2
kk
pik
ik k
ik
pik
pik
V
V t
dQ
dV
U
\
)

'
!
1
1
1
1
1
+
1
1
1
1
1
'
1

!!
1
1
1
1
1
+
1
1
1
1
1

(57)
dU
dt
E
V
V V U
fik
fik
k ik fik
=

= +
{ }

(58)
dU
dt
E
V
V U
xik
xik
ik xik
=

= +
{ }
(59)
dU
dt
E
V
P P V
k
k
Dk Lk pik
i
N
= +

= +
=
1
(60)
dU
dt
E
V
V F
k
k
fik
i
N
Dk
= +

=
=
1
(61)
dU
dt
E
V
V X V t Q
ik
ik
xik i fil l il
l
k
= +

= +
( )
=
0
1

(62)
where
= +
=
P
V
B V B
Lk
pik
ij pjk
j
N
i
2
1
0
(63)
= +
Q
V
e fV
ik
pik
i i pik
2 (64)
U
pik
, U
fik
, U
xik
inputs of continuous neurons
corresponding to the outputs V
pik
, V
fik
and V
xik
,
respectively;
U
k
, U
k
, U
ik
inputs of multiplier neurons
corresponding to the outputs V
k
, V
k
and V
ik
,
respectively.
75
at iteration n is as follows:
U U
E
V
pik
n
pik
n
p
pik
( ) ( )
=

1
(65)
U U
E
V
fik
n
fik
n
f
fik
( ) ( )
=

1
(66)
U U
E
V
xik
n
xik
n
x
xik
( ) ( )
=

1
(67)
U U
E
V
k
n
k
n
k

( ) ( )
= +

1
(68)
U U
E
V
k
n
k
n
k

( ) ( )
= +

1
(69)
U U
E
V
ik
n
ik
n
ik

( ) ( )
= +

1
(70)
where
p
,
f
,
x
continuous neuron updating step
sizes;
multiplier neuron updating step sizes.

The outputs of neurons representing for output
power, fuel delivery and fuel storage of units are
determined by:
V g U
P P
U
pik c pik
i i
pik
= =

\
)
( )
l
l
( ) tanh
,max ,min
2
1
ll
P
i,min

(71)
V g U F F
U
fik c fik i i
fik
= = ( )
+ ( )
( )
tanh
,max ,min
1
2
++ F
i,min

(72)
V g U F F
U
fik c fik i i
fik
= = ( )
+ ( )
( )
tanh
,max ,min
1
2
++ F
i,min

(73)
The outputs of multiplier neurons are deter-
mined by:
V
k m k k
g U U = = ( ) (74)
V
k m k k
g U U = = ( ) (75)
V
ik m ik ik
g U U = = ( ) (76)
The maximum error for the neural network at
iteration n is determined as follows:
Err P F X V V V
k ik ik pik fik xik max
max , , , , , = { }

(77)
where
P P P V
k Dk Lk pik
i
N
= +
=
1
(78)
F V F
ik fik
i
N
Dk
=
=
1
(79)
X V X V t Q
ik xik i fil l il
l
k
= +
( )
=
0
1
(80)
V V V
pik pik
n
pik
n
=
( ) ( ) 1
(81)
V V V
fik fik
n
fik
n
=
( ) ( ) 1
(82)
V V V
xik xik
n
xik
n
=
( ) ( ) 1
(83)
76
Numerical Results
The test system includes five thermal generat-
ing units remaining online for a period of three
weeks. The system data and demand are given in
Tables 2 and 3, respectively. The system power
loss is neglected.
The parameters of HLN for the problem are
selected after tuning as follows: = 100,
p
=
3.5x10
-5
,
f
=
x
=
= 1.25,
= 2x10
-6
and
=
3.5x10
-7
. The maximum number of iterations and
the maximum error for the neural network are set
to 2500 and 10
-4
, respectively.
In this problem, the optimal solution can be
affected by the initial conditions of fuel storage.
For considering the effects of the initial storage
conditions to the optimal solution, the problem
is considered for 3 cases with 3 different initial
storages as follows.
Case 1: The initial storage for 5 units is X
i0
=
[2000 5000 5000 8000 8000]
T
tons. This is also
called full case since the fuel stored at the begin-
ning of the schedule time is enough for generating
units to operate in certain duration before fuel
delivered. In this case, the optimal solution for the
problem is rather easily found since the problem
constraints are not so restricted. The total cost and
emission obtained by the HLN method for the
problem are $1,044,728.25 and 534,488.74 kg,
respectively. The solution for this case is given
in Table 4. As shown in the table,
k
represents
for energy production cost,
k
represents for fuel
Table 2. Data for five-unit system
Unit 1 2 3 4 5
a
fi
($/h) 25 60 100 120 40
b
fi
($/MWh) 2.0 1.8 2.1 2.2 1.8
c
fi
($/MW
2
h) 0.008 0.003 0.001 0.004 0.002
a
ei
(kg/h) 80 50 70 45 30
b
fi
(kg/MWh) -0.805 -0.555 -0.955 -0.600 -0.555
c
ei
(kg/MW
2
h) 0.018 0.015 0.012 0.008 0.012
d
i
(ton/h) 0.83612 2.00669 3.34448 4.01338 1.33779
e
i
(ton/MWh) 0.066890 0.060200 0.070230 0.073580 0.060200
f
i
(ton/MW
2
h) 0.00026756 0.00010033 0.00004013 0.00013378 0.00005017
P
i,max
(MW) 75 125 175 250 300
P
i,min
(MW) 20 20 30 40 50
F
i,max
(tons) 1000 1000 2000 3000 3000
F
i,min
(tons) 0 0 0 0 0
X
i,max
(tons) 10000 10000 20000 30000 30000
X
i,min
(tons) 0 0 0 0 0
Table 3. Demand of five-unit system
Subinterval 1 2 3
Duration t
k
(h) 168 168 168
Load demand P
Dk
(MW) 700 800 650
Fuel demand F
Dk
(tons) 7000 7000 7000
77
delivery cost, and
ik
represents for fuel storage
cost for each unit for each sub-interval. The
computational time for this case is 0.13 seconds.
i0
=
[2000; 5000; 5000; 500; 8000]
T
tons. This is also
called shortage case since the initial fuel storage
for unit 4 can only guaranty it to operate in a very
short term. The solution for this case is more dif-
ficult to be found since the constraint condition
is more restricted than the full case. In this case,
the total cost and emission obtained by the pro-
posed method are $1,044,701.09 and 534,515.92
kg, respectively. The solution for this case is
given in Table 5. Due to the initial fuel shortage,
the fuel delivered to unit 4 in this case is more
than that in Case 1 while fuel delivered to other
units is less. For this case, the computational time
is 0.17 seconds.
i0
=
[2000; 2500; 2500; 8000; 500]
T
tons. This case
is also a shortage case with the initial fuel short-
age for unit 5 and lower initial fuel storage than
that in Case 1 for units 2 and 3. This case is more
restricted than Case 2 since the initial storage is
lower. Therefore, the optimal solution is also more
difficult to be found. For this case, the proposed
method gives a total cost of $ 1,044,738.45 and
an emission of 534,478.54 kg with a computa-
tional time of 0.27 seconds. The obtained solution
for this case is given in Table 6. In this case, the
fuel delivered to unit 5 is also increased while
fuel delivered to the others is reduced.
The results from the three cases show that the
final solutions are little affected by the initial
conditions of fuel storage. The proposed HLN
could find the corresponding optimal solution for
each case. In all cases, the obtained total costs
and emissions are not much difference from each
other.
In this multi-objective optimization problem,
the single objective optimization is also considered
as an option for decision maker. In this case, the
single objective problems are also considered for
purity of fuel cost dispatch and purity of emission
dispatch with the initial storage from Case 1. In
these single objective cases, the optimal solution
is easier to be found than the cases with multi-
objectives since there is no conflict between the
objectives. The parameters of HLN for these cases
are chosen as follows:
p
= 10
-5
,
f
=
x
=
= 0.1,
and
= 10
-7
. The remaining parameters are
Table 4. Solution for Case 1
Sub
interval
k
($/MWh)
k
($/ton-h) Unit 1 2 3 4 5
1 942.6 0.0012 P
ik
(MW) 75.00 21.26 175.00 167.07 161.67
F
ik
(10
3
tons) 0.7816 0.5425 1.5275 2.1183 2.0300
X
ik
(10
3
tons) 1.7950 3.9770 3.8996 7.3750 8.1688
ik
($/ton-h) 0.0076 0.0021 0.0071 0.0056 0.0049
2 1,148.1 0.0023 P
ik
(MW) 75.00 125.00 175.00 218.03 206.97
F
ik
(10
3
tons) 0.7697 0.5971 1.5804 2.1278 1.9250
X
ik
(10
3
tons) 1.5781 2.9707 2.8520 6.1286 7.7741
ik
($/ton-h) 0.0084 0.0043 0.0090 0.0068 0.0053
3 884.1 0.0031 P
ik
(MW) 75.00 111.58 162.07 152.57 148.78
F
ik
(10
3
tons) 0.7739 0.6619 1.6544 2.0978 1.8120
X
ik
(10
3
tons) 1.3653 2.1652 2.0312 5.6629 7.8554
ik
($/ton-h) 0.0092 0.0064 0.0109 0.0073 0.0052
78
the same as the ones selected for the three cases
above.
Fuel Cost Objective Only: In this case, there
only fuel cost objective is minimized while the
emission objective is neglected. The obtained
total cost by HLN is $1,002,468.61 and the emis-
sion 744,675.06 kg. It is obvious that when the
fuel cost objective is priority to be considered,
the total cost is much lower than that in the case
of bi-objective case meanwhile the emission in
this case is much higher and all constraints are
satisfied. The obtained solution for the economic
dispatch is given in Table 7.
Emission objective only: This case is opposite
to the case considering only fuel cost objective.
The obtained emission level in this case is
Sub
interval
k
($/MWh)
k
($/ton-h)
Unit 1 2 3 4 5
1 942.6 0.0039 P
ik
(MW) 75.00 121.25 175.00 167.09 161.66
F
ik
(10
3
tons) 0.6141 0.6259 1.5930 1.5408 2.6262
X
ik
(10
3
tons) 1.6275 1.5604 1.4650 6.7974 1.2652
ik
($/ton-h) 0.0082 0.0084 0.0127 0.0061 0.0156
2 1148.1 0.0057 P
ik
(MW) 75.00 125.00 175.00 218.04 206.96
F
ik
(10
3
tons) 0.5890 0.7311 1.7417 1.5148 2.4234
X
ik
(10
3
tons) 1.2298 0.6881 0.5788 4.9378 1.3690
ik
($/ton-h) 0.0098 0.0130 0.0176 0.0081 0.0152
3 884.2 0.0067 P
ik
(MW) 75.00 111.57 162.04 152.60 148.79
F
ik
(10
3
tons) 0.5990 0.9113 1.9567 1.4531 2.0800
X
ik
(10
3
tons) 0.8421 0.1320 0.0606 3.8269 1.7181
ik
($/ton-h) 0.0119 0.0216 0.0290 0.0096 0.0140
Sub
interval
k
($/MWh)
k
($/ton-h)
Unit 1 2 3 4 5
1 942.7 0.0039 P
ik
(MW) 75.00 21.27 175.00 167.34 161.69
F
ik
(10
3
tons) 0.6894 0.4206 1.3338 2.8664 1.6898
X
ik
(10
3
tons) 1.7027 3.8549 3.7059 0.6236 7.8284
ik
($/ton-h) 0.0079 0.0023 0.0074 0.0193 0.0052
2 1148.3 0.0057 P
ik
(MW) 75.00 125.00 175.00 217.98 207.02
F
ik
(10
3
tons) 0.6673 0.4636 1.3914 2.9414 1.5363
X
ik
(10
3
tons) 1.3834 2.7151 2.4694 0.1914 7.0445
ik
($/ton-h) 0.0091 0.0049 0.0098 0.0252 0.0059
3 884.2 0.0067 P
ik
(MW) 75.00 111.59 162.08 152.53 148.80
F
ik
(10
3
tons) 0.6828 0.5441 1.5212 2.8310 1.4209
X
ik
(10
3
tons) 1.0795 1.7916 1.5152 0.4593 6.7345
ik
($/ton-h) 0.0106 0.0076 0.0125 0.0208 0.0062
79
521,793.49 kg and the total cost $1,072,673.58.
The obtained total cost in this case is much
higher than the case with only fuel cost objective
while the emission level is opposite to the previ-
ous case. In fact, the performance of emission
dispatch is the same manner with the economic
dispatch except the different objective coefficients.
The solution for this emission dispatch is given
in Table 8. It can be observed from the obtained
result that the power outputs of generating units
in this case are different to those from the eco-
nomic dispatch since they are decided by the
different objective coefficients.
The obtained test results from the proposed
HLN above are better than those from the con-
ventional Hopfield network in (Basu, 2002) for
Table 7. Solution for fuel cost objective only
Sub
interval
k
($/MWh)
k
($/ton-h)
Unit 1 2 3 4 5
1 451.7 0.0011 P
ik
(MW) 43.00 125.00 175.00 61.03 295.97
F
ik
(10
3
tons) 0.7470 0.5495 1.5324 1.9964 2.1747
X
ik
(10
3
tons) 2.1214 3.9461 3.9045 8.5664 6.9541
ik
($/ton-h) 0.0066 0.0021 0.0071 0.0046 0.0060
2 537.7 0.0022 P
ik
(MW) 74.98 125.00 175.00 125.02 300.00
F
ik
(10
3
tons) 0.7368 0.6065 1.5881 1.8905 2.1780
X
ik
(10
3
tons) 1.8718 2.9491 2.8647 8.2345 5.8707
ik
($/ton-h) 0.0073 0.0044 0.0089 0.0049 0.0071
3 435.6 0.0029 P
ik
(MW) 37.00 125.00 175.00 49.04 263.96
F
ik
(10
3
tons) 0.6925 0.6888 1.6836 1.7422 2.1929
X
ik
(10
3
tons) 2.0064 2.0345 1.9204 8.6951 5.1670
ik
($/ton-h) 0.0069 0.0068 0.0112 0.0045 0.0078
Table 8. Solution for emission objective only
Sub
interval
k
($/MWh)
k
($/ton-h)
Unit 1 2 3 4 5
1 468.3 0.0012 P
ik
(MW) 75.00 111.41 162.68 211.66 139.25
F
ik
(10
3
tons) 0.7809 0.5321 1.5110 2.1715 2.0046
X
ik
(10
3
tons) 0.7679 0.5853 1.5614 2.2237 1.8617
ik
($/ton-h) 0.0076 0.0019 0.0069 0.0061 0.0047
2 612.4 0.0024 P
ik
(MW) 75.00 125.00 175.00 250.00 175.00
F
ik
(10
3
tons) 0.7710 0.6377 1.6159 2.2506 1.7249
X
ik
(10
3
tons) 1.7943 4.0664 4.0284 6.8761 8.3703
ik
($/ton-h) 0.0084 0.0041 0.0087 0.0077 0.0049
3 421.9 0.0032 P
ik
(MW) 75.00 102.19 150.67 194.39 127.75
F
ik
(10
3
tons) 1.5755 3.0482 2.9619 5.3297 8.2359
X
ik
(10
3
tons) 1.3599 2.3136 2.2371 4.4988 8.4429
ik
($/ton-h) 0.0092 0.0060 0.0104 0.0087 0.0047
80
all cases considered. This shows that the proposed
method is better than the conventional Hopfield
neural network in dealing with complicated prob-
lems.
HLN for Solving Hydrothermal
ELD Problem
Problem Formulation
The objective of hydrothermal ELD (HELD)
problem is to minimize the total fuel cost of
thermal generators while satisfying hydraulic,
power balance, and generator operating limits con-
straints. Mathematically, the HELD problem for
a hydrothermal system with N
1
thermal units and
N
2
hydro units scheduled in M time sub-intervals
with t
k
hours for each is formulated as follows:
Min F t a b P c P
k i i ik i ik
i
N
k
M
= + +
( )
= =

2
1 1
1
(84)
subject to:
Power balance constraints
P P P P
ik
i
N
hk
h
N
Lk Dk
= =

+ =
1 1
1 2
0 ; k = 1,, M
(85)
P P B P B P B
Lk pk pq qk
q
N N
p
N N
p pk
p
N N
= + +
=
+
=
+
=
+

1 1
0
1
00
1 2 1 2 1 2

(86)
Continuity of reservoir head constraints
d d
t
f
r q
hk hk
k
h
hk hk
= + ( )
1
; h = 1, , N
2

(87)
q P d
hk hk hk
= ( ) ( ) (88)
( ) P a b P c P
hk ph ph hk ph hk
= + +
2
(89)
( ) d a b d c d
hk dh dh hk dh hk
= + +
2
(90)
P P P
i ik i ,min ,max
; i = 1, , N
1
; k = 1, , M
(91)
P P P
h hk h ,min ,max
; h = 1, , N
2
; k = 1, ,
M (92)
In terms of water availability, the Equation
(87) can be rewritten as follows:
t q r d d f W
k hk hk
k
M
h hM h h
( ) = ( ) =
=
1
0
(93)
where a
i
, b
i
, c
i
cost coefficients for thermal unit i;
a
ph
, b
ph
, c
ph
water discharge coefficients for
hydro unit h;
a
dh
, b
dh
, c
dh
reservoir head variation coefficients
for hydro unit h;
B
pq
, B
0p
, B
00
coefficients for the system;
d
h0
, d
hM
the initial and final height of the reser-
voir head of hydro unit h, in ft;
d
hk
the height of the reservoir head of hydro unit
h in interval k, in ft;
f
h
the surface of the vertical sided tank of hydro
unit h;
P
ik
generation output of thermal unit s during
sub-interval k, in MW;
P
hk
generation output of hydro unit h during
subinterval k, in MW;
81
P
h
min
, P
h
max
hydro unit h, in MW;
P
i
min
, P
i
max
thermal unit s, in MW;
P
Dk
load demand of the system during subinter-
val k, in MW;
P
Lk
transmission loss of the system during sub-
interval k, in MW;
q
hk
rate of water flow from hydro unit h in inter-
val k, in acre-ft per hour or MCF per hour;
r
hk
reservoir inflow for hydro unit h in interval
k, in acre-ft per hour or MCF per hour;
W
h
volume of water available for generation by
hydro unit h during the scheduling period;
(P
hk
) water discharge function for hydro unit
h at subinterval k;
(P
hk
) reservoir head variation function for
hydro unit h at subinterval k.
When effect of the height of the reservoir head
is neglected, the problem is called fixed-head
HELD with the constraints (87) and (89) neglected.
In contrast, the problem is called variable-head
HELD when the variation of the reservoir head
is included.
HLN Implemented to the
HELD Problem
The Lagrange function L is formulated as follows:
L
t a b P c P
P P P P
k i i ik i ik
i
N
k
M
k Lk Dk ik
i
N
hk
h
=

( )

= =
= =

2
1 1
1
1
1
11 1
1 1
2
2
N
k
M
h k hk hk
k
M
h
h
N
t q r W

\
)
( )
l
l
l
l
=
= =
''
!
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1
'
!
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1

(94)
To implementation in Hopfield Lagrange
model, (N
1
+N
2
)M continuous neurons and N
2
+M
multiplier neurons are required. The energy func-
tion E of the problem is described in terms of
neurons as follows:
E
t a bV cV
V P P V V
k i i pik i pik
i
N
k
M
k Lk Dk pik
s
N
=

( )

= =
=

2
1 1
1
1
1
pphk
h
N
k
M
h k hk hk
k
M
h
V t q r W
= =
=

\
)
( )
l
l
l
l
1 1
1
2
hh
N
c
V
i
N
c
V
h
N
g V dV g V dV
pik phk
=
1
1
0
1
1
0
1
2
1 2
( ) ( )
\\
)
'
!
1
1
1
1
1
1
1
1
1
1
1
1
1
1
+
1
1
1
1
1
1
1
1
1
1
1
1
1
1
'
!
1
1
1
1
1
1
1
1
1
1
1
=
k
M
1
11
1
1
+
1
1
1
1
1
1
1
1
1
1
1
1
1
1

(95)
where V
pik
output of continuous neuron ik repre-
senting P
ik
;
V
phk
output of continuous neuron hk represent-
ing P
hk
;
V
k
, V
h
outputs of the multiplier neurons associ-
ated with power balance and water constraint,
respectively.
The dynamics of Hopfield Lagrange model
for updating neuron inputs based on the previous
information are follows:
82
dU
dt
E
V
t b cV V
P
V
pik
pik
k i i pik k
Lk
pik
=

= ( )

\
)
2 1
'
!
1
11
+
1
1
1
'
!
1
11
+
1
1
1
U
pik

(96)
dU
dt
E
V
V
P
V
V t
q
V
phk
phk
k
Lk
phk
h k
hk
p
=

\
)

1
hhk
phk
U
'
!
1
11
+
1
1
1
'
!
1
11
+
1
1
1

(97)
dU
dt
E
V
P P V V
k
k
Dk Lk pik
i
N
phk
h
N
= +

= +
= =

1 1
1 2

(98)
dU
dt
E
V
t q r W
h
h
k hk hk
k
M
h
= +

= ( )
=
1
(99)
where
= + +
= =

P
V
B V B V B
Lk
pik
ip ppk
p
N
ih phk
h
N
i
2 2
1 1
0
1 2

(100)
= + +
= =

P
V
B V B V B
Lk
phk
hi pik
i
N
hq pqk
q
N
h
2 2
1 1
0
1 2

(101)
= = +
( )
q
V
d
d V
dV
d b c V
hk
phk
hk
phk
phk
hk ph ph phk

( )
( )
( ) 2

(102)
B
ip
loss coefficients related to thermal plants;
B
hq
loss coefficients related to hydro plants;
B
ih
, B
hi
loss coefficients between thermal and
hydro plants, B
ih
= B
hi
T
;
U
pik
, U
phk
inputs of the neurons ik and hk, re-
spectively;
U
k
, U
h
inputs of the multiplier neurons.
at step n is as follows:
U U
E
V
pik
n
pik
n
i
pik
( ) ( )
=

1
(103)
U U
E
V
phk
n
phk
n
h
phk
( ) ( )
=

1
(104)
U U
E
V
k
n
k
n
k

( ) ( )
=

1
(105)
U U
E
V
h
n
h
n
h

( ) ( )
=

1
(106)
where
i
,
h
updating step sizes for continuous
neurons;
updating step sizes for multiplier neurons.

The outputs of continuous neurons are calcu-
lated by a sigmoid function:
V g U P P
U
pik c pik i i
pik
= = ( )
+ ( )
( )
tanh
,max ,min
1
2
++ P
i,min

(107)
V g U P P
U
phk c phk h h
phk
= = ( )
+ ( )
( )
tanh
,max ,min
1
2
++P
h,min

(108)
Since multiplier neurons are unconstrained
outputs, the outputs are defined as below:
V
k
= g
m
(U
k
) = U
k
(109)
V
h
= g
m
(U
h
) = U
h
(110)
The maximum error for the neural network is:
Err P W V V
k h pik phk max
max , , , =
{ }

(111)
where
83
P P P V V
k Dk Lk pik
i
N
phk
h
N
= +
= =

1 1
1 2
(112)
W t q r W
h k hk hk
k
M
h
= ( )
=
1
(113)
V V V
pik pik
n
pik
n
=
( ) ( ) 1
(114)
V V V
phk phk
n
phk
n
=
( ) ( ) 1
(115)
Numerical Results
Fixed-Head HELD Problem
In this case, the reservoir head variation of hydro
units is neglected; therefore water discharge of
each hydro unit in (88) is a function of only its
generation output.
The test system has two thermal and two hydro
plants with their characteristics given as follows:
F
1
(P
1
) = 380 + 6.75P
1
+ 0.00225P
1
2
$/h
47.5 MW P
1
450 MW
F
2
(P
2
) = 600 + 5.28P
2
+ 0.0055P
2
2
$/h
100 MW P
2
1000 MW
q
3
(P
3
) = 260 + 8.5P
3
+ 0.00986P
3
2
acre-ft/h
0 MW P
3
250 MW
q
4
(P
4
) = 250 + 9.8P
4
+ 0.0114P
4
2
acre-ft/h
0 MW P
4
500 MW
where F
1
and F
2
are fuel cost functions of thermal
power plants 1 and 2, respectively and q
3
and q
4

are water discharge functions for hydro power
plants 3 and 4, respectively.
Water inflow of reservoirs during schedule
period is supposed to be zeros. Transmission loss
coefficient matrix is given below, per MW:
B =
4 0 . 1.0 1.5 1.5
1.0 3.5 1.0 1.2
1.5 1.0 3.99 2.0
1.5 1.2 2.0 4.9

10
5

The schedule time horizon is 48h which is
divided in four sub-periods with 12 hours for
each supplying to the load demand of [1200 1500
1400 1700] MW. The allowable volumes of water
for hydro plants 3 and 4 for the whole period are
given by:
W
3
= 125,000 acre-ft
W
4
= 286,000 acre-ft
The parameters of the HLN for the problem
are selected after tuning as follows: = 100,
i

=
h
= 310
-4
,
= 10
-2
and
= 7.510
-7
. The
maximum number of iterations and the maximum
error for the neural network are set to 2,500 and
10
-4
, respectively.
When the system power loss is neglected,
the total power generation from the thermal and
hydro units is balanced to only load demand.
The proposed method provides a total cost of
$ 353,444.60 with a computational time of 0.8
seconds. The water discharge cost for each hydro
plant for the whole schedule time is [0.73 0.51]
T
$/acre-ft. Obviously, this water discharge cost
reflects to the practice that the marginal cost for
hydro is small and negligible in the operating
cost calculation. Therefore, the energy produc-
tion cost for the system is mainly based on the
energy production cost from thermal units. The
solution obtained by the HLN method for this case
is given Table 9. As observed from this result, all
constraints including power balance and water
discharge are satisfied.
84
The obtained result by HLN method for this
case is better than that from the conventional
Hopfield neural network in (Basu, 2003). This
confirms that the proposed HLN is more efficient
than the conventional Hopfield neural network in
approach to complicated problems.
When the system power loss is included, the
total power generation from both thermal and
hydro units balances to load demand plus power
loss in the system. The HLN method obtains a
total cost of $ 375,933.65 with the water discharge
cost for each hydro plant for the whole period is
[0.76 0.52]
T
$/acre-ft for a computational time of
0.11 seconds. The total cost and energy production
cost in this case are obviously higher than those
in the case without power loss. This is because
the allowable volumes of water for hydro units
are fixed, thus the power outputs from thermal
units are increased to compensate to power loss
in the system leading to more fuel consumption.
The final solution for this case is given in Table
10. This result is also satisfies all constraints of
power balance and water discharge.
Variable-Head HELD Problem
This case considers all constraints as in the
problem formulation. The test system consists
of two thermal and two hydro plants. Their data
is given below:
Table 10. Solution for fixed-head HELD problem with power loss
Subinterval 1 2 3 4
Duration t
k
(h) 12 12 12 12
Load demand P
Dk
(MW) 1200 1500 1400 1700
Production cost
k
($/MWh) 111.09 129.20 122.93 147.42
Power loss P
Lk
(MW) 31.42 48.92 42.57 63.69
Thermal unit 1 P
1k
(MW) 442.16 450.00 450.00 450.00
Thermal unit 2 P
2k
(MW) 326.19 445.10 404.21 563.69
Hydro unit 3 P
3k
(MW) 160.56 247.12 217.39 250.00
Hydro unit 4 P
4k
(MW) 302.52 406.69 370.96 500.00
Water discharge q
3k
(acre-ft) 22547.18 35552.36 30885.46 36015.00
Water discharge q
4k
(acre-ft) 1095.33 73453.53 65451.15 96000.00
Table 9. Solution for fixed-head HELD problem neglecting power loss
Subinterval 1 2 3 4
Duration t
k
(h) 12 12 12 12
Load demand P
Dk
(MW) 1200 1500 1400 1700
Production cost
k
($/MWh) 103.53 116.55 111.66 129.67
Thermal unit 1 P
1k
(MW) 417.01 450.00 450.00 450.00
Thermal unit 2 P
2k
(MW) 304.37 403.01 366.06 502.38
Hydro unit 3 P
3k
(MW) 167.84 243.09 214.92 250.00
Hydro unit 4 P
4k
(MW) 310.78 403.91 369.02 497.62
Water discharge q
3k
(acre-ft) 23572.79 34907.03 30507.11 36015.00
Water discharge q
4k
(acre-ft) 52760.45 72817.82 65025.60 95395.30
85
F
1
(P
1
) = 25 + 3.2P
1
+ 0.0025P
1
2
$/h
F
2
(P
2
) = 30 + 3.4P
2
+ 0.0008P
2
2
$/h
3
(P
3
) = 0.1980 + 0.306P
3
+ 0.000216P
3
2

MCF/h
3
(d
3
) = 0.90 - 0.0030d
3
+ 0.00001d
3
2
ft
4
(P
4
) = 0.9360 + 0.612P
4
+ 0.000360P
4
2

MCF/h
4
(d
4
) = 0.95 - 0.0025d
4
+ 0.00002d
4
2
ft
d
30
= 300 ft; d
40
= 250 ft
f
3
= 1000 M square ft; f
4
= 400 M square ft
W
3
= 2850 MCF; W
4
= 2450 MCF
The transmission loss coefficient matrix is:
B =
1 40 . 0.10 0.15 0.15
0.15 0.60 0.10 0.13
0.15 0.10 0.68 0.65
0.15 0.13 0.65 0.70

10
4

The schedule time for this problem is 24 hours
with zero water inflow during the scheduled
period. The load demand for the whole schedule
time is given in Figure 8.
The parameters of the HLN are selected for
the problem after tuning as follows: = 100,
i
=
h
= 210
-2
,
= 2.510
-4
and
= 1.2510
-4
. The
maximum number of iterations and maximum
error for the neural network are set to 2,500 and
10
-4
, respectively.
Figure 8. Load demand for variable-head HELD problem
86
When power loss is neglected, the total power
generation from the thermal and hydro plants is
balanced to only load demand at each interval for
the whole schedule time horizon. The obtained
total cost by HLN is $ 62,839.58 and the water
discharge cost of hydro plants [10.23 3.93]
T
$/
MCF with a computational time of 0.13 seconds.
The solution by the HLN for this case is given
in Table 11. As observed from the table, based
on the obtained schedule, hydro plant 4 is off at
hours 2-4 for low load demand and generates high
power at the high load demand hours 9-20 so as
its total power generation is fitted to the pre-fixed
allowable discharge water volume. The power
generation from hydro plant 3 during the schedule
time is similar to hydro plant 4. However, hydro
plant 3 is not off at the hours of low load demand
since the allowable water discharge volume for
this plant is higher than that for plant 4 which can
guarantee it to operate for the whole scheduled
time. The water head of the hydro plants changes
according to the water discharge for each time-
interval of one hour. With the contribution of the
hydro power generation in the system, the energy
production cost is lower than the case with all
thermal unit system. The power generation outputs
of plants for this case for the whole schedule time
are given in Figure 9.
Table 11. Solution for variable-head HELD problem neglecting power loss
Hr. P
D
(MW) ($/MWh) P
1
(MW) P
2
(MW) P
3
(MW) P
4
(MW) d
3
(ft) d
4
(ft)
1 800 3.9349 147.44 337.17 281.47 33.92 299.91 249.91
2 700 3.8566 132.03 289.42 262.04 16.50 299.82 249.87
3 600 3.7770 116.42 241.31 242.27 0 299.74 249.87
4 600 3.7768 116.38 241.16 242.46 0 299.66 249.86
5 600 3.7766 116.33 241.02 242.65 0 299.59 249.86
6 650 3.8166 124.18 265.17 252.99 7.65 299.50 249.84
7 800 3.9331 147.08 336.06 282.65 34.21 299.41 249.75
8 1000 4.0878 177.52 430.73 321.99 69.76 299.30 249.57
9 1330 4.3427 227.64 586.94 386.73 128.68 299.17 249.23
10 1350 4.3566 230.37 595.45 390.74 133.44 299.03 248.89
11 1450 4.4326 245.24 641.81 410.44 152.50 298.88 248.48
12 1500 4.4695 252.45 664.28 420.33 162.93 298.73 248.05
13 1300 4.3123 221.68 568.35 381.11 128.86 298.60 247.72
14 1350 4.3494 228.96 591.06 390.99 138.98 298.47 247.36
15 1350 4.3477 228.63 590.01 391.06 140.30 298.33 246.99
16 1370 4.3615 231.32 598.42 395.04 145.22 298.19 246.62
17 1450 4.4216 243.10 635.14 410.79 160.96 298.04 246.20
18 1570 4.5128 260.87 690.49 434.45 184.19 297.89 245.71
19 1430 4.4019 239.24 623.09 406.90 160.78 297.75 245.29
20 1350 4.3381 226.73 584.11 391.23 147.93 297.61 244.91
21 1270 4.2745 214.26 545.22 375.59 134.93 297.48 244.56
22 1150 4.1804 195.75 487.53 352.15 114.58 297.36 244.27
23 1000 4.0635 172.73 415.83 322.85 88.58 297.25 244.05
24 900 3.9853 157.35 367.98 303.36 71.31 297.15 243.88
87
When power loss included in the problem, the
total power generation from the thermal and hy-
dro plants has to satisfy load demand plus power
loss in the system. The total cost is $ 67,952.42
with the water discharge cost of [10.40 3.99]
T
$/
MCF for each hydro plant. The total computa-
tional time of HLN for this case is 0.15 seconds.
The final solution is given in Table 12. The sched-
ule for this system in this case is also similar to
the case neglecting power loss except higher
thermal power generation to compensate the
power loss leading to higher energy production.
The power generation outputs of power plants for
this case and the comparison of energy production
costs for both cases during the whole schedule
time are given in Figures 10 and 11, respectively.
For the both cases, the variations of power
generation outputs and energy production costs
are corresponding to the variation of the load
demand in the whole schedule time horizon.
Future Research Directions
For further research directions based on the HLN
method, more developments of HLN and its
broader implementations to optimization problems
in power systems will be considered as follows:
Adaptive updating mechanism will be con-
sidered to replace the current updating step
sizes since they need to be tuned in the
HLN method for each problem.
Other ELD problems for thermal units such
as combined heat and power economic dis-
patch, economic dispatch with piecewise
fuel const function, economic dispatch
with prohibited operating zones etc will be
considered and solved by implementation
of the proposed HLN method.
Figure 9. Power generation of plants for variable-head HELD problem neglecting power loss
88
The HLN method will be also applied for
solving hydrothermal economic dispatch
with cascaded hydro plants.
The implementation of the HLN for solv-
ing large-scale optimization problems in
power systems will be also studied due to
its fast convergence to optimal solution.
CONCLUSION
In this chapter, the proposed HLN method has been
efficiently implemented for solving different ELD
problems including basic economic load dispatch,
fuel constrained economic load dispatch and hy-
drothermal economic load dispatch. By directly
using Lagrangian function as the energy function
of continuous Hopfield network in the HLN, it
is not necessary to pre-define an energy function
for the problem and map the problem into neural
network like the conventional Hopfield network.
Moreover, the HLN method can simultaneously
process all variables and constraints, so it can
quickly converge to optimal solution. The ob-
tained results from the test cases have shown that
the proposed HLN is reliable for finding optimal
solutions of the considered problems. The obvi-
ous advantages of the proposed HLN method for
Table 12. Solution for variable-head HELD problem with power loss
Hr. P
D
(MW)
($/MWh)
P
L
(MW)
P
1
(MW)
P
2
(MW)
P
3
(MW)
P
4
(MW)
d
1
(ft)
d
2
(ft)
1 800 4.2028 22.31 151.45 364.54 276.36 29.97 299.91 249.92
2 700 4.0849 16.97 135.06 314.40 255.75 11.77 299.83 249.89
3 600 3.9627 12.40 117.71 261.49 233.19 0 299.75 249.89
4 600 3.9624 12.40 117.67 261.38 233.35 0 299.68 249.88
5 600 3.9622 12.40 117.63 261.26 233.51 0 299.60 249.88
6 650 4.0262 14.59 126.79 289.16 246.10 2.53 299.52 249.87
7 800 4.2010 22.31 151.17 363.72 277.39 30.04 299.43 249.79
8 1000 4.4428 35.37 184.06 464.41 319.54 67.36 299.32 249.62
9 1330 4.8671 64.19 239.46 633.66 390.54 130.52 299.19 249.28
10 1350 4.8922 66.24 242.61 643.27 394.87 135.49 299.05 248.92
11 1450 5.0270 77.03 259.47 694.60 416.83 156.13 298.90 248.51
12 1500 5.0946 82.77 267.76 719.81 427.84 167.37 298.75 248.07
13 1300 4.8190 61.18 233.17 614.55 383.66 129.80 298.61 247.73
14 1350 4.8841 66.24 241.42 639.73 394.51 140.58 298.47 247.37
15 1350 4.8821 66.24 241.14 638.89 394.43 141.78 298.33 247.00
16 1370 4.9072 68.33 244.27 648.45 398.73 146.89 298.19 246.62
17 1450 5.0145 77.04 257.67 689.27 416.24 163.86 298.05 246.19
18 1570 5.1813 91.20 278.02 750.99 442.93 189.26 297.89 245.69
19 1430 4.9820 74.81 253.53 676.74 411.49 163.06 297.74 245.26
20 1350 4.8712 66.26 239.54 634.14 393.80 148.77 297.60 244.88
21 1270 4.7630 58.27 225.67 591.84 376.32 134.44 297.47 244.54
22 1150 4.6058 47.34 205.20 529.34 350.44 112.36 297.36 244.25
23 1000 4.4164 35.39 180.01 452.28 318.55 84.56 297.25 244.04
24 900 4.2935 28.46 163.36 401.32 297.55 66.24 297.15 243.88
89
Figure 10. Power generation of plants for variable-head HELD problem with power loss
Figure 11. Energy production costs for both cases with and without power loss
90
solving the economic dispatch problems in power
systems as well as optimization problems are its
ability to efficiently deal with nonlinear objectives
and constraints, to properly handle time-coupling
constraints by Lagrangian function and variable
limits by sigmoid function of continuous neurons
in Hopfield neural network, to quickly find the
optimal solution for the problems, and to deal with
very large-scale problems with multiple schedule
periods. In addition, unlike the population based
methods such as evolutionary programming, dif-
ferential evolution or particle swarm optimization,
the proposed method needs only one run to obtain
the optimal solution that does not depend on the
initially assumed solution for the algorithm. One
more highlighted characteristic of HLN is that its
energy function is simultaneously minimized with
respect to continuous neurons and maximized
with respect to multiplier neurons which satisfy
the Lagrangian function condition. Therefore,
this is a contribution for development of a new
computational tool for solving ELD problems
in power systems in particular and optimization
problems in general. On the other hand, a draw-
back of the HLN is that the updating step sizes
for the continuous neurons have to be tuned by
experiments for different problems. However,
these parameters can be easily tuned by starting
from small values and then gradually increasing
them until the continuous neurons produce the
solution at their lower and upper limits, thus the
obtained parameter values can be used.
REFERENCES
Asgarpoor, S. (1994). Comparison of linear,
nonlinear, and network flow programming tech-
niques in fuel scheduling. Electric Power Systems
Research, 30(3), 169174. doi:10.1016/0378-
7796(94)00851-5
Attaviriyanupap, P., Kita, H., Tanaka, E., &
Hasegawa, J. (2002). A hybrid EP and SQP for
dynamic economic dispatch with nonsmooth
fuel cost function. IEEE Transactions on Power
Systems, 17(2), 411416. doi:10.1109/TP-
WRS.2002.1007911
Bakirtzis, A., Petridis, V., & Kazarlis, S. (1994).
Genetic algorithm solution to the economic dis-
patch problem. IEE Proceedings. Generation,
Transmission and Distribution, 141(4), 377382.
doi:10.1049/ip-gtd:19941211
Basu, M. (2002). Fuel constrained economic emis-
sion load dispatch using Hopfield neural networks.
Electric Power Systems Research, 63(1), 5157.
doi:10.1016/S0378-7796(02)00090-1
Basu, M. (2003). Hopfield neural networks for op-
timal scheduling of fixed head hydrothermal power
systems. Electric Power Systems Research, 64(1),
1115. doi:10.1016/S0378-7796(02)00118-9
Chowdhury, B. H., & Rahman, S. (1990). A review
of recent advances in economic dispatch. IEEE
doi:10.1109/59.99376
Coelho, L. S., & Mariani, V. C. (2006). Combin-
ing of chaotic differential evolution and quadratic
programming for economic dispatch optimization
with valve-point effect. IEEE Transactions on
TPWRS.2006.873410
da Silva, I. N., Nepomuceno, L., & Bastos, T. M.
(2004). An efficient Hopfield network to solve
economic dispatch problems with transmission
system representation. International Journal
of Electrical Power & Energy Systems, 26(9),
733738. doi:10.1016/j.ijepes.2004.05.007
Fisher, M. L. (1973). Optimal solution of sched-
uling problems using Lagrange multipliers:
Part I. Operations Research, 21, 11141127.
doi:10.1287/opre.21.5.1114
91
Happ, H. H. (1977). Optimal power dispatch
A comprehensive survey. IEEE Transactions
on Power Apparatus and Systems, PAS-96(3),
841854. doi:10.1109/T-PAS.1977.32397
Hopfield, J. J. (1982). Neural networks and
physical systems with emergent collective com-
putational abilities. Proceedings of the National
Academy of Science, USA: Vol. 79 (pp. 2554-2558).
Hopfield, J. J. (1984). Neurons with graded re-
sponse have collective, computational properties
like those of two-state neuron. Proceedings of
the National Academy of Sciences of the United
States of America, 81, 30883092. doi:10.1073/
pnas.81.10.3088
Irving, M. R., & Sterling, M. J. H. (1985). Eco-
nomic dispatch of active power by quadratic
programming using a sparse linear comple-
mentary algorithm. International Journal of
Electrical Power & Energy Systems, 7(1), 26.
doi:10.1016/0142-0615(85)90002-X
Jeyakumar, D. N., Jayabarathi, T., & Raghuna-
than, T. (2006). Particle swarm optimization for
various types of economic dispatch problems.
International Journal of Electrical Power &
Energy Systems, 28(1), 3642. doi:10.1016/j.
ijepes.2005.09.004
Lee, K. Y., Nuroglu, F. M., & Sode-Yome, A.
(2000). Real power optimization with load flow
using adaptive Hopfield neural network. Engineer-
ing Intelligent Systems, 8(1), 5358.
Lee, K. Y., Sode-Yome, A., & Park, J. H. (1998).
Adaptive Hopfield neural networks for economic
load dispatch. IEEE Transactions on Power Sys-
tems, 13(2), 519526. doi:10.1109/59.667377
Liang, Z.-X., & Glover, J. D. (1992). A zoom
feature for a dynamic programming solution to
economic dispatch including transmission losses.
544550. doi:10.1109/59.141757
Nomana, N., & Iba, H. (2008). Differential evolu-
tion for economic load dispatch problems. Electric
Power Systems Research, 78(8), 13221331.
doi:10.1016/j.epsr.2007.11.007
Ongsakul, W., Dechanupaprittha, S., & Ngam-
roo, I. (2004). Parallel tabu search algorithm for
constrained economic dispatch. IEE Proceed-
ings. Generation, Transmission and Distribution,
151(2), 157166. doi:10.1049/ip-gtd:20040460
Park, J. H., Kim, Y. S., Eom, I. K., & Lee, K. Y.
(1993). Economic load dispatch for piecewise
quadratic cost function using Hopfield neural
network. IEEE Transactions on Power Systems,
8(3), 10301038. doi:10.1109/59.260897
Ponnambalam, K., Quintana, V. H., & Vanelli,
A. (1992). A fast algorithm for power system
optimization problems using an interior point
method. IEEE Transactions on Power Systems,
7(2), 892899. doi:10.1109/59.141801
Ruangpayoongsak, N., Ongsakul, W., & Rungger-
atigul, S. (2002). Constrained economic dispatch
by combined genetic and simulated annealing
algorithm. Electric Power Components & Systems,
30(9), 917931. doi:10.1080/15325000290085235
Somasundaram, P., Lakshmiramanan, R., & Kup-
pusamy, K. (2005). Hybrid algorithm based on EP
and LP for security constrained economic dispatch
problem. Electric Power Systems Research, 76(1-
3), 7785. doi:10.1016/j.epsr.2005.04.005
Song, Y. H., Chou, C. S., & Stonham, T. J. (1999).
Combined heat and power economic dispatch
by improved ant colony search algorithm. Elec-
tric Power Systems Research, 52(2), 115121.
doi:10.1016/S0378-7796(99)00011-5
Song, Y. H., Wang, G. S., Wang, P. Y., & Johns,
A. T. (1997). Environmental/economic dispatch
using fuzzy logic controlled genetic algorithms.
IEE Proceedings. Generation, Transmission and
Distribution, 144(4), 377382. doi:10.1049/ip-
gtd:19971100
92
Stahl, E. C. M. (1930). Load division in intercon-
nections. Electrical World, 95, 434438.
Stahl, E. C. M. (1931). Economic loading of
generating stations. Electrical Engineering, 50,
722727.
Su, C.-T., & Chiou, G.-J. (1997). A fast-compu-
tation Hopfield method to economic dispatch of
power systems. IEEE Transactions on Power Sys-
tems, 12(4), 17591764. doi:10.1109/59.627888
Su, C.-T., & Chiou, G.-J. (1997). An enhanced
Hopfield model for economic dispatch consider-
ing prohibited zones. Electric Power Systems
Research, 42(1), 7276. doi:10.1016/S0378-
7796(96)01208-4
Su, C.-T., & Lin, C.-T. (2000). New approach
with a Hopfield modeling framework to economic
dispatch. IEEE Transactions on Power Systems,
15(2), 541545. doi:10.1109/59.867138
van den Berg, J., & Bioch, J. C. (1993). Con-
strained optimization with a continuous Hop-
field-Lagrange model. (Technical report EUR-
CS-93-10), Erasmus University Rotterdam, Comp.
Sc. Dept., Faculty of Economics.
Victoire, T. A. A., & Jeyakumar, A. E. (2004). Hy-
brid PSOSQP for economic dispatch with valve-
point effect. Electric Power Systems Research,
71(1), 5159. doi:10.1016/j.epsr.2003.12.017
Wells, D. W. (1968). Method for economic secure
loading of a power system. Proceedings IEE,
115(8), 11901194.
Wong, K. P., & Fung, C. C. (1993). Simulated
annealing based economic dispatch algorithm.
IEE Proceedings. Generation, Transmission
ip-c.1993.0074
Wong, K. P., & Yuryevich, J. (1998). Evolution-
ary-programming-based algorithm for environ-
mentally constrained economic dispatch. IEEE
doi:10.1109/59.667339
Yalcinoz, T., & Short, M. J. (1997). Large-scale
economic dispatch using an improved Hopfield
neural network. IEE Proceedings. Generation,
Transmission and Distribution, 144(2), 181185.
doi:10.1049/ip-gtd:19970866
ADDITIONAL READING
Book Chapters
El-Hawary, M. E., & Christensen, G. S. (1979).
Optimal economic operation of electric power sys-
tems. London: Academic Press, Inc. Abdelaziza,
Kothari, D. P., & Dhillon, J. S. (2006). Power
system optimization. New Delhi: Prentice Hall
of India Private Limited.
Kothari, D. P., & Nagrath, I. J. (2003). Modern
power system analysis (3rd ed.). Boston: Mc
Graw-Hill.
Saadat, H. (2002). Power system analysis (2nd
ed.). Boston: Mc Graw-Hill.
Wood, A. J., & Wollenberg, B. F. (1996). Power
generation operation and control (2nd ed.). John
Wiley & Sons, Inc.
Wadhwa, C. L. (2006). Electrical power systems
(4th ed.). New Age International Publisher.
Zhu, J. (2009). Optimization of power system
operation. John Wiley & Sons, Inc., Publication.
doi:10.1002/9780470466971
93
Journal Articles
A. Y., Mekhamera, S. F., Badra, M. A. L., &
Kamh, M. Z. (2008). Economic dispatch using
an enhanced Hopfield neural network. Electric
Power Components and Systems, 36(7), 719732.
doi:10.1080/15325000701881969
Benyahia, M., Benasla, L., & Rahli, M. (2008).
Application of Hopfield neural networks to
economic environmental dispatch (EED). Acta
Electrotehnica, 49(3), 323327.
Dieu, V. N., & Ongsakul, W. (2009). Augmented
Lagrange Hopfield network for economic load
dispatch with combined heat and power. Int. J.
Electric Power Components and Systems, 37(12),
12891304. doi:10.1080/15325000903054969
Dieu, V. N., & Ongsakul, W. (2010). Economic
dispatch with emission and transmission con-
straints by augmented Lagrange Hopfield network.
Global Journal on Technology and Optimization,
1, 7783.
Elmetwally, M. M., Aal, F. A., Awad, M. L., &
Omran, S. (2008). A Hopfield neural network
approach for integrated transmission network
expansion planning. Journal of Applied Sciences
Research, 4(11), 13871394.
Haque, M. T., & Kashtiban, A. M. (2005). Ap-
plication of neural networks in power systems: A
review. World Academy of Science. Engineering
and Technology, 6, 5357.
King, T. D., El-Hawary, M. E., & El-Hawary,
F. (1995). Optimal environmental dispatch-
ing of electric power systems via an improved
Hopfield neural network model. IEEE Trans-
doi:10.1109/59.466488
Mekhamera, S. F., Abdelaziza, A. Y., Kamha, M.
Z., & Badr, M. A. L. (2009). Dynamic economic
dispatch using a hybrid Hopfield neural network
quadratic programming based technique. Electric
Power Components and Systems, 37(3), 253264.
doi:10.1080/15325000802454344
Mishra, D., Shukla, A. & Kalra, P. K. (2006).
OR-Neuron based Hopfield neural network for
solving economic load dispatch problem. Neural
Information Processing Letters and Reviews,
10(11), 249-259.
Naresh, R., & Sharma, J. (1999). Two-phase neural
network based solution technique for short term
hydrothermal scheduling. IEE Proceedings. Gen-
eration, Transmission and Distribution, 146(6),
657663. doi:10.1049/ip-gtd:19990855
Naresh, R., & Sharma, J. (2002). Short term hydro
scheduling using two-phase neural network. Elec-
trical Power & Energy Systems, 24(7), 583590.
doi:10.1016/S0142-0615(01)00069-2
Sharma, V., Jha, R., & Naresh, R. (2004). Optimal
multi-reservoir network control by two-phase
neural network. Electric Power Systems Research,
68(3), 221228. doi:10.1016/j.epsr.2003.06.002
Swarup, K. S., & Simi, P. V. (2006). Neural com-
putation using discrete and continuous Hopfield
networks for power system economic dispatch
and unit commitment. Neurocomputing, 70(1-
3), 119129. doi:10.1016/j.neucom.2006.05.002
Walsh, M. P., & OMalley, M. J. (1997). Augment-
ed Hopfield network for unit commitment and eco-
nomic dispatch. IEEE Transactions on Power Sys-
tems, 12(4), 17651774. doi:10.1109/59.627889
Yalcinoz, T., & Altun, H. (2000). Comparison
of simulation algorithms for the Hopfield neural
network: An application of economic dispatch.
Turkish Journal of Electrical Engineering &
Computer Sciences, 8(1), 6780.
94
Yalcinoz, T., Cory, B. J., & Short, M. J. (2001).
Hopfield neural network approaches to economic
dispatch problems. Int. J. Electrical and Energy
Systems, 23(6), 435442. doi:10.1016/S0142-
0615(00)00084-3
Yalcinoz, T., & Short, M. J. (1998). Neural net-
works approach for solving economic dispatch
problem with transmission capacity constraints.
307313. doi:10.1109/59.667341
Yalcinoz, T., Short, M. J., & Cory, B. J. (1999).
Security dispatch using the Hopfield neural net-
work. IEE Proceedings. Generation, Transmission
ip-gtd:19990506
KEY TERMS AND DEFINITIONS
Economic Load Dispatch: One of the optimi-
zation problems in power systems. In this problem,
it is assumed that the generating units are online
during the considered schedule time. The objective
of the economic dispatch problem is usually to
minimize the total cost of thermal generating units
while satisfying the unit and system constraints
such as power balance, power generation limits,
ramp rate constraints, etc.
Energy Function: A function of the outputs
of neurons defined for Hopfield neural network.
In the Hopfield neural network, any change in
the status of continuous neurons always leads to
minimization of the energy function.
Fuel Constrained Economic Load Dispatch:
The economic dispatch problem with more con-
straints of fuel delivery and fuel storage for each
generating unit.
Hopfield Lagrange Network: A continuous
Hopfield neural network with its energy func-
tion based on Lagrangian function. By using the
Lagrangian relaxation for the energy function,
the Lagrange Hopfield network is simpler, easier,
and more efficient than the conventional Hopfield
neural network in solving optimization problems.
Hydrothermal Economic Load Dispatch:
An economic dispatch problem for both thermal
and hydro generating units. In this problem, the
hydro constraints are also taken into consideration
in addition to the thermal and system constraints.
This problem is more complicated than the eco-
nomic dispatch problem due to integration of
time-coupling constraints.
Lagrangian Function: A function used to
convert a constrained optimization problem to an
unconstrained optimization problem so that the
problem could be easier to be solved.
Sigmoid Function: A nonlinear function used
to calculate the outputs of continuous neurons
based on their inputs. In the sigmoid function, its
slope can be adjusted leading to the change of the
function shape and the outputs will be changed
accordingly.
Transfer Function: A linear function used to
calculate the outputs of multiplier neurons based
on their inputs. The slope of this function cannot
be usually adjusted and the outputs of neurons
are equally set to their inputs.
95
Chapter 3
DOI: 10.4018/978-1-61350-138-2.ch003
INTRODUCTION
Globally, buildings are responsible for approxi-
mately 40% of the total world annual energy con-
sumption. Most of this energy is for the provision
of lighting, heating, cooling, and air conditioning.
Increasing awareness of the environmental im-
pact of CO
2
, NO
x
and CFCs emissions triggered
a renewed interest in environmentally friendly
cooling, and heating technologies. Under the 1997
Montreal Protocol, governments agreed to phase
out chemicals used as refrigerants that have the
potential to destroy stratospheric ozone. It was
therefore considered desirable to reduce energy
consumption and decrease the rate of depletion
of world energy reserves and pollution of the en-
vironment. One way of reducing building energy
consumption is to design buildings, which are
Abdeen Mustafa Omer
Energy Research Institute, UK
Renewable Energy and
Sustainable Development
ABSTRACT
The increased availability of reliable and effcient energy services stimulates new development alter-
natives. This article discusses the potential for such integrated systems in the stationary and portable
power market in response to the critical need for a cleaner energy technology. Anticipated patterns of
future energy use and consequent environmental impacts (acid precipitation, ozone depletion and the
greenhouse effect or global warming) are comprehensively discussed in this chapter. Throughout the
theme several issues relating to renewable energies, environment, and sustainable development are ex-
amined from both current and future perspectives. It is concluded that green energies like wind, solar,
ground-source heat pumps, and biomass must be promoted, implemented, and demonstrated from the
economic and/or environmental point view.
96
Renewable Energy and Sustainable Development
more economical in their use of energy for heat-
ing, lighting, cooling, ventilation and hot water
supply. Passive measures, particularly natural or
hybrid ventilation rather than air-conditioning,
can dramatically reduce primary energy con-
sumption (Omer, 2009a). However, exploitation
of renewable energy in buildings and agricultural
greenhouses can, also, significantly contribute
towards reducing dependency on fossil fuels.
Therefore, promoting innovative renewable ap-
plications and reinforcing the renewable energy
market will contribute to preservation of the
ecosystem by reducing emissions at local and
global levels. This will also contribute to the
amelioration of environmental conditions by
replacing conventional fuels with renewable ener-
gies that produce no air pollution or greenhouse
gases (during their use). The provision of good
indoor environmental quality while achieving
energy and cost efficient operation of the heating,
ventilating and air-conditioning (HVAC) plants
(devices) in buildings represents a multi variant
problem (Omer, 2009b). The comfort of building
occupants is dependent on many environmental
parameters including air speed, temperature, rela-
tive humidity and air quality in addition to lighting
and noise. The overall objective is to provide a
high level of building performance (BP), which
can be defined as indoor environmental quality
(IEQ), energy efficiency (EE) cost efficiency (CE),
and environmental performance (EP).
Indoor environmental quality is the per-
ceived condition of comfort that building
occupants experience due to the physical
and psychological conditions to which
they are exposed by their surroundings.
The main physical parameters affecting
IEQ are air speed, temperature, relative
humidity and air quality.
Energy effciency is related to the provi-
sion of the desired environmental condi-
tions while consuming the minimal quan-
tity of energy.
Cost effciency is the fnancial expenditure
on energy relative to the level of environ-
mental comfort and productivity that the
building occupants attained. The overall
cost effciency can be improved by im-
proving the indoor environmental quality
and the energy effciency of a building.
Several definitions of sustainable development
have been put forth, including the following com-
mon one: development that meets the needs of
the present without compromising the ability of
future generations to meet their own needs. The
World Energy Council (WEC) study found that
without any change in our current practice, the
world energy demand in 2020 would be 50-80%
higher than 1990 levels (WEC, 2009). According
to the USA Department of Energy (DoE) report,
annual energy demand will increase from a current
capacity of 363 million kilowatts to 750 million
kilowatts by 2020 (DoE, 2009). The worlds energy
consumption today is estimated to 22 billion kWh
per year, 53 billion kWh by 2020 (WEC, 2009).
Such ever-increasing demand could place signifi-
cant strain on the current energy infrastructure and
potentially damage world environmental health
by CO, CO
2
, SO
2
, NO
x
effluent gas emissions
and global warming (ASHRAE, 2005). Achiev-
ing solutions to environmental problems that we
face today requires long-term potential actions for
sustainable development. In this regards, renew-
able energy resources appear to be the one of the
most efficient and effective solutions since the
intimate relationship between renewable energy
and sustainable development. More rational use
of energy is an important bridge to help transition
from todays fossil fuel dominated world to a world
powered by non-polluting fuels and advanced
technologies such as photovoltaics (PVs) and fuel
cells (FCs) (Abdeen, 2008a).
An approach is needed to integrate renewable
energies in a way to meet high building perfor-
mance. However, because renewable energy
sources are stochastic and geographically diffuse,
97
their ability to match demand is determined by
adoption of one of the following two approaches
(EUO, 2000): the utilisation of a capture area
greater than that occupied by the community to
be supplied, or the reduction of the communitys
energy demands to a level commensurate with the
locally available renewable resources.
For a northern European climate, which is
characterised by an average annual solar irradi-
ance of 150 Wm
-2
, the mean power production
from a photovoltaic component of 13% conver-
sion efficiency is approximately 20 Wm
-2
(Duffie
and Beckman, 1980). For an average wind speed
of 5 ms
-1
, the power produced by a micro wind
turbine will be of a similar order of magnitude,
though with a different profile shape. In the UK,
for example, a typical office building will have
a demand in the order of 300 kWhm
-2
yr
-1
(EUO,
2000). This translates into approximately 50 Wm
-2

of faade, which is twice as much as the available
renewable energies (Lysen, 1983). Thus, the aim
is to utilise energy efficiency measures in order to
reduce the overall energy consumption and adjust
the demand profiles to be met by renewable ener-
gies. For instance, this approach can be applied to
greenhouses, which use solar energy to provide
indoor environmental quality. The greenhouse
effect is one result of the differing properties of
heat radiation when it is generated at different
temperatures. Objects inside the greenhouse, or
any other building, such as plants, re-radiate the
heat or absorb it. Because the objects inside the
greenhouse are at a lower temperature than the
sun, the re-radiated heat is of longer wavelengths,
and cannot penetrate the glass. This re-radiated
heat is trapped and causes the temperature inside
the greenhouse to rise. Note that the atmosphere
surrounding the earth, also, behaves as a large
greenhouse around the world. Changes to the
gases in the atmosphere, such as increased carbon
dioxide content from the burning of fossil fuels,
can act like a layer of glass and reduce the quantity
of heat that the planet earth would otherwise radi-
ate back into space (Brain, and Mark, 2007). This
particular greenhouse effect, therefore, contributes
to global warming. The application of greenhouses
for plants growth can be considered one of the
measures in the success of solving this problem.
Maximising the efficiency gained from a green-
house can be achieved using various approaches,
employing different techniques that could be ap-
plied at the design, construction and operational
stages. The development of greenhouses could be
a solution to farming industry and food security
(Abdeen, 2008b).
Energy security, economic growth and envi-
ronment protection are the national energy policy
drivers of any country of the world. As world
populations grow, many faster than the growth
rate of 2%, the need for more and more energy
is exacerbated (Figure 1). Enhanced lifestyle and
energy demand rise together and the wealthy in-
dustrialised economics, which contain 25% of the
worlds population, consume 75% of the worlds
energy supply (WEC, 2009). The worlds energy
consumption today is estimated to 22 billion kWh
per year (WEC, 2009). About 6.6 billion metric
tons carbon equivalent of greenhouse gas (GHG)
emission are released in the atmosphere to meet
this energy demand (WEC, 2009). Approximately
80% is due to carbon emissions from the com-
bustion of energy fuels (Abdeen, 2008c). At the
current rate of usage, taking into consideration
population increases and higher consumption of
energy by developing countries, oil resources,
natural gas and uranium will be depleted within a
few decades. People could depend on new nuclear
technologies that will enable much slower uranium
depletion in the future. As for coal, it may take
two centuries or so. Technological progress has
dramatically changed the world in a variety of
ways. It has, however, also led to developments
e.g., environmental problems, which threaten man
and nature. Build-up of carbon dioxide and other
GHGs is leading to global warming with unpre-
dictable but potentially catastrophic consequences.
When fossil fuels burn, they emit toxic pollutants
that damage the environment and peoples health
98
with over 700,000 deaths resulting each year, ac-
cording to the World Bank review of 2000. At the
current rate of usage, taking into consideration
population increases and higher consumption of
energy by developing countries, oil resources, and
natural gas will be depleted within a few decades.
A Figure 2 shows the annual and estimated world
population and energy demand, and Figure 3 the
world oil productions in the next 10-20 years. As
for coal, it may take two centuries or so. One must
therefore endeavour to take precautions today for
a viable world for coming generations.
Research into future alternatives has been and
still being conducted aiming to solve the complex
problems of this recent time e.g., rising energy
requirements of a rapidly and constantly growing
world population and global environmental pol-
lution. Therefore, options for a long-term and
environmentally friendly energy supply have to
be developed leading to the use of renewable
sources (water, sun, wind, biomass, geothermal,
hydrogen production by electrolysis of water and
fuel cells. Renewables could shield a nation from
the negative effect in the energy supply, price and
related environment concerns.
Hydrogen for fuel cells and the sun for PV have
been considered for many years as a likely and
eventual substitute for oil, gas, and coal. The sun
is the most abundant element in the universe. The
use of solar thermal energy or solar photovoltaic
(PVs) for the everyday electricity needs has a
distinct advantage: avoid consuming resources
and degrading the environment through pollut-
ing emissions, oil spills and toxic by-products.
A one-kilowatt PV system producing 150 kWh
each month prevents 75 kg of fossil fuel from
being mined (WEC, 2009), and 150 kg of CO
2

from entering the atmosphere and keeps 473
litres of water from being consumed (Abdeen,
2008d). Electricity from fuel cells can be used
in the same way as grid power: to run appliances
and light bulbs and even to power cars since each
gallon of gasoline produced and used in an inter-
nal combustion engine releases roughly 12 kg of
CO
2
, a GHG that contributes to global warming.
People, Power and Pollution
Over millions of years ago plants covered the earth,
converting the energy of sunlight into living tissue,
Figure 1. Annual and estimated world population and energy demand in [Million of barrels per day of
oil equivalent (MBDOE)] (Omer, 2008a)
99
some of which was buried in the depths of the earth
to produce deposits of coal, oil and natural gas.
The past few decades, however, have experienced
many valuable uses for these complex chemical
substances, manufacturing from them plastics,
textiles, fertiliser and the various end products of
the petrochemical industry. Indeed, each decade
sees increasing uses for these products. Renew-
able energy is the term used to describe a wide
range of naturally occurring, replenishing energy
sources. Coal, oil and gas, which will certainly
be of great value to future generations, as they
are to ours, are non-renewable natural resources.
This is particularly true now as it is, or soon will
be, technically and economically feasible to sup-
ply all of mans needs from the most abundant
energy source of all, the sun. The sunlight is not
only inexhaustible, but, moreover, it is the only
energy source, which is completely non-polluting.
Figure 2. World oil productions in the next 10-20 years (Omer, 2008a)
Figure 3. Volume of oil discovered worldwide (Omer, 2008a)
100
Industrys use of fossil fuels has been blamed
for warming the climate. When coal, gas and oil
are burnt, they release harmful gases, which trap
heat in the atmosphere and cause global warm-
ing (Bos et al, 1994). However, there has been an
ongoing debate on this subject, as scientists have
struggled to distinguish between changes, which
are human induced, and those, which could be
put down to natural climate variability. Never-
theless, industrialised countries have the highest
emission levels, and must shoulder the greatest
responsibility for global warming. However, ac-
tion must also be taken by developing countries
to avoid future increases in emission levels as
their economies develop and populations grow, as
clearly captured by the Kyoto Protocol (Abdeen,
2008e). Notably, human activities that emit carbon
dioxide (CO
2
), the most significant contributor to
potential climate change, occur primarily from
fossil fuel production. Consequently, efforts to
control CO
2
emissions could have serious, negative
consequences for economic growth, employment,
investment, trade and the standard of living of
individuals everywhere. Scientifically, it is dif-
ficult to predict the relationship between global
temperature and GHG concentrations. The climate
system contains many processes that will change
if warming occurs. Critical processes include heat
transfer by winds and tides, the hydrological cycle
involving evaporation, precipitation, runoff and
groundwater and the formation of clouds, snow,
and ice, all of which display enormous natural
variability (UNECA, 2003b).
The equipment and infrastructure for energy
supply and use are designed with long lifetimes,
and the premature turnover of capital stock in-
volves significant costs. Economic benefits occur
if capital stock is replaced with more efficient
equipment in step with its normal replacement
cycle. Likewise, if opportunities to reduce future
emissions are taken in a timely manner, they
should be less costly. Such a flexible approach
would allow society to take account of evolving
scientific and technological knowledge, while
gaining experience in designing policies to address
climate change (Abdeen, 2009a).
The World Summit on Sustainable Develop-
ment in Johannesburg in 2002 committed itself
to encourage and promote the development of
renewable energy sources to accelerate the shift
towards sustainable consumption and produc-
tion. Accordingly, it aimed at breaking the link
between resource use and productivity. This can
be achieved by the followings:
Trying to ensure economic growth does
not cause environmental pollution.
Improving resource effciency.
Examining the whole life-cycle of a
product.
Enabling consumers to receive more infor-
mation on products and services.
Examining how taxes, voluntary agree-
ments, subsidies, regulation and informa-
tion campaigns, can best stimulate inno-
vation and investment to provide cleaner
technology.
The energy conservation scenarios include
rational use of energy policies in all economy
sectors and the use of combined heat and power
systems, which are able to add to energy savings
from the autonomous power plants. Electricity
from renewable energy sources is by definition
the environmental green product. Hence, a renew-
able energy certificate system, as recommended
by the World Summit, is an essential basis for
all policy systems, independent of the renewable
energy support scheme. It is, therefore, important
that all parties involved support the renewable
energy certificate system in place if it is to work
as planned. Moreover, existing renewable energy
technologies (RETs) could play a significant miti-
gating role, but the economic and political climate
will have to change first. The change in climate is
real. It is happening now, and GHGs produced by
human activities are significantly contributing to
it. The predicted global temperature increase of
101
between 1.5 and 4.5
o
C could lead to potentially
catastrophic environmental impacts (DEFRA,
2006). These include sea level rise, increased
frequency of extreme weather events, floods,
droughts, disease migration from various places
and possible stalling of the Gulf Stream. This has
led scientists to argue that climate change issues
are not ones that politicians can afford to ignore,
and policy makers tend to agree (DEFRA, 2006).
However, reaching international agreements on
climate change policies is no trivial task as the
difficulty in ratifying the Kyoto Protocol has
proved (UNECA, 2004).
Therefore, the use of renewable energy sources
and the rational use of energy, in general, are the
fundamental inputs for any responsible energy
policy. However, the energy sector is encounter-
ing difficulties because increased production and
consumption levels entail higher levels of pollu-
tion and eventually climate change, with possibly
disastrous consequences. At the same time, it is
important to secure energy at an acceptable cost
in order to avoid negative impacts on economic
growth. To date, renewable energy contributes
as much as 20% of the global energy supplies
worldwide (Abdeen, 2009b). Over two thirds of
this comes from biomass use, mostly in develop-
ing countries, some of it unsustainable. Yet, the
potential for energy from sustainable technologies
is huge. On the technological side, renewables
have an obvious role to play. In general, there is
no problem in terms of the technical potential of
renewables to deliver energy. Moreover, there
are very good opportunities for RETs to play an
important role in reducing emissions of GHGs
into the atmosphere, certainly far more than have
been exploited so far. However, there are still some
technical issues to address in order to cope with
the intermittency of some renewables, particularly
wind and solar. Yet, the biggest problem with rely-
ing on renewables to deliver the necessary cuts in
GHG emissions is more to do with politics and
policy issues than with technical ones (DEFRA,
2006). For example, the single most important step
governments could take to promote and increase
the use of renewables is to improve access for
renewables to the energy market. This access to
the market needs to be under favourable conditions
and, possibly, under favourable economic rates as
well. One move that could help, or at least justify,
better market access would be to acknowledge
that there are environmental costs associated with
other energy supply options and that these costs
are not currently internalised within the market
price of electricity or fuels. This could make a
significant difference, particularly if appropriate
subsidies were applied to renewable energy in
recognition of the environmental benefits it offers.
Similarly, cutting energy consumption through
end-use efficiency is absolutely essential. This
suggests that issues of end-use consumption of
energy will have to come into the discussion in
the foreseeable future (Levine et al, 2005).
However, RETs have the benefit of being
environmentally benign when developed in a
sensitive and appropriate way with the full involve-
ment of local communities. In addition, they are
diverse, secure, locally based and abundant. In
spite of the enormous potential and the multiple
benefits, the contribution from renewable energy
still lags behind the ambitious claims for it due
to the initially high development costs, concerns
about local impacts, lack of research funding and
poor institutional and economic arrangements
(IPCC, 2001).
Hence, an approach is needed to integrate
renewable energies in a way that meets high
building performance requirements. However,
because renewable energy sources are stochastic
and geographically diffuse, their ability to match
demand is determined by adoption of one of the
following two approaches (Parikn et al, 1999):
the utilisation of a capture area greater than that
occupied by the community to be supplied, or the
reduction of the communitys energy demands to
a level commensurate with the locally available
renewable resources.
102
Energy and Population Growth
Urban areas throughout the world have increased
in size during recent decades. About 50% of the
worlds population and approximately 7.6% in
more developed countries are urban dwellers
(UNIDO, 2007). Even though there is evidence
to suggest that in many advanced industrialised
countries there has been a reversal in the rural-to-
urban shift of populations, virtually all population
growth expected between 2000 and 2030 will be
concentrated in urban areas of the world (UN,
2002b). With an expected annual growth of 1.8%,
the worlds urban population will double in 38
years (UNIDO, 2007).
With increasing urbanisation in the world,
cities are growing in number, population and
complexity. At present, 2% of the worlds land
surface is covered by cities, yet the people living
in them consume 75% of the resources consumed
by mankind (WRI, 2004). Indeed, the ecological
footprint of cities is many times larger than the
areas they physically occupy. Economic and social
imperatives often dictate that cities must become
more concentrated, making it necessary to increase
the density to accommodate the people, to reduce
the cost of public services, and to achieve required
social cohesiveness. The reality of modern urbani-
sation inevitably leads to higher densities than in
traditional settlements and this trend is particularly
notable in developing countries (Omer, 2010a).
Generally, the world population is rising rap-
idly, notably in the developing countries. Historical
trends suggest that increased annual energy use per
capita, which promotes a decrease in population
growth rate, is a good surrogate for the standard of
living factors. If these trends continue, the stabi-
lisation of the worlds population will require the
increased use of all sources of energy, particularly
as cheap oil and gas are depleted. The improved
efficiency of energy use and renewable energy
sources will, therefore, be essential in stabilising
population, while providing a decent standard of
living all over the world (UN, 2002a). Moreover,
energy is the vital input for economic and social
development of any country. With an increase in
industrial and agricultural activities the demand
for energy is also rising. It is, however, a well-
accepted fact that commercial energy use has to be
minimised. This is because of the environmental
effects and the availability problems. Consequent-
ly, the focus has now shifted to non-commercial
energy resources, which are renewable in nature.
This is bound to have less environmental effects
and also the availability is guaranteed. However,
even though the ideal situation will be to encour-
age people to use renewable energy resources,
there are many practical difficulties, which need
to be tackled. The people groups who are using
the non-commercial energy resources, like urban
communities, are now becoming more demanding
and wish to have commercial energy resources
made available for their use (UNEP, 2000). This
is attributed to the increased awareness, improved
literacy level and changing culture (Abdeen,
2009c). The quality of life practiced by people is
usually represented as being proportional to the
per capita energy use of that particular country.
It is not surprising that people want to improve
their quality of life. Consequently, it is expected
that the demand for commercial energy resources
will increase at a greater rate in the years to come
(WRI, 2004). Because of this emerging situation,
the policy makers are left with two options: either
to concentrate on renewable energy resources and
have them as substitutes for commercial energy
resources or to have a dual approach in which
renewable energy resources will contribute to
meet a significant portion of the demand whereas
the conventional commercial energy resources
would be used with caution whenever necessary
(UNECA, 2002). Even though the first option is
the ideal one, the second approach will be more
appropriate for a smooth transition (UN, 2001).
103
Energy and Environmental Problems
Technological progress has dramatically changed
the world in a variety of ways. It has, however, also
led to developments of environmental problems,
which threaten man and nature (UNECA, 2003a).
During the past two decades the risk and reality
of environmental degradation have become more
apparent. Growing evidence of environmental
problems is due to a combination of several fac-
tors since the environmental impact of human
activities has grown dramatically because of the
sheer increase of world population, consumption,
industrial activity, etc., throughout the 1970s most
environmental analysis and legal control instru-
ments concentrated on conventional effluent gas
pollutants such as SO
2
, NO
x
, particulates, and CO
(Table 1). Recently, environmental concerns has
extended to the control of micro or hazardous
air pollutants, which are usually toxic chemical
substances and harmful in small doses, as well to
that of globally significant pollutants such as CO
2
.
Aside from advances in environmental science,
developments in industrial processes and struc-
tures have led to new environmental problems.
For example, in the energy sector, major shifts
to the road transport of industrial goods and to
individual travel by cars has led to an increase in
road traffic and hence a shift in attention paid to
the effects and sources of NO
x
and volatile organic
compound (VOC) emissions.
Environmental problems span a continuously
growing range of pollutants, hazards and ecosys-
tem degradation over wider areas. The main areas
of environmental problems are: major environ-
mental accidents, water pollution, maritime pol-
lution, land use and sitting impact, radiation and
radioactivity, solid waste disposal, hazardous air
pollutants, ambient air quality, acid rain, strato-
spheric ozone depletion and global warming
(greenhouse effect, global climatic change) (Table
2). The four more important types of harm from
mans activities are global warming gases, ozone
destroying gases, gaseous pollutants and micro-
biological hazards (Table 3). Notably, human
activities that emit carbon dioxide (CO
2
), the most
significant contributor to potential climate change,
occur primarily from fossil fuel production. Con-
sequently, CO
2
emissions could have serious,
negative consequences for economic growth,
employment, investment, trade and the standard
of living of individuals everywhere. The earth is
warmer due to the presence of gases but the
global temperature is rising. This could lead to
Table 1. EU criteria pollutant standards in the ambient air environment (Omer, 2008a)
Pollutant EU limit
CO 30 mg/m
2
; 1h
NO
2
200 g/m
2
; 1h
O
3
235 g/m
2
; 1h
SO
2
250-350 g/m
2
; 24 h
80-120 g/m
2
; annual
PM
10
250 g/m
2
; 24 h
80 g/m
2
; annual
SO
2
+ PM
10
100-150 g/m
2
; 24 h
40-60 g/m
2
; annual
Pb 2 g/m
2
; annual
Total suspended particulate (TSP 260 g/m
2
; 24 h
HC 160 g/m
2
; 3 h
104
the sea level rising at the rate of 60 mm each
decade with the growing risk of flooding in low-
lying areas (Figure 4).
At the United Nations Earth Summit at Rio in
June 1992 some 153 countries agreed to pursue
sustainable development (Boulet, 1987). A main
aim was to reduce emission of carbon dioxide and
other GHGs. Reduction of energy use in buildings
is a major role in achieving this. Carbon dioxide
targets are proposed to encourage designers to
look at low energy designs and energy sources.
Problems with energy supply and use are
related not only to global warming that is taking
place, due to effluent gas emission mainly CO
2
,
but also to such environmental concerns as air
pollution, acid precipitation, ozone depletion,
forest destruction and emission of radioactive
substances (Table 4). These issues must be taken
into consideration simultaneously if humanity is
to achieve a bright energy future with minimal
environmental impacts. Much evidence exists,
which suggests that the future will be negatively
impacted if humans keep degrading the environ-
ment.
During the past century, global surface tem-
peratures have increased at a rate near 0.6
o
C/
century and the average temperature of the At-
lantic, Pacific and Indian oceans (covering 72%
of the earth surface) have risen by 0.06
o
C since
1995. Global temperatures in 2001 were 0.52
o
C
Table 2. Significant EU environmental directives in water, air and land environments (Omer, 2008a)
Environment Directive name
Water Surface water for drinking
Sampling surface water for drinking
Drinking water quality
Quality of freshwater supporting fish
Shellfish waters
Bathing waters
Dangerous substances in water
Groundwater
Urban wastewater
Nitrates from agricultural sources
Air Smokes in air
Sulphur dioxide in air
Lead in air
Large combustion plants
Existing municipal incineration plants
New municipal incineration plants
Asbestos in air
Sulphur content of gas oil
Lead in petrol
Emissions from petrol engines
Air quality standards for NO
2
Emissions from diesel engines
Land Protection of soil when sludge is applied
105
Table 3. The external environment (Omer, 2009)
Damage Manifestation Design
NO
x
, SO
x
Irritant Low NO
x
burners
Acid rain land damage Low sulphur fuel
Acid rain fish damage Sulphur removal
Global warming Thermal insulation
CO
2
Rising sea level Heat recovery
Drought, storms Heat pumps
O
3
destruction Increased ultra violet No CFCs or HCFCs
Skin cancer Minimum air conditioning
Crop damage Refrigerant collection
Legionnellosis Pontiac fever Careful maintenance
Legionnaires Dry cooling towers
Figure 4. Change in global sea level (Omer, 2008a)
Table 4. Global emissions of the top fourteen nations by total CO
2
volume (billion of tones) (Omer, 2009)
Rank Nation CO
2
Rank Nation CO
2
Rank Nation CO
2
1
2
3
4
USA
Russia
China
Japan
1.36
0.98
0.69
0.30
6
7
8
9
India
UK
Canada
Italy
0.19
0.16
0.11
0.11
11
12
13
14
Mexico
Poland
S. Africa
S. Korea
0.09
0.08
0.08
0.07
106
above the long-term 1880-2000 average (the
1880-2000 annually averaged combined land and
ocean temperature is 13.9
o
C). Also, according to
the USA Department of Energy, world emissions
of carbon are expected to increase by 54% above
1990 levels by 2015 making the earth likely to
warm 1.7-4.9
o
C over the period 1990-2100, as
shown in Figure 5. Such observation and others
demonstrate that interests will likely increase
regarding energy related environment concerns
and that energy is one of the main factors that
must be considered in discussions of sustainable
development.
New and renewable sources of energy can
make an increasing contribution to the energy
supply mix of the world in view of favourable
renewable energy resource endowments, limita-
tions and uncertainties of fossil fuel supplies,
adverse balance of payments and the increasing
pressure on environment from conventional en-
ergy generation.
Among the renewable energy technologies, the
generation of mechanical and electrical power by
wind machines has emerged as a techno-economi-
cal viable and cost-effective option (Omer, 2010b).
Environmental Transformations
In recent years a number of countries have adopted
policies aimed at giving a greater role to private
ownership in the natural resource sector. For ex-
ample, in the UK the regional water companies
have been privatised and have been given a con-
siderable degree of control over the exploitation
of the nations regional water resources. Similar
policies have been followed in France and other
European countries. Typically, a whole range of
new regulatory instruments such as technological
standards accompanies such privatisation on water
treatment plants, minimum standards on drinking
water quality, price controls and maximum with-
drawal quotas. While some of these instruments
address problems of monopolistic behaviour and
other forms of imperfect competition, the bulk of
regulatory measures is concerned with establishing
good practices aimed at maintaining the quality
of the newly privatised resources as a shorthand.
Society has to meet the freshwater demands of its
population and its industry by extracting water
from the regional water resources that are provided
by the natural environment (lakes, rivers, aquifers,
Figure 5. Global mean temperature changes over the period of 1990-2100 and 1990-2030 (Omer, 2009)
107
etc.). These water resources are renewable but
potentially destructible resources. While moder-
ate amounts of human water extractions from
a given regional water system can be sustained
for indefinite periods. Excessive extractions will
change the geographical and climatic conditions
supporting the water cycle and will diminish
the regenerative capacity of the regional water
system, thereby reducing the potential for future
withdrawals. Typically, recovery from any such
resource degradation will be very slow and dif-
ficult, if not impossible; resource degradation is
partially irreversible (Erreygers, 1996).
To make sustainable water extraction economi-
cally viable, the sustainable policy has to break
even (all costs are covered by revenues) while
unsustainable policy has to be unprofitable (costs
exceed revenues):
(1+r) vt
-1
= 5y
t
+ v
t
(1)
Where: r is the interest rate, t=year, y
t
is the
revenue, v
t
is initial costs recovered by revenue,
and vt
-1
is all costs are covered by revenues.
(1+r) vt
-1
> 105y
t
(2)
(1+r) vt
-1
< [105/(105-5)] v
t
(3)
The term [105/(105-5)] is to define the natural
productivity factor of the water resource as (1+g)
= [105/(105-5)]; g is the natural productivity rate.
Rate g will be close to zero if the sustainable
extraction level is much smaller than the unsustain-
able level. Using g, the equation can be as follows:
v
t >
(1+r)/(1+g) v
t-1
(4)
Regulatory measures that prevent resource
owners from adopting certain unsustainable
extraction policies are a necessary pre-condition
for the effective operation of a privatised natural
resource sector. Unregulated water privatisation
would result in an inflationary dynamics whose
distributional effects would threaten the long-
term viability of the economy. This inflation-
ary dynamics is not due to any form of market
imperfection but is a natural consequence of the
competitive arbitrage behaviour of unregulated
private resource owners.
Sustainability Concept
Absolute sustainability of electricity supply is a
simple concept: no depletion of world resources
and no ongoing accumulation of residues. Relative
sustainability is a useful concept in comparing the
sustainability of two or more generation technolo-
gies. Therefore, only renewables are absolutely
sustainable, and nuclear is more sustainable than
fossil. Energy used to produce devices and plants
for renewable energy is not sustainable. How-
ever, any discussion about sustainability must
not neglect the ability or otherwise of the new
technologies to support the satisfactory opera-
tion of the electricity supply infrastructure. The
electricity supply system has been developed to
have a high degree of resilience against the loss
of transmission circuits and major generators, as
well as unusually large and rapid load changes.
It is unlikely that consumers would tolerate any
reduction in the quality of the service, even if
this were the result of the adoption of otherwise
benign generation technologies. Renewables are
generally weather-dependent and as such their
likely output can be predicted but not controlled.
The only control possible is to reduce the output
below that available from the resource at any given
time. Therefore, to safeguard system stability and
security, renewables must be used in conjunction
with other, controllable, generation and with large-
scale energy storage. There is a substantial cost
associated with this provision (Abdeen, 2009d).
It is useful to codify all aspects of sustain-
ability, thus ensuring that all factors are taken into
account for each and every development proposal.
Therefore, with the intention of promoting debate,
the following considerations are proposed:
108
1. Long-term availability of the energy source
or fuel.
2. Price stability of energy source or fuel.
3. Acceptability or otherwise of by-products
of the generation process.
4. Grid services, particularly controllability of
real and reactive power output.
5. Technological stability, likelihood of rapid
technical obsolescence.
6. Knowledge base of applying the technology.
7. Life of the installation a dam may last more
than 100 years, but a gas turbine probably
will not.
8. Maintenance requirement of the plant.
Environmental Aspects
Environmental pollution is a major problem facing
all nations of the world. People have caused air
pollution since they learned to how to use fire,
but man-made air pollution (anthropogenic air
pollution) has rapidly increased since industriali-
sation began. Many volatile organic compounds
and trace metals are emitted into the atmosphere
by human activities. The pollutants emitted into
the atmosphere do not remain confined to the
area near the source of emission or to the local
environment, and can be transported over long
distances, and create regional and global envi-
ronmental problems. The privatisation and price
liberalisation in energy fields has been secured
to some extent (but not fully). There should be
availability and adequate energy supplies to the
major productive sectors. The result is that, the
present situation of energy supplies is for better
than ten years ago.
A great challenge facing the global community
today is to make the industrial economy more like
the biosphere, that is, to make it a more closed
system. This would save energy, reduce waste
and pollution, and reduce costs. In short, it would
enhance sustainability. Often, it is technically
feasible to recycle waste in one of several dif-
ferent ways. For some wastes there are powerful
arguments for incineration with energy recovery,
rather than material recycling. Cleaner produc-
tion approach and pollution control measures
are needed in the recycling sector as much as
in another. The industrial sector world widely is
responsible for about one third of anthropogenic
emissions of carbon dioxide, the most important
greenhouse gas. Industry is also an important
emitter of several other greenhouse gases. And
many of industrys products emit greenhouse gases
as well, either during use or after they become
waste. Opportunities exist for substantial reduc-
ing industrial emissions through more efficient
production and use of energy: fuel substitutions,
the use of alternative energy technologies, process
modification, and by revising materials strategies
to make use of less energy and greenhouse gas
intensive materials. Industry has an additional
role to play through the design of products that
use less energy and materials and produce lower
greenhouse gas emissions. Table 5 summarises
the classification of data requirements.
Development in the environmental sense is a
rather recent concern relating to the need to man-
age scarce natural resources in a prudent manner-
because human welfare ultimately depends on
ecological services. The environmental interpre-
tation of sustainability focuses on the overall
viability and health of ecological systems- defined
in terms of a comprehensive, multiscale, dy-
namic, hierarchical measure of resilience, vigour
and organisation. Natural resource degradation,
pollution and loss of biodiversity are detrimental
because they increase vulnerability, undermine
system health, and reduce resilience. The envi-
ronmental issues include:
Global and transnational (climate change,
ozone layer depletion).
Natural habitats (forests and other
ecosystems).
Land (agricultural zones).
Water resources (river basin, aquifer, water
shed).
109
Urban-industrial (metropolitan area,
air-shed).
Environmental sustainability depends on sev-
eral factors, including:
Climate change (magnitude and frequency
of shocks).
Systems vulnerability (extent of impact
damage).
System resilience (ability to recover from
impacts).
Economic importance of environmental issue
is increasing, and new technologies are expected
to reduce pollution derived both from productive
processes and products, with costs that are still
unknown. This is due to market uncertainty, weak
appropriability regime, lack of a dominant design,
and difficulties in reconfiguring organisational
routines. The degradation of the global environ-
ment is one of the most serious energy issues.
Various options are proposed and investigated to
mitigate climate change, acid rain or other envi-
ronmental problems. Additionally, the following
aspects play a fundamental role in developing
environmental technologies, pointing out how
technological trajectories depend both on exog-
enous market conditions and endogenous firm
competencies:
1. Formulating regulations concerning intro-
duction of zero emission vehicles (ZEV),
create market demand and business develop-
ment for new technologies.
2. Each stage of technology development
requires alternative forms of division and
coordination of innovative labour, upstream
and downstream industries are involved in
new forms of inter-firm relationships, caus-
ing a reconfiguration of product architectures
and reducing effects of path dependency.
3. Product differentiation increases firm capa-
bilities to plan at the same time technology
reduction and customer selection, while
meeting requirements concerning network
externalities.
4. It is necessary to find and/or create alter-
native funding sources for each research,
development and design stage of the new
technologies.
Action areas for producers:
Management and measurement tools-
adopting environmental management sys-
tems appropriate for the business.
Performance assessment tools- making use
of benchmarking to identify scope for im-
pact reduction and greater eco-effciency
in all aspects of the business.
Table 5. Classifications of data requirements (Omer, 2008b)
Plant data System data
Existing data Size
Life
Cost (fixed and variation
Operation and Maintenance)
Forced outage
Maintenance
Efficiency
Fuel
Emissions
Peak load
Load shape
Capital costs
Fuel costs
Depreciation
Rate of return
Taxes
Future data All of above, plus
Capital costs
Construction trajectory
Date in service
System lead growth
Fuel price growth
Fuel import limits
Inflation
110
Best practice tools- making use of free help
and advice from government best practice
programmes (energy effciency, environ-
mental technology, resource savings).
Innovation and ecodesign- rethinking the
delivery of value added by the business,
so that impact reduction and resource ef-
fciency are frmly built in at the design
stage.
Cleaner, leaner production processes- pur-
suing improvements and savings in waste
minimisation, energy and water consump-
tion, transport and distribution, as well as
reduced emissions. Tables (6-8) indicate
energy conservation, sustainable develop-
ment and environment.
Supply chain management- specifying
more demanding standards of sustainabil-
ity from upstream suppliers, while sup-
porting smaller frms to meet those higher
standards.
Product stewardship- taking the broad-
est view of producer responsibility and
working to reduce all the downstream ef-
fects of products after they have been sold
on to customers.
Openness and transparency- publicly re-
porting on environmental performance
against meaningful targets; actively us-
ing clear labels and declarations so that
customers are fully informed; building
stakeholder confdence by communicat-
ing sustainability aims to the workforce,
the shareholders and the local community
(Figure 6 and Table 9).
With the debate on climate change, the prefer-
ence for real measured data has been changed.
The analyses of climate scenarios need an hourly
weather data series that allows for realistic
changes in various weather parameters (REN21,
2007). By adapting parameters in a proper way,
Table 6. Classification of key variables defining facility sustainability
Criteria Intra-system impacts Extra-system impacts
Stakeholder satisfaction Standard expectations met.
Relative importance of standard expectations.
Covered by attending to extra-system resource base
and ecosystem impacts
Resource base impacts Change in intra-system resource bases.
Significance of change.
Resource flow into/out of facility system.
Unit impact exerted by flow on source/sink system.
Significance of unit impact.
Ecosystem impacts Change in intra-system ecosystems.
Significance of change.
Resource flows into/out of facility system.
Unit impact exerted by how on source/sink system.
Significance of unit impact.
Table 7. Energy and sustainable environment
Technological criteria Energy and environment criteria Social and economic criteria
Primary energy saving in regional scale Sustainability according to greenhouse
gas pollutant emissions
Labour impact
Technical maturity, reliability Sustainable according to other pollutant
emissions
Market maturity
Consistence of installation and maintenance re-
quirements with local technical known-how
Land requirement Compatibility with political, legislative and
administrative situation
Continuity and predictability of performance Sustainability according to other envi-
ronmental impacts
Cost of saved primary energy
111
Table 8. Positive impact of durability, adaptability and energy conservation on economic, social and
environment systems
Economic system Social system Environmental system
Durability Preservation of cultural values Preservation of resources
Meeting changing needs of economic de-
velopment
Meeting changing needs of individuals and
society
Reuse, recycling and preservation of resources
Energy conservation and saving Savings directed to meet other social needs Preservation of resources, reduction of pol-
lution and global warming
Figure 6. Link between resources and productivity
Table 9. The basket of indicators for sustainable consumption and production
Economy-wide decoupling indicators
1. Greenhouse gas emissions.
2. Air pollution.
3. Water pollution (river water quality).
4. Commercial and industrial waste arisings and household waste not cycled.
Resource use indicators
5. Material use.
6. Water extraction.
7. Homes built on land not previously developed, and number of households.
Decoupling indicators for specific sectors
8. Emissions from electricity generation.
9. Motor vehicle kilometres and related emissions
10. Agricultural output, fertiliser use, methane emissions and farmland bird populations.
11. Manufacturing output, energy consumption and related emissions..
12. Household consumption, expenditure energy, water consumption and waste generated
112
data series can be generated for the site. Weather
generators should be useful for:
Calculation of energy consumption (no ex-
treme conditions are required)
Design purposes (extremes are essential),
and
Predicting the effect of climate change
such as increasing annually average of
temperature.
This results in the following requirements:
Relevant climate variables should be gen-
erated (solar radiation: global, diffuse, di-
rect solar direction, temperature, humidity,
wind speed and direction) according to the
statistics of the real climate.
The average behaviour should be in accor-
dance with the real climate.
Extremes should occur in the generated
series in the way it will happen in a real
warm period. This means that the gener-
ated series should be long enough to assure
these extremes, and series based on aver-
age values from nearby stations.
Growing concerns about social and environ-
mental sustainability have led to increased interest
in planning for the energy utility sector because of
its large resource requirements and production of
emissions (Roriz, 2001). A number of conflicting
trends combine to make the energy sector a major
concern, even though a clear definition of how to
measure progress toward sustainability is lacking.
These trends include imminent competition in the
electricity industry, global climate change, expect-
ed long-term growth in population and pressure to
balance living standards (including per capital en-
ergy consumption). Designing and implementing
a sustainable energy sector will be a key element
of defining and creating a sustainable society. In
the electricity industry, the question of strategic
planning for sustainability seems to conflict with
the shorter time horizons associated with market
forces as deregulation replaces vertical integration.
Sustainable low-carbon energy scenarios for the
new century emphasise the untapped potential
of renewable resources. Rural areas can benefit
from this transition. The increased availability of
reliable and efficient energy services stimulates
new development alternatives. It is concluded
that renewable environmentally friendly energy
must be encouraged, promoted, implemented, and
demonstrated by full-scale plant especially for use
in remote rural areas (CEC, 2000).
This is the step in a long journey to encourage a
progressive economy, which continues to provide
us with high living standards, but at the same time
helps reduce pollution, waste mountains, other
environmental degradation, and environmental
rationale for future policy-making and interven-
tion to improve market mechanisms. This vision
will be accomplished by:
Decoupling economic growth and envi-
ronmental degradation. The basket of indi-
cators illustrated shows the progress being
made. Decoupling air and water pollution
from growth, making good headway with
CO
2
emissions from energy, and transport.
The environmental impact of our own in-
dividual behaviour is more closely linked
to consumption expenditure than the econ-
omy as a whole.
Focusing policy on the most important en-
vironmental impacts associated with the
use of particular resources, rather than on
the total level of all resource use.
Increasing the productivity of material and
energy use that are economically effcient
by encouraging patterns of supply and de-
mand, which are more effcient in the use
of natural resources (the aim is to promote
innovation and competitiveness) and in-
vesting in areas such as energy effciency,
water effciency and waste minimisation.
113
Encouraging and enabling active and
informed individual and corporate
consumers.
On some climate change issues (such as global
warming), there is no disagreement among the
scientists. The greenhouse effect is unquestion-
ably real; it is essential for life on earth. Water
vapour is the most important GHG; next is carbon
dioxide (CO
2
). Without a natural greenhouse ef-
fect, scientists estimate that the earths average
temperature would be 18
o
C instead of its present
14
o
C. There is also no scientific debate over the fact
that human activity has increased the concentration
of the GHGs in the atmosphere (especially CO
2

from combustion of coal, oil and gas) (Leszek and
Jakub, 2009). The greenhouse effect is also being
amplified by increased concentrations of other
gases, such as methane, nitrous oxide, and CFCs
as a result of human emissions. Most scientists
predict that rising global temperatures will raise
the sea level and increase the frequency of intense
rain or snowstorms. Climate change scenarios
sources of uncertainty and factors influencing the
future climate are:
The future emission rates of the GHGs.
The effect of this increase in concentration
on the energy balance of the atmosphere.
The effect of these emissions on GHGs
concentrations in the atmosphere, and
The effect of this change in energy balance
on global and regional climate.
Wastes
Waste is defined as an unwanted material that
is being discarded. Waste includes items being
taken for further use, recycling or reclamation.
Waste produced at household, commercial and
industrial premises are control waste and come
under the waste regulations. Waste Incineration
Directive (WID) emissions limit values will favour
efficient, inherently cleaner technologies that do
not rely heavily on abatement. For existing plant,
the requirements are likely to lead to improved
control of:
NO
x
emissions, by the adoption of infur-
nace combustion control and abatement
techniques.
Acid gases, by the adoption of abate-
ment techniques and optimisation of their
control.
Particulate control techniques, and their
optimisation, e.g., of bag flters and elec-
trostatic precipitators.
Lifecycle analysis of several ethanol feedstocks
shows the emissions per ton of feedstock are high-
est for corn stover and switchgrass (about 0.65
tons of CO
2
per ton of feedstock) and lowest for
corn (about 0.5 ton). Emissions due to cultivation
and harvesting of corn and wheat are higher than
those for lignocellulosics, and although the lat-
ter have a far higher process energy requirement
(Figure 8). GHG emissions are lower because this
energy is produced from biomass residue, which
is carbon neutral.
The waste and resources action programme
has been working hard to reduce demand for
virgin aggregates and market uptake of recycled
and secondary alternatives (Figure 7). The pro-
gramme targets are:
To deliver training and information on the
role of recycling and secondary aggregates
in sustainable construction for infuences
in the supply chain, and
To develop a promotional programme to
highlight the new information on websites.
Sulphur in Fuels and Its
Environmental Consequences
Organic sulphur is bonded within the organic
structure of the coal in the same way that sulphur
114
is bonded in simple thio-organics, e.g., thiols.
Sulphur contents of coals vary widely, and Table
10 gives some examples.
Control of SO
2
Emissions
Emissions will also, of course, occur from
petroleum-based or shale-based fuels, and in
heavy consumption, such as in steam raising.
There will frequently be a need to control SO
2

emissions. There are, broadly speaking, three
ways of achieving such control:
Pre-combustion control: involves carrying
out a degree of desulphurisation of the fuel.
Combustion control: incorporating into the
combustion system something capable of
trapping SO
2
.
Post-combustion control: removing SO
2

from the fue gases before they are dis-
charged into the atmosphere.
Table 11 gives brief details of an example of
each.
The Control of NO
x
Release by
Combustion Processes
Emission of nitrogen oxides is a major topic in
fuel technology. It has to be considered even in the
total absence of fuel nitrogen if the temperature
is high enough for thermal NO
x
, as it is in very
many industrial applications. The burnt gas from
the flame is recirculated in two ways:
Internally, by baffing and restricting fow
of the burnt gas away from the burner, re-
sulting in fame re-entry of part of it.
Externally, by diverting up to 10% of the
fue gas back into the fame.
Particles
Some of the available control procedures for
particles are summarised in Table 12. Figure 9
shows the variation of distribution factor with
particle size.
Figure 7. Comparison of thermal biomass usage options, CHP displacing natural gas as a heat source
(Omer, 2009) Billion tones (Bton) Scenarios are (1) household, (2) commercial, (3) agriculture and (4)
industrial
Table 10. Representative sulphur contents of coals
(Meffe et al, 1996)
Source Rank Sulphur content
(%)
Ayrshire, Scotland
Lancs. /Cheshire, UK
S. Wales, UK
Victoria, Australia
Pennsylvania, USA
Natal, S. Africa
Bulgaria
Bituminous
Bituminous
Anthracite
Lignite
Anthracite
Bituminous
Lignite
0.6
Up to 2.4
Up to 1.5
Typically 0.5
0.7
Up to 4.2
2.5
115
Figure 8. The lifecycle energy balance of corn and Switchgrass reveal a paradox: corn, as an ethanol
feedstock requires less energy for production, i.e., more of the original energy in starch is retained in
the ethanol fuel. Nevertheless, the Switchgrass process yields higher GHG emissions. This is because
most of the process energy for Switchgrass process is generated from biomass residue.
116
Figure 9. The variation of distribution factor against particle size for coal undersizes in a classifier. The
sizes correspond to mid-point for ranges (Omer, 2008c)
Table 11. Examples of SO
2
control procedures (Meffe et al, 1996)
Type of control Fuel Details
Pre-combustion Fuels from crude oil Alkali treatment of crude oil to convert thiols, RSSR, disulphides; solvent
removal of the disulphides
Post-combustion Coal or fuel oil Alkali scrubbing of the flue gases with CaCO
3
/CaO
Combustion Coal Limestone, MgCO
3
and/or other metallic compounds used to fix the
sulphur as sulphates
Table 12. Particle control techniques (Meffe et al, 1996)
Technique Principle Application
Gravity settlement Natural deposition by gravity of particles from
a horizontally flowing gas, collection in hoppers
Removal of coarse particles (>50 m) from
a gas stream, smaller particles removable
in principle but require excessive flow
distances
Cyclone separator Tangential entry of a particle-laden gas into
a cylindrical or conical enclosure, movement
of the particles to the enclosure wall and from
there to a receiver
Numerous applications, wide range of
particles sizes removable, from = 5 m to
= 200 m, poorer efficiencies of collection
for the smaller particles
Fabric filters Retention of solids by a filter, filter materials in-
clude woven cloth, felt and porous membranes
Used in dust removal for over a century
Electrostatic precipitation Passage of particle-laden gas between elec-
trodes, application of an electric field to the
gas, resulting in acquisition of charge by the
particles and attraction to an electrode where
coalescence occurs, electrical resistivity of the
dust an important factor in performance
Particles down to 0.01 m removable, ex-
tensive application to the removal of flyash
from pulverised fuel (pf) combustion
117
Ground Source Heat Pumps
The ground is as universal as air and solar radiation
(Mortal, 2002). Over the past twenty years, as the
hunt for natural low-carbon energy sources has
intensified, there has been an increased endeavour
to investigate and develop both earth and ground
water thermal energy storage and usage (Lund,
Freeston, and Boyd, 2005). Geothermal energy
solutions, although well known, are another in
our armoury of renewable energy sources that are
within our immediate grasp to use and integrate
with an overall energy policy (Huttrer, 2001).
For high temperature heat storage with tem-
peratures in excess of 50
o
C the particular concerns
were:
Clogging of wells and heat exchangers due
to fnes and precipitation of minerals.
Water treatment to avoid operational prob-
lems resulting from the precipitation of
minerals.
Corrosion of components in the groundwa-
ter system.
Automatic control of the ground water
system.
Three main techniques that are used to ex-
ploit the heat available are geothermal aquifers,
hot dry rocks and GSHPs. Geothermal energy is
the natural heat that exists within the earth and
that can be absorbed by fluids occurring within,
or introduced into the crystal rocks. Heat pump
technology can be used for heating only, or for
cooling only, or be reversible and used for heat-
ing and cooling depending on the demand. More
generally, there is still potential for improvement
in the performance of heat pumps.
As consumers in less-developed countries
increase their capacity of electricity and green
power, developed nations are starting to realise
the benefits of using low-grade thermal energy
for green heat applications that do not require
high-grade electricity. This shift will not only
benefit renewable energies that are designed for
space conditioning, but also will contribute to
the global mix of green power and green heat
capacity. Earth energy (also called geothermal
or ground source heat pumps or GeoExchange),
which transfers absorbed solar heat from the
ground into a building for space heating or water
heating. The same system can be reversed to reject
heat from the interior into the ground in order to
provide cooling. A typical configuration buries
polyethylene pipe below the frost line to serve as
the head source (or sink), or it can use lake water
and aquifers as the heat medium (Omer, 2008c).
Effects of Urban Density
Compact development patterns of buildings can
reduce infrastructure demands and the need to
travel by car. As population density increases,
transportation options multiply and dependence
areas, per capita fuel consumption is much lower
in densely populated areas because people drive
much less. Few roads and commercially viable
public transport are the major merits. On the other
hand, urban density is a major factor that deter-
mines the urban ventilation conditions, as well as
the urban temperature. Under given circumstances,
an urban area with a high density of buildings
can experience poor ventilation and strong heat
island effect. In warm-humid regions these fea-
tures would lead to a high level of thermal stress
of the inhabitants and increased use of energy in
air-conditioned buildings (Reddy, Williams, and
Johansson, 2007).
However, it is also possible that a high-density
urban area, obtained by a mixture of high and low
buildings, could have better ventilation condi-
tions than an area with lower density but with
buildings of the same height. Closely spaced or
high-rise buildings are also affected by the use
of natural lighting, natural ventilation and solar
energy. If not properly planned, energy for elec-
tric lighting and mechanical cooling/ventilation
may be increased and application of solar energy
118
systems will be greatly limited. Table 13 gives a
summary of the positive and negative effects of
urban density. All in all, denser city models require
more careful design in order to maximise energy
efficiency and satisfy other social and develop-
ment requirements. Low energy design should
not be considered in isolation, and in fact, it is
a measure, which should work in harmony with
other environmental objectives. Hence, building
energy study provides opportunities not only for
identifying energy and cost savings, but also for
examining the indoor and outdoor environment
(Aroyeun, et al., 2009).
Energy Efficiency and Architectural
Expression
The focus of the worlds attention on environmen-
tal issues in recent years has stimulated response
in many countries, which have led to a closer
examination of energy conservation strategies
for conventional fossil fuels. Buildings are im-
portant consumers of energy and thus important
contributors to emissions of greenhouse gases
into the global atmosphere. The development and
adoption of suitable renewable energy technol-
ogy in buildings has an important role to play.
A review of options indicates benefits and some
problems (BS, 1989). There are two key elements
to the fulfilling of renewable energy technology
potential within the field of building design; first
the installation of appropriate skills and attitudes
in building design professionals and second the
provision of the opportunity for such people to
demonstrate their skills. This second element may
only be created when the population at large and
clients commissioning building design in particu-
lar, become more aware of what can be achieved
and what resources are required. Terms like passive
cooling or passive solar use mean that the cooling
of a building or the exploitation of the energy of
the sun is achieved not by machines but by the
buildings particular morphological organisation
((EFC, 2000). Hence, the passive approach to
themes of energy savings is essentially based on
the morphological articulations of the construc-
tions. Passive solar design, in particular, can realize
significant energy and cost savings. For a design to
be successful, it is crucial for the designer to have a
good understanding of the use of the building. Few
of the buildings had performed as expected by their
designers. To be more precise, their performance
Table 13. Effects of urban density on citys energy demand
Positive effects Negative effects
Transport:
Promote public transport and reduce the need for, and length of, trips by
private cars.
Infrastructure:
Reduce street length needed to accommodate a given number of inhabit-
ants.
Shorten the length of infrastructure facilities such as water supply and
sewage lines, reducing the energy needed for pumping.
Thermal performance:
Multi-story, multiunit buildings could reduce the overall area of the build-
ings envelope and heat loss from the buildings.
Shading among buildings could reduce solar exposure of buildings during
the summer period.
Energy systems:
District cooling and heating system, which is usually more energy ef-
ficiency, is more feasible as density is higher.
Ventilation:
A desirable flow pattern around buildings may be obtained by proper ar-
rangement of high-rise building blocks.
Transport:
Congestion in urban areas reduces fuel efficiency of
vehicles.
Vertical transportation:
High-rise buildings involve lifts, thus increasing the need
for electricity for the vertical transportation.
Ventilation:
A concentration of high-rise and large buildings may
impede the urban ventilation conditions.
Urban heat island:
Heat released and trapped in the urban areas may in-
crease the need for air conditioning.
The potential for natural lighting is generally reduced in
high-density areas, increasing the need for electric lighting
and the load on air conditioning to remove the heat result-
ing from the electric lighting.
Use of solar energy:
Roof and exposed areas for collection of solar energy are
limited.
119
had been compromised by a variety of influences
related to their design, construction and opera-
tion. However, there is no doubt that the passive
energy approach is certainly the one that, being
supported by the material shape of the buildings
has a direct influence on architectural language
and most greatly influences architectural expres-
siveness (Lazzarin et al, 2002). Furthermore, form
is a main tool in architectural expression. To give
form to the material things that one produces is
an ineluctable necessity. In architecture, form,
in fact, summarises and gives concreteness to
its every value in terms of economy, aesthetics,
functionality and, consequently, energy efficiency
(David, 2003). The target is to enrich the expres-
sive message with forms producing an advantage
energy-wise. Hence, form, in its geometric and
material sense, conditions the energy efficiency
of a building in its interaction with the environ-
ment. It is, then, very hard to extract and separate
the parameters and the elements relative to this
efficiency from the expressive unit to which they
belong. By analysing energy issues and strategies
by means of the designs, of which they are an inte-
gral part, one will, more easily, focus the attention
on the relationship between these themes, their
specific context and their architectural expres-
siveness. Many concrete examples and a whole
literature have recently grown up around these
subjects and the wisdom of forms and expedients
that belong to millennia-old traditions has been
rediscovered. Such a revisiting, however, is only,
or most especially, conceptual, since it must be
filtered through todays technology and needs;
both being almost irreconcilable with those of the
past. Two among the historical concepts are of
special importance. One is rooted in the effort to
establish rational and friendly strategic relations
with the physical environment, while the other
recognises the interactions between the psyche and
physical perceptions in the creation of the feeling
of comfort. The former, which may be defined
as an alliance with the environment deals with
the physical parameters involving a mixture of
natural and artificial ingredients such as soil and
vegetation, urban fabrics and pollution (Zuatori,
2005). The most dominant outside parameter is, of
course, the suns irradiation, our planets primary
energy source. All these elements can be measured
in physical terms and are therefore the subject
of science. Within the second concept, however,
one considers the emotional and intellectual en-
ergies, which are the prime inexhaustible source
of renewable power (Anne et al, 2005). In this
case, cultural parameters, which are not exactly
measurable, are involved. However, they repre-
sent the very essence of the architectural quality.
Objective scientific measurement parameters tell
us very little about the emotional way of perceiv-
ing, which influences the messages of human are
physical sensorial organs. The perceptual reality
arises from a multitude of sensorial components;
visual, thermal, acoustic, olfactory and kinaesthet-
ics. It can also arise from the organisational quality
of the space in which different parameters come
together, like the sense of order or of serenity.
Likewise, practical evaluations, such as useful-
ness, can be involved too. The evaluation is a
wholly subjective matter, but can be shared by a
set of experiencing persons (Randal et al, 1998).
Therefore, these cultural parameters could be dif-
ferent in different contexts in spite of the inexorable
levelling on a planet- wide scale. However, the
parameters change in the anthropological sense,
not only with the cultural environment, but also in
relation to function. The scientifically measurable
parameters can, thus, have their meanings very
profoundly altered by the non-measurable, but
describable, cultural parameters.
However, the low energy target also means to
eliminate any excess in the quantities of material
and in the manufacturing process necessary for the
construction of our built environment. This claims
for a more sober, elegant and essential expression,
which is not jeopardising at all, but instead enhanc-
ing, the richness and preciousness of architecture,
while contributing to a better environment from
an aesthetic viewpoint (Yadav et al, 1997). Argu-
120
ably, the most successful designs were in fact the
simplest. Paying attention to orientation, plan
and form can have far greater impact on energy
performance than opting for elaborate solutions
(EIBI, 1999). However, a design strategy can fail
when those responsible for specifying materials
for example, do not implement the passive solar
strategy correctly. Similarly, cost-cutting exercises
can seriously upset the effectiveness of a design
strategy. Therefore, it is imperative that a designer
fully informs key personnel, such as the quantity
surveyor and client, about their design and be
prepared to defend it. Therefore, the designer
should have an adequate understanding of how
the occupants or processes, such as ventilation,
would function within the building (Lam, 2000).
Thinking through such processes in isolation
without reference to others can lead to conflicting
strategies, which can have a detrimental impact
upon performance. Likewise, if the design intent of
the building is not communicated to its occupants,
there is a risk that they will use it inappropriately,
thus, compromising its performance. Hence, the
designer should communicate in simple terms
the actions expected of the occupant to control
the building. For example, occupants should be
well informed about how to guard against summer
overheating. If the designer opted for a simple,
seasonally adjusted control; say, insulated sliding
doors were to be used between the mass wall and
the internal space. The lesson here is that designers
must be prepared to defend their design such that
others appreciate the importance and interrelation-
ship of each component (IEA, 2008). A strategy
will only work if each individual component is
considered as part of the bigger picture. Failure to
implement a component or incorrect installation,
for example, can lead to failure of the strategy
and consequently, in some instances, the build-
ing may not liked by its occupants due to its poor
performance.
Energy Efficiency
Energy efficiency is the most cost-effective way
of cutting carbon dioxide emissions and improve-
ments to households and businesses. It can also
have many other additional social, economic and
health benefits, such as warmer and healthier
homes, lower fuel bills and company running costs
and, indirectly, jobs. Britain wastes 20 per cent
of its fossil fuel and electricity use (Witte, et al.,
2002). This implies that it would be cost-effective
to cut 10 billion a year off the collective fuel bill
and reduce CO
2
emissions by some 120 million
tones. Yet, due to lack of good information and
advice on energy saving, along with the capital
to finance energy efficiency improvements, this
huge potential for reducing energy demand is not
being realised (Paul, 2001). Traditionally, energy
utilities have been essentially fuel providers and
the industry has pursued profits from increased
volume of sales. Institutional and market arrange-
ments have favoured energy consumption rather
than conservation. However, energy is at the centre
of the sustainable development paradigm as few
activities affect the environment as much as the
continually increasing use of energy. In addition,
more than three quarters of the worlds consump-
tion of these fuels is used, often inefficiently,
by only one quarter of the worlds population.
Without even addressing these inequities or the
precious, finite nature of these resources, the scale
of environmental damage will force the reduction
of the usage of these fuels long before they run
out (WB, 2003b).
Throughout the energy generation process
there are impacts on the environment on local,
national and international levels, from opencast
mining and oil exploration to emissions of the
potent greenhouse gas carbon dioxide in ever
increasing concentration. Recently, the worlds
leading climate scientists reached an agreement
that human activities, such as burning fossil
fuels for energy and transport, are causing the
temperature to rise. The Intergovernmental Panel
121
on Climate Change has concluded that the bal-
ance of evidence suggests a discernible human
influence on global climate. It predicts a rate
of warming greater than any one seen in the last
10,000 years, in other words, throughout human
history. The exact impact of climate change is
difficult to predict and will vary regionally. It
could, however, include sea level rise, disrupted
agriculture and food supplies and the possibility
of more freak weather events such as hurricanes
and droughts. Indeed, people already are wak-
ing up to the financial and social, as well as the
environmental, risks of unsustainable energy
generation methods that represent the costs of the
impacts of climate change, acid rain and oil spills.
The insurance industry, for example, concerned
about the billion dollar costs of hurricanes and
floods, has joined sides with environmentalists
to lobby for greenhouse gas emissions reduction.
Friends of the earth are campaigning for a more
sustainable energy policy, guided by the principle
of environmental protection and with the objec-
tives of sound natural resource management and
long-term energy security. The key priorities of
such an energy policy must be to reduce fossil fuel
use, move away from nuclear power, improve the
efficiency with which energy is used and increase
the amount of energy obtainable from sustainable,
renewable sources (WB, 2004). Efficient energy
use has never been more crucial than it is today,
particularly with the prospect of the imminent
introduction of the climate change levy (CCL).
Establishing an energy use action plan is the es-
sential foundation to the elimination of energy
waste. A logical starting point is to carry out an
energy audit that enables the assessment of the
energy use and determine what actions to take. The
actions are best categorised by splitting measures
into the following three general groups:
1. High priority/low cost:
These are normally measures, which require
minimal investment and can be implemented
quickly. The followings are some examples of
such measures:
A. Good housekeeping, monitoring energy use
and targeting waste-fuel practices.
B. Adjusting controls to match requirements.
C. Improved greenhouse space utilisation.
D. Small capital item time switches, thermo-
stats, etc.
E. Carrying out minor maintenance and repairs.
F. Staff education and training.
G. Ensuring that energy is being purchased
through the most suitable tariff or contract
arrangements.
2. Medium priority/medium cost:
Measures, which, although involve little or no
design, involve greater expenditure and can take
longer to implement. Examples of such measures
are listed below:
A. New or replacement controls.
B. Greenhouse component alteration, e.g.,
insulation, sealing glass joints, etc.
C. Alternative equipment components, e.g.,
energy efficient lamps in light fittings, etc.
3. Long term/high cost:
These measures require detailed study and design.
They can be best represented by the followings:
A. Replacing or upgrading of plant and
equipment.
B. Fundamental redesign of systems, e.g., com-
bined heat and power (CHP) installations.
This process can often be a complex experience
and therefore the most cost-effective approach is
to employ an energy specialist to help.
122
Policy Recommendations for a
Sustainable Energy Future
Sustainability is regarded as a major consideration
for both urban and rural development. People
have been exploiting the natural resources with
no consideration to the effects, both short-term
(environmental) and long-term (resources crunch).
It is also felt that knowledge and technology have
not been used effectively in utilising energy re-
sources (Mildred, and Trevor, 2009). Energy is the
vital input for economic and social development
of any country. Its sustainability is an important
factor to be considered. The urban areas depend,
to a large extent, on commercial energy sources.
The rural areas use non-commercial sources like
firewood and agricultural wastes (WB, 2003a).
Sustainability is regarded as a major con-
sideration for both urban and rural develop-
ment. People have been exploiting the natural
resources with no consideration to the effects,
both short-term (environmental) and long-term
(resources crunch). It is also felt that knowledge
and technology have not been used effectively
in utilising energy resources. Energy is the vital
input for economic and social development of
any country. Its sustainability is an important
factor to be considered. The urban areas depend,
to a large extent, on commercial energy sources.
The rural areas use non-commercial sources like
firewood and agricultural wastes. With the present
day trends for improving the quality of life and
sustenance of mankind, environmental issues are
considered highly important (Felice and Alessio,
2010). In this context, the term energy loss has
no significant technical meaning. Instead, the
exergy loss has to be considered, as destruction
of exergy is possible. Hence, exergy loss mini-
misation will help in sustainability. In the process
of developing, there are two options to manage
energy resources: (1) End use matching/demand
side management, which focuses on the utilities.
The mode of obtaining this is based on economic
terms. It is, therefore, a quantitative approach. (2)
Supply side management, which focuses on the
renewable energy resource and methods of utiliz-
ing it. This is decided based on thermodynamic
consideration having the resource-user tempera-
ture or exergy destruction as the objective criteria.
It is, therefore, a qualitative approach. The two
options are explained schematically in Figure 10.
The exergy-based energy, developed with supply
side perspective is shown in Figure 11.
The following policy measures had been iden-
tified:
Clear environmental and social objectives
for energy market liberalisation, includ-
ing a commitment to energy effciency and
renewables.
Economic, institutional and regulatory
frameworks, which encourage the transi-
tion to total energy services.
Economic measures to encourage utility
investment in energy effciency (e.g., lev-
ies on fuel bills).
Incentives for demand side management,
including grants for low-income house-
holds, expert advice and training, stan-
dards for appliances and buildings and tax
incentives.
Research and development funding for re-
newable energy technologies not yet com-
mercially viable.
Continued institutional support for new re-
newables (such as standard cost-refective
payments and obligation on utilities to
buy).
Ecological tax reform to internalise exter-
nal environmental and social costs within
energy prices.
Planning for sensitive development and
public acceptability for renewable energy.
Energy resources are needed for societal devel-
opment. Their sustainable development requires
123
a supply of energy resources that are sustainably
available at a reasonable cost and can cause no
negative societal impacts. Energy resources such
as fossil fuels are finite and lack sustainability,
while renewable energy sources are sustainable
over a relatively longer term. Environmental
concerns are also a major factor in sustainable
development, as activities, which degrade the
environment, are not sustainable (Jeremy, 2005).
Hence, as much as environmental impact is as-
sociated with energy, sustainable development
requires the use of energy resources, which cause
as little environmental impact as possible. One
way to reduce the resource depletion associated
with cycling is to reduce the losses that accompany
the transfer of exergy to consume resources by in-
creasing the efficiency of exergy transfer between
resources i.e. increasing the fraction of exergy
Figure 11. Exergy based optimal energy model
Figure 10. Supply side and demand side management approach for energy
124
removed from one resource that is transferred to
another (Erlich, 1991).
As explained above, exergy efficiency may
be thought of as a more accurate measure of
energy efficiency that accounts for quantity and
quality aspects of energy flows. Improved exergy
efficiency leads to reduced exergy losses (WB,
2006). Most efficiency improvements produce
direct environmental benefits in two ways. First,
operating energy input requirements are reduced
per unit output, and pollutants generated are
correspondingly reduced. Second, consideration
of the entire life cycle for energy resources and
technologies suggests that improved efficiency
reduces environmental impact during most stages
of the life cycle (Dragana, 2008). Quite often, the
main concept of sustainability, which often inspires
local and national authorities to incorporate en-
vironmental consideration into setting up energy
programmes have different meanings in different
contexts though it usually embodies a long-term
perspective (WB, 2007). Future energy systems
will largely be shaped by broad and powerful
trends that have their roots in basic human needs.
Combined with increasing world population, the
need will become more apparent for success-
ful implementation of sustainable development
(White, and Robinson, 2008).
Heat has a lower exergy, or quality of energy,
compared with work. Therefore, heat cannot be
converted into work by 100% efficiency. Some
examples of the difference between energy and
exergy are shown in Table 14.
The terms used in Table 14 have the following
meanings:
Carnot Quality Factor (CQF) = (1-T
o
/T
s
) (5)
Exergy = Energy (transferred) x CQF (6)
Where T
o
is the environment temperature (K)
and T
s
is the temperature of the stream (K).
Various parameters are essential to achieving
sustainable development in a society. Some of
them are as follows:
Public awareness
Information
Environmental education and training
Innovative energy strategies
Renewable energy sources and cleaner
technologies
Financing
Monitoring and evaluation tools
The development of a renewable energy in a
country depends on many factors. Those important
to success are listed below:
1. Motivation of the population
The population should be motivated towards
awareness of high environmental issues, rational
use of energy in order to reduce cost. Subsidy
programme should be implemented as incentives
to install renewable energy plants. In addition,
image campaigns to raise awareness of renew-
able technology.
2. Technical product development
Table 14. Qualities of various energy sources (Omer, 2008a)
Source Energy (J) Exergy (J) CQF
Water at 80
o
C 100 16 0.16
Steam at 120
o
C 100 24 0.24
Natural gas 100 99 0.99
Electricity/work 100 100 1.00
125
To achieve technical development of renewable
energy technologies the following should be ad-
dressed:
Increasing the longevity and reliability of
renewable technology.
Adapting renewable technology to house-
hold technology (hot water supply).
Integration of renewable technology in
heating technology.
Integration of renewable technology in ar-
chitecture, e.g., in the roof or faade.
Development of new applications, e.g., so-
lar cooling.
Cost reduction.
3. Distribution and sales
Commercialisation of renewable energy technol-
ogy requires:
Inclusion of renewable technology in the
product range of heating trades at all levels
of the distribution process (wholesale, and
retail).
Building distribution nets for renewable
technology.
Training of personnel in distribution and
sales.
Training of feld sales force.
4. Consumer consultation and installation
To encourage all sectors of the population to
participate in adoption of renewable energy tech-
nologies, the following has to be realised:
Acceptance by craftspeople, marketing by
them.
Technical training of craftspeople, initial
and follow-up training programmes.
Sales training for craftspeople.
Information material to be made available
to craftspeople for consumer consultation.
5. Projecting and planning
Successful application of renewable technologies
also requires:
Acceptance by decision makers in the
building sector (architects, house technol-
ogy planners, etc.).
Integration of renewable technology in
training.
Demonstration projects/architecture
competitions.
Renewable energy project developers
should prepare to participate in the carbon
market by:
Ensuring that renewable energy proj-
ects comply with Kyoto Protocol
requirements.
Quantifying the expected avoided
emissions.
Registering the project with the re-
quired offces.
Contractually allocating the right to
this revenue stream.
Other ecological measures employed on
the development include:
Simplifed building details.
Reduced number of materials.
Materials that can be recycled or
reused.
Materials easily maintained and
repaired
Materials that do not have a bad in-
fuence on the indoor climate (i.e.,
non-toxic).
Local cleaning of grey water.
Collecting and use of rainwater for
outdoor purposes and park elements.
Building volumes designed to give
maximum access to neighbouring
park areas.
All apartments have visual access to
both backyard and park.
126
6. Energy saving measures
The following energy saving measures should
also be considered:
Building integrated solar PV system.
Day-lighting.
Ecological insulation materials.
Natural/hybrid ventilation.
Passive cooling.
Passive solar heating.
Solar heating of domestic hot water.
Utilisation of rainwater for fushing.
Improving access for rural and urban low-
income areas in developing countries through
energy efficiency and renewable energies will be
needed. Sustainable energy is a prerequisite for
development. Energy-based living standards in
developing countries, however, are clearly below
standards in developed countries. Low levels of
access to affordable and environmentally sound
energy in both rural and urban low-income areas
are therefore a predominant issue in developing
countries. In recent years many programmes for
development aid or technical assistance have
been focusing on improving access to sustainable
energy, many of them with impressive results.
Apart from success stories, however, experi-
ence also shows that positive appraisals of many
projects evaporate after completion and vanishing
of the implementation expert team. Altogether,
the diffusion of sustainable technologies such
as energy efficiency and renewable energies for
cooking, heating, lighting, electrical appliances
and building insulation in developing countries
has been slow.
Energy efficiency and renewable energy
programmes could be more sustainable and pilot
studies more effective and pulse releasing if the
entire policy and implementation process was
considered and redesigned from the outset. New fi-
nancing and implementation processes are needed
which allow reallocating financial resources and
thus enabling countries themselves to achieve a
sustainable energy infrastructure. The links be-
tween the energy policy framework, financing and
implementation of renewable energy and energy
efficiency projects have to be strengthened and
capacity building efforts are required.
FUTURE RESEARCH DIRECTIONS
Energy constitutes the motive force of the civiliza-
tion and it determines, in a high degree, the level
of economy development as a whole. Despite the
increase use of different type of energy, particu-
larly, renewable energy sources, fossil fuels will
continue dominating the energy combinations in
the world near future. However, oil reserves are
declining and this situation would have a nega-
tive impact in the future economic development
of many countries all over the world.
Climate change issues, the reduced world re-
serves of fossils, and higher and higher fuel prices
play an important role in the development of clean
technologies, such as biohydrogen, biodiesel and
bioethanol, for producing renewable energy. This
research gathers and presents current research
from across the globe in the study of clean energy
resources, their production and developments.
In Asia, the import energy dependency is ris-
ing. Unless Europe can make domestic energy
more competitive in the next 20 to 30 years,
around 70% of the Asians energy requirements,
compared to 50% today, will be met by imported
products some of them from regions threatened
by insecurity. Now, the energy requirements of
the different countries are so high that, for the
first time in the humanitys history, there is a need
to consider different types of available energy
sources and their reserves to plan the economic
development of the countries. At the same time,
there is also a need to use these sources in the
most efficient possible manner in order to sustain
that development.
127
Sustainable energy is the provision of energy
such that it meets the needs of the present without
compromising the ability of future generations to
meet their needs. A broader interpretation may al-
low inclusion of fossil fuels and nuclear fission as
transitional sources while technology develops, as
long as new sources are developed for future gen-
erations to use. A narrower interpretation includes
only energy sources, which are not expected to be
depleted in a time frame relevant to the human race.
Sustainable energy sources are most often regarded
as including all renewable sources, such as biofuels,
solar power, wind power, wave power, geothermal
power and tidal power. It usually also includes tech-
nologies that improve energy efficiency. This new
and important handbook gathers the latest research
from around the globe in the study of sustainable
energy and highlights such topics as: monitoring
sustainable energy development; methane; energy
and territory; biodiesel production; electrochemi-
cal hydrogen storage; environmental policies in an
electricity sector and others).
The move towards a de-carbonised world,
driven partly by climate science and partly by the
business opportunities it offers, will need the pro-
motion of environmentally friendly alternatives,
if an acceptable stabilisation level of atmospheric
carbon dioxide is to be achieved. This requires the
harnessing and use of natural resources that produce
no air pollution or greenhouse gases and provides
comfortable coexistence of human, livestock,
and plants. This study reviews the energy-using
technologies based on natural resources, which are
available to and applicable in the farming industry.
Integral concept for buildings with both excel-
lent indoor environment control and sustainable
environmental impact are reported in the present
communication.
CONCLUSION
There is strong scientific evidence that the average
temperature of the earths surface is rising. This is
a result of the increased concentration of carbon
dioxide and other GHGs in the atmosphere as re-
leased by burning fossil fuels. This global warming
will eventually lead to substantial changes in the
worlds climate, which will, in turn, have a major
impact on human life and the built environment.
Therefore, effort has to be made to reduce fossil
energy use and to promote green energies, particu-
larly in the building sector. Energy use reductions
can be achieved by minimising the energy demand,
by rational energy use, by recovering heat and the
use of more green energies. This article was a step
towards achieving that goal. The adoption of green
or sustainable approaches to the way in which
society is run is seen as an important strategy in
finding a solution to the energy problem. The key
factors to reducing and controlling CO
2
, which is
the major contributor to global warming, are the
use of alternative approaches to energy generation
and the exploration of how these alternatives are
used today and may be used in the future as green
energy sources. Even with modest assumptions
about the availability of land, comprehensive
fuel-wood farming programmes offer significant
energy, economic and environmental benefits.
These benefits would be dispersed in rural areas
where they are greatly needed and can serve as
linkages for further rural economic development.
The nations as a whole would benefit from savings
in foreign exchange, improved energy security, and
socio-economic improvements. With a nine-fold
increase in forest plantation cover, a nations
resource base would be greatly improved. The
international community would benefit from
pollution reduction, climate mitigation, and the
increased trading opportunities that arise from
new income sources. The non-technical issues,
which have recently gained attention, include: (1)
Environmental and ecological factors e.g., carbon
sequestration, reforestation and revegetation. (2)
Renewables as a CO
2
neutral replacement for fossil
fuels. (3) Greater recognition of the importance
of renewable energy, particularly modern bio-
mass energy carriers, at the policy and planning
128
levels. (4) Greater recognition of the difficulties
of gathering good and reliable renewable energy
data, and efforts to improve it. (5) Studies on
the detrimental health efforts of biomass energy
particularly from traditional energy users.
ACKNOWLEDGMENT
A special thanks to my spouse Kawthar Abdelhai
Ali for her support and her unwavering faith in me.
Her intelligence, humour, spontaneity, curiosity
and wisdom added to this article.
REFERENCES
Abdeen, M. O. (2008a). Renewable building
energy systems and passive human comfort solu-
tions. Renewable & Sustainable Energy Reviews,
12(6), 15621587. doi:10.1016/j.rser.2006.07.010
Abdeen, M. O. (2008b). People, power and pollu-
tion. Renewable & Sustainable Energy Reviews,
12(7), 18641889. doi:10.1016/j.rser.2006.10.004
Abdeen, M. O. (2008c). Energy, environment
and sustainable development. Renewable &
Sustainable Energy Reviews, 12(9), 22652300.
doi:10.1016/j.rser.2007.05.001
Abdeen, M. O. (2008d). Focus on low carbon
technologies: The positive solution. Renewable &
doi:10.1016/j.rser.2007.04.015
Abdeen, M. O. (2008e). Development of inte-
grated bioenergy for improvement of quality of
life of poor people in developing countries. In
Magnusson, F. L., & Bengtsson, O. W. (Eds.),
Energy in Europe: Economics, policy and strategy
(pp. 341373). New York, NY: NOVA Science
Publishers.
Abdeen, M. O. (2009a). Environmental and
socio-economic aspect of possible development
in renewable energy use. In Proceedings of the
4
th
International Symposium on Environment,
Athens, Greece, 21-24 May 2009.
Abdeen, M. O. (2009b). Energy use, environment
and sustainable development. In Proceedings of
the 3
rd
International Conference on Sustainable
Energy and Environmental Protection (SEEP
2009), Paper No.1011, Dublin, Republic of Ire-
land, 12-15 August 2009.
Abdeen, M. O. (2009c). Energy use and environ-
mental: Impacts: A general review. Journal of
Renewable and Sustainable Energy, 1(5), 129.
Abdeen, M. O. (2009d). Energy use, environ-
ment and sustainable development. In Mancuso,
R. T. (Ed.), Environmental cost management
(pp. 129166). New York, NY: NOVA Science
Publishers.
Aroyeun, S. O. (2009). Reduction of aflatoxin
B1 and Ochratoxin A in cocoa beans infected
with Aspergillus via Ergosterol Value. World
Review of Science. Technology and Sustain-
able Development, 6(1), 7590. doi:10.1504/
WRSTSD.2009.022459
ASHRAE. (2005). Commercial/institutional
ground source heat pump engineering manual.
Atlanta, GA: American Society of Heating, Re-
frigeration and Air-conditioning Engineers, Inc.
Barton, A. L. (2007). Focus on sustainable devel-
opment research advances (pp. 189205). New
York, NY: NOVA Science Publishers, Inc.
Bos, E., My, T., Vu, E., & Bulatao, R. (1994). World
population projection: 1994-95. Baltimore, MD
& London, UK: John Hopkins University Press,
World Bank edition.
Boulet, T. (1987). Controlling air movement: A
manual for architects and builders (pp. 85138).
New York, NY: McGraw-Hill.
129
Brain, G., & Mark, S. (2007). Garbage in, energy
out: Landfill gas opportunities for CHP projects.
Cogeneration and On-Site Power, 8(5), 3745.
BS 5454. (1989). Storage and exhibition archive
documents. British Standard Institute. London.
Commission of the European Communities (CEC).
(2000). Towards a European strategy for the se-
curity of energy supply. (Green Paper, Brussels,
29 November 2000 COM 769).
Davenport, A., Galbraith, A., Stockdale, M., Wil-
son, S., Mitchell, R., & Hewitson, R. Woodley,
M. (2005). Building and land management for
law students, 5
th
edition. Oxford: UK.
David, E. (2003). Sustainable energy: Choices,
problems and opportunities. The Royal Society
of Chemistry, 19, 1947.
DeCarlo, F., & Bassano, A. (2010). Freshwa-
ter ecosystems and aquaculture research (pp.
63105). New York, NY: Nova Science Publish-
ers, Inc.
DEFRA. (2006). Energy resources. Doncaster,
UK: Sustainable Development and Environment.
Department of Energy (DoE). (2009). Annual
energy demand. USA.
DETR. (1994). Best practice programme-intro-
duction to energy efficiency in buildings. Don-
caster, UK. UK: Department of the Environment,
Transport and the Regions.
Duffie, J. A., & Beckman, W. A. (1980). Solar
engineering of thermal processes. New York, NY:
J. Wiley and Sons.
EIBI (Energy in Building and Industry). (1999).
Constructive thoughts on efficiency, building
regulations, inside committee limited. Inside
Energy: Magazine for Energy Professional, 13-
14. UK: KOPASS.
Energy use in offices (EUO). (2000). Energy
consumption guides 19 (ECG019) Energy ef-
ficiency best practice programme. London, UK:
UK Government.
Erlich, P. (1991). Forward facing up to climate
change. In Wyman, R. C. (Ed.), Global climate
change and life on Earth. London, UK: Chapman
and Hall.
Erreygers, G. (1996). Sustainability and stability
in a classical model of production. In Faucheux,
S., Pearce, D., & Proops, J. (Eds.), Models of
sustainable development. Cheltenham.
Farm Energy Centre (EFC). (2000). Helping
agriculture and horticulture through technology,
energy efficiency and environmental protection.
Warwickshire.
Hindsworth, M. F., & Lang, T. B. (2009). Com-
munity participation and empowerment (pp.
235261). New York, NY: NOVA Science Pub-
lishers, Inc.
Huttrer, G. (2001). The status of world geothermal
power generation 1995-2000. Geothermics, 30,
127. doi:10.1016/S0375-6505(00)00042-0
IEA. (2008). Combined heat and power: Evalu-
ating the benefits of greater global investment.
IPCC. (2001). Climate change 2001 (3 volumes).
United Nations International Panel on Climate
Change. UK: Cambridge University Press.
Jeremy, L. (2005). The energy crisis, global
warming and the role of renewables. Renewable
Energy World, 8(2).
Kowalczyk, L., & Piotrowski, J. (2009). Energy
costs, international developments and new direc-
tions (pp. 137). New York, NY: NOVA Science
Publishers, Inc.
Lam, J. C. (2000). Shading effects due to nearby
buildings and energy implications. Energy Con-
version and Management, 47(7), 4759.
130
Lazzarin, R., DAscanio, A., & Gaspaella, A.
(2002). Utilisation of a green roof in reducing
the cooling load of a new industrial building. In
Proceedings of the 1
st
International Conference
on Sustainable Energy Technologies (SET), (pp.
32-37). Porto, Portugal. 12-14 June 2002.
Levine, M., & Hirose, M. (2005). Energy efficiency
improvement utilising high technology: An as-
sessment of energy use in industry and buildings.
Report and Case Studies. London, UK: World
Energy Council.
Lund, J. W., Freeston, D. H., & Boyd, T. L. (2005).
Direct application of geothermal energy: 2005
Worldwide Review. Geothermics, 34, 691727.
doi:10.1016/j.geothermics.2005.09.003
Lysen, E. H. (1983). Introduction to wind energy
(pp. 1550). The Netherlands: CWD.
Meffe, S., Perkson, A., & Trass, O. (1996). Coal
beneficiation and organic sulphur removal. Fuel,
75, 2530. doi:10.1016/0016-2361(95)00171-9
Mortal, A. (2002). Study of solar powered heat
pump for small spaces. Portugal.
Omer, A. M. (2008a). Green energies and the
environment. Renewable & Sustainable En-
ergy Reviews, 12, 17891821. doi:10.1016/j.
rser.2006.05.009
Omer, A. M. (2008b). On the wind energy re-
sources of Sudan. Renewable & Sustainable
Energy Reviews, 12(8), 21172139. doi:10.1016/j.
rser.2006.10.010
Omer, A. M. (2008c). Energy demand for heating
and cooling equipment systems and technology
advancements. In White, J. R., & Robinson, W. H.
(Eds.), Natural resources: Economics, manage-
ment and policy (pp. 131165).
Omer, A. M. (2009a). Energy use and environ-
mental impacts: A general review. Renewable and
Sustainable Energy, 1, 129.
Omer, A. M. (2009b, October-November). Prin-
ciple of low energy building design: Heating,
ventilation and air conditioning. [Mumbai, India.].
Cooling India: Indias Premier Magazine on the
Cooling Industry, 5(4), 2646.
Omer, A. M. (2010a). Development of sustainable
energy research and applications. In Lee, W. H.,
& Cho, V. G. (Eds.), Handbook of sustainable
energy (pp. 385418). New York, NY: NOVA
Science Publishers, Inc.
Omer, A. M. (2010b). The crux of matter: Water in
the Republic of the Sudan (pp. 150). New York,
NY: NOVA Science Publishers, Inc.
Parikn, J., Smith, K., & Laxmi, V. (1999). In-
doors air pollution: A reflection on gender bias.
Economic and Political Weekly, 34(9).
Paul, F. (2001). Indoor hydroponics: A guide to
understanding and maintaining a hydroponic
nutrient solution. UK. 2001.
Randal, G., & Goyal, R. (1998). Greenhouse
technology. New Delhi, India: Narosa Publish-
ing House.
REN21. (2007). Renewables global status report.
Retrieved from www.ren21.net
Reddy, A., Williams, R., & Johansson, T.
(2007). Energy after Rio: Prospects and chal-
lenges. United Nations Development Programme
(UNDP). Retrieved from http://www.undp.org/
seed/energy/-exec-en.html
Roriz, L. (2001). Determining the potential energy
and environmental effects reduction of air con-
ditioning systems. Commission of the European
Communities DG TREN.
Shao, S. (2002). Thermodynamic analysis on heat
pumps with economiser for cold regions. China.
131
UNEP. (2003). Handbook for the international
treaties for the protection of the ozone layer.
Nairobi, Kenya: United Nations Environment
Programme.
UNIDO. (2007). Changing courses sustainable
industrial development, as a response to agenda
21. Vienna.
United Nations. (UN). (2001). World urbanisation
prospect: The 1999 revision. New York, NY: The
United Nations Population Division.
United Nations. (UN). (2002a). Science and
technology as a foundation for SD. Summary by
the Scientific and Technological Community for
the Multi-Stakeholder Dialogue Segment of the
fourth session of the Commission on SD acting as
the preparatory committee for the World Summit
on SD. Note by the Secretary-General. Commis-
sion on SD acting as the preparatory committee
for the World Summit on SD Fourth Preparatory
Session, 27 May7 June, 2002.
United Nations. (UN). (2002b). Global challenge
global opportunity: Trends in sustainable devel-
opment. Department of Economics and Social,
World Summit on Sustainable Development,
Johannesburg, SA.
United Nations. (UN). (2002c). Implementation of
the United Nations millennium declaration. Report
of the Secretary-General, United Nations General
Assembly. Retrieved from http://www.un.org
United Nations Economic Commission for Africa
(UNECA). (2002). Address by Josu Dion:
Science and technology policies for sustainable
development and Africas global inclusion. Sus-
tainable Development Division ATPS Conference,
11 November, Abuja, Nigeria.
(UNECA). (2003). Making science and technol-
ogy work for the poor and for SD in Africa. Paper
prepared by the SD Division with the assistance
of a senior international consultant, Akin Adubifa,
January.
(UNECA). (2003a). The state of food security in
Africa. Progress Report of the 3rd Meeting of the
Committee on Sustainable Development, 710
October, Addis Ababa, Ethiopia.
United Nations Economic Commission for Europe
(UNECE). (2004). Note by the ECE Secretariat,
Steering Group on Sustainable Development.
Second Meeting of the 2003/2004 Bureaus,
Conference of European Statisticians, Statistical
Commission, Geneva, Switzerland.
United Nations Under-Secretary General and the
United Nations Environment Programme (UNEP).
(2000). Overview: Outlook and recommendations.
Global Environment Outlook. Retrieved from
http://grid.cr.usgs.gov/geo2000/ov-e/0012.htm
Vilinac, D. (2008). Plant medicines: An herbalists
perspective. World Review of Science. Technology
and Sustainable Development, 3(2), 140151.
doi:10.1504/WRSTSD.2008.018556
White, J. R., & Robinson, W. H. (2008). Natural
resources: Economics, management and policy
(pp. 149). New York, NY: NOVA Science Pub-
lishers, Inc.
Witte, H. (2002). Comparison of design and op-
eration of a commercial UK ground source heat
pump project. Groenholland BV.
World Bank. (WB). (2003a). World development
report. New York, NY: Oxford University Press.
132
World Bank. (WB). (2003b). Global economic
prospects: Realizing the development promise of
the Doha agenda. Washington, DC: The World
Bank, World Bank, The International Bank for
Reconstruction and Development.
World Bank. (WB). (2004). World developments
report 2004: Making services work for poor
people. Washington, DC: World Bank.
World Bank. (WB). (2006). Sustainable de-
velopment in the 21st century. Retrieved from
http://lnweb18.worldbank.org/ESSD/sdvext.
nsf/43ByDocName/SustainableDevelopmentint
World Bank. (WB). (2007). World Bank sustain-
able development reference guide. Retrieved from
http://www.WorldBankSustainableDevelopmen-
tReferenceGuide
World Energy Council (WEC). (2009). The world
energy demand in 2020.
World Energy Outlook (WEO). (1995). Inter-
national Energy Agency. Paris, France: OECD
Publications.
WRI (World Resource Institute). (2004). World
resources: A guide to the global environment-
People and the environment. Washington, DC.
Yadav, I., & Chauadhari, M. (1997). Progressive
floriculture (pp. 15). Bangalore, India: The
House of Sarpan.
Zuatori, A. (2005). An overview on the national
strategy for improving the efficiency of energy
use. Jordanian Energy Abstracts, 9(1), 3132.
ADDITIONAL READING
Abdeen, M. O. (2008a). Renewable building en-
ergy systems and passive human comfort solutions,
Renewable and Sustainable Energy Reviews,
Vol.12, No.6, p.1562-1587, United Kingdom,
August 2008.
Abdeen, M. O. (2008b). People, power and pollu-
tion, Renewable and Sustainable Energy Reviews,
Vol.12 No.7, p.1864-1889, United Kingdom,
September 2008.
Abdeen, M. O. (2008c). Energy, environment and
sustainable development, Renewable and Sustain-
able Energy Reviews, Vol.12, No.9, p.2265-2300,
United Kingdom, December 2008.
Abdeen, M. O. (2008d). Focus on low carbon
technologies: the positive solution, Renewable
and Sustainable Energy Reviews, Vol.12, No.9,
p.2331-2357, United Kingdom, December 2008.
Abdeen, M. O. (2008e). Development of integrated
bioenergy for improvement of quality of life of
poor people in developing countries. In Magnus-
son, F. L., & Bengtsson, O. W. (Eds.), Energy
in Europe: Economics, Policy and Strategy- IB
(pp. 341373). New York, USA: NOVA Science
Publishers, Inc.
Abdeen, M. O. (2009a). Environmental and
socio-economic aspect of possible development
in renewable energy use, In: Proceedings of the
4
th
International Symposium on Environment,
Athens, Greece, 21-24 May 2009.
Abdeen, M. O. (2009b). Energy use, environment
and sustainable development, In: Proceedings of
the 3
rd
International Conference on Sustainable
Energy and Environmental Protection (SEEP
2009), Paper No.1011, Dublin, Republic of Ire-
land, 12-15 August 2009.
Abdeen, M. O. (2009c). Energy use and envi-
ronmental: impacts: a general review, Journal
of Renewable and Sustainable Energy, Vol.1,
No.053101, p.1-29, United State of America,
September 2009.
Abdeen, M. O. (2009d). Energy use, environ-
ment and sustainable development. In Mancuso,
R. T. (Ed.), Environmental Cost Management
Publishers, Inc.
133
Aroyeun, S. O. (2009). Reduction of aflatoxin
B1 and Ochratoxin A in cocoa beans infected
with Aspergillus via Ergosterol Value. World
Review of Science. Technology and Sustain-
able Development, 6(1), 7590. doi:10.1504/
WRSTSD.2009.022459
Barton, A. L. (2007). Focus on Sustainable De-
velopment Research Advances (pp. 189205).
New York, USA: NOVA Science Publishers, Inc.
Brain, G., & Mark, S. (2007). Garbage in, energy
out: landfill gas opportunities for CHP projects.
Cogeneration and On-Site Power, 8(5), 3745.
Commission of the European Communities (CEC).
(2000). Towards a European strategy for the se-
curity of energy supply. Green Paper, Brussels,
29 November 2000 COM (2000) 769.
Farm Energy Centre (EFC). (2000). Helping
agriculture and horticulture through technology,
energy efficiency and environmental protection.
Warwickshire. 2000.
Felice DeCarlo and Alessio Bassano. (2010).
Freshwater Ecosystems and Aquaculture Research
(pp. 63105). New York, USA: Nova Science
Publishers, Inc.
Huttrer, G. (2001). The status of world geothermal
power generation 1995-2000. Geothermics, 30,
127. doi:10.1016/S0375-6505(00)00042-0
IEA. (2008). Combined heat and power: evaluating
the benefits of greater global investment. 2008.
Jeremy, L. (2005). The energy crisis, global warm-
ing and the role of renewables. Renewable Energy
World 2005; 8 (2).
Lam, J. C. (2000). Shading effects due to nearby
buildings and energy implications. Energy Con-
version and Management, 47(7), 4759.
Leszek Kowalczyk and Jakub Piotrowski. (2009).
Energy Costs, International Developments and
New Directions (pp. 137). New York, USA:
NOVA Science Publishers, Inc.
Lund, J. W., Freeston, D. H., & Boyd, T. L. (2005).
Direct application of geothermal energy: 2005
Worldwide Review. Geothermics, 34, 691727.
doi:10.1016/j.geothermics.2005.09.003
Mildred, F., & Trevor, B. (2009). Lang (pp.
235261). New York, USA: Community Par-
ticipation and Empowerment. NOVA Science
Publishers, Inc.
Omer, A. M. (2010). The crux of matter: water in
the Republic of the Sudan. 2010 NOVA Science
Publishers, Inc., p.1-50, New York, USA.
Paul, F. (2001). Indoor hydroponics: A guide
to understanding and maintaining a hydroponic
nutrient solution. UK. 2001.
REN21. (2007). Renewables (2007) global status
report. www.ren21.net.
Reddy, A., Williams, R., & Johansson, T. (2007).
Energy after Rio: prospects and challenges. United
Nations Development Programme (UNDP). http://
www.undp.org/seed/energy/-exec-en.html. 2007.
Roriz, L. (2001). Determining the potential energy
and environmental effects reduction of air con-
ditioning systems. Commission of the European
Communities DG TREN.
Shao, S. (2002). Thermodynamic analysis on heat
pumps with economiser for cold regions. China.
UNEP. (2003). Handbook for the International
Treaties for the Protection of the Ozone Layer.
Nairobi, Kenya: United Nations Environment
Programme.
United Nations. (UN). (2001). World Urbanisa-
tion Prospect: The 1999 Revision. New York.
The United Nations Population Division. 2001.
134
United Nations. (UN). (2002a) Science and Tech-
nology as a Foundation for SD. Summary by the
Scientific and Technological Community for the
Multi-Stakeholder Dialogue Segment of the fourth
session of the Commission on SD acting as the
preparatory committee for the World Summit on
SD. Note by the Secretary-General. Commission
on SD acting as the preparatory committee for the
World Summit on SD Fourth Preparatory Session,
27 May7 June, 2002.
United Nations. (UN) (2002b) Global challenge
global opportunity: trends in sustainable devel-
opment, Department of Economics and Social,
World Summit on Sustainable Development,
Johannesburg, SA.
(UNECA). (2002) Address by Josu Dion:
science and technology policies for sustainable
development and Africas global inclusion, Sus-
tainable Development Division ATPS Conference,
11 November, Abuja, Nigeria.
(UNECA) (2003) Making Science and Technology
Work for the Poor and for SD in Africa, Paper
prepared by the SD Division with the assistance
of a senior international consultant, Akin Adubifa,
January.
(UNECA). (2003a) The state of food security in
Africa, Progress Report of the 3rd Meeting of the
Committee on Sustainable Development, 710
October, Addis Ababa, Ethiopia.
United Nations Economic Commission for Europe
(UNECE). (2004) Note by the ECE Secretariat,
Steering Group on Sustainable Development.
Second Meeting of the 2003/2004 Bureau,
Conference of European Statisticians, Statistical
Commission, Geneva, Switzerland.
United Nations Under-Secretary General and the
United Nations Environment Programme (UNEP).
(2000) Overview: outlook and recommenda-
tions, Global Environment Outlook, http://grid.
cr.usgs.gov/geo2000/ov-e/0012.htm, Earthscan,
1999, London.
Vilinac, D. (2008). Plant medicines: an herbalists
perspective. World Review of Science. Technology
and Sustainable Development, 5(2), 140151.
doi:10.1504/WRSTSD.2008.018556
White, J. R., & Robinson, W. H. (2008). Natural
Resources: Economics, Management and Policy
Publishers, Inc.
Witte, H. (2002). Comparison of design and op-
eration of a commercial UK ground source heat
pump project. Groenholland BV.
World Bank. (WB) (2003a) World Development
Report, Oxford University Press, New York.
World Bank. (WB). (2003b) Global economic
prospects: realizing the development promise of
the Doha agenda, The International Bank for
Reconstruction and Development, The World
Bank, World Bank: Washington, DC.
World Bank. (WB). (2004) World Development
Report 2004: Making Services Work for Poor
People, World Bank: Washington, DC.
World Bank. (WB). (2006). Sustainable De-
velopment in the 21st Century.http://lnweb18.
worldbank.org/ESSD/sdvext.nsf/43ByDocName/
SustainableDevelopmentint
World Bank. (WB). (2007). World Bank Sus-
tainable Development Reference Guide, http://
www. WorldBankSustainableDevelopmentRef-
erenceGuide.
135
Renewable Energy: Renewable energy is
energy generated from natural resources such as
sunlight, wind, rain, tides, and geothermal heat,
which are renewable (naturally replenished).
Energy obtained from sources that are essentially
inexhaustible (unlike, for example the fossil fuels,
of which there is a finite supply). Energy sources
that are, within a short time frame relative to the
Earths natural cycles, sustainable, and include
non-carbon technologies such as solar energy,
hydropower, and wind, as well as carbon-neutral
technologies.
Solar Energy: Energy from the sun that is
converted into thermal or electrical energy; the
amount of energy falling on the earth is given by
the solar constant, but very little use has been made
of solar energy. Energy derived ultimately from
the sun. It can be divided into direct and indirect
categories. Most energy sources on Earth are forms
of indirect solar energy, although we usually do
not think of them in that way. Solar energy uses
semiconductor material to convert sunlight into
electric currents. Although solar energy only pro-
vides 0.15% of the worlds power and less than
1% of US energy, experts believe that sunlight
has the potential to supply 5,000 times, as much
energy as the world currently consumes.
Biomass Energy: The energy embodied in
organic matter (biomass) that is released when
chemical bonds are broken by microbial digestion,
combustion, or decomposition. Biofuels are a wide
range of fuels, which are in some way derived
from biomass. The term covers solid biomass,
liquid fuels and various biogases. Biofuels are
gaining increased public and scientific attention,
driven by factors such as oil price spikes and the
need for increased energy security.
Wind Energy: Kinetic energy present in wind
motion that can be converted to mechanical en-
ergy for driving pumps, mills, and electric power
generators. Wind power is the conversion of wind
energy into a useful form of energy, such as using
wind turbines to make electricity, wind mills for
mechanical power, wind pumps for pumping water
or drainage, or sails to propel ships.
Hydropower: Hydropower, hydraulic power
or waterpower is power that is derived from the
force or energy of moving water, which may be
harnessed for useful purposes. Hydropower is
using water to power machinery or make elec-
tricity. Water constantly moves through a vast
global cycle, evaporating from lakes and oceans,
forming clouds, precipitating as rain or snow, and
then flowing back down to the ocean.
Geothermal Energy: Geothermal power
(from the Greek roots geo, meaning earth, and
thermos, meaning heat) is power extracted from
heat stored in the earth. This geothermal energy
originates from the original formation of the
planet, from radioactive decay of minerals, and
from solar energy absorbed at the surface. Heat
transferred from the earths molten core to under-
ground deposits of dry steam (steam with no water
droplets), wet steam (a mixture of steam and water
droplets), hot water, or rocks lying fairly close to
the earths surface.
Resource Management: Efficient incident
management requires a system for identifying
available resources at all jurisdictional levels to
enable timely and unimpeded access to resources
needed to prepare for, respond to, or recover from
an incident. Resource management is the efficient
and effective deployment for an organizations
resources when they are needed. Such resources
may include financial resources, inventory, hu-
man skills, production resources, or information
technology (IT).
Sustainable Development: Development,
which seeks to produce sustainable economic
growth while ensuring future generations ability
to do the same by not exceeding the regenerative
capacity of the nature. In other words, its trying
to protect the environment. A process of change
in which the resources consumed (both social and
ecological) are not depleted to the extent that they
cannot be replicated. Environmentally friendly
136
forms of economic growth activities (agriculture,
logging, manufacturing, etc.) that allow the con-
tinued production of a commodity without damage
to the ecosystem (soil, water supplies, biodiversity
or other surrounding resources).
Environment: The natural environment,
commonly referred to simply as the environment,
encompasses all living and non-living things oc-
curring naturally on Earth or some region thereof.
The biophysical environment is the symbiosis be-
tween the physical environment and the biological
life forms within the environment, and includes
all variables that comprise the Earths biosphere.
Greenhouse Gases: Greenhouse gases are
gases in an atmosphere that absorb and emit
radiation within the thermal infrared range. This
process is the fundamental cause of the greenhouse
effect. The main greenhouse gases in the Earths
atmosphere are water vapour, carbon dioxide,
methane, nitrous oxide, and ozone. Changes in
the concentration of certain greenhouse gases,
due to human activity such as fossil fuel burning,
increase the risk of global climate change.
137
Chapter 4
DOI: 10.4018/978-1-61350-138-2.ch004
Fouad Kamel
University of Southern Queensland, Australia
Marwan Marwan
Queensland University of Technology, Australia
Demand-Side Response
Smart Grid Technique for
Optimized Energy Use
ABSTRACT
The chapter describes a dynamic smart grid concept that enables electricity end-users to be acting on
controlling, shifting, or curtailing own demand to avoid peak-demand conditions according to informa-
tion received about electricity market conditions over the Internet. Computer-controlled switches are
used to give users the ability to control and curtail demand on a users premises as necessary, follow-
ing a preset users preferences. The computerized switching provides the ability to accommodate local
renewable energy sources as available. The concept offers further the ability to integrate charging
electrical vehicles during off-peak periods, helping thus substantially improving the utilization of the
whole electricity system. The approach is pursuing improved use of electrical energy associated with
improved energy management, reduced electricity prices and reduced pollution caused by excessive
use of combustion engine in transport. The technique is inherently restricted to take effect in frame of
energy tariff regimes based on real-time price made to encourage and reward conscious users being
proactively participating in holistic energy management strategies.
138
Demand-Side Response Smart Grid Technique for Optimized Energy Use
INTRODUCTION
The traditional user-supplier rapport in the elec-
trical energy market has historically evolved
following a strategy implying whenever a load is
switched on it is expected to be fulfilled by the
supplier at the expected time and quality. Grow-
ing electrical demands followed by constantly
growing supply led to troubled electrical services
manifested mainly by daily and seasonal excessive
peak and low demands. Those chronic peaks on
electrical networks are usually associated with
compromised quality, risk of forced outages and
high-priced energy supply; while low-demands
on the other side might be driving some power
plants to be operating at critical economic vi-
ability. Demand-side-response techniques are
helping electricity users to become proactively
participating in averting detrimental conditions
presently prevailing in the electricity sector
(Kamel, 2009b). Coordinated strategies shall
help achieving improved use of electrical power
plants and electricity infrastructure, besides in-
tegrated use of different types of energy sources.
(Chua-Liang Su & Kirschen, 2009) proposed a
day-ahead market-clearing mechanism that allows
consumers to submit complex bids. Those bids
are expected to give consumers the opportunity
to specify constraints on their hourly and daily
consumptions in the same way as generators can
specify the operating constraints on their generat-
ing units. It is a day-ahead market with complex
bids and offers whose objective is to maximize the
social welfare. The social welfare is described as
the difference between the value that consumers
attach to the electrical energy that they buy and the
cost of producing this energy. Not all consumers
have the ability or the motivation to adjust their
demand as a function of price. Part of the demand
will therefore remain perfectly inelastic. There-
fore, consumers were classified into two types,
price-taking and price-sensitive. Price-taking is
those consumers who have, in theory, an infinite
marginal value as otherwise the consumers would
have placed price responsive bids with a finite
marginal value attached.
Plant Capacity Factor
The plant capacity factor (PCF) or the so also
called utilization factor of a power plant is, by
definition, the proportion of the actual electrical
energy generated yearly by the plant to the quantity
of electrical energy, which would be generated
if the plant was operated at rated power for full
years time (8760 hours) as reported by Brinkmann
(1980). The factor has a direct influence on the
energy cost as can be deduced from the following
equation of the fixed charge method according
to De-Meo (1978), Leonard (1977), Chobotov
(1978) and Clorefeine (1980):
c
E
= c
tr
FCR / (T
o
PCF) + c
op
(1)
where c
E
is the cost of energy generated, c
tr
cost
of installed power including taxes during the
installation period, FCR fixed charge rate of the
capital, normally 1518% a year according to
Leonard (1977) and Leonard (1978), T
o
=8760
(h) the hours per year, PCF plant capacity factor
and c
op
the operation and maintenance cost of the
plant. For plants operating 24 hour/day, 7 days a
week, i.e. 8760 hour/year PCF is a unity, which
produces the least possible energy cost and best
economic conditions. For power plants operat-
ing any less than 8760 hour/year the PCF will
respectively be lower (below unity) what drives
the cost of the produced energy to be accordingly
higher, equation (1).
Figure 1 illustrates the impact of the plant ca-
pacity factor on the cost of the produced energy.
The calculation is made on the basis of the cost
of the installed power c
tr
= $1000/kW, capital
fixed charge rate FCR = 0.17 and the operation
and maintenance cost of the plant c
op
= $0.02/
kWh. It is evident that a power plant operated at
low plant capacity factor e.g. PCF = 0.1 (this is
2.4 hour/day) will be producing energy at $150/
139
MWh, while operated continuously for 24 hour
will produce energy at a cost of $25/MWh.
Figure 1 depicts the importance of operating
power plants and electrical network infrastructure
at elevated plant capacity factor, close to the
unity, in order to verify best economic perfor-
mance. Various efforts on avoiding peak demands
on the electrical network are mainly aiming at
leveling demand throughout the year in order to
achieve as high plant capacity factors as possible
for all electrical power components. Additionally,
leveling demands is pursued to avert or delay the
urgency to expand generation capacity and net-
work infrastructure to cope with consistently
rising peak demands. Important methods to avert
or delay peak demands are represented in demand-
side response activities, both on the supplier and
on the users side.
Demand Side Response and
Smart Grid Technologies
Demand Side Response (DSR)
Demand Side Response (DSR), as described by
Albadi & El Saadany (2007) can be defined as
the changes in electricity usage by end-use cus-
tomers from their normal consumption patterns
in response to changes in the price of electricity
over time. Parvania & Fotuhi-Firuzabad (2010)
describe DSR as tariff or program established
to motivate change in electric consumption by
end-use customers in response to change in the
price of electricity over time. Further on, DSR
programs provide means for utilities to reduce the
power consumption and save energy, maximize
utilizing the current capacity of the distribution
system infrastructure, reducing or eliminating
the need for building new lines and expanding
the system as described by Dam, Mohagheghi &
Stoupis (2008).
Vos (2009) described demand response as an
integral part of the smart grid, is a cost effective,
rapidly deployed resource that provides benefits
to utilities and customers. Some advantages of
DSR according to Greening (2010) are: increased
economic efficiency of the electricity infrastruc-
ture, enhanced reliability of the system, relief of
power congestions and transmission constraints,
reduced energy prices and mitigated potential
market power. Further on, demand response can
help reduce peak demands and therefore reduce
spot price volatility as illustrated by Nguyen
(2010). In addition, Hyung Seon & Thomas (2008)
Figure 1. Impact of the plant capacity factor on the cost of the produced energy (Kamel, 2009b)
140
described demand-side-response participation
would help electricity power markets operate in a
more efficient way. Based on a review of current
utility programs, the Electric Power Research
Institute (EPRI) estimated that demand response
has the potential to reduce peak demand in the
U.S. by 45,000 MW as reported by Walawalkar,
Blumsack, Apt & Fernands (2008). Therefore,
the implementation of DSR programs is expected
to improve economic efficiency in the electricity
market.
In the United Kingdom, various techniques
have been used to develop load electricity man-
agement. One of the methods is called responsive
demand or demand side management program and
was developed in the early 1960s (Hamidi, Li &
Robinson, 2009). This system served to maintain
the security of electricity supply and limited the
facilities for electricity generation, transmission
and distribution. This program has been partici-
pating in improving the economy, security and
reliability of the electricity industry as well as
eliminating the environmental concerns (Hamidi,
Li & Robinson, 2009). However, later in 2007
the British Government initiated the building of
the Energy Demand Research Project which
focused on the actual benefits of demand response
for consumers (Torriti, Hassan & Leach, 2009).
The British Government is currently con-
sidering the economic benefits of the Demand
Side Response program, as this system requires
a high implementation cost. Besides that, the
government must first conduct a reform of the
electricity industry to support this program, for
instance: by restructuring the electricity price
and market, transmission and distribution as well
as the retail sector. According to Torriti, Hassan
& Leach (2009) much of the debate around the
economic potential of Demand Response focuses
on the actual benefits of DR for consumers, and it
provides some benefits and weaknesses for both
the government and the user. Hence, there are
five technology specifications that can potentially
compromise such as: a minimum meter specifi-
cation, smart enables meters that substitute old
meters, dumb meters combined with smart boxes,
retrofit devices and clip-on costumer display unit
(Torriti, Hassan & Leach, 2009).
Similar to what has occurred in UK, in Finland,
Interruptible Programmes as a part of demand
side response model have been used as distur-
bance reserve for several years (Torriti, Hassan &
Leach, 2009). Utilisation of the demand response
program is more effective to overcome peak
load, breakdown and manage electricity supply
to all customers. This plan is not just applied by
small-costumers but also has been used by large-
scale industry. Therefore, in 2005 total Demand
Response potential in Finnish large-scale industry
was estimated at about 1280 MW, which represents
9% of the Finnish power demand peak (Stam,
2005). Following that, in 2008 the Finnish main
electricity utility invested in an advance metering
reading system to automatically read, control and
manage all 60,000 of its customer metering points
(Torriti, Hassan & Leach, 2009).
In Korea, the Demand Side Management
program has been used for several years. In the
1970s, several programs were introduced in Load
Management, for instance: night thermal-storage
per rates program (1972), inverted block pro-
gram (1974), the seasonal tariff (1977) and the
time of use tariff (1977) (Jin-Ho, Tae-Kyung &
Kwang-Seok, 2009). However, this program has
not reached the maximum results to control load
demand for peak demand sessions. Therefore,
2006, after the revision of the law, the government
announced its 3
rd
National Electricity Demand
Forecast and Supply Plan which addressed the
governments main concerns about the Demand
Side Management (Jin-Ho, Tae-Kyung & Kwang-
Seok, 2009). In China, demand side management
started from 1990s (Zhong, Kang & Liu 2010).
In Australia, implementation of the DSR
program has been conducted several years ago.
In late 2002, the Energy Users Association of
141
Australia (EUAA) conducted a trial to demonstrate
the benefits of a DSR aggregation process which
would enable electricity consumers to respond to
both the extreme prices and extreme peak demand
(Fraser, 2005). This experiment was conducted by
end-users to determine the value of an effective
DSR for electricity consumers and its impact in
terms of supporting an energy saving program.
This trial was supported by the Victorian, New
South Wales and Commonwealth Government, as
well as the CSIRO, to implement a Demand Side
Response Facility Trial (Energy User Association
of Australia Document, 2010).
In the experiment described above, the Aus-
tralian Government through the EUAA involved
customer to participate in the DSR trial. This
experiment was conducted in three regions that
fall under the National Electricity Market op-
eration, New South Wales, South Australia and
Victoria (Jones, 2004). These areas are regarded
to represent the electricity load in Australia, and
the results obtained show some significant ben-
efits of using DSR for consumer and electricity
providers. Hence, in December 2003 the Minis-
terial Council for Energy advised the Council of
Australian Governments (COAG) on the need for
further reform of the energy market to enhance
active energy user participation (Jones, 2004).
The energy users association of Australia tar-
geting a demand-side-response action, according
to Fraser (2005), summarizes that, for example,
South Australian electricity consumers only use
the highest 10% of their maximum electrical
demand on the network less than 0.5% of the
time per year, i.e., for about 40 hours per year.
The report is stating further: while the electric-
ity consumers are insulated from price volatility
by flat electricity prices, they are also paying
a significant and undisclosed (hard to evaluate)
premium in their retail electricity prices to cover
the retail suppliers costs of managing the risks
of the extreme price volatility.
Demand Side Response Models
Many different economic models are used to
represent Demand Side Response programs. In
the report of the strategic plan of the International
Energy Agency (2010) DSR is divided into two
basic categories, namely the time based program
and the incentives based program. According to
Aalami H.A, Moghaddam M.P & Yousefi G.R
(2008) the specific types of time based program
are: time of use, real time pricing (RTP) and
critical peak pricing; Federal Energy Regulatory
Commission (2006) reports, while the specific
types of incentive based program consist of di-
rect load control (DLC), Interruptible/curtailable
(I/C), demand bidding (DB), emergency demand
response program (EDRP), capacity market (CAP)
and ancillary service markets (A/S) programs.
An overview of selected DSR models: I/C pro-
gram, the EDRP, TOU and the proposed model,
as presented in Figure 2. In the following a brief
description of three popular market available
models: I/C program, the EDRP and TOU.
Interruptible and Curtailable Program
The interruptible/Curtailable program has
traditionally been one of the most common
demand-side-management (DSM) tools used by
electric-power utilities, which customers sign an
interruptible-load contract with a utility to reduce
their demand at a fixed time during the systems
peak-load period or at any time requested by the
utility (Yu, Zhang, Chung & Wong, 2005). This
service provides incentives/rewords to customers
participating to curtail electricity demand. The
electricity provider sends directives to the user
for following this program at certain times. The
user must obey those directives to curtail their
electricity when being notified from the utility or
face penalties. For example: the customer must
curtail their electricity consumption starting from
6:00 pm 7:00 pm; those customers who are fol-
lowing will get a financial bonus/reword to their
142
electricity bill from the utility. In California the
incentive of I/C program was $700/MWh/month
in 2001 as reported in (Aalami H.A, Moghaddam
M.P & Yousefi G.R, 2009).
Emergency Demand
Response Program
Emergency demand response program is energy-
efficient program that provides incentives to
customers who can reduce electricity usage for a
certain time; this is usually conducted at the time
of limited availability of electricity. According to
Covino (2003) emergency demand response pro-
gram provide participant with significant intensive
to reduce load. To participate on this program, all
customers are expected to reduce their energy
consumption during the events. Tyagi & Black
(2010) described this program will determine
which houses must be included in the event to
minimize cost and disruption, while alleviating
the overload condition. When asked to curtail,
and verified to have performed, the resource is
paid the higher of $500/MWh (Rahimi & Ipakchi,
2010). In New York, emergency demand response
program allows participant to be paid for reducing
Figure 2. Models of DSR programs: a) Interruptible/Curtailable, b) Emergency Demand Response, c)
Time Of Use and d) the Proposed Model (Marwan & Kamel, 2010b)
143
their energy consumption upon notice from the
New York Independent System Operator that an
operating reserves deficiency or major emergency
exists (Lawrence & Neenan, 2003).
Time of Use Program
According to Na & Ji-Lai (2006) Time of Use is
one of the important demand side management
methods, TOU demand side will response to the
price and will change the shape of the demand
curve. Further on, Time of Use rate is the most
obvious strategy developed for the management of
the peak demand in the world, which is designed
to encourage the consumer to modify the pattern
of electricity usage (Wen-Chen, Yi-Ping & Tzu-
Hao, 2007). For applying this program, the utility
does not provide reward or penalty to customer. To
participate, all customers are required to remove
their energy consumption during peak session to
off-peak session as soon as their receipt informa-
tion from the utility. Kirschen suggested (2003)
in this type of contract, the rate is fixed for the
duration of the contract but depends on the time
of the day. As compared to the flat rate contract,
some of the risk is shifted from the retailer to the
consumer because the consumer has an incentive to
consume during periods when the rates are lower.
Calculation of Energy Consumption
for Maximized Benefit
Different models are used in Demand Side Re-
sponse program planning as described by the
Federal Energy Regulatory Commission (2006). In
the report of the strategic plan of the International
Energy Agency (2010), DSR is divided into two
basic categories, namely, the time based program
and the incentives based program. In Aalami H.A,
Moghaddam M.P & Yousefi G.R (2008) the fol-
lowing approach is used to calculate the savings
in electricity expenditure based on price, demand,
incentives and costumer benefits associated with
the DSR program used.
The change in energy consumption d(t) at
the time t when the user changes demand from
d(t) to do(t) is:
d(t)= d(t)- do(t) (2)
For participating in the DSR program, the total
incentives Pd(t), when A(t) is paid as incentive
to the costumer at the time t for each kWh load
reduction, can be calculated as the following:
P(d(t))=A(t).d(t) (3)
The total penalty PENd(t) when the consumer
does not commit to the obligations as a member of
the DSR program, pen(t) is the penalty per kWh
at the time t and D(t) is the DSR program contract
level of consumption in kWh, will be accounted
as the following:
PEN(d(t))=pen(t).{D(t)-[d(t)-do(t)]} (4)
The total customers benefit S a member of
the DSR program can make at a certain time t is
made up of income and expenditures. The contract
could be mentioning a benefit from joining the
program so that e.g.B(d(t)) is the customer income
during that period from the use of d(t) kWh, at
the same time the customer could be receiving
additional incentives P(d(t)) as described in
Equation(2). The cost of the consumed electricity
and any penalty for not following the program,
if applicable, will be deducted from the income
as the following:
S=B(d(t))+P(d(t))-d(t).r(t)-PEN(d(t))
(5)
where,r(t) is the rate the customer pays per kWh
electricity at that time.
144
To maximize the customer benefit the slope
S/d(t) should be equal to zero, accordingly:
=
S
d t
B d t
d t
P d t
d t
r t
PEN
d t ( )
( ( ))
( )
( ( ))
( )
( )
( )
0

(6)
This leads to:
= +
B d t
d t
r t A t pen t
( ( ))
( )
( ) ( ) ( ) (7)
The benefit function most often used according
to Schweppe, Caramanis, Tabors & Bohn (1988)
is the quadratic benefit function:
= + +

B d t Bo t ro t d t do t
d t do t
t do t
( ( )) ( ) ( ).{ ( ) ( )}.
( ) ( )
( ) ( )
1
2

(8)
Where:
Bo(t) Benefit when d(t) =do(t) ($)
ro(t) Nominal rate for electricity consumption
($/kWh)
(t) Elasticity parameter
( , )
( )
( )
.
( )
( )
t h
ro h
do t
d t
r h
=

(9)
The elasticity parameter is a unit-less factor
indicating how strong the energy demand depends
on energy price, i.e. the effect the energy price on
demand. The multiplier
ro h
do t
( )
( )
helps transforming
the parameter into a unit-less factor by referring
to initial known conditions.
By differentiating Equation (8) and solving
for
B d t
d t
( ( ))
( )
and substituting the result in (7)
r t A t pen t ro t
d t do t
t do t
( ) ( ) ( ) ( )
( ) ( )
( ) ( )
+ = +

1
2

(10)
When the customer is participating in a DSR
program, the customer consumption d(t) for a
maximized benefit can be calculated from Equa-
tion (10) as the following:
d t do t t t
r t ro t A t pen t
ro t
( ) ( ) ( , )
[ ( ) ( ) ( ) ( )
( )
= +
+
(11)
Integration Renewable Energy
Sources and Electricity Vehicles
to DSR Smart Grid System
Integrating renewable energy with power genera-
tion is a new way to improve reliability, sustain-
ability and cost effectiveness on the electrical
network. According to Hammons (2006) the
key challenges that need to be considered in the
composition of future network include distributed
generation and the integration of renewable energy
sources, such as biomass, wind and solar. The
utilization of renewable energy is expected to be
leading to the harmony between humans and nature
with low pollution and sustainable accessibility
to resources as described in Figure 3.
Some countries have applied smart grid tech-
nology for renewable energy utilization. In Japan,
using solar power generation systems in every
ordinary house is the most active project in smart
grid researches as described by Li & Yao (2010).
The significant amount of installed wind power
in the German power system in 2008 produced
more than 22 GW producing between 1500 - 7700
GWh/month as reported in Hammons (2006).
On the other hand, electricity vehicles (EV)
technology brings impacts to the electrical distri-
bution grid. According to Clement-Nyns, Haesen
& Driesen ((2010) the vehicle can not only charge,
but also discharge and thus inject energy to the
145
grid. In addition, there are social, environmental
and economic advantages in switching to elec-
tricity vehicles as described by Anna Cain, Iain
MacGill & Bruce (2010).
The Users DSR Concept
This concept helps developing the technical tools
to be independently implemented and managed by
end-users to enable control energy consumption
accustomed to users preferences. The proposed
DSR Smart-Grid scheme is targeting flattening
load profiles by averting periods of peak de-
mands helping thus redistributing the load profile
throughout the day/year. The scheme helps electri-
cal generating plants and electrical infrastructure
to be operated 24 hour/day achieving thus higher
utilization factor, enhanced efficiency and con-
siderably reduced energy price. The presented
users side DSR concept is providing the needed
balance in the electrical system to complement
the efforts undertaken by suppliers to mitigate
peak demands and improve supply reliability and
stability, Figure 4.
The concept is presenting a low-cost DSR
technique implemented at users premises, which
assists electricity end-users to be shifting loads
around the clock averting peak-demand periods
and making use of on-site renewable energy
sources as appropriate. This shall help users to be
engaged in mitigating peak demands on the elec-
tricity network. The proposed concept comprises
a technical set-up of a programmable internet
relay, a router, solid state switches in addition to
the suitable software to control electricity demand,
Figure 5. The softwares on appropriate multime-
dia tool (CD Rom) offers users optimized control
of energy consumption.
The concept enables commercial and indus-
trial customers on fluctuating energy prices to be
achieving immediate financial savings. For resi-
dential customers on flat-rate tariffs, in contrast,
users are gaining financial benefits from reducing
energy consumptions at peak-demand periods.
Residential customers on different tariffs, where
energy price differs with day time and network
conditions (e.g. night tariffs), they are gaining
financial benefits also from shifting loads from
day- to night-times, when electricity is cheaper.
The scheme uses a router and a programmable
internet relay and solid-state switches to control
electrical demand at the users premises. The relay
Figure 3. Smart grid network diagram (Energy Efficiency and Conservation Authority, 2009)
146
Figure 4. The proposed users smart grid demand-side-response scheme (DSR) balancing conventional
suppliers smart grid DSR programs
Figure 5. The users DSR concept
147
is programmed to receive and act upon information
received from the AEMO on the internet about
demand/price conditions. Figure 6 illustrates the
control regime, where three appliances are con-
trolled by three solid-state switches receiving on/
off signals from the relay.
Consumers use local computers to set-up their
preferences for appliance profile usage and pri-
orities, e.g. Table 1. The profile of appliances
identifies when an appliance is run according to
electricity price or network conditions (national
demand). Pursuant to the order from the relay to
a solid-state switch, household appliances con-
nected to that switch can be turned on/off.
All control systems above are implemented
by a shell script under a Linux operation system.
Figure 7 shows the pseudo code of the controller
that is executed with each interaction.
Table 1 illustrates an example of an appliance
profile. All control systems above is implement-
ed by a shell script under a Linux operation
system.
ELECTRICITY INDUSTRY
DEVELOPMENT IN AUSTRALIA
(CASE STUDY)
In Eastern and Southern states of Australia the
Australian Energy Market Operator is manag-
ing the power flow across the Australian Capital
Territory, New South Wales, Queensland, South
Australia, Victoria and Tasmania. The AEMO
is regularly updating energy demand and prices
current situations publicly on the internet. The
presented concept is using programmable internet
relays and controllable electronic switches to be
proactively responding to demand/price peaks and
congestions conditions. The technique is globally
valid in other electricity markets under similar
operating conditions. Figure 8 depicts an example
of an actual energy demand and prices situation.
The price curve is closely following the demand
curve. Electricity prices are typically at their
lowest level at night during times of low demand
(off-peak). Prices are rising daily according to two
Figure 6. Control Regime
Table 1. Example of appliance profile
Appliance Start After Finish Before Session Time
Kettle 08.00 AM 04.00 PM Off-peak Session
Washing Machine 06.00 AM 10.00 AM Off-peak Session
Air Condition 06.00 AM 04.00 PM Off-peak Session
148
major peak demands in the morning and evening.
It is to be noted that electricity prices in Figure 8
are wholesale regional references prices (RRP)
i.e. prices traded to electricity suppliers. Electrical
energy at this stage still needs to be transmitted
to the different localities then distributed to end-
users. End-users could be industrial, commercial
or residential. Usually industrial and commercial
end-users are contracting electricity suppliers on
special agreements to provide the service satis-
fying their requirements. For most residential
electricity customers, electricity pricing typically
follow one set price flat-rate. Residential users
are often also offered a night tariff, where elec-
tricity prices are substantially reduced. The night
tariff corresponds off-peak times.
Figure 9 illustrates the frequency of electric-
ity demand supplied in Queensland during the
year 2008 as extracted from data of the Australian
Energy Market Operator (2009). The figure indi-
cates mainly the fact that the higher the load above
the base load the lesser likely the extent of their
duration will be. Base load power stations are
those operated twenty four hours/day throughout
the year corresponding to 8760 h/year, a utiliza-
tion factor of 1. Those power stations are provid-
ing the most economic operation at the least
possible energy price. Any loads exceeding the
base load are usually covered by other power
plants operated for shorter periods of time at
utilization factor lesser than 1, thus generating
energy at high prices. Accordingly, the intermittent
operation of the expensive to-run power plants
makes their operation even more expensive.
Figure 10 illustrates the occurrence of the
regional reference wholesale price RRP in
Queensland during the year 2008; extracted from
the Australia Energy Market Operator (2009). The
figure indicates mainly that low-priced supplies
are taking place at very high occurrences of more
than 80% a year, while high prices happen at
lower occurrences. For instance, prices around
AUD $20/MWh are occurring at frequencies of
about 80%, while prices of over $50/MWh have
occurrences of less than 10%.
Based on data of 31 December 2008,
Queensland total electricity generating capacity
was 12487 MW; coal-fired power stations pro-
Figure 7. Pseudo code of the control loop
149
Figure 8. Wholesale electricity price in AUD $/MWh and demand in MW for a typical day in Queensland
on 5th May 2009(Australia Energy Market Operator, 2009)
Figure 9. Occurrence of electrical energy demand Queensland during 2008. Peak demand 8413 MW,
base-load 4100 MW and total supplied electrical energy 52.18 TWh (Kamel, 2009b)
150
vided 70% of this total capacity, while gas-fired
electricity accounted for 17% and renewable
energy accounted for around 5% as stated by the
Department of Employment Economic Develop-
ment and Innovation (2010). These power gen-
erations are used to provide electrical energy for
all consumers in the Queensland area: residential,
commercial and industrial. However, the amounts
of energy produced from various generators de-
pend on market demand, price and availability of
sources. Figure 11 illustrates electricity generation
in Queensland according to the Department of
Employment Economic Development and Innova-
tion (2010).
Most of the power stations are directly con-
nected to the transmission system. The Queensland
electricity transmission system is provided by
Powerlink, licensed to operate more than 12,000
kilometres of Queensland high voltage transmis-
sion network, transporting electricity from the
generators to the distribution networks as by
(Department of Employment Economic Develop-
ment and Innovation, 2009). The distribution
network is carrying electricity from the transmis-
sion system to consumers. In Queensland, EN-
ERGEX and ERGON energy are purchasing
electrical energy from the Energy Market and
distributing it to the customer. ERGON e.g. pro-
vides energy at several tariff options to end users.
For example, Tariff 11 for all domestic consump-
tion 18.84 /kWh, while the night rate Tariff 31
for all consumption 7.7 /kWh and the economy
Tariff 33 for all consumption 11.32 /kWh (Ergon
Energy, 2010).
According to Queensland Competition Au-
thority (2000), the total energy consumption in
Australian grew at an annual rate of 2.6% for
the 25 years to 1997/1998. In the 2007- 2008
period according to Department of Employment
Economic Development and Innovation (2009a)
the annual electricity consumption in Queensland
has grown by over 29% or approximately 10500
GWh, making Queensland the second highest
consumer of electricity in Australia. This indicated
that Queensland has a significantly greater number
Figure 10. Electricity wholesale price RRP in Queensland in 2008 (Kamel, 2009b).
151
of high energy users than any other state, most of
these in regional Queensland.
Narayan & Smith (2005) describes since the
beginning of the 1990s, Australias electric power
industry has undergone a series of structural
reforms. In Queensland, the electricity industry
was restructured on 1 July 1998 to prepare the
industry for participating in the competitive Na-
tional Electricity Market, which is responsible for
structure, rules and regulations in the delivery of
energy to customers (Department of Employment
Economic Development and Innovation, 2009b).
The National Electricity Market Management
Company (NEMMCO) Limited was the Whole-
sale Market and Power System Operator for the
Australian NEM. NEMMCO was established in
1996 to administer and manage the NEM, develop
the market and continually improve its efficiency
and as of 1 July 2009 was replaced by the Aus-
tralian Energy Market Operator.
To improve governance, and enhance the reli-
ability and sustainability of the States electricity
system, the Australian Government has created a
collaborative electricity and gas industry in the
form of the Australian Electricity Market Opera-
tor (Australian Energy Market Operator, 2010),
which commenced operation on 1 July 2009.
The AEMO is managing power flows across the
Australian Capital Territory, New South Wales,
Queensland, South Australia, Victoria and Tasma-
nia. Western Australia and the Northern Territory
are not currently connected to this market primarily
because of their geographic distance from the rest
of the market. AEMOs responsibilities include
wholesale and retail energy market operation,
infrastructure and long term market planning
demand forecasting data and scenario analysis as
described in (Australian Energy Market Opera-
tor, 2010). The electricity market comprises of a
wholesale sector and a competitive retail sector.
All electricity dispatched in the market must be
traded through the central spot market. The Market
structure of NEMMCO / AEMO can be presented
as in Figure 12.
Figure 11. Electricity generation in Queensland (Department of Employment Economic Development
and Innovation, 2010)
152
Figure 12. The Market structure of NEMMCO/AEMO (Department of Resources Energy and Tourism,
2009)
Figure 13. Fluctuation of electricity price in Queensland (Marwan & Kamel, 2010a)
153
Marwan & Kamel (2010a) summarized in
Figure 13 an example of classic fluctuations in
electricity price in Queensland, from 22 May 2008
to 22 May 2009. This graph illustrates that the
average price during that time was in the range
of $50/MWh (5/kWh) Regional Reference
Wholesale Price (RRP), however, extreme prices
occurred exceeding $500/MWh (50/kWh). The
graph indicates also that excessive demands are
occurring regularly in all states on the intercon-
nected power network. Fraser (2005) stated that
customers, even those bound by flat-rate contracts,
must bear the additional cost for managing the
corresponding extreme prices.
Controlled Scenario Using
the Users DSR Scheme
In order to evaluate the effect of the proposed
scheme on electricity energy saving the electricity
price/demand in Queensland for the period 10
th
-
12
nd
May 2010 has been used. In the following,
ten scenarios have been formulated to demonstrate
the results as presented in Figure 14 and sum-
marized in Table 2.
Scenario 1, In this scenario users are shifting
375 MWh peak electricity usage occurring be-
tween 17:00 pm-19:00 pm towards the time pe-
riod 19:00 pm-21:30 pm when energy demand
and prices are low. All participants are suggested
to set-up the electricity profile to stop chosen
appliance from running during that time. For
example, kettle, washing machine and air condi-
tion could be effectively operated at optional times
of the day. No savings in energy cost due to ap-
plicable day-time tariffs. However, the scheme
was still able to remove congestions out of peak
demand areas.
Scenarios 2, Users are shifting peak demand
of 375 MWh occurring between 17:00 pm-19:00
pm to the period between 21:30 pm to 23:30 pm.
Achievable savings $21149 per day.
Scenario 3, Users are shifting peak demand
Figure 14. Controlled Scenarios
154
pm to the period between 23:30 pm to 01:00 am.
pm to the period between 01:00 am to 03:00 am.
Achievable savings $28200 per day
Scenario 8, Users are shifting peak demand of
1730 MWh occurring between 10:30 am-19:30
pm. All participants are suggested to set-up the
electricity profile to stop some appliance to run
during this time. User can run chosen appliances
between 19:30 pm to 23:30 pm. Achievable sav-
ings in energy cost $48785.
Scenario 9, Users are shifting peak demand of
1730 MWh occurring between 10:30 am-19:30
pm to the period between 23:30 pm to 01:30 am.
of 1730 MWh occurring between 10:30 am-19:30
Economic Model
While the concept is designed to be targeting flat-
tening the national electrical demand throughout
the year the concept involves an economic model
based on the maximization of financial benefits
to electricity users. The scheme is applicable in
regions managed by the Australian Energy Man-
agement Operator (AEMO) and other regions
under similar conditions. Usually the electricity
price is high during peak demands and low at
off-peak periods. The concept allows customers
controlling consumption to avoid peak demands
periods. In case the user is on other DSR program
with the supplier, the scheme is still allowing ad-
ditional savings besides the benefits and saving
already achievable through the DSR agreement.
For commercial and industrial customers on
fluctuating energy prices implementing the scheme
will enable achieving immediate financial savings.
For residential customers on flat-rate tariffs, in
Table 2. Results of some scenarios
Scenario NR Time to Curtail
Load
Time to Reconnect
Load
Load to
Curtail
Day Tarrif
18.84 kWh
Night Tarrif
11.32 kWh
Saving
($)
1 17.00 pm
to
19.00 pm
19:00 pm to 21:30 pm 375 70650 NA NA
2 21:30 pm to 23:30 pm 375 70650 49501 21149
3 23:30 pm to 1:00 am 375 70650 42450 28200
4 1:00 am to 3:00 am 375 70650 42450 28200
5 3:00 am to 5:30 am 375 70650 42450 28200
6 5:30 am to7:00 am 375 70650 42450 28200
7 6:30 am to 10:30 am 375 70650 67125 3525
8 10.30 am
to
19.30 pm
19:30 am to 23:30 am 1730 325923 277147 48785
9 23:30 pm to 01:30 am 1730 325923 195836 130096
10 1.30 am to 4:00 am 1730 325923 195836 130096
155
contrast, users are gaining financial benefits from
reducing energy consumptions at certain times a
day; mainly averting peak-load periods or using
off-peak night tariffs.
Multimedia Tool
The concept includes a multimedia tool (CD-
Rom) run on local computers at users premises
to allow implementing and operating the DSR
model continuously. The CD-Rom is containing
an introductory part to address the electricity
peak demand issue to the user, a programming
part, where the user will be recording operation
profiles for the different appliances and an execu-
tive part, where the information is transferred to
the programmable relay, which on turn sending
signals to electronic switches to operate the dif-
ferent appliances.
Significance
Figure 15 shows customers fully curtailing energy
withdrawals at any energy price above $55/MWh
as example. Figure 16 shows achievable energy
savings in Queensland by curtailing energy de-
mand over a certain energy prices.
Referring to Figure 16, the technique is able
to remove 6.1 TWh/year from a total of 52.18
TWh/year if users are setting switches to curtail
own loads at any regional reference prices above
$50/MWh; this is a percentage of 11.7% of the
total demand. In case users chose to curtail loads
at $40/MWh, the savings will be 11.1 TWh/year;
a percentage of 21.2%, and 24.8 TWh/year at a
$30/MWh curtailment; a 47.5%.
Figure 17 depicts the case where coordinated
strategies are able to lead customers to defer
loads from times of peak-demand to times of low-
demands. Such a procedure shall help flatten the
total energy demand to meet a constant average of
5941 MW for Queensland, achieving considerable
improvement in the system utilization and thus in
the whole system economics. In such a procedure
the technique enables deferring 3.26 TWh/year
from peak to off-peak times.
Queensland Government (2009) is describing:
transport is the fourth largest source of
Queenslands greenhouse gas emissions, contrib-
uting 10.4 per cent to Queenslands total emission
profile and the Queensland Government will invest
Figure 15. Day electricity demand curtailed for wholesale regional reference price not to exceed AUD
$55/MWh in Queensland in 2008 (Kamel 2009a)
156
$1.4 million to undertake a trial of low-emission
diesel-electric buses in the public transport fleet.
The Department of Resources Energy and Tourism
(2009) is indicating 1359 PJ (377.5 TWh) were
used for transport in Australia in 2006-07 while
1695 PJ (460.83 TWh) for electricity generation
(1TWh = 3.6 PJ) as illustrated in Figure 18.
The technique offers further the present elec-
trical supply system to be supplying the transport
sector for the use of charging electric vehicles at
low-demand times (at night). Figure 19 depicts
the possibility to utilize 21.52 TWh/year of elec-
trical capacity, mainly of peak-load power stations,
otherwise not used. The procedure helps enhanc-
Figure 16. Achievable energy savings by curtailing energy demand over a certain energy prices in
Queensland (Kamel, 2009a)
Figure 17. Occurrence and average electrical energy demand Queensland (Kamel, 2009a)
157
ing the utilization of present electrical power
stations to approach a plant capacity factor close
to the unity, achieving thus an optimal use of
power plants.
FUTURE RESEARCH DIRECTION
For future research direction on DSR method, more
development of modeling both on-site renewable
energy sources and electricity vehicles connected
to the distribution grid on the Australian national
electricity market, will be considered as follows:
Figure 18. Energy consumption in Australia (Department of Resources Energy and Tourism, 2009)
Figure 19. Electrical energy demand Queensland Vs maximum utilization of generating capacity (Ka-
mel, 2009a)
158
Optimization of on-site renewable energy
sources to distribution grid under DSR
method in Australian national electricity
market.
Deployment of plug-in electricity vehicles
and their impacts to the Australian national
electricity market.
CONCLUSION
The concept is aiming to achieve moderated energy
demand, reduced energy prices and curbed price
volatility, improved grid usability and reliability,
and reduced energy consumption. The concept
is making use of the internet and modern com-
munication technologies to maximize benefit for
users and suppliers. Additionally, the concept is
providing additional capacity more quickly and
more efficiently than new supplies. The flexibility
provided lowers the likelihood and consequences
of forced outages as well. By reducing significant
peaks, the concept is averting the need to use the
most costly-to-run power plants, driving electric-
ity costs down for all electricity users. And most
importantly, by enabling end-users to observe
electricity prices and congestions on the electrical
network it allows them to be positively sharing
responsibility by reducing and optimizing energy
consumption and realizing electricity savings.
The concept can be considered a complementary
effort to concurrent energy suppliers efforts to
mitigate electrical peak demands and the associ-
ated technical and economic detriments. It allows
electricity end-users to smoothen out significant
peaks by curtailing or shifting demand, avoiding
or delaying investments in new infrastructure.
A wide deployment of the scheme will allow a
quite flattened load profile representing thus an
optimized use of the electricity generation and
distribution infrastructure.
REFERENCES
Aalami, H. A., Moghaddam, M. P., & Yousefi, G.
R. (2008). Demand response model considering
EDRP and TOU programs. Paper presented at the
The Transmission and Distribution Conference
and Exposition.
Aalami, H. A., Moghaddam, M. P., & Yousefi, G.
R. (2009). Demand response modeling considering
interruptible/curtailable loads and capacity mar-
ket programs. Applied Energy, 87(1), 243250.
doi:10.1016/j.apenergy.2009.05.041
Albadi, M. H., & El-Saadany, E. F. (2007). De-
mand response in electricity markets: An overview.
Paper presented at the Power Engineering Society
General Meeting, 2007. IEEE.
Australia Energy Market Operator. (2009). Cur-
rent trading interval price and demand graph
Queensland. Retrieved from http://www.aemo.
com.au/data/GRAPH_30QLD1.html
Australian Energy Market Operator. (2010). About
AEMO. Australian Government. Retrieved from
http://www.aemo.com.au/aboutaemo.html
Brinkmam, K. (1980). Introduction in the eco-
nomic of electrical energy. Braunsching, Ger-
many: Vieweg Verlag.
Cain, A., MacGill, I., & Bruce, A. (2010). Assess-
ing the potential impacts of electricity vehicles
on the electricity distribution network. Paper
presented at the The 48th Australian Solar Energy
Society (AuSES), Solar 2010, Canberra.
Chobotov, Y. (1978). Analysis of photovoltaic
total energy system. Paper presented at the Proc
IEEE Photovoltaic Specialist Conference, Was-
ington DC.
Chua-Liang, S., & Kirschen, D. (2009). Quan-
tifying the effect of demand response on elec-
tricity markets. IEEE Transactions on Power
Systems, 24(3), 11991207. doi:10.1109/TP-
WRS.2009.2023259
159
Clement-Nyns, K., Haesen, E., & Driesen, J.
(2010). The impact of vehicle-to-grid on the dis-
tribution grid. Electric Power Systems Research,
81(1), 185192. doi:10.1016/j.epsr.2010.08.007
Clorefeine, A. S. (1980). Economic feasibility of
photovoltaic energy system. Paper presented at
the 14th IEEE Photovoltaic Specialist Conf., San
Diego, California.
Covino, S. (2003). Demand side response 21st
century style. Paper presented at the Power En-
gineering Society General Meeting, 2003, IEEE.
Dam, Q. B., Mohagheghi, S., & Stoupis, J.
(2008). Intelligent demand response scheme for
customer side load management. Paper presented
at the Energy 2030 Conference, 2008. ENERGY
2008. IEEE.
De-Meo, E. A. (1978). Prespectives on util-
ity central station photovoltaic applications.
Solar Energy, 21, 177192. doi:10.1016/0038-
092X(78)90020-8
ment and Innovation. (2009). Transmission and
distribution. Brisbane, Australia: Queensland
Government. Retrieved from http://www.dme.
qld.gov.au/Energy/transmission_and_distribu-
tion.cfm
Department of Employment Economic Devel-
opment and Innovation. (2009a). Electricity.
Brisbane, Australia: Queensland Government.
Retrieved from http://www.dme.qld.gov.au/En-
ergy/electricity.cfm
ment and Innovation. (2009b). National electricity
market. Brisbane, Australia: Queensland Govern-
ment. Retrieved from http://www.dme.qld.gov.au/
Energy/national_electricity_market.cfm
Department of Employment Economic De-
velopment and Innovation. (2010). Electricity
generation in Queensland. Brisbane, Australia:
Queensland Government. Retrieved from http://
www.energyfutures.qld.gov.au/electricity_gen-
eration_in_queensland.cfm
Department of Employment Economic De-
velopment and Innovation. (2010). Electricity
generation in Queensland. Brisbane, Australia:
Queensland Government. Retrieved from http://
www.energyfutures.qld.gov.au/electricity_gen-
eration_in_queensland.cfm
Department of Resources Energy and Tourism.
(2009). Energy in Australia. Canberra, Australia:
Abare. Retrieved from http://www.ret.gov.au/
energy/Documents/facts%20statistics%20publi-
cations/energy_in_aus_2009.pdf
Energy Efficiency and Conservation Authority.
(2009). Domestic-scale distributed generation
guidance for local government. Wellington.
Energy User Association of Australia Document.
(2010). Energy User Association of Australia
document. Australian Government. Retrieved
from http://www.euaa.com.au/index.htm
Ergon Energy. (2010). Domestic tariff. Retrieved
from http://www.ergon.com.au/home/electric-
ity_for_your_home/ep_domestic.asp
Federal Energy Regulatory Commission. (2006).
Assesment of demand response and advanced
metering. Washington, DC. Retrieved from
http://www.ferc.gov/legal/staff-reports/demand-
response.pdf
Fraser,R. S.(2005). Demand side response in the
national electricity market case studies: End use
customer awareness program.
Greening, L. A. (2010). Demand response re-
sources: Who is responsible for implementation
in a deregulated market? Energy, 35, 15181525.
doi:10.1016/j.energy.2009.12.013
160
Hamidi, V., Li, F., & Robinson, F. (2009). Demand
response in the UKs domestic sector. Electric
Power Systems Research, 79(12), 17221726.
doi:10.1016/j.epsr.2009.07.013
Hammons, T. J. (2006). Integrating renewable
energy sources into European Grids. Proceed-
ings of the 41st International Universities Power
Engineering Conference, 2006. UPEC 06.
Hyung Seon, O., & Thomas, R. J. (2008). Demand-
side bidding agents: Modeling and simulation.
10501056. doi:10.1109/TPWRS.2008.922537
International Energy Agency. (2010). Strategic
plan for the International Energy Agency demand-
side management program 2004-2009.
Jin-Ho, K., Tae-Kyung, H., & Kwang-Seok, Y.
(2009). Roadmap for demand response in the
Korean electricity market. Paper presented at the
Power & Energy Society General Meeting, 2009.
PES 09. IEEE.
Jones, T. E. (2004). Australian example of demand
side management actions. Australia.
Kamel, F. (2009a). Sharing communication
network resources for a user-controlled electri-
cal energy consumption. Paper presented at the
QUESTnet, Gold Coast Australia.
Kamel, F. (2009b). User-controlled energy con-
sumption in a transparent electricity system. Paper
presented at the 47th Annual Conference of the
Australian and New Zealand Solar Energy Society,
Townsville Queensland Australia.
Kirschen, D. S. (2003). Demand-side view
of electricity markets. IEEE Transactions on
TPWRS.2003.810692
Lawrence, D. J., & Neenan, B. F. (2003). The status
of demand response in New York. Paper presented
at the Power Engineering Society General Meet-
ing, 2003, IEEE.
Leonard, S. L. (1977). Mission analysis of photo-
voltaic-major mission for the mid-term 1988-2000.
(Solar Energy Conversion-San 1101/PA 8 - 1/3, 3).
Leonard, S. L. (1978). Central station power plant
application for photovoltaic. Paper presented at the
Solar Energy Conversion 13th IEEE Photovoltaic
Sp Conference.
Li, Z., & Yao, T. (2010). Renewable energy bas-
ing on Smart Grid. Paper presented at the Wire-
less Communications Networking and Mobile
Computing (WiCOM), 2010 6th International
Conference on IEEE.
Marwan, M., & Kamel, F. (2010a). Demand-side
response load management modelling encoun-
tering electrical peak demands in Eastern and
Southern Australia - Smart Grid tools. Paper
presented at the Australasian Universities Power
Engineering Conference AUPEC 2010, Christ-
church New Zealand.
Marwan, M., & Kamel, F. (2010b). User-con-
trolled electrical energy consumption towards
optimized usage of electricity infrastructure. Paper
presented at the Southern Region Engineering
Conference, Toowoomba Australia.
Na, Y., & Ji-Lai, Y. (2006). Optimal TOU deci-
sion considering demand response model. Paper
presented at the Power System Technology, 2006.
PowerCon 2006. International Conference on
IEEE.
Narayan, P. K., & Smyth, R. (2005). Electricity
consumption, employment and real income in
Australia evidence from multivariate Granger
causality tests. Energy Policy, 33(9), 11091116.
doi:10.1016/j.enpol.2003.11.010
Nguyen, D. T. (2010). Demand response for
domestic and small business consumers: A new
challenge. Paper presented at the Transmission
and Distribution Conference and Exposition,
2010 IEEE PES.
161
Parvania, M., & Fotuhi-Firuzabad, M. (2010).
Demand response scheduling by stochastic SCUC.
IEEE Transactions on Smart Grid, 1(1), 8998.
doi:10.1109/TSG.2010.2046430
Queensland Competition Authority. (2000). Elec-
tricity demand forecast. NSW, Australia. Retrieved
from http://www.qca.org.au/files/QLDElectrici-
tyDemandForecast.pdf
Queensland Government. (2009). ClimateQ:
Toward a greener Queensland -Transport moving
towards a low carbon future. Retrieved from http://
www.climatechange.qld.gov.au/__data/assets/
pdf_file/0013/24061/ClimateQ_Report_chap-
ter15.pdf
Rahimi, F., & Ipakchi, A. (2010). Overview of
demand response under the smart Grid and mar-
ket paradigms. Paper presented at the Innovative
Smart Grid Technologies (ISGT) 2010, IEEE.
Schweppe, F. C., Caramanis, M. C., Tabors, R.
D., & Bohn, R. E. (1988). Spot electricity price.
Boston, MA: Kluwer Academic.
Stam, E. (2005). Demand response activities in
Finlad. Paper presented at the A Nordic Confer-
ence on Enhancing and Developing Demand
Response in the Energy Market, Copenhagen.
Torriti, J., Hassan, M. G., & Leach, M. (2009).
Demand response experience in Europe: Policies,
programmes and implementation. Energy, 35(4),
15751583. doi:10.1016/j.energy.2009.05.021
Tyagi, R., & Black, J. W. (2010). Emergency
demand response for distribution system contin-
gencies. Paper presented at the Transmission and
Distribution Conference and Exposition, 2010
IEEE PES.
Vos, A. (2009). Effective business models for de-
mand response under the Smart Grid paradigm.
Paper presented at the Power Systems Conference
and Exposition, 2009. PSCE 09. IEEE/PES.
Walawalkar, R., Blumsack, S., Apt, J., & Fernands,
S. (2008). Analyzing PJMs economic demand
response program. Paper presented at the Power
and Energy Society General Meeting - Conver-
sion and Delivery of Electrical Energy in the 21st
Century, 2008 IEEE.
Wen-Chen, C., Yi-Ping, C., & Tzu-Hao, L. (2007).
The competitive model based on the demand re-
sponse in the off-peak period for Taipower System.
Paper presented at the Industrial & Commercial
Power Systems Technical Conference, 2007. ICPS
2007. IEEE/IAS.
Yu, C. W., Zhang, S., Chung, T. S., & Wong, K. P.
(2005). Modelling and evaluation of interruptible-
load programmes in electricity markets. IEEE
Proceedings Generation. Transmission and
gtd:20045138
Zhong, J., Kang, C., & Liu, K. (2010). Demand
side management in China. Paper presented at
the Power and Energy Society General Meeting,
2010 IEEE.
ADDITIONAL READING
Almeida, A., & Rosenfeld, A. (1988). Demand-
side management and electricity end-use effi-
ciency. Kluwer Academic Publishers.
Americas Energy Future Panel on Electricity
from Renewable Resources, & Council, N. R.
(2010). Electricity from Renewable Resources:
Status, Prospects, and Impediments: National
Academies Press.
Barnett, D., & Bjornsgaard, K. (2000). Electric
power generation: a nontechnical guide. Pen-
nWell Books.
162
Barsali, S., Ceraolo, M., & Possenti, A. (2002).
Techniques to control the electricity generation
in a series hybrid electrical vehicle. Energy Con-
version. IEEE Transactions on, 17(2), 260266.
Books
Budde, P., & Whittle, R. (2007). Smart grid:
energy management and broadband. Paul Budde
Communication.
Caves, D. W., Herriges, J. A., Hanser, P., &
Windle, R. J. (1988). Load impact of interruptible
and curtailable rate programs: evidence from ten
utilities [tariff incentives]. Power Systems. IEEE
Transactions on, 3(4), 17571763.
Chen, C. S., & Leu, J. T. (1990). Interruptible
load control for Taiwan Power Company. Power
Systems. IEEE Transactions on, 5(2), 460465.
Chua-Liang, S., & Kirschen, D. (2009). Quantify-
ing the Effect of Demand Response on Electricity
Markets. Power Systems. IEEE Transactions on,
24(3), 11991207.
Collins, M. M., & Mader, G. H. (1983). The tim-
ing of EV recharging and its effect on utilities.
Vehicular Technology. IEEE Transactions on,
32(1), 9097.
Daryanian, B., Bohn, R. E., & Tabors, R. D. (1989).
Optimal demand-side response to electricity spot
prices for storage-type customers. Power Systems.
IEEE Transactions on, 4(3), 897903.
Deep, U. D., Petersen, B. R., & Meng, J. (2009).
A Smart Microcontroller-Based Iridium Satellite-
Communication Architecture for a Remote Re-
newable Energy Source. Power Delivery. IEEE
Transactions on, 24(4), 18691875.
Fox-Penner, P. (2010). Smart Power: Climate
Change, the Smart Grid, and the Future of Electric
Utilities. Island Press.
Garg, H. (1987). Advances in Solar Energy
Technology: Collection and storage systems. D.
Reidel Pub. Co.
Goswami, D. (2007). Advances in Solar Energy:
An Annual Review of Research And Development:
Earthscan in association with The American Solar
Energy Society.
Hajimiragha, A., Caizares, C. A., Fowler, M.
W., & Elkamel, A. (2010). Optimal Transition
to Plug-In Hybrid Electric Vehicles in Ontario,
Canada, Considering the Electricity-Grid Limita-
tions. Industrial Electronics. IEEE Transactions
on, 57(2), 690701.
Hamilton, K., & Gulhar, N. (2010). Taking De-
mand Response to the Next Level. Power and
Energy Magazine, IEEE, 8(3), 6065. doi:10.1109/
MPE.2010.936352
Hori, Y. (2004). Future vehicle driven by electric-
ity and Control-research on four-wheel-motored
UOT electric march II. Industrial Electronics.
Ibitoye, F. I. & Adenikinju, A. (2007) Future
demand for electricity in Nigeria. Applied En-
ergy, 84(5), 492-504. doi: DOI: 10.1016/j.apen-
ergy.2006.09.011
Jazayeri, P., Schellenberg, A., Rosehart, W. D.,
Doudna, J., Widergren, S., & Lawrence, D. (2005).
A Survey of Load Control Programs for Price and
System Stability. Power Systems. IEEE Transac-
tions on, 20(3), 15041509.
Journal Articles
Kreith, F., & Goswami, D. (2007). Handbook of
energy efficiency and renewable energy. CRC
Press.
Masiello, R. (2010). Demand Response the other
side of the curve [Guest Editorial]. Power and
Energy Magazine, IEEE, 8(3), 1818. doi:10.1109/
MPE.2010.936206
163
Molina-Garcia, A., Bouffard, F. & Kirschen, D. S.
(2010) Decentralized Demand-Side Contribution
to Primary Frequency Control. Power Systems,
IEEE Transactions on, PP(99), 1-1.
Momoh, J. (2011). Smart Grid: Fundamentals of
Design and Analysis. John Wiley & Sons, Limited.
Ochoa, L. F. & Harrison, G. P. (2010) Minimiz-
ing Energy Losses: Optimal Accommodation
and Smart Operation of Renewable Distributed
Generation. Power Systems, IEEE Transactions
on, PP(99), 1-1.
Roos, J. G., & Lane, I. E. (1998). Industrial power
demand response analysis for one-part real-time
pricing. Power Systems. IEEE Transactions on,
13(1), 159164.
Rotering, N. & Ilic, M. (2010) Optimal Charge
Control of Plug-In Hybrid Electric Vehicles in
Deregulated Electricity Markets. Power Systems,
IEEE Transactions on, PP(99), 1-1.
Sheen, J. N., Chen, C. S., & Yang, J. K. (1994).
Time-of-use pricing for load management pro-
grams in Taiwan Power Company. Power Systems.
Sovacool, B. (2008). The dirty energy dilemma:
whats blocking clean power in the United States.
Praeger Publishers.
Talukdar, S., & Gellings, C. (1987). Load man-
agement. IEEE Press.
Warkentin, D., & Warkentin-Glenn, D. (1998).
Electric power industry in nontechnical language.
PennWell.
Smart Grid: is a form of electricity network
utilizing digital and computer technology. A
smart grid delivers electricity from suppliers to
consumers using two-way digital communications
to control appliances at consumers homes; this
saves energy, reduces costs and increases reli-
ability and transparency.
Market Conditions: are characteristics of
electricity market, such as electricity price and de-
mand, which could be introduced to all customers.
Switch: is an electrical component that can
break an electrical circuit, interrupting the cur-
rent or diverting it from one conductor to another.
On-Site Renewable Energy: is an energy
sources generated from natural resources, such as
biodiesel, photovoltaic and wind. These energies
are provided by customers.
Electrical Vehicle: is as an electric driven
vehicle, uses one or more electric motors for
propulsion. Electric vehicles include electric cars,
electric trains, electric lorries, electric airplanes,
electric boats, electric motorcycles and scooters
and electric spacecrafts.
Electrical Generation: is the process of creat-
ing electricity from other forms of energy.
Energy Management: is a system of com-
puter-aided tools used by operators of electric
utility grids to monitor, control, and optimize the
performance of the generation and/or transmis-
sion system.
164
Chapter 5
INTRODUCTION
Energy is a vital issue for human society and also
an important topic for economical development.
Energy consumption has physically started with
the industrial revolution. In the initial phase of
the industrial revolution, steam machine has been
utilized to obtain power by using coal. Because of
the hard structure of steam machine and environ-
mentally negative effects of coal, new fossil fuels
emerged such as gas and crude oil. Comparing with
the coal, gas has provided cleaner burning power
Selcuk Cebi
Karadeniz Technical University, Turkey
Cengiz Kahraman
Istanbul Technical University, Turkey
hsan Kaya
Yldz Technical University, Turkey
Soft Computing and
Computational Intelligent
Techniques in the Evaluation of
Emerging Energy Technologies
ABSTRACT
The global warming and energy need requires developing emerging energy technologies for the electric-
ity, heat, and transport markets. The emerging energy technologies aim at increasing effciency of energy
utilization processes from energy sources and diminish CO
2
exhalation. The main aim of the chapter
is to exhaustively present soft computing and computational intelligent techniques in the evaluation of
emerging energy technologies. In the scope of the chapter, classifcation of emerging energy technolo-
gies, their application trends in the literature, a brief explanation for soft computing and computational
intelligent techniques, and literature survey of related techniques on both emerging energy technologies
and energy planning are included. Moreover, technique for order performance by similarity to ideal
solution, analytic hierarchy processes, and their fuzzy structures are introduced.
DOI: 10.4018/978-1-61350-138-2.ch005
165
Soft Computing and Computational Intelligent Techniques
plants and cleaner heating of homes. In addition,
crude oil made possible new transportation options
such as road vehicle and aircraft by invention of the
internal combustion engine (Vanek and Albright,
2008). Nowadays, fossil fuels play an important
role in the transport and stationary. However, it is
thought that current energy systems are not sustain-
able since most of the world primary energy use
is from fossil fuels (Kajikawa et al., 2007). There
are two serious hazards with fossil fuels; the first
one is that the production of fossil fuel has been
predicted to diminish at the middle of this century
(Kajikawa et al., 2007; Jefferson, 2006) and the
other is that fossil fuels have caused emission of
greenhouse gases into the atmosphere and also
global warming. Global warming and fossil fuel
depletion are two of the most important issues of
this century. The considerations of energy security
and climate change force increased societal interest
in technologies that enable a reduction in the use
of fossil fuels. It has been well-recognized that
an effective solution to these issues is to develop
non-carbon-dioxide-emitting and inexhaustible
energy resources and energy technologies (Chen et
al., 2009). Recently, discovering of nuclear power
have provide both to diminish our dependence on
fossil fuel resources, and also to provide electricity
without any emissions of harmful air pollutants.
Although nuclear power is cleaner than many
other forms of energy production and although
nuclear energy can be produced in large quanti-
ties over short periods of time, nuclear power
generates harmful radiation and throwing out of
nuclear waste which is produced by nuclear power
plants is difficult and expensive. Negative effects
of both fossil and nuclear technology, renewable
energy technologies became more advanced and
the range of their applications became broader
(Vanek and Albright, 2008). Therefore, sustainable
and renewable energy sources such as sun, wind,
geothermal, biomass, wave etc. and emerging
energy technologies have been attracted greater
interest as an important concept while energy
planning of a country. In addition, it is thought
that it is an urgent need to develop highly ef-
ficient energy utilization processes from energy
sources effectively and substitute energy sources
since the emerging energy technologies are still
in an early phase of development (Jacobsson and
Bergek, 2004; Kajikawa et al., 2007). Therefore,
recent budgets for governmental research and
development (R&D) for energy technologies
have increased to support emerging energy re-
searches (Hultman and Koomey, 2007). Moreover,
European Union is committed in supporting the
development of emerging energy technologies,
in improving the use of renewable energy, and in
increasing the energy efficiency, to reach global
objectives of sustainability, competitiveness, and
security of energy supply (Segurado et al., 2009).
When any investment or design decision about
energy systems is required, a number of goals or
criteria that are local, regional, or global must be
taken into account. It is possible to classify these
into three categories; (1) Physical goals which meet
physical requirements that make it possible for the
system to operate. (2) Financial goals which are
monetary objectives related to the energy system.
(3) Environmental goals which are the objectives
related to the way in which the energy system im-
pacts the natural environment. Regional or global
impacts include the emissions of greenhouse gases
that contribute to climate change, air pollutants
that degrade air quality and physical effects from
extracting resources used either for materials or
energy (Vanek and Albright, 2008). Therefore,
the emerging technologies have high degree of
uncertainty and it represents the large variety of
opportunities that a new technology has to offer.
This uncertainty creates opportunities for inves-
tors to engage in emerging technologies. Thus, the
relation between uncertainty and the decision of
investors to engage in emerging technologies is
very complex (Mejer et al., 2007). Furthermore,
traditional techniques or conventional (hard)
computing models may not present an effective
solution dealing with the problems in which the
dependencies between variables are complex or
166
ill-defined. And, these solutions may not satisfy
the decision-makers expectations. The selection
of the most suitable sustainable energy technol-
ogy for implementation is a complex problem that
includes multiple conflicting goals or criteria. In
particular, the difficulties dealing with information
about qualitative criteria such as social, cultural
etc. during the evaluation of many conflicting cri-
teria makes the problem complex. For example, a
new energy technology may need to provide good
value for money, low maintenance costs, while
at the same time having a large and stable energy
output and positive social and environmental
effects. Multicriteria decision making methods
(MCDM) among the soft computing methods are
useful and effective tools in order to take into ac-
count simultaneously all the basic aspects of the
decision making problems while other decision-
support tools, such as costbenefit analysis are
not well effective (Burton and Hubacek, 2007;
Karakosta et al., 2010). Therefore, soft computing
(SC) and computational intelligent (CI) techniques
have been widely used in the literature to solve
complex problems.
SC and CI techniques which are based on
copying ability of human mind under uncer-
tainty and imprecision are emerging approaches
(Konar, 2007). The main characteristics of SC
are representation and processing of human mind
and knowledge, qualitative and approximate
reasoning, computing with words, and biologi-
cal models of problem solving and optimization,
and are directly related to intelligent systems and
applications (Karray and Silva, 2004). In this re-
spect, these techniques differ from the respective
conventional computing techniques in that they are
tolerant of imprecision, uncertainty, partialtruth,
and approximation. The soft computing techniques
comprises of fuzzy logic, artificial neural net-
works, probabilistic reasoning and meta-heuristic
techniques such as genetic algorithm, tabu search,
etc. (Altun and Yalcinoz, 2008). The main aim of
this section is to introduce the most known and
the most used SC and CI techniques for complex
problems in emerging energy technologies and
energy science that cannot be satisfactorily solved
using conventional crisp computing techniques. In
particular, multicriteria decision making methods
dealing with decision making problems in emerg-
ing energy technologies and energy science have
been emphasized.
The organization of the chapter is as follows:
Section 2 presents the classification of emerging
energy technologies. Section 3 introduces the
most known MCDM techniques in the literature.
Section 4 includes the most used MCDM methods
(AHP and TOPSIS) and information axiom, newly
presented to literature. Section 5 presents the lit-
erature survey for emerging energy technologies.
The trends for the emerging energy technologies
and SC and CI techniques are given. Finally,
concluding remarks are given in Section 5.
EMERGING ENERGY
TECHNOLOGIES
The emerging energy technologies (EETs) are
classified into six main groups in terms of usage
area. These are building technology, industry
technology, transportation technology, coal
technology, non-conventional fuel technology,
and biomass technology. The classification is
presented in Figure 1(Lee et al., 2009). Emerg-
ing energy technologies for building includes
lighting, air conditioning, building envelop, and
building system technologies. Industrial emerg-
ing energy technologies are classified into waste
heat technologies, common technologies, and
petroleum refinery and fine chemical technolo-
gies. Transportation technologies involve fuel
efficiency improvement technologies, hybrid
electronic technologies, electric, hydrogen fuel
cell, and biodiesel technologies. Coal emerging
energy technologies are classified into direct uti-
lization technology and conversion technologies.
Non-conventional fuel technologies are oil shale/
oil sand technologies and gasification technolo-
167
gies of waste. Biomass technologies are direct
utilization and conversion technologies.
Figure 2 and Figure 3 presents the trends for
emerging energy technologies based on years and
application areas of EETs, respectively. The years
between 1990 and 2010 are divided into four
periods as 1990-1995, 1996-2000, 2001-2005,
and 2006-2010 and numbers of publications in
SCI index are analyzed. Figure 2 presents the
attractiveness of emerging energy technologies
in the literature while Figure 3 presents the most
attractive years for each technology.
According to Figure 2, biomass technology is
the most attractive topic for researchers between
the years 1990-1995. In the same period, building
technology is the second attractive topic. Although
the attractiveness of biomass technology is the
first at the second period, its popularity decreas-
es and the popularity of building and coal tech-
nologies have increased according to previous
period. At the third period, while the popularity
of biomass has continues, the popularity of non-
conventional fuel, transportation, and industry
technologies have increased. Finally, the popular-
ity of coal and transportation technologies have
dramatically increases at the last period.
According to the Figure 3, popularity of
the each emerging energy technology has been
clearly increasing year by year except for non-
conventional fuel technology.
SOFT COMPUTING
AND COMPUTATIONAL
INTELLIGENT TECHNIQUES
Soft computing (SC) and computational intelligent
(CI) techniques are used to obtain the closest solu-
tions to computationally-hard problems. The basis
of the SC and CI techniques are founded on bio-
logical or behavioral phenomena related to humans
or animals, and analogues of these technologies
Figure 1. Classification of EETs
168
exist in many human and animal systems (Uhrig
and Tsoukalas, 1999). SC and CI techniques are
vitally practical tools for many complex problems
since they can tolerant of imprecision, uncertainty,
partial truth, and approximation. However, tra-
ditional hard computing methods are often too
cumbersome for complex problems. They need
a precisely stated analytical model and often a lot
of computational time (Zadeh, 1965). Following
the hardware technology advances, SC and CI
techniques have been intensely studied and im-
proved in the last years, and nowadays practical
applications become a reality. Such techniques
present several advantages when compared to
traditional ones, such as: (i) acquisition of better
results in the optimization processes when no
prior knowledge is available, (ii) possibility of
application to problems to which the conventional
Figure 2. The popularity of EETs with respect to years
Figure 3. Trends for EETs
169
methods are not suitable, (iii) simulation of hu-
man cognition processes, instead of trying to
solve deterministically what is not deterministic
(Schirru et al, 1999). The term soft computing
and computational intelligence in its broadest
sense, encompasses a number of technologies
that includes, but is not limited to, artificial neural
networks (ANN), genetic algorithms (GA), fuzzy
logic models, ant colony techniques (AC), tabu
search (TS), expert systems (ExS), multicriteria
decision making methods etc. These are the
well known soft computing and computational
intelligence techniques in the literature. Beside
of these techniques, genetic programming (GP),
artificial immune system (AIS), harmony search
(HS), scatter search (SC), variable neighborhood
search (VNS), pattern search (PS), differential
evolution (DE), evolutionary programming (EP),
evolutionary strategies (ES), simulated anneal-
ing (SA), particle swarm optimization (PSO),
swarm intelligence (SI) etc. are the relatively new
developed tools. However, these techniques are
currently not as popular especially with regard to
the emerging energy technology. Thus, only the
main tools are briefly highlighted in this section.
The artificial neural networks (ANN) were
first introduced by McCulloch et al. (1943), who
suggested that the biological function of the
human brain could be emulated by a simplified
computational model (Saridakis and Dentsoras,
2008). The technique is a computational model
and it is inspired of biological neural networks.
The ANNs are structured with a set of inter con-
nected layers, each of them composed of nodes;
the typology of connections and nodes (called
neurons) characterizes the different typologies of
neural networks (Bertini et al. 2010). Outputs of
neurons in a given layer are the inputs of neurons
in the next layer. First and last layers are called
input and output layers, respectively, while those in
between are hidden layers. Neurons are character-
ized by a transfer function, which is applied to an
appropriate function of inputs. The main advantage
of ANN is that it does not need any mathemati-
cal model, since it learns from historical data to
recognize non-evident relations and patterns in a
set of inputoutput variables, without any prior
assumption about their nature (Pena et al., 2010).
The structure of ANN is given in Figure 4.
Genetic algorithms (GAs) are global search
and optimization techniques motivated by the
process of natural selection in biological system
(Gen and Cheng, 2000; Kaya, 2009). GA as a
field of study was initiated and developed in the
early 1970s by John Holland (Holland, 1975,
1992) and his students, but its applications to
real-world practical problems was almost three
decades in developing. GA approaches are good
at solving the ill-posed problems such as non-
convex functions, non-differentiable functions,
domains not connected, badly behaved functions,
multiple local optima, and multiple objectives
Figure 4. The structure of ANN
170
(Miranda et al., 1998). The main advantages of
GA is that it present an approximate solution in
relatively short time compared with other random
searching methods, such as simulated annealing
or dynamic programming (Won and Park, 2003).
The first population of GA is a randomly selected
initial solution set. To obtain an optimum solution,
a search is conducted by moving from the initial
population of individuals to a new population
using genetics-like operators such as selection,
crossover and mutation, which are inspired from
the mechanics of natural selection and genetics
encountered in natural life. Each individual rep-
resents a candidate to the optimization solution
and is modeled by a value called chromosome.
The GA operators perform task on the chromo-
some, in the reproduction process, in order to
produce new generations so that solution at the
global optimum may be obtained. The operation
is based on a selective nature, i.e. the best candi-
dates in terms of fitness are chosen as parent so
that the new generation holds best genetic heritage.
For this purpose, a fitness function assigns a fit-
ness value to each individual within the population.
This fitness value is the measure for the quality
of an individual. The basic optimization procedure
involves nothing more than processing highly fit
individuals in order to produce better individuals
as the search progresses. A typical genetic algo-
rithm cycle involves four major processes of
fitness evaluation, selection, recombination and
creation of a new population. Based on fitness
criterion, poorer performing individuals are
gradually taken out, and better individuals have
a greater possibility of conveying genetic informa-
tion to the next generation (Altun and Yalcinoz,
2008). The cycle of GA is illustrated in Figure 5.
Ant colony (ACO) which is used for solving
combinatorial optimization problems is a coop-
erative search algorithm inspired by the behavior
of ants in finding paths from the nest to food (Yang
and Zhuang, 2010). In the early 1992, ACO was
proposed to literature as a metaheuristic optimiza-
tion tool as by Dorigo (1992). High concentrations
of pheromones indicate more favorable paths that
other members should follow in order to reach
the optimal solution (Al-Rashidi and El-Hawary,
2009). ACO is based on the indirect communica-
tion of a colony of simple agents, called ants,
mediated by (artificial) pheromone trails. The
pheromone trails in ACO serve as distributed,
numerical information which the ants use to
Figure 5. The cycle of genetic algorithms (Konar, 2000)
171
probabilistically construct solutions to the problem
being solved and which the ants adapt during the
algorithms execution to reflect their search ex-
perience (Dorigo and T. Sttzle, 2003).
Tabu search (TS) is another algorithm which is
used for the solution of combinatorial optimization
problems like the traveling salesman problem.
TS method originally proposed by Fred Glover
(Glover, 1986) is based on neighborhood search
procedure such that the algorithm iteratively
moves from a solution to another solution in the
related neighborhood, until it reaches any stopping
criterion. The basic principle of TS is to pursue lo-
cal search whenever it encounters a local optimum
by allowing non-improving moves; cycling back
to previously visited solutions is prevented by the
use of memories, called tabu lists, which record
the recent history of the search (Grandeu, 2003).
TS is a powerful algorithmic approach that has
been applied with great success to many difficult
combinatorial problems. The best feature of TS
is that it easily handle complicating constraints.
Thus, TS heuristics generally find good solutions
very early in the search. Both depth and breadth
need to be achieved in the searching process.
Although depth is usually not a problem for TS,
breadth can be a critical issue (Gendreau and
Potvin, 2010).
MCDM techniques are also known as SC and
CI techniques. MCDM techniques are classified
into two groups as Multiple Objective Decision
Making (MODM) and Multiple Attribute Deci-
sion Making (MADM). The difference between
MADM and MODM is that MADM is associated
with problems of which numbers of alternatives
have been predetermined. The Decision Maker
(DM) is to select/rank a finite number of courses of
action. On the other hand, MODM is not associated
with the problems in which alternatives have been
predetermined (Lai and Hwang, 1994). In other
words, MODM techniques present optimization
of an alternative or alternatives on the bases of
prioritized objectives while MADM techniques
present selection of an alternative from a set of
alternatives based on prioritized attributes of the
alternatives.
MCDM Techniques
The well-known and the most used MADM tech-
niques in the literature are Analytic Hierarchical
Process (AHP) (Saaty, 1980), Technique for
Order Performance by Similarity to Ideal Solu-
tion (TOPSIS) (Hwang and Yoon, 1981), Simple
Additive Weighting method (SAW) (Hwang and
Yoon, 1981), Elimination By Aspects (EBA)
(Tversky, 1972), ELimination and Choice Ex-
pressing REality (ELECTRE) (Bernard, 1968),
Preference Ranking Organisation METHod for
Enrichment Evaluations (PROMETHEE) (Brans
and Vincke, 1985). Besides these methodologies,
relatively new decision making methodologies
such as information axiom method (Suh, 1990),
VlseKriterijumska Optimizacija I Kompromisno
Resenje (VIKOR) (Opricovic and Tzeng, 2004)
etc. or integrated methods like Hierarchical TOP-
SIS (Kahraman et al, 2007), have been proposed
to solve complex decision making problems (Cebi
and Kahraman, 2010a). In the literature, there
are also fuzzy MCDM techniques such as fuzzy
TOPSIS (Chen, 2000), fuzzy AHP (Buckley, 1985,
Chang, 1996; Laarhoven and Pedrycz, 1983),
fuzzy VIKOR (Opricovic and Tzeng, 2004), fuzzy
information axiom (Kulak and Kahraman, 2005a;
2005b; Kulak et al., 2005; Kahraman and Cebi,
2009) etc. in order to make decision makers cope
with incomplete and vague information.
In this chapter AHP, TOPSIS, Fuzzy AHP,
Fuzzy TOPSIS and information axiom among
the MCDM techniques are presented.
Analytic Hierarchy Process
The analytic hierarchy process (AHP) is based on
pairwise comparisons. This provides an advantage
when there is no any quantitative information
about the problem. The mains steps of the AHP
is as follows (nt and Soner, 2008);
172
Step 1. The pairwise comparison matrix is
constructed. Let C
1
,C
2
,,C
n
symbolize a set of
elements, while a
ij
represents a quantified judg-
ment on a pair of elements C
i
and C
j
. The relative
importance of two elements is rated using a scale
with the values 1, 3, 5, 7 and 9, where refers to
equally important, slightly more important,
strongly more important, demosrably more
important, and absolutely more important,
respectively (Saaty, 1980). This produces a nn
square matrix A as follows:
A a
C
C
C
a a
a
a
a a
ij
n
n
n
n n
=

1
2
12 1
12
2
1 2
1
1
1
1 1
1
.
(1)
where a
ij
=1 when i=j, and a
ij
=1 / a
ij
for i,j=1,2,,n.
Step 2. The comparison matrix is normalized
and weights are obtained.
A
a
t
a
t
a
t
a
t
a
t
a
t
n
n
n
n
n n
=
1
1
1
12
2
1
21
1
2
1
1
2
2
(2)
t a
j ij
i
n
=
=
1
(3)
where i and j represents row and column number,
respectively. Then, a set of numerical weights,
w
1
, w
2
, , w
n
are obtained by averaging of rows.
Step 3. The consistency analysis is done;
A w w
i i
* *
max
= (4)
and the consistency index is obtained as follows:
CI
n
n
=

( )
max
1
(5)
CR
CI
RI
= (6)
where RI represents the average consistency index
over numerous entries of same order reciprocal
matrices (Saaty, 1980). If CR0.1the matrix is
accepted as consistent, otherwise the evaluation
procedure is repeated until consistency is satisfied.
TOPSIS Method
In TOPSIS method, the distance to both positive
and negative ideal solution is calculated and an
alternative is the best if the alternative has the
shortest distance from the positive ideal-solution
and the longest distance from the negative-ideal
solution (Yoon and Hwang, 1981). The TOPSIS
method can be summarized as follows (nt and
Soner, 2008):
Let A
1
,A
2
,,A
j
be the j different alternatives.
For alternative A
j
the rating of the ith aspect is
denoted by f
ij
which is the value of the ith criterion
function for the alternative A
j
, and n is the number
of the criteria. Then
Step 1. The normalized decision matrix is
obtained by;
r
f
f
ij
ij
ij
j
n
=
=
2
1
, j=1,2,,J and i=1,2,,n
(7)
Step 2. The weighted normalized value v
ij
is
calculated as:
v w r
ij i ij
= j=1,2,,J and i=1,2,,n (8)
where w
i
is the weight of the ith attribute or cri-
terion, and
173
w
i
i
n
=
=
1
1. (9)
Step 3. The positive ideal solution A*and the
negative ideal solution A
-
are determined as:
A v v v i I v i I
i
j
ij
j
ij

= =
{
} { , ..., } (max ), (min )
' ''
1

(10)
A v v v i I v i I
i
j
ij
j
ij

= =
{
} { , ..., } (min ), (max )
' ''
1

(11)
where I is a set of benefit criteria and I is a set
of cost criteria.
Step 4. The distance to positive ideal solution
is calculated by;
D v v
j ij i
i
n

=
=
( )
2
1
, j=1,2,,J (12)
Similarly, the distance to the negative-ideal
solution is obtained by;
D v v
j ij i
i
n

=
=
( )
2
1
, j=1,2,,J (13)
Step 5. The relative closeness coefficient of
the alternative A
j
is defined by;
C
D
D D
j
j
j j

=
+
, j=1,2,,J (14)
Step 6. The alternatives are ranked with respect
to closeness coefficients.
Fuzzy Analytic Hierarchy Process
In the literature, there are at least three different
fuzzy AHP algorithms. The first algorithm in fuzzy
AHP was proposed by van Laarhoven and Pedrycz
(1983), which compared fuzzy ratios described by
triangular membership functions. Then, Buckley
(1985) presented fuzzy priorities of comparison
ratios whose membership functions trapezoidal.
And the last one is proposed by Chang (1996) with
the use of triangular fuzzy numbers for pairwise
comparison scale of fuzzy AHP, and the use of
the extent analysis method for the synthetic ex-
tent values of the pairwise comparisons. In this
chapter, Buckleys (1985) fuzzy AHP approach
is presented in detail.
Step 1. Pairwise comparison matrices are
constructed. Each element( c
ij
) of the pairwise
comparison matrix (C) is a linguistic terms pre-
senting which is the more important of two cri-
teria. The pairwise comparison matrix is given
by;

C
c c
c c
c c
k
n
n
n n
=
1
1
1
12 1
21 2
1 2
, k=1,2,3,,K
(15)
where
C
k
is a pairwise comparison matrix belongs
to k
th
expert for FR
m
. For the evaluation procedure,
the linguistic terms given in Table 1a are used.
Arithmetic mean is used to aggregate expert
opinions.
Step 2. Weights are calculated. At first, the
fuzzy weight matrix is calculated by Buckleys
Method as follows (Hsieh et al., 2004);
r c c c
i i i in
n
= ( ... )
/
1 2
1
(16)
w r r r r
i i n
= + + +

( ... )
1 2
1
(17)
where r
i
is the geometric mean of fuzzy com-
parison value and w
i
indicated by triangular fuzzy
numbers w L M U
i i i i
( , , ) is fuzzy weight of i
th

criterion.
174
Step 3. After the fuzzy relative weight matrix is
obtained, defuzzification process which converts
a fuzzy number into a crisp value is utilized. At
first, fuzzy numbers will be defuzzified into crisp
values and then normalization procedure will be
applied. For the defuzzification process, centroid
method, which provides a crisp value based on
the center of the gravity, is selected since it is
the most commonly used method (Opricovic and
Tzeng, 2004).
w
w
w
L M U
w
i
i
j
j
n
i i i
j
j
n
= =
+ +
= =

1 1
(18)
Fuzzy TOPSIS
Fuzzy TOPSIS methodology consists of four
main steps which are listed below (Chen, 2000):
Step 1. Evaluation values are normalized be-
cause of two different scales. To avoid the com-
plicated normalization formula used in classical
TOPSIS, the linear scale transformation is used
to obtain normalized fuzzy decision matrix de-
noted by
R .
R r r
a
c
b
c
c
c
ij mxn ij
ij
j
ij
j
ij
j
= =
[ ] , ,
* * *
(19)
where c c
j
i
ij
*
max = if criterion is benefit. Other-
wise, if criterion is cost, following equation is
used.
r
a
c
a
b
a
a
ij
j
ij
j
ij
j
ij
=

, , where a a
j
i
ij
= min (20)
Step 2. The weighted normalized fuzzy deci-
sion matrix is constructed as follows:

V v i m j n
v r w
ij mxn
ij ij j
= = =
=
[ ] , , , , , ,
.
1 2 1 2

(21)
Step 3. Then, the distances (d
i
*
, d
i
) of each
alternative from fuzzy positive-ideal solution
(FPIS, A*) and fuzzy negative-ideal solution
(FNIS, A
-
) are calculated, respectively.
A v v v
where v
A v v v
n
j
n
* * * *
*
( , , ...., )
( , , )
( , , ...., )
=
=
=

1 2
1 2
1 1 1
wwhere v
j
=( , , ) 0 0 0
(22)
d d v v i m
d d v v i m
i ij j
j
n
i ij j
j
* *
( , ) , , ,
( , ) , , ,
= =
= =
=

1
1 2
1 2

==
1
n
(23)
Table 1a. Linguistic scale for weight matrix (Hsieh et al., 2004)
Linguistic scales Scale of fuzzy number
(1,1,3) Equally important (Eq)
(1,3,5) Weakly important (Wk)
(3,5,7) Essentially important (Es)
(5,7,9) Very strongly important (Vs)
(7,9,9) Absolutely important (Ab)
175
Step 4. A closeness coefficient (CC
i
) is calcu-
lated by using d
i
*
and d
i
in Equation (10).
CC
d
d d
i m
i
i
i i
=
+
=
*
, , , 1 2 (24)
The alternatives are ranked via CC
i
. An alter-
native which is closest to the FPIS (A*) and the
farthest from FNIS (A
-
) among all alternatives is
the best alternative. CC
i
value of the best alterna-
tive is the biggest and it approaches to 1.
Information Axiom
Information axiom is the second axiom of the
axiomatic design methodology and it is used for
decision making tool. The second axiom is used
to select the best alternative when two or more
alternatives satisfy the first axiom. The informa-
tion axiom states that the alternative having the
highest probability of success is the best design.
In another word, the alternative having the least
information content is the best (Suh, 2001). In-
formation content (I
j
) is defined in terms of prob-
ability of satisfying FR
j
(p
j
), where j
th
functional
requirement. The information content is given by
Equation(25).
I
p
j
j
= log
2
1
(25)
The logarithmic function is chosen so that the
information content can be additive when there
are many FRs that must be satisfied simultane-
ously (Suh, 1990). If there is more than one FR,
the information content of a system (I
system
) is
calculated by Equation 26.
I p p
system j
j
m
j
j
m
= =
= =

log log ( / )
2
1
2
1
1
(26)
The probability of success (p
j
) is calculated
by Equation 27 (Figure 6).
p
common range
system range
= (27)
where system range and common range are defined
by the area of system range and by the intersection
area of the system range and design range which
is determined by a functional requirement of the
design, respectively.
Kulak and Kahraman (2005a, 2005b) devel-
oped the information axiom to be used under
fuzzy environment for the solution of the complex
decision making problems. The main difference
Figure 6. System, design and common ranges
176
between the conventional information axiom and
the fuzzy information axiom is that the fuzzy in-
formation axiom uses fuzzy numbers. Kulak et al.
(2005) developed unweighted and weighted multi
attribute axiomatic design approaches includ-
ing both crisp and fuzzy criteria and applied the
methodology to an equipment selection problem.
Kulak (2005) developed a decision support system
for the selection of a material handling system.
Then, Kahraman and Cebi (2009) extended the
usability of the fuzzy information axiom for vari-
ous decision making problems.
LITERATURE SURVEY
FOR EMERGING ENERGY
TECHNOLOGIES
In this section, a literature survey has been pre-
sented. At first, the general applications of SC and
CI techniques and trends on energy problems in
last five years are presented. Then, the applications
on EETs have been given. The literature survey
is divided into two categories since the applica-
tions of SC and CI techniques on the evaluation
of emerging energy technologies do not yield a
huge material in the literature; one of them is the
studies related to MCDM and the other is related
to rest of SC and CI techniques. Some studies for
the last five years deal with the applications of SC
and CI techniques on energy planning have been
given in Table 1. Table 2 presents the applications
of SC and CI techniques on EETs. The number of
presented publications on energy problems does
not include all in the literature. Since we aimed
to demonstrate the usefulness and possible ap-
plications of SC and CI techniques, the papers
published in last five years are presented.
Except MCDM methods, the other SC and CI
techniques are widely used in the literature to
forecast energy demand or energy supply. In
particular, ANN is the most widely used technique
in all to predict energy demand or supply (Table
1b). In addition, there are a few studies in the
literature to take into account EETs based on SC
and CI techniques. For instance, ANN and GA
techniques are generally used for sizing photo-
voltaic (PV) technology, since these methods
present a good solution (Table 2). The performance
of the PV systems depends upon several factors
such as solar radiation, ambient temperature and
wind speed. In order to size a PV system so that
it can work properly, efficiently and economi-
cally to meet the desired load requirements under
the local meteorological conditions, the charac-
teristic performance of each component in the PV
system is required (Mellit et al. 2008). According
to Table 2, the most used technique is ANN too
for sizing problem of PV system. Table 2 indicates
that SC and CI techniques have an increasing
popularity on the sizing of PV-systems. In par-
ticular, the numbers of the applications using SC
and CI technologies are
In the literature, MCDM methods have become
increasingly popular in decision-making for en-
ergy planning. Table 3 illustrates applications of
MCDM methods on the solution of energy plan-
ning in last five years.
Furthermore, the applications of MCDM
methods on emerging energy technology are more
than other SC and CI techniques. These studies
are summarized as follows;
Beccali et al. (2003) used ELECTRE III method
for the selection of suitable energy technologies
in renewable energy technology diffusion plan.
In the study, technological alternatives given in
Table 4 are evaluated under the criteria which
are classified into four main groups such as
technological criteria, energy and environmental
criteria, social and economic criteria (Table 5).
The evaluations are implemented in three differ-
ent scenarios; environmentaloriented scenario,
economy-oriented scenario, and energy saving
and rational use scenario.
Pohekar and Ramachandran (2006a, 2006b)
assessed the utility of parabolic solar cooker (PSC)
under the criteria techno-economic, social, be-
havioral and commercial comparing with other
177
contemporary cooking energy devices. In the
paper, the alternatives; chulha, improved chulha,
kerosene stove, biogas stove, lpg stove, micro
wave oven, electric oven, solar box cooker, para-
bolic solar cooker are evaluated under thirty
criteria. The criteria are categorized under five
main criteria such as technical, economical, social,
behavioral, and commercial (Table 6).
Burton and Hubacek (2007) investigated a
local case study of different scales of renewable
energy provision for local government in the UK.
In the study, the perceived social, economic, and
environmental cost of the small-scale energy were
compared with the large-scale alternatives. In
order to investigate whether the energy could have
been generated at a lower social, economic, and
environmental cost if large-scale projects had
been available, a multi-criteria decision making
(MCDM) methodology, MACBETH, was used
to compare the advantages and disadvantages of
a number of different renewable energy tech-
nologies. MACBETH method proposed by (Bana
Table 1b. The applications of SC and CI techniques related to energy planning for last five years
Author ANN ACO GA TS Other
Eynard et al. (2010)
Paolli et al. (2010)
Li and Shi (2010)
Li and Su (2010)
Cinar et al. (2010)
Azadeh et al (2010) Network based fuzzy inference system
Pao (2009)
nler (2008) SI
Neto and Fiorelli (2008)
Abdel-Aal (2008)
Azadeh et al (2008a)
Gonzlez-Romera et al. (2008)
Azadeh et al (2008b) Fuzzy System
Sozen and Arcaklioglu (2007)
Toksari (2007)
Pao (2007)
Hamzaebi (2007)
Azadeh and Tarvendian (2007)
Ediger and Akar (2007) ARIMA
Gareta et al. (2006)
Murat and Ceylan (2006)
Ozturk et al. (2005)
Ceylan et al. (2005),
Haldenbilen and Ceylan (2005)
Sozen et al. (2005a, 2005b)
Gonzlez and Zamarreo (2005)
Dong et al (2005) Vector machine
178
e Costa and Vansnick, 1997). In the basis of the
MACBETH, a series of pairwise comparisons,
where a decision-maker is asked to specify the
difference in attractiveness between all of the
alternatives is included. In the study eight renew-
able energy technologies of differing scales are
considered such as solar photovoltaic, micro-wind,
micro-hydro, large-scale wind, large-scale hydro,
energy from waste, landfill gas and biomass (wood
chippings) based on the definition of renewable
energy used by the UK government under eight
criteria such as; capital cost, operation and main-
tenance cost, generation capacity, lifespan, carbon
emissions, noise, impact upon the natural environ-
ment and social effects. These criteria were se-
lected in order to consider the viability of renew-
able energy developments and the need to have
a breadth of criteria covering social, economic
and environmental issues (Burton & Hubacek,
2007).
Doukas et al. (2007) presented a direct and
transparent MCDM approach, using linguistic
variables, to assist policy makers in formulating
technological energy priorities towards a sustain-
able energy system. In the paper, technologies
given in Table 7 are handled under following
criteria; economic (including investment cost
criterion and economic visibility using payback
period criterion), environmental (including con-
tribution to confrontation of the climate change
phenomenon criterion and effects on natural
environment criterion), technological (includ-
ing efficiency rate criterion and knowledge of
the innovative technology criterion), and social
(including contribution to employment oppor-
tunities creation criterion and contribution to
Table 2. SC and CI techniques for sizing problem of PV technology
ANN GA Fuzzy Logic Other
Ben Salah and Ouali (2011) Fuzzy Systems
Mellit et al. (2010)
Mellit (2010)
Liao(2010) Genetic K means algorithm
Venayagamoorthy and Welch (2010)
Chaouachi et al. (2010)
Thiaux et al. (2010)
Chang(2009)
Ashhap (2008)
Mellit and Benghanem (2007)
Dufo-Lopez et al. (2007)
Senjyua et al.(2007)
Mellit et al. (2007)
Karatepe (2006)
Hontaria et al. (2005)
Mellit et al.(2005)
Zang and Bai (2005)
Dufo-Lopez, Bernal-Agustin. (2005)
Bahgat et al. (2004)
Hussein et al. (2004) Learning networks
Benlarbi et al. (2004)
179
Table 3. The applications of MCDM methods on energy planning
AHP TOPSIS OTHER APPLICATION
Ma et al. (2005) Determining the land-suitability assessment of potential
energy systems
Madlener and Stagl (2005) PROMETHEE Designing the renewable energy policy instruments
Lee et al. (2007) Determining the priorities in technology development for the
energy efficiency and greenhouse gas control plans
Madlener et al. (2007) PROMETHEE Determining the best energy planning
Georgiou et al. (2008) ELECTRE III Evaluation of projects on clean technologies
Lee et al. (2008) Evaluation of hydrogen energy technologies
Mrz (2008) ELECTRE III Determining the most compromise scenarios of the commu-
nity heating system modernization and development
Papadopoulos and Karagi-
annidis (2008)
ELECTE III Determining the achievable penetration of renewable energy
sources
Thakker et al. (2008) Selection of wave energy extraction turbine blade material
Buchholz et al. (2009) Multi-criteria
analysis
Design and implementation of sustainable bioenergy projects
Cavallaro (2009) Multi-criteria
analysis
Assessment of concentrated solar thermal technologies
Kahraman et al. (2009) Axiomatic Design Selection of the best renewable alternative
Kowalski et al. (2009) Multi-criteria
analysis
Evaluation of renewable energy scenarios
Lee et al. (2009) Determining the priority the weights of energy Technologies
Madlener et al. (2009) ELECTRE III Comparison of the renewable energy conversion plants
Rivire and Marlair (2009) A new method Ranking the risks pertaining for biofuel chains
Cavallaro (2010a) ELECTRE III Selection of production processes of thin-film solar technol-
ogy
Cavallaro (2010b) Comparing different heat transfer fluids
Ghafghazi et al. (2010) PROMETHEE To evaluate and rank energy sources
Heo et al. (2010) Evaluation of factors for renewable energy
Kahraman and Kaya
(2010)
Determining the best energy policy
Kaya and Kahraman
(2010)
VIKOR Determining the best renewable energy alternative and energy
production site for Istanbul
Lee et al. (2010) Data envelopment
analysis
Evaluation of relative efficiency of the research and develop-
ment performance in the national hydrogen energy technol-
ogy development
Rovere et al. (2010) Multi-criteria
analysis
Selection of the best renewable energy source
Lee et al. (2010) Prioritizing the weights of hydrogen energy technologies
Nixon et al. (2010) Determining the best solar thermal collection technology for
electricity generation
Sadeghzadeh and Salehi
(2010)
Determining the strategic technologies of fuel cells as con-
verters in the automotive industry
180
Table 4. Energy technologies (Beccali et al., 2003)
Energy source Technology/Action
Solar energy 1 Domestic solar water heaters
2 Solar water heating for large demands at low levels of temperature
3 PV roofs: Grid connected system generating electric energy (without storage)
Wind energy 4 Wind turbines (grid connected)
Hydraulic energy 5 Hydro plants in derivation schemes
6 Hydro plants in existing water distribution networks
Biomass 7 High efficiency wood boilers
8 CHP plants fed by agricultural wastes or energy crops
Animal manure 9 CHP plants fed by biogas
Energy saving in residential and industry sectors 10 Building insulation
Combined Heat and Power (CHP) 11 High efficiency lighting
12 High efficiency electric householders appliances
13 High efficiency boilers
14 Plants coupled with refrigerating adsorption machines
Table 5. Main and sub criteria used for evaluation of energy technologies (Beccali et al., 2003)
Main Criteria Sub-criteria
Technological criteria Targets of primary energy saving in regional scale
Technical maturity, reliability
Consistence of installation and maintenance requirements with local technical know-how
Continuity and predictability of performances
Cost of saved primary energy
Energy and environmental criteria Sustainability according to greenhouse pollutant emissions
Sustainability according to other pollutant emissions
Land requirement
Sustainability according to other environmental impacts
Social and economic criteria Labor impact
Market maturity
Compatibility with political legislative and administrative situation
181
regional development criterion). In the paper a
two-staged method was developed using linguistic
ordered weighted averaging (LOWA) and ordered
weighted maximum (OWMAX) operators for the
technologies assessment. The energy technologies
are firstly identified and the most promising is
chosen based on the countrys specific priorities
and objectives.
Lee et al. (2007) determined the priorities for
energy technology development in the sectors of
energy efficiency improvement and greenhouse
gas (GHG) control plans (EGCP) for a new na-
tional energy and resource technology R&D plan
(NERP)by using the analytic hierarchy process
(AHP). They focused on the areas of energy ef-
ficiency improvement and GHG control. In the
paper, in order to determine priorities of energy
technologies, 9 energy technologies from 3 sectors
given in Table 8 in terms of GHG control and 34
energy technologies from six sectors given in
Table 9 in terms of energy efficiency improvement
are evaluated under a set of criteria. The hierarchy
of the criteria consists of five main criteria and
five sub criteria. The main criteria are United
Nations framework convention on climate change,
economic spin-off, technical spin-off, urgency of
technology development, and quantity of energy
use. The sub-criteria are possibility of developing
technologies (domestic technical level and pos-
sibility of commercialization), potential quantity
of energy saving (quantity of energy saving and
quantity CO2 saving), market size (domestic
market size, potential export market size, and
effect of generating hiring), investment cost, and
ease of energy use (ease of product, applicable
area of other technologies).
In another study, Lee et al. (2009) prioritized
the energy technologies against high oil prices in
the energy technology roadmap (ETRM) in order
to allocate R&D budget strategically. The fuzzy
analytic hierarchy process, which integrates the
fuzzy theory into the classical AHP approach, is
utilized to generate the weights of energy technol-
ogy against high oil prices of the ETRM. In the
paper, four criteria which are economical spin-off,
possibility of commercialization, inner capacity,
and technical spinoff are handled. Energy tech-
nologies against high oil prices such as building
Table 6. Classification of criteria (Pohekar & Ramachandran, 2006a, 2006b)
Main Criteria Sub-criteria Main Criteria Sub-criteria
Technical Fuel consumption Social Pollution hazards
Cooking time Human drudgery
Durability Overall safety
Quality, reliability Behavioral Aesthetics
Sophistication level Motivation to buy
Size/weight Taste of food
Ruggedness Cleanliness of utensils
Continuity of use Ease of operation
Need for tracking Type of dishes cooked
Nutrition value of food Need for additional cooking system
Economic Initial cost Commercial Improvement in models
Fuel cost per month Spares and after sales service
Maintenance cost per year Distribution network
Available subsidy Market research
Rate of interest on loan Need for user training
182
technology, industry technology, transportation
technology, coal technology, non-conventional
fuel technology, and biomass technology are as-
sessed.
Lee et al. (2008; 2010a) developed an algorithm
by integrating fuzzy analytic hierarchy process
(Fuzzy AHP) and the data envelopment analysis
(DEA) in order to measure the relative efficiency
of the R&D performance in the national hydrogen
energy technology development. On the first stage,
the fuzzy AHP was used to reflect the vagueness of
human thought. On the second stage, the DEA ap-
proach was used to measure the relative efficiency
of the national R&D performance in the sector of
hydrogen energy technology development with
economic viewpoints. In the paper, the following
criteria was used for the assessment; technological
status, hydrogen technology infrastructure, R&D
human resources, R&D budgets
Lee et al. (2010b) determined the priorities
for hydrogen energy technologies by using fuzzy
AHP. In the study, four criteria which are economic
impact, commercial potential, inner capacity, and
technical spin-off are handled to evaluate and
determine the weights of five hydrogen energy
technologies which are hydrogen production,
hydrogen separation and storage, polymer elec-
trolyte membrane fuel cell, direct ethanol fuel
cell, and solid oxide fuel cell.
Table 7. Energy technologies (Doukas et al., 2007)
Main Group Sub Group
The natural fossil fuels technologies Pressurized Fluidized Bed Combustion
Pressurized pulverized coal combustion
Natural Gas Combined Cycle
The hydrogen technologies Molten Carbonate Fuel Cell;
Fuel Cell/Turbine Hybrids
Renewable energy technologies Biomass Co-firing;
Biomass Gasification;
Off-shore Wind farms;
Large scale Wind farms;
Building Integrated Photovoltaics
Table 8. Technologies for GHG (Lee et al., 2007)
Sectors Technologies
GHG tech CO2 capture storage and conversion tech
Non-CO2 gas tech
Clean fossil tech Advanced combustion tech
Next-generation clean coal tech
Clean petroleum and conversion tech
DME tech
GTL tech
Gas hydrate
GHG policy GHG mitigation policy
183
Table 9. Technologies for energy efficiency improvement (Lee et al., 2007)
Sectors Technologies
Industry High-efficiency drying tech
Fine chemical processing
Energy conversion tech
Unutilized energy tech
Energy material tech
High-efficiency dying tech
Cold storage and freezing tech
Process automation and intelligence tech
Supercritical fluid process tech
Evaporation and distillation tech
Adsorption separation tech
Membrane separation tech
Crystallization tech
Building Green building tech
Building renovation tech
High-efficiency HVAC tech
CHP tech
Efficiency policy Energy efficiency improvement policy
Transportation High efficiency low emission vehicles tech
Electricity Superconductor tech
Electric power conversion tech
High-efficiency electric heating tech
Energy storage tech
Standby power saving tech
Common utilities Heat exchange tech
Boiler tech
High-efficiency furnace tech
Burner tech
Motor tech
Lighting tech
Fluid machine tech
6 major appliances
DSM tech
184
Klemes et al. (2009) developed software
named Early Market Introduction of New En-
ergy Technologies (EMINENT). The software
was developed to analyze the potential impact
of new and underdeveloped energy technolo-
gies in different sectors emerging from different
countries. The software involved two databases;
new technologies (which contains renewable
electricity generation, renewable heating and
cooling technologies, production and distribution
of liquid and gaseous bio-fuels, eco-buildings,
poly-generation, energy demand management
and renewable energy supply in high performance
communities, and alternative motor fuels) and
sectoral energy supplies and demands (which
contains information of the number of consum-
ers per sector, type of demand, typical quality
of the energy required and the consumption and
installed capacity per end-user). The main aim
of the program is to evaluate the market potential
of energy-related early stage energy technolo-
gies in various energy supply chains, and their
performance in terms of CO
2
emissions, costs of
energy supply, use of primary fossil energy, and
in different subsectors of society.
Segurado et al. (2009) compared EMINENT
with other tools which are carbon dioxide tech-
nology database (CO2DB), MARket Allocation
(MARKAL), IKARUS, and energy emission
economy database (E3database) already on the
market for energy technology assessment. The
main conclusion of the comparison is that EMI-
NENT is the only energy technology assessment
tool that targets early stage energy technology.
Karakosta et al. (2010) presented the priorities
of sustainable electricity generation technologies
for five developing countries, namely Chile, China,
Israel, Kenya and Thailand by using ELECTRE III.
In the paper, energy technologies given in Table
10 were analyzed under accordance with strate-
gic/developmental planning, local and regional
economic development, co
2
emissions reduction,
minimization of the negative effects on the natural
environment at nationalregional level, contribu-
tion to the employment, and contribution to energy
independence developing country criteria.
Sadegzadeh and Salehi (2010) used TOPSIS
methodology to rank the attractiveness and im-
portance of the stack of fuel cells as a sub-system.
The technologies which are taken into consider-
ation as follows; the situation of professional
manpower on the industrial and semi-industrial
scales, the situation of professional manpower on
the laboratory scale, the situation of know-how
on the industrial and semi-industrial scales, the
situation of know-how on the laboratory scales,
the situation of hardware on the industrial and
semi-industrial scales, the situation of hardware
on the laboratory scale. The criteria used in the
evaluation are as follow; production of platinum
catalyst powder-carbon, placing catalyst on the
carbon base, production of gas penetration layer,
production of polymer membrane of ion exchange,
construction of membrane collection-electrode
Table 10. Examined technologies (Karakosta et al., 2010)
Clean coal Wind
Steal boiler upgrading Solar (PV)
Coal to gas Mini/micro hydro (rivers)
Oil steam improvement Biomass (forest/agriculture) boiler
Coal steam improvement Biogas for generator
Methane combustion Mini/micro decentralized
Geothermal Solar towers
Hydro (dams) Coal Mine Methane (CMM) for generator
185
with low heat, plates of current field, technology
of waterproofing stack of fuel cell with low heat,
and technology of engineering designing of the
collection (stack) of fuel cell. Finally, the priorities
of allocating power and capital for the develop-
ment of technology of fuel cells in automotive
industry are as follows; hardware on the labora-
tory scale, know-how on the laboratory scale,
know-how on the industrial scale, professional
manpower on the laboratory scale of laboratory,
and professional manpower on the industrial scale.
According to literature survey, the interest on EET
has been monotonically increasing day by day. In
Figure 7, interest on all EETs has a pick point in
the last period except for non-conventional fuel
technologies. In particular, in the last period, the
researchers interest on transportation technologies
has the biggest increase. And the coal and industry
technologies are also the second and the third at-
tractive technologies, respectively. These indicate
that the developments on EETs for transportation,
coal, and industry keep going on and these topics
will be highlighted in near future.
In addition, the utilization of SC and CI tech-
niques on development of EET technologies has
been grown up. And application of SC and CI
techniques on energy problems has been also
increasing fast. In particular, the application of
MCDM on EETs and Energy problems are the
most attractive topics in all and it is clearly un-
derstood that their popularity will be increase in
near future. The percentiles of publications with
respect to last five years have been presented in
Figure 8.
CONCLUSION
The main aim of the study is to present SC and CI
techniques utilized in the evaluation of emerging
energy technologies exhaustively. Therefore, this
chapter has presented an extensive review of the
literature on SC and CI techniques and their ap-
plications of emerging energy technologies. In this
purpose, emerging energy technologies have been
classified and applications of emerging energy
technologies in the literature have been illustrated.
Figure 7. Distrubiton of publication percentiles with respect to the determined periods
186
Then, SC and CI techniques have been briefly
explained and various numbers of publications
related to their applications on energy planning
and emerging energy technologies in the literature
have been reviewed. From the literature, following
conclusions are obtained;
The interest of researchers in emerging en-
ergy technologies has increased in recent
years. In particular, the biggest increase
has been in transportation and coal tech-
nologies. Although, the interest in biomass
technology is the top until the last period,
the interest in transportation and coal tech-
nologies has left behind the interest in bio-
mass technology.
For energy planning studies, ANN and
AHP are the most used techniques in the
literature. Except for MCDM methods, the
other SC and CI techniques are usually
used to forecast energy demand/supply.
Although applications of SC and CI tech-
niques on EETs are narrow, the numbers of
publications have been increasing in recent
years. Hence, SC and CI techniques have
become increasingly popular in application
of emerging energy technologies. In par-
ticular, MCDM methods have been widely
used in decision making of both emerging
energy technology and energy planning
because of the complexity of the problem.
Although, there are a few studies deal with
applications of SC and CI techniques on
emerging energy technologies, ANN, GA
and AHP techniques are widely used meth-
odologies among the published articles in
the literature.
ANN and GA is the most used techniques
for sizing problem of PV systems while
AHP is used to obtain a decision related to
EET.
In MCDM applications, social, technical,
economical, and commercial criteria are
the most used criteria during evaluation of
emerging energy technologies.
Although the published literature on the
EETs based on SC and CI techniques indicates
that the popularity of SC and CI techniques has
been increasing day by day, there are two gabs
in the literature; the first one, there are not any
application of other SC and CI techniques such as
TS, ACO, SI. The second one is that none of the
MCDM studies in the literature takes into account
the interdependencies among the criteria related
to EETs. For further research, a decision making
Figure 8. The percentiles of application of both SC and CI techniques and MCDM methods on energy
and EET field
187
tool such as analytic network process, Choquet
Integral, etc. may be used for the evaluation of
emerging energy technologies to handle interde-
pendencies among the criteria.
It should be noted that the findings given in
this chapter are based on the data collected from
articles published in scientific journals, which
do not include conference proceeding papers,
masters theses, doctoral dissertations, textbooks
deal with the literature. It is possible to extend
this study by including these sources.
REFERENCES
Abdel-Aal, R. E. (2008). Univariate modeling and
forecasting of monthly energy demand time series
using abductive and neural networks. Comput-
ers & Industrial Engineering, 54(4), 903917.
doi:10.1016/j.cie.2007.10.020
Al-Rashidi, M. R., & El-Hawary, M. E. (2009).
Applications of computational intelligence tech-
niques for solving the revived optimal power flow
problem. Electric Power Systems Research, 79(4),
694702. doi:10.1016/j.epsr.2008.10.004
Altun, H., & Yalcinoz, T. (2008). Implementing
soft computing techniques to solve economic
dispatch problem in power systems. Expert
Systems with Applications, 35(4), 16681678.
doi:10.1016/j.eswa.2007.08.066
Ashhab, M. S. S. (2008). Optimization and mod-
eling of a photovoltaic solar integrated system
by neural networks. Energy Conversion and
Management, 49(11), 33493355. doi:10.1016/j.
enconman.2007.10.036
Azadeh, A., Asadzadeh, S. M., & Ghanbari, A.
(2010). An adaptive network-based fuzzy infer-
ence system for short-term natural gas demand
estimation: Uncertain and complex environments.
Energy Policy, 38(3), 15291536. doi:10.1016/j.
enpol.2009.11.036
Azadeh, A., Ghaderi, S. F., & Sohrabkhani, S.
(2008a). Annual electricity consumption forecast-
ing by neural network in high energy consum-
ing industrial sectors. Energy Conversion and
Management, 49(8), 22722278. doi:10.1016/j.
Azadeh, A., Saberi, M., Ghaderi, S. F., Gitiforouz,
A., & Ebrahimipour, V. (2008b). Improved estima-
tion of electricity demand function by integration
of fuzzy system and data mining approach. Energy
Conversion and Management, 49(8), 21652177.
doi:10.1016/j.enconman.2008.02.021
Azadeh, A., & Tarverdian, S. (2007). Integration of
genetic algorithm, computer simulation and design
of experiments for forecasting electrical energy
consumption. Energy Policy, 35(10), 52295241.
doi:10.1016/j.enpol.2007.04.020
Bahgat, A. B. G., Helwa, N. H., Ahamd, G. E.,
& El Shenawy, E. T. (2004). Estimation of the
maximum power and normal operating power
of a photovoltaic module by neural networks.
Renewable Energy, 29(3), 443457. doi:10.1016/
S0960-1481(03)00126-5
Bana e Costa, C. A., & Vansnick, J. C. (1997).
Applications of the MACBETH approach
in the framework of an additive aggrega-
tion model. Journal of Multi-Criteria Deci-
sion Analysis, 6(2), 107114. doi:10.1002/
(SICI)1099-1360(199703)6:2<107::AID-
MCDA147>3.0.CO;2-1
Beccali, M., Cellura, M., & Mistretta, M. (2003).
Decision-making in energy planning: Application
of the Electre method at regional level for the
diffusion of renewable energy technology. Renew-
able Energy, 28(13), 20632087. doi:10.1016/
S0960-1481(03)00102-2
188
Ben Salah, C., & Ouali, M. (2011). Comparison
of fuzzy logic and neural network in maximum
power point tracker for PV systems. Electric Power
Systems Research, 81(1), 4350. doi:10.1016/j.
epsr.2010.07.005
Benlarbi, K., Mokrani, L., & Nait-Said, M. S.
(2004). A fuzzy global efficiency optimization
of a photovoltaic water pumping system. Solar
Energy, 77(2), 203216. doi:10.1016/j.sole-
ner.2004.03.025
Bernard, R. (1968). Classement et choix en
prsence de points de vue multiples (la mthode
ELECTRE). la Revue dInformatique et de Re-
cherche Oprationelle, 8, 5775.
Bertini, I., Ceravolo, F., Citterio, M., De Felice,
M., Di Pietra, B., & Margiotta, F. (2010). Ambi-
ent temperature modelling with soft computing
techniques. Solar Energy, 84(7), 12641272.
doi:10.1016/j.solener.2010.04.003
Brans, J. P., & Vincke, P. (1985). A preference
ranking organization method: The PROMETHEE
method for MCDM. Management Science, 31(6),
647656. doi:10.1287/mnsc.31.6.647
Buchholz, T., Rametsteiner, E., Volk, T. A., &
Luzadis, V. A. (2009). Multi criteria analysis for
bioenergy systems assessments. Energy Policy,
37(2), 484495. doi:10.1016/j.enpol.2008.09.054
Buckley, J. J. (1985). Fuzzy hierarchical
analysis. Fuzzy Sets and Systems, 17, 233247.
doi:10.1016/0165-0114(85)90090-9
Burton, J., & Hubacek, K. (2007). Is small beauti-
ful? A multicriteria assessment of small-scale en-
ergy technology applications in local governments.
enpol.2007.08.002
Cavallaro, F. (2009). Multi-criteria decision
aid to assess concentrated solar thermal tech-
nologies. Renewable Energy, 34(7), 16781685.
doi:10.1016/j.renene.2008.12.034
Cavallaro, F. (2010a). A comparative assessment
of thin-film photovoltaic production processes
using the electre iii method. Energy Policy, 38(1),
463474. doi:10.1016/j.enpol.2009.09.037
Cavallaro, F. (2010b). Fuzzy topsis approach for
assessing thermal-energy storage in concentrated
solar power (CSP) systems. Applied Energy, 87(2),
496503. doi:10.1016/j.apenergy.2009.07.009
Cebi, S., & Kahraman, C. (2010a). Developing
a group decision support system based on fuzzy
information axiom. Knowledge-Based Systems,
23(1), 316. doi:10.1016/j.knosys.2009.07.005
Ceylan, H., Ozturk, H. K., Hepbasli, A., & Utlu,
Z. (2005). Estimating energy and exergy produc-
tion and consumption values using three different
genetic algorithm approaches. Part 2: Application
and scenarios. Energy Sources, 27, 629639.
doi:10.1080/00908310490448631
Chang, D. Y. (1996). Applications of the ex-
tent analysis method on fuzzy-AHP. European
Journal of Operational Research, 95, 649655.
doi:10.1016/0377-2217(95)00300-2
Chang, Y.-P. (2009). Optimal design of discrete-
value tilt angle of PV using sequential neural-net-
work approximation and orthogonal array. Expert
Systems with Applications, 36(3), 60106018.
doi:10.1016/j.eswa.2008.06.105
Chaouachi, A., Kamel, R. M., & Nagasaka, K. (in
press). A novel multi-model neuro-fuzzy-based
mppt for three-phase grid-connected photovoltaic
system. [forthcoming]. Solar Energy.
189
Chen, T.-C. (2000). Extensions of the TOPSIS for
group decision-making under fuzzy environment.
Fuzzy Sets and Systems, 114, 19. doi:10.1016/
S0165-0114(97)00377-1
Chen, T. Y., Yu, O. S., Hsu, G. J. Y., Hsu, F. M.,
& Sung, W. N. (2009). Renewable energy tech-
nology portfolio planning with scenario analysis:
A case study for Taiwan. Energy Policy, 37(8),
29002906. doi:10.1016/j.enpol.2009.03.028
Cinar, D., Kayakutlu, G., & Daim, T. (2010).
Development of future energy scenarios with
intelligent algorithms: Case of hydro in Turkey.
Energy, 35(4), 17241729. doi:10.1016/j.en-
ergy.2009.12.025
Dong, B., Cao, C., & Lee, S. E. (2005). Apply-
ing support vector machines to predict building
energy consumption in tropical region. Energy
and Building, 37(5), 545553. doi:10.1016/j.
enbuild.2004.09.009
Dorigo, M. (1992). Optimization, learning and
natural algorithms. PhD thesis, Dipartimento di
Elettronica, Politecnico di Milano, Italy. (in Italian)
Dorigo, M., & Sttzle, T. (2004). Ant colony
optimization. Cambridge, MA & London, UK:
MIT Press.
Doukas, H. C., Andreas, B. M., & Psarras, J.
E. (2007). Multi-criteria decision aid for the
formulation of sustainable technological energy
priorities using linguistic variables. European
Journal of Operational Research, 182, 844855.
doi:10.1016/j.ejor.2006.08.037
Dufo-Lopez, R., & Bernal-Agustin, J. L. (2005).
Design and control strategies of PVdiesel systems
using genetic algorithms. Solar Energy, 79, 3346.
doi:10.1016/j.solener.2004.10.004
Dufo-Lopez, R., Bernal-Agustin, J. L., & Contre-
ras, J. (2007). Optimization of control strategies
for stand-alone renewable energy systems with
hydrogen storage. Renewable Energy, 32(7),
11021126. doi:10.1016/j.renene.2006.04.013
Ediger, V. S., & Akar, S. (2007). Arima forecast-
ing of primary energy demand by fuel in Turkey.
enpol.2006.05.009
Eynard, J., Grieu, S., & Polit, M. (in press). Wave-
let-based multi-resolution analysis and artificial
neural networks for forecasting temperature and
thermal power consumption. [forthcoming]. En-
gineering Applications of Artificial Intelligence.
Gareta, R., Romeo, L. M., & Gil, A. (2006).
Forecasting of electricity prices with neural
networks. Energy Conversion and Management,
47(13-14), 17701778. doi:10.1016/j.encon-
man.2005.10.010
Gen, M., & Cheng, R. (2000). Genetic algorithms
and engineering optimization. New York, NY:
John Wiley and Sons.
Gendr eau, M. ( 2003) . Handbook of
metaheurstcs:An introducton to tabu search
(Glover, F., & Kochenberger, G. A., Eds.). New
York, NY/ Boston, MA/ Dordrecht, The Neth-
erlands/ London, UK/ Moscow, Russia: Kluwer
Academic Publishers.
Gendreau, M., & Potvin, J.-Y. (2010). Handbook
of metaheuristics. Springer Science+Business
Media. LLC.
Georgiou, P., Tourkolias, C., & Diakoulaki, D.
(2008). A roadmap for selecting host countries
of wind energy projects in the framework of the
clean development mechanism. Renewable &
doi:10.1016/j.rser.2006.11.001
190
Ghafghazi, S., Sowlati, T., Sokhansanj, S., &
Melin, S. (2010). A multicriteria approach to
evaluate district heating system options. Applied
Energy, 87(4), 11341140. doi:10.1016/j.apen-
ergy.2009.06.021
Glover, F. (1986). Future paths for integer pro-
gramming and links to artificial intelligence.
Computational Opeation. Research, 13, 533549.
Gonzlez, P. A., & Zamarreo, J. M. (2005).
Prediction of hourly energy consumption in
buildings based on a feedback artificial neural
network. Energy and Building, 37(6), 595601.
doi:10.1016/j.enbuild.2004.09.006
Gonzlez-Romera, E., Jaramillo-Morn, M. A., &
Carmona-Fernndez, D. (2008). Monthly electric
energy demand forecasting with neural networks
and Fourier series. Energy Conversion and
Management, 49(11), 31353142. doi:10.1016/j.
Haldenbilen, S., & Ceylan, H. (2005). Genetic
algorithm approach to estimate transport energy
demand in Turkey. Energy Policy, 33, 8998.
doi:10.1016/S0301-4215(03)00202-7
Hamzaebi, C. (2007). Forecasting of Turkeys net
electricity energy consumption on sectoral bases.
enpol.2006.03.014
Heo, E., Kim, J., & Boo, K.-J. (2010). Analysis
of the assessment factors for renewable energy
dissemination program evaluation using fuzzy
AHP. Renewable & Sustainable Energy Reviews,
14(8), 22142220. doi:10.1016/j.rser.2010.01.020
Holland, J. H. (1975). Adaptation in natural and
artificial systems. Ann Arbor, MI: University of
Michigan Press.
Holland, J. H. (1992). Genetic algorithms.
Scientific American, 267, 6672. doi:10.1038/
scientificamerican0792-66
Hontoria, L., Aguilera, J., & Zufiria, P. (2005).
A new approach for sizing stand alone photo-
voltaic systems based in neural networks. Solar
Energy, 78(2), 313319. doi:10.1016/j.sole-
ner.2004.08.018
Hsieh, T. Y., Lu, S. T., & Tzeng, G. T. (2004). Fuzzy
MCDM approach for planning and design tenders
selection in public office buildings. International
Journal of Project Management, 22, 573584.
doi:10.1016/j.ijproman.2004.01.002
Hultman, N. E., & Koomey, J. G. (2007). The
risk of surprise in energy technology costs.
Environmental Research Letters, 2(3), 16.
doi:10.1088/1748-9326/2/3/034002
Hussein, A., Hirasawa, K., & Hu, J. (2004).
A robust control method for a pv-supplied dc
motor using universal learning networks. Solar
Energy, 76(6), 771780. doi:10.1016/j.sole-
ner.2003.12.011
Hwang, C. L., & Yoon, K. (1981). Multiple at-
tribute decision making: Methods and applica-
tions. Berlin/Heidelberg, Germany/New York,
NY: Springer-Verlag.
Jacobsson, S., & Bergek, A. (2004). Transforming
the energy sector: The evolution of technological
systems in renewable energy technology. Indus-
trial and Corporate Change, 13(5), 815849.
doi:10.1093/icc/dth032
Jefferson, M. (2006). Sustainable energy de-
velopment: Performance and prospects. Re-
newable Energy, 31, 571582. doi:10.1016/j.
renene.2005.09.002
Kahraman, C., & Cebi, S. (2009). A new multi-
attribute decision making method: Hierarchical
fuzzy axiomatic design. Expert Systems with
Applications, 36(3), 48484861. doi:10.1016/j.
eswa.2008.05.041
191
Kahraman, C., evik, S., Ates, N. Y., & Glbay,
M. (2007). Fuzzy multi-criteria evaluation of
industrial robotic systems. Computers & Indus-
trial Engineering, 52(4), 414433. doi:10.1016/j.
cie.2007.01.005
Kahraman, C., & Kaya, I. (2010). A fuzzy
multicriteria methodology for selection among
energy alternatives. Expert Systems with Ap-
plications, 37(9), 62706281. doi:10.1016/j.
eswa.2010.02.095
Kahraman, C., Kaya, I., & Cebi, S. (2009). A
comparative analysis for multiattribute selection
among renewable energy alternatives using fuzzy
axiomatic design and fuzzy analytic hierarchy pro-
cess. Energy, 34(10), 16031616. doi:10.1016/j.
energy.2009.07.008
Kajikawa, Y., Yoshikawa, J., Takeda, Y., & Mat-
sushima, K. (2008). Tracking emerging technolo-
gies in energy research: Toward a roadmap for
sustainable energy. Technological Forecasting
and Social Change, 75, 771782. doi:10.1016/j.
techfore.2007.05.005
Karakosta, C., Doukas, H., & Psarras, J. (2010).
EU-MENA energy technology transfer under the
CDM: Israel as a frontrunner? Energy Policy, 38(5),
24552462. doi:10.1016/j.enpol.2009.12.039
Karatepe, E., Boztepe, M., & Colak, M. (2006).
Neural network based solar cell model. Energy
Conversion and Management, 47(9-10), 1159
1178. doi:10.1016/j.enconman.2005.07.007
Karray, F. O., & Silva, C. (2004). Soft computing
and intelligent systems design. New York, NY:
Pearson Education Limited.
Kaya, I. (2009). A genetic algorithm approach
to determine the sample size for control charts
with variables and attributes. Expert Systems with
Applications, 36(5), 87198734. doi:10.1016/j.
eswa.2008.12.011
Kaya, T., & Kahraman, C. (2010). Multicriteria
renewable energy planning using an integrated
fuzzy VIKOR & AHP methodology: The case of Is-
tanbul. Energy, 35(6), 25172527. doi:10.1016/j.
energy.2010.02.051
Klemes, J., Bulatov, I., & Koppejan, J. (2009).
Novel energy saving technologies evaluation tool.
Computers & Chemical Engineering, 33(3), 751
758. doi:10.1016/j.compchemeng.2008.07.005
Konar, A. (2007). Artificial intelligence and soft
computing: Behavioral and cognitive modeling
of the human brain. New York, NY: CRC Press.
Kowalski, K., Stagl, S., Madlener, R., & Omann, I.
(2009). Sustainable energy futures: Methodologi-
cal challenges in combining scenarios and partici-
patory multi-criteria analysis. European Journal
of Operational Research, 197(3), 10631074.
doi:10.1016/j.ejor.2007.12.049
Kulak, O. (2005). A decision support system for
fuzzy multi-attribute selection of material handling
equipments. Expert Systems with Applications,
29(2), 310319. doi:10.1016/j.eswa.2005.04.004
Kulak, O., Durmusoglu, M. B., & Kahraman,
C. (2005). Fuzzy multi-attribute equipment se-
lection based on information axiom. Journal of
Materials Processing Technology, 169, 337345.
doi:10.1016/j.jmatprotec.2005.03.030
Kulak, O., & Kahraman, C. (2005a). Fuzzy
multi-attribute selection among transportation
companies using axiomatic design and analytic
hierarchy process. Information Sciences, 170,
191210. doi:10.1016/j.ins.2004.02.021
Kulak, O., & Kahraman, C. (2005b). Multi-
attribute comparison of advanced manufacturing
systems using fuzzy vs. crisp axiomatic design
approach. International Journal of Produc-
tion Economics, 95, 415424. doi:10.1016/j.
ijpe.2004.02.009
192
Laarhoven, P. J. M., & Pedrycz, W. (1983). A
fuzzy extension of Saatys priority theory. Fuzzy
Sets and Systems, 11, 229241.
Lai, Y.-J., & Hwang, C.-L. (1994). Fuzzy multi
objective decision making methods and applica-
tions. Berlin, Germany: Springer-Verlag.
Lee, S. K., Mogi, G., & Kim, J. W. (2008). The
competitiveness of Korea as a developer of hy-
drogen energy technology: The AHP approach.
enpol.2007.12.003
Lee, S. K., Mogi, G., & Kim, J. W. (2009). Deci-
sion support for prioritizing energy technologies
against high oil prices: A fuzzy analytic hierar-
chy process approach. Journal of Loss Preven-
tion in the Process Industries, 22(6), 915920.
doi:10.1016/j.jlp.2009.07.001
Lee, S. K., Mogi, G., & Kim, J. W. (2009). Deci-
sion support for prioritizing energy technologies
against high oil prices: A fuzzy analytic hierar-
chy process approach. Journal of Loss Preven-
tion in the Process Industries, 22(6), 915920.
doi:10.1016/j.jlp.2009.07.001
Lee, S. K., Mogi, G., Kim, J. W., & Gim, B. J.
(2008). A fuzzy analytic hierarchy process ap-
proach for assessing national competitiveness
in the hydrogen technology sector. International
Journal of Hydrogen Energy, 33(23), 68406848.
doi:10.1016/j.ijhydene.2008.09.028
Lee, S. K., Mogi, G., Lee, S. K., Hui, K. S., &
Kim, J. W. (2010). Econometric analysis of the
R&D performance in the national hydrogen en-
ergy technology development for measuring rela-
tive efficiency: The fuzzy AHP/DEA integrated
model approach. International Journal of Hy-
drogen Energy, 35(6), 22362246. doi:10.1016/j.
ijhydene.2010.01.009
Lee, S. K., Mogi, G., Lee, S. K., Hui, K. S., &
Kim, J. W. (2010). Econometric analysis of the
R&D performance in the national hydrogen en-
ergy technology development for measuring rela-
tive efficiency: The fuzzy AHP/DEA integrated
model approach. International Journal of Hy-
drogen Energy, 35(6), 22362246. doi:10.1016/j.
ijhydene.2010.01.009
Lee, S. K., Mogi, G., Lee, S. K., & Kim, J. W.
(in press). Prioritizing the weights of hydrogen
energy technologies in the sector of the hydro-
gen economy by using a fuzzy AHP approach.
International Journal of Hydrogen Energy. doi:.
doi:10.1016/j.ijhydene.2010.01.035
Lee, S. K., Yoon, Y. J., & Kim, J. W. (2007). A
study on making a long-term improvement in
the national energy efficiency and GHG control
plans by the AHP approach. Energy Policy, 35(5),
28622868. doi:10.1016/j.enpol.2006.09.019
Li, G., & Shi, J. (2010). On comparing three arti-
ficial neural networks for wind speed forecasting.
Applied Energy, 87(7), 23132320. doi:10.1016/j.
apenergy.2009.12.013
Li, K., & Su, H. (2010). Forecasting building
energy consumption with hybrid genetic algo-
rithm-hierarchical adaptive network-based fuzzy
inference system. Energy and Building, 42(11),
20702076. doi:10.1016/j.enbuild.2010.06.016
Liao, C.-C. (2010). Genetic k-means algorithm
based RBF network for photovoltaic MPP pre-
diction. Energy, 35(2), 529536. doi:10.1016/j.
energy.2009.10.021
Ma, J., Scott, N. R., Degloria, S. D., & Lembo, A.
J. (2005). Siting analysis of farm-based centralized
anaerobic digester systems for distributed genera-
tion using GIS. Biomass and Bioenergy, 28(6),
591600. doi:10.1016/j.biombioe.2004.12.003
193
Madlener, R., Antunes, C. H., & Dias, L. C. (2009).
Assessing the performance of biogas plants with
multi-criteria and data envelopment analysis. Eu-
ropean Journal of Operational Research, 197(3),
10841094. doi:10.1016/j.ejor.2007.12.051
Madlener, R., Kowalski, K., & Stagl, S. (2007).
New ways for the integrated appraisal of national
energy scenarios: The case of renewable energy
use in Austria. Energy Policy, 35(12), 60606074.
doi:10.1016/j.enpol.2007.08.015
Madlener, R., & Stagl, S. (2005). Sustainability-
guided promotion of renewable electricity gen-
eration. Ecological Economics, 53(2), 147167.
doi:10.1016/j.ecolecon.2004.12.016
McCulloch, W. S., & Pitts, W. A. (1943). A logical
calculus of the ideas immanent in neural nets. The
Bulletin of Mathematical Biophysics, 5, 115133.
doi:10.1007/BF02478259
Meijer, I. S. M., Hekkert, M. P., & Koppenjan, J. F.
M. (2007). The influence of perceived uncertainty
on entrepreneurial action in emerging renewable
energy technology; biomass gasification projects
in the Netherlands. Energy Policy, 35, 58365854.
doi:10.1016/j.enpol.2007.07.009
Mellit, A. (2010). Ann-based GA for generating
the sizing curve of stand-alone photovoltaic sys-
tems. Advances in Engineering Software, 41(5),
687693. doi:10.1016/j.advengsoft.2009.12.008
Mellit, A., & Benghanem, M. (2007). Sizing of
stand-alone photovoltaic systems using neural
network adaptive model. Desalination, 209(1-3),
6472. doi:10.1016/j.desal.2007.04.010
Mellit, A., Benghanem, M., Arab, A. H., & Gues-
soum, A. (2005). An adaptive artificial neural
network model for sizing stand-alone photovol-
taic systems: Application for isolated sites in
Algeria. Renewable Energy, 30(10), 15011524.
doi:10.1016/j.renene.2004.11.012
Mellit, A., Benghanem, M., & Kalogirou, S. A.
(2007). Modeling and simulation of a stand-alone
photovoltaic system using an adaptive artificial
neural network: Proposition for a new sizing
procedure. Renewable Energy, 32(2), 285313.
doi:10.1016/j.renene.2006.01.002
Mellit, A., Kalogirou, S. A., & Drif, M. (2010).
Application of neural networks and genetic algo-
rithms for sizing of photovoltaic systems. Renew-
able Energy, 35(12), 28812893. doi:10.1016/j.
renene.2010.04.017
Mellit, A., Kalogirou, S. A., Hontoria, L., & Shaari,
S. (2009). Artificial intelligence techniques for
sizing photovoltaic systems: A review. Renewable
& Sustainable Energy Reviews, 13(2), 406419.
doi:10.1016/j.rser.2008.01.006
Miranda, V., Srinivasan, D., & Proenc, A. M.
(1998). Evolutionary computation in power sys-
tems. Electric Power Systems Research, 20(2),
8998. doi:10.1016/S0142-0615(97)00040-9
Mrz, T. M. (2008). Planning of community
heating systems modernization and development.
Applied Thermal Engineering, 28(14-15), 1844
1852. doi:10.1016/j.applthermaleng.2007.11.020
Murat, Y. S., & Ceylan, H. (2006). Use of artifi-
cial neural networks for transport energy demand
modeling. Energy Policy, 34(17), 31653172.
doi:10.1016/j.enpol.2005.02.010
Neto, A. H., & Fiorelli, F. A. S. (2008). Com-
parison between detailed model simulation and
artificial neural network for forecasting building
energy consumption. Energy and Building, 40(12),
21692176. doi:10.1016/j.enbuild.2008.06.013
Nixon, J. D., Dey, P. K., & Davies, P. A. (in
press). Which is the best solar thermal collection
technology for electricity generation in north-west
India? Evaluation of options using the analytical
hierarchy process. [forthcoming]. Energy.
194
nt, S., & Soner, S. (2008). Transshipment site
selection using the AHP and TOPSIS approaches
under fuzzy environment. Waste Management
(New York, N.Y.), 28(9), 15521559. doi:10.1016/j.
wasman.2007.05.019
Opricovic, S., & Tzeng, G. H. (2004). Compro-
mise solution by MCDM methods: a comparative
analysis of VIKOR and TOPSIS. European Jour-
nal of Operational Research, 156(2), 445455.
doi:10.1016/S0377-2217(03)00020-1
Ozturk, H. K., Ceylan, H., Canyurt, O. E., &
Hepbasli, A. (2005). Electricity estimation us-
ing genetic algorithm approach: a case study of
Turkey. Energy, 30, 10031012. doi:10.1016/j.
energy.2004.08.008
Pao, H. T. (2007). Forecasting electricity market
pricing using artificial neural networks. Energy
Conversion and Management, 48(3), 907912.
doi:10.1016/j.enconman.2006.08.016
Pao, H. T. (2009). Forecasting energy consump-
tion in Taiwan using hybrid nonlinear models.
Energy, 34(10), 14381446. doi:10.1016/j.en-
ergy.2009.04.026
Paoli, C., Voyant, C., Muselli, M., & Nivet, M.-
L. (in press). Forecasting of preprocessed daily
solar radiation time series using neural networks.
[forthcoming]. Solar Energy.
Papadopoulos, A., & Karagiannidis, A. (2008).
Application of the multi-criteria analysis method
ELECTRE III for the optimisation of decen-
tralised energy systems. Omega, 36(5), 766776.
doi:10.1016/j.omega.2006.01.004
Pea, B., Teruel, E., & Dez, L. I. (in press). Soft-
computing models for soot-blowing optimization
in coal-fired utility boilers. [forthcoming]. Applied
Soft Computing.
Pohekar, S. D., & Ramachandran, M. (2006a).
Multi-criteria evaluation of cooking devices with
special reference to utility of parabolic solar cooker
(PSC) in India. Energy, 31(8-9), 12151227.
doi:10.1016/j.energy.2005.04.012
Pohekar, S. D., & Ramachandran, M. (2006b).
Utility assessment of parabolic solar cooker as
a domestic cooking device in India. Renewable
Energy, 31(11), 18271838. doi:10.1016/j.re-
nene.2005.09.014
Rivire, C., & Marlair, G. (2009). Biosafuel, a
pre-diagnosis tool of risks pertaining to biofuels
chains. Journal of Loss Prevention in the Pro-
cess Industries, 22(2), 228236. doi:10.1016/j.
jlp.2008.09.014
Rovere, E. L. L., Soares, J. B., Oliveira, L. B.,
& Lauria, T. (2010). Sustainable expansion of
electricity sector: Sustainability indicators as an
instrument to support decision making. Renewable
& Sustainable Energy Reviews, 14(1), 422429.
doi:10.1016/j.rser.2009.07.033
Saaty, T. L. (1980). The analytic hierarchy process.
New York, NY: McGraw Hill.
Sadeghzadeh, K., & Salehi, M. B. (in press). Math-
ematical analysis of fuel cell strategic technologies
development solutions in the automotive industry
by the TOPSIS multi-criteria decision making
method. [forthcoming]. International Journal of
Hydrogen Energy.
Saridakis, K. M., & Dentsoras, A. J. (2008). Soft
computing in engineering design - A review. Ad-
vanced Engineering Informatics, 22(2), 202221.
doi:10.1016/j.aei.2007.10.001
Schirru, R., Martinez, A. S., Pereira, C. M. N. A.,
Domingos, R. P., Machado, M. D., & Machado,
L. (1999). Intelligent soft computing in nuclear
engineering in Brazil. Progress in Nuclear
Energy, 35(3-4), 367391. doi:10.1016/S0149-
1970(99)00022-0
195
Segurado, R., Pereira, S., Pipio, A., & Alves,
L. (2009). Comparison between EMINENT
and other energy technology assessment tools.
Journal of Cleaner Production, 17(10), 907910.
doi:10.1016/j.jclepro.2009.02.002
Senjyua, T., Hayashia, D., Yonaa, A., Urasakia, N.,
& Funabashib, T. (2007). Optimal configuration
of power generating systems in isolated island
with renewable energy. Renewable Energy, 32,
19171933. doi:10.1016/j.renene.2006.09.003
Sozen, A., & Arcaklioglu, E. (2007). Prospects
for future projections of the basic energy sources
in Turkey. Energy Sources, Part B. Economics,
Planning, and Policy, 2(2), 183201.
Sozen, A., Arcaklioglu, E., zalp, M., & aglar,
N. (2005a). Forecasting based on neural network
approach of solar potential in Turkey. Renew-
able Energy, 30(7), 10751090. doi:10.1016/j.
renene.2004.09.020
Sozen, A., Arcaklioglu, E., & Ozkaymak, M.
(2005b). Modelling of the Turkeys net energy
consumption using artificial neural network.
International Journal of Computer Applications
in Technology, 22(2/3), 130136. doi:10.1504/
IJCAT.2005.006944
Suh, N. P. (1990). The principles of design. New
York, NY: Oxford University Press Inc.
Suh, N. P. (2001). Axiomatic design: Advances
and applications. New York, NY: Oxford Uni-
versity Press.
Thakker, A., Jarvis, J., Buggy, M., & Sahed, A.
(2008). A novel approach to materials selection
strategy case study: Wave energy extraction im-
pulse turbine blade. Materials & Design, 29(10),
19731980. doi:10.1016/j.matdes.2008.04.022
Thiaux, Y., Seigneurbieux, J., Multon, B., & Ben
Ahmed, H. (2010). Load profile impact on the
gross energy requirement of stand-alone photovol-
taic systems. Renewable Energy, 35(3), 602613.
doi:10.1016/j.renene.2009.08.005
Toksari, M. D. (2007). Ant colony optimization
approach to estimate energy demand in Turkey.
Energy Policy, 35, 39843990. doi:10.1016/j.
enpol.2007.01.028
Tversky, A. (1972). Elimination by aspects: A
theory of choice. Psychological Review, 79,
281299. doi:10.1037/h0032955
Uhrig, R. E., & Tsoukalas, L. H. (1999). Soft
computing technologies in nuclear engineering
applications. Progress in Nuclear Energy, 34(1),
1375. doi:10.1016/S0149-1970(97)00109-1
nler, A. (2008). Improvement of energy demand
forecasts using swarm intelligence: The case of tur-
key with projections to 2025. Energy Policy, 36(6),
19371944. doi:10.1016/j.enpol.2008.02.018
Vanek, F. M., & Albright, L. D. (2008). Energy
systems engineering: Evaluation and implementa-
tion. New York, NY: McGraw Hill.
Venayagamoorthy, G. K., & Welch, R. L. (2010).
Energy dispatch controllers for a photovoltaic
system. Engineering Applications of Artificial
Intelligence, 23(2), 249261. doi:10.1016/j.en-
gappai.2009.11.001
Won, J. R., & Park, Y. M. (2003). Economic
dispatch solutions with piecewise quadratic cost
functions using improved genetic algorithm.
Energy Systems, 25(5), 355361. doi:10.1016/
S0142-0615(02)00098-4
Yang, J., & Zhuang, Y. (2010). An improved ant
colony optimization algorithm for solving a com-
plex combinatorial optimization problem. Applied
Soft Computing, 10(2), 653660. doi:10.1016/j.
asoc.2009.08.040
196
Zadeh, L. A. (1965). Fuzzy sets. Information
and Control, 8, 338353. doi:10.1016/S0019-
9958(65)90241-X
Zhang, L., & Fei Bai, Y. (2005). Genetic algorithm-
trained radial basis function neural networks for
modelling photovoltaic panels. Engineering Appli-
cations of Artificial Intelligence, 18(7), 833844.
doi:10.1016/j.engappai.2005.02.004
ADDITIONAL READING
Aliev, R. A., & Aliev, R. (2001). Soft Computing
& Its Applications. Singapore: World Scientific.
Castillo, O., Melin, P., Ross, O. M., Cruz, R. S.,
& Pedrycz, W. (2007). Theoretical Advances and
Applications of Fuzzy Logic and Soft Computing.
Berlin: Springer. doi:10.1007/978-3-540-72434-6
Cebi, S., & Kahraman, C. (2010). Developing a
group decision support system based on fuzzy
information axiom. Knowledge-Based Systems,
23(1), 316. doi:10.1016/j.knosys.2009.07.005
Cebi, S., & Kahraman, C. (2010). Developing a
group decision support system based on fuzzy
information axiom. Knowledge-Based Systems..
doi:10.1016/j.knosys.2009.07.005
Celik, M., Cebi, S., Kahraman, C., & Er, D.
(2009a). Application of axiomatic design and
TOPSIS methodologies under fuzzy environment
for proposing competitive strategies on Turkish
container ports in maritime transportation net-
work. Expert Systems with Applications, 36(3),
1, 45414557. doi:10.1016/j.eswa.2008.05.033
Celik, M., Cebi, S., Kahraman, C., & Er, I. D.
(2009b). An integrated fuzzy QFD model pro-
posal on routing of shipping investment decisions
in crude oil tanker market. Expert Systems with
Applications, 36(3), 2, 62276235. doi:10.1016/j.
eswa.2008.07.031
Celik, M., Kahraman, C., Cebi, S., & Er, I. D.
(2009c). Fuzzy axiomatic design-based perfor-
mance evaluation model for docking facilities in
shipbuilding industry: the case of Turkish ship-
yards. Expert Systems with Applications, 36(1),
599615. doi:10.1016/j.eswa.2007.09.055
Engin, O., elik, A., & Kaya, I. (2008). A fuzzy
approach to define sample size for attributes con-
trol chart in multistage processes: An application
in engine valve manufacturing process. Applied
Soft Computing, 8(4), 16541663. doi:10.1016/j.
asoc.2008.01.005
Kahraman, C. (2008). Fuzzy Multi-Criteria Deci-
sion Making: Theory and Applications with Recent
Developments. Berlin: Springer.
Kahraman, C., Ruan, D., & Dogan, I. (2003).
Fuzzy group decision-making for facility location
selection. Information Sciences, 157, 135153.
doi:10.1016/S0020-0255(03)00183-X
Kaliszewski, I. (2006). Soft Computing for Com-
plex Multiple Criteria Decision Making. USA:
Springer.
Kaya, I. (2009). A genetic algorithm approach to
determine the sample size for attribute control
charts. Information Sciences, 179(10), 15521566.
doi:10.1016/j.ins.2008.09.024
Kaya, I., & Engin, O. (2007). A new approach to
define sample size at attributes control chart in
multistage processes: An application in engine
piston manufacturing process. Journal of Ma-
terials Processing Technology, 183(1), 3848.
doi:10.1016/j.jmatprotec.2006.09.022
Konar, A. (1999). Artificial Intelligence and Soft
Computing: Behavioral and Cognitive Model-
ing of the Human Brain. Boca Raton/ London/
New York/ Washington, D.C.: CRC Press.
doi:10.1201/9781420049138
197
Rayward-Smith, V. J., Osman, I. H., Reeves, C. R.,
& Smith, G. D. (1996). Modern Heuristic Search
Methods. New York, Toronto: Wiley.
Schniederjans M. J., Hamaker J.L., Schnieder-
jans A. M., Information Technology Investment:
Decision-Making Methodology,
Tettamanzi, A., & Tomassini, M. (2001). Soft
Computing: Integrating Evolutionary, Neural,
and Fuzzy Systems. Berlin: Springer.
Triantaphyllou, E. (2000) Multi-Criteria Decision
Making Methods: A comparative Study, Kluwer
Academic Publishers, Baton Rogue/Louisiana.
Emerging Energy Technology (EET): EET
is the technology that improves the utilization
performance of energy sources.
Multi Criteria Decision Making (MCDM):
MCDM is to give the best decision under the multi
objective or multi attribute.
Multi Attribute Decision Making (MADM):
MADM is to select the best alternative under
multiple criteria.
Multi Objective Decision Making (MODM):
MODM is to present the best opportunity by
satisfying multiple objectives.
Analytic Hierarchy Process (AHP): AHP
is a MCDM tool which is based on pairwise
comparison when there is not any quantitative
information.
Technique for Order Performance by
similarity to Ideal Solution (TOPSIS): TOPSIS
is another MCDM method which is based on
measuring the distance to positive and negative
ideal solutions.
Information Axiom: Information axiom
which is based on satisfying level of functional
requirements is used to determine the best alterna-
tive under multiple criteria.
Fuzzy Sets: Fuzzy is based on the membership
function definition of an element. In classical sets,
either an element is a member of the defined set or
not while each member is defined by a member-
ship value in a fuzzy sets.
198
Chapter 6
INTRODUCTION
Conventional energy sources such as fossil fuels
and uranium reserves are limited and adversely
impacts on environment, therefore greet interest
for utilization of renewable energy has been es-
tablished. For recent expansion of renewable en-
ergy applications, wind energy generation among
other renewable energies has been experiencing
a rapid growth. As the use of wind power units
increases worldwide, there is a rising interest on
their impacts on power system dynamic/control
and finding appropriate solutions.
Mohammad Saleh
University of Kurdistan, Iran
Hassan Bevrani
Dynamic Analysis and Stability
Improvement Concerning the
Integration of Wind Farms:
Kurdistan Electric Network Case Study
ABSTRACT
This chapter presents an overview of key issues and technical challenges in a regional electric net-
work, following the integration of a considerable amount of wind power. A brief survey on wind power
system, the present status of wind energy worldwide, common dynamic models, and control loops for
wind turbines are given. In this chapter, the Kurdistan electric network in the Northwest part of Iran is
introduced as a case study system, and an analytical approach is conducted to evaluate the potential of
wind power installation, overall capacity estimation, and economic issues, based on the practical data.
Then, the impact of high penetration wind power on the system dynamic and performance for various
wind turbine technologies is presented. The stability of integrated system is analyzed, and the need for
revising of conventional controls and performance standards is emphasized. Finally, a STATCOM-based
control approach is addressed to improve the system stability.
DOI: 10.4018/978-1-61350-138-2.ch006
199
Dynamic Analysis and Stability Improvement Concerning the Integration of Wind Farms
The recent investigation studies indicate that
relatively large scale wind generation affects the
power system frequency and voltage regulation,
as well as other control and operation issues. This
impact may increase at the penetration rates that
are expected to be high in the next several years.
On the other hand, most of existing wind turbine
technologies cannot provide necessary control
capabilities for the regulation issue. The power
system control of the future will require a high
degree of flexibility and intelligence to ensure that
it can continuously balance fluctuating power and
regulate frequency/voltage deviation caused by
renewable energy sources such as wind (Bevrani,
et al., 2011).
This chapter presents an overview of new dy-
namical challenges in regional electric networks,
following a high penetration of wind power. The
Kurdistan electric network in Iran is considered
as a case study. Mountainous environment, costly
process for electricity production from conven-
tional sources, and numerous windy areas make
Kurdistan as an appropriate region for installa-
tion of wind farms. In this work, an analytical
approach is conducted to evaluate the potential
of wind power installation and overall capacity
estimation, and to study economic issues based
on the practical data.
The impact of high penetration wind power on
the system dynamic and performance for different
wind turbine technologies including fixed-speed
induction generator (FSIG), doubly-fed induction
generator (DFIG) and permanent magnet syn-
chronous generator (PMSG) is presented. Using
DIgSILENT simulation software, the stability of
the integrated system is re-analyzed, and the need
for revising of conventional controls and perfor-
mance standards is emphasized. Finally, a control
approach to improve the system stability using
static synchronous compensator (STATCOM) and
energy storage devices is addressed. This work
is supplemented by some nonlinear simulations
on the Kurdistan power system case study using
real data and parameters.
In the next section, a background with a brief
literature review is presented. In section 3, an over-
view of wind energy status around the world and
Iran is provided. Section 4 presents a discussion
about wind power systems and the main control
schemes. Section 5 determines the potential of
Kurdistan province for wind power generation.
In section 6, a preliminary study on wind energy
costs in Kurdistan is performed. Section 7 pres-
ents a dynamic analysis on the impact of a high
wind power penetration on the Kurdistan electric
network and introduces an appropriate control
solution for its stability improvement. Finally,
conclusion and future research directions are
presented in sections 8 and 9, respectively.
BACKGROUND
In order to clarify the interaction behavior between
wind farm(s) and the power system, building of an
effective dynamic model for wind power systems
(WPSs) is needed. Model simplifications and some
comparisons between different types of WPSs
and wind farm equivalent models are presented
in recent performed research works (Mansouri,
et al., 2004; Ekanayake, et al., 2003; Slootweg,
et al., 2003; Akhmatov, et al., 2006; Fernandeza,
et al., 2006; Ledesma & Usaola, 2005).
The role of WPS control strategy to qualify
system output and stability augmentation is stud-
ied in many papers. Optimization control, power
smoothing and voltage control of WPSs are most
important topics of related new research areas
(Senjyu, et al., 2006; Wang & Chang, 2004).
Increasing the penetration of wind turbine gen-
erators in a power system may affect the system
security/stability limits, frequency, voltage and
dynamic behavior (Muyeen, et al., 2009; Bev-
rani, 2009; Slootweg, 2003; Bevrani, Tikdari, &
Hiyama, 2010). This effect can be mostly caused
by fluctuation of wind power. The impacts of
wind turbines on the power system frequency
and voltage have been studied in many research
200
works (Jowder, 2009; Radics & Bartholy, 2008;
Bevrani & Tikdari, 2010). Power system frequency
response model in the presence of high wind power
penetration, frequency control issue, and a com-
prehensive survey with some new perspectives
are already well addressed (Bevrani & Hiyama,
2011; Bevrani, Ghosh, & Ledwich, 2010; Bevrani,
Daneshfar, & Daneshmand, 2010).
The effects of DFIG and induction generator
type of WPSs on the voltage transient behaviors
are explained and the disadvantages of the induc-
tion generator type are shown in (Nunes, et al.,
2004). The loadability of various types of WPSs
is studied and it is shown that the DFIG has larger
loadability than induction generators (Bevrani &
Tikdari, 2010). Frequency nadir in the presence
of different types of the WPSs has been also
compared in (Erlich, et al., 2006; Gillian, et al.,
2005). As argued in the mentioned references, wind
turbines affect frequency behavior because they
add amount of inertia to the power system. Both
stator and rotor windings of induction generator
type of WPSs are directly connected to the power
grid, but in DFIG type, only stator is directly con-
nected and the rotor is linked through a power
electronic type converter. The induction generator
WPS in turn adds much inertia than DFIG in the
power system; and in conclusion, the induction
generator WPS frequency response is better than
systems with DFIG type in the same conditions.
Continuous increase of installed wind power
during recent years has forced the system opera-
tors and responsible organizations to tighten the
performance standards and connection rules
known as grid code - in order to limit the effects
of wind power penetration on the power system
performance and stability.
Interconnection procedures and standards
need to be reviewed to ensure that the new op-
erating control schemes and their responses are
in a consistent manner to all power generation
technologies, including wind generating units as
variable generation technologies.
The revised operating performance standards
require that most type of power plants support
the electricity network throughout their opera-
tion. Important key issues can be considered as
steady state and dynamic active/reactive power
capability, continuously acting frequency/voltage
control and fault ride through behavior. Some
commonly used turbine designs have some limits
in terms of achieving grid code compliance in
several countries. For the wind farms containing
these turbines, additional equipments are needed
(Maibach, et al., 2007).
Variable generation technologies generally
refer to generating technologies whose primary
energy source varies over time and cannot
reasonably be stored to address such variation.
Uncertainty and variability are two major factors
of a variable generator that distinguish it in con-
ventional forms of generation and may impact the
overall system planning and operations (Bevrani
& Hiyama, 2011).
In order to specify wind power potential in
a particular site, a long-term record of wind
speed has to be statistically analyzed. There are
several studies related to the determination of
wind characteristics and wind power potential in
many countries over the whole world (Radics &
Bartholy, 2008; Elamouri, Ben, & Amar, 2008;
Al-Abbadi, 2005; Jowder, 2009; Ucar & Balo,
2009; Weigt, 2009)
State of Wind Power Generation
At present, wind power has effective impact on
energy markets. In 2009, more than 38.3 GW of
new wind power capacity was installed around the
world that bringing the total installed capacity up to
158.5 GW. The main markets driving this growth
rate are Asia, North America and Europe. The top
five countries in wind energy installed capacity
in 2009 were US with 35064 MW (22.1%), PR
China with 25805 MW (16.3%), Germany with
25777 MW (16.3%), Spain with 19149 MW
(12.1%), and India with 10926 MW (6.9%); while
201
the total installed wind power in the rest of world
was 41784 MW (26.4%) (GWEC, 2010).
Similar to the most countries in the Middle
East and Africa, contribution of the wind energy
for production of electricity in Iran is relatively
low. The total installed capacity of wind turbine
in Iran by 2010 is less than 200 MW. There are
many suitable areas for installing wind turbines
in Iran; however as depicted in Table 1, the major
installed wind power is centralized in Manjil and
Binalud areas.
Because of existing high sources of oil and
gas in a relatively low price in Iran, most of elec-
tricity has been produced by fossil fuel in the past.
However, nowadays for many reasons, as well as
other countries there is a great concern towards
renewable energies like wind and solar. The gov-
ernment and other responsible organizations have
put some efforts to expand wind and solar farms
in different parts of country. Currently, the poten-
tial for wind power generation is estimated to be
more than 6500MW.
Wind Power Controls
A WPS transforms the energy presented in the
belonging wind into electrical energy. A general
scheme of this system is shown in Figure 1a. Wind
energy is transformed into mechanical energy by
wind turbine units. Based on rotational speed, the
wind turbines can be split into two types:
1. Fixed speed wind turbine (FSWT)
2. Variable speed wind turbine (VSWT)
Major characteristics of the FSWT are brush-
less and rugged construction, low cost and simplic-
ity. The main advantage of the VSWT is that more
energy can be extracted for a specific wind speed
regime. In addition, the mechanical stress is less;
because the rotor acts as a flywheel (Slootweg,
2003). Common VSWT structures are known as
DFIG and the PMSG. A FSWT is usually directly
equipped with a grid coupled squirrel cage induc-
tion generator whose speed variations are limited.
The power extracted from the wind energy by
a wind turbine can be expressed as follows (Heier,
1998; Bansal, et al., 2002):
P AV C
m w p
=
1
2
3
( , ) (1)
where, P
m
is the power extracted from the wind,
p is the air density (Kg/m
3
), A is the rotor disc
area (m
2
), V
w
is the wind speed (m/s), and C
p
is
a power coefficient which is a function of the tip
speed ratio and the pitch angle of rotor blades
. The tip speed ratio is defined by

=
r
w
R
V
. (2)
Table 1. Wind power in Binalud and Manjil areas
Area No. of turbines Power (KW)
27 300
2 500
Binalud 18 550
1 600
64 660
Total 112 61840
Manjil 43 660
Total 43 28300
202
where,
r
is mechanical angular velocity of the
turbine rotor and R is the blade radius of the wind
turbine.
As shown in Fig. 1b, the drive train model of
wind turbines is usually represented by two mass
models (Slootweg, Haan, Polinder, & Kling, 2003;
Slootweg, Polinder, & Kling, 2003):
T T J
d
dt
w m r
r
=

(3)
T D K dt
m mc r g mc r g
= +
( )
( )
(4)
T T J
d
dt
m g g
g
=

(5)
where, J
r
, and J
g
are inertia of wind turbine and
generator,
g
is the rotor speed, T
m
is the me-
chanical torque from the generator shaft, T
g
is
the generator electrical torque. Finally, K
mc
and
D
mc
are the stiffness and damping of mechanichal
coupling, repectively.
Power extracted from a wind turbine can be
controlled in two states, in above and below rated
wind speed of wind turbine. In the above rated
wind speed, a blade pitch angle controller reduces
the power coefficient and thus the power extracted
from the wind. The pitch controller limits the
generators speed to a rated value (
gen, rated
) by
adjusting the pitch angle ().
Second control state (below rated wind speed)
exists only for the VSWT generator type. The aim
is to control the rotational speed to follow the maxi-
mum power point trajectory (MPPT), when wind
speed is in change. Since, precise measurement
of wind speed is difficult, for maximum power
point tracking operation, it is better to use the rotor
speed as a control input instead of wind speed.
Figure 1. Wind energy conversion system; a) General scheme of WPS, and b) Drive train model
203
Modeling and Control of DFIG
The control strategy that generally applied to
control of VSWT is based on vector control tech-
niques. An overview of dynamic model for DFIG
wind turbine and the associated control system is
shown in Figure 2a (Hansen, et al., 2003; Hansen,
et al., 2004).
The rotor-side converter operates in a stator
flux reference frame that decomposes the rotor
current into active power (q-axis) and reactive
power (d-axis) components. A fast inner current
control loop controls rotor current in d- and q-
axis and a slower outer control loop regulates
active and reactive powers. The MPPT unit pro-
vides the reference signal P
Grid,ref
for the active
power, while the reactive power Q
Grid,ref
is typi-
cally fixed at zero.
The grid-side converter controller operates in
a grid side converter voltage oriented reference
frame (DIgSILENT GmbH, 2003). Active and
reactive components of the grid-side converter
currents are controlled by the fast inner control
loop. The slower outer control loop determines the
q-current set point, which regulates the DC-voltage
to a pre-defined value. For achieving unity power
factor operation of converter, it is sufficient that
q-current to be regulated to zero.
Figure 2. Overall structure and main control loops of VSWT systems a) DFIG, and b) PMSG
204
Modeling and Control of PMSG
An overview for dynamic model of the PMSG wind
turbine and its control system is shown in Figure
2b. The control structures and related concepts
are well discussed in the recent published works
(Conroy & Watson, 2008).
At the generator-side converter, AC-voltage
and active power are regulated (reactive power
regulation is optional). The grid-side converter
operates in a stator voltage oriented reference
frame. A fast inner control loop regulates the d-
and q-axis current components of the grid-side
PWM-converter. Current references are defined by
a slower outer control-loop regulating DC-voltage
of the intermediate DC-circuit and reactive power.
Wind Energy Potential
Assessment in Kurdistan
In this section, to determine the potential of wind
power generation the hourly measured wind speed
data over a period of almost 5 years between 2004
and 2008 from 6 stations in Kurdistan, at 10 m
height that obtained from Kurdistan Meteorologi-
cal Organization are statically analyzed. Extrapo-
lation of the 10 m data, using the power law, is
used to determine the wind data at upper heights.
The power law used in this study is as follow:
V
V
Ln
H
Z
Ln
H
Z
H
ref
ref
=
0
0
(6)
where, V
H
is the wind speed at height H, V
ref
is
the wind speed at height H
ref
, and Z
0
is surface
roughness length.
Gilbert (2004) explains the roughness clas-
sifications and roughness lengths. The roughness
length for water surface (Class 0), open areas with
a few windbreaks (Class 1), farm land with some
windbreaks more than 1 km apart (Class 2), urban
districts and farm land with many windbreaks
(Class 3), and dense urban or forest (Class 4) are
determined as 0.0002 m, 0.03 m, 0.1 m, 0.4 m,
and 1.6 m, respectively. In this work, since the
studied stations are located in near cities or even
inside cities and due to mountainous environment
of Kurdistan, the roughness lengths for most sta-
tions are fixed at 1 (Class 1).
Site Selection
Gillbert (2004) presented a standard for using
in site selection in a wind farm installation pro-
cedure. Based on this standard and according
to the annual mean wind speed from 2004 to
2008, among considered six cities of Kurdistan
(Bijar, Qorveh, Marivan, Saqez, Sanandaj, and
Divandarreh), regions of Bijar and Divandarreh
are determined as more fair places for wind farm
installation (Table 2).
Monthly Variation of Mean Wind Speed
Figure 3 shows monthly variation of mean wind
speed for selected two sites (Divandarreh and
Bijar). In both stations, the highest monthly wind
speed occurs in March. In Bijar, the lowest wind
speed happens in January; while for Divandarreh
it happens in December.
Wind Rose Diagram
The direction of the wind is taken into consider-
ation for the sake of installing the wind turbines
in a wind farm. The wind rose diagram illustrates
the wind direction. Figure 4 shows the wind rose
diagram for Bijar and Divandarreh, using the
WRPLOT software. Based on these diagrams,
the wind mainly blows to the north side of city
in Bijar, however for region of Divandarreh, the
wind blows mainly in direction of east.
205
Wind Power Installation and
Economic Issues
In this section, the economic evaluation for
installing four wind turbines in capacity of 0.8,
1.5, 2 and 3 MW, for Bijar and Divandarreh are
estimated using the levelised cost of electricity
(LCOE) method. For this purpose, the weibul
distribution is obtained for these sites. Wind fre-
quency distributions for Bijar and Divandarreh at
60 m height are shown in Figure 5.
Table 2. Annual mean speed at 50 m height from the ground for different locations
Wind power class Annual mean wind
speed (
m
/s)
Height from sea
level (m)
Longitude Latitude Location
Min Deg Min Deg
Class 3 6.73 1883.4 37 47 53 35 Bijar
Class 1 5.58 1906.0 48 47 10 35 Qorveh
Class 1 3.06 1286.8 12 46 31 35 Marivan
Class 1 4.25 1522.8 16 46 15 36 Saqez
Class 1 3.40 1373.4 0 47 20 35 Sanandaj
Class 3 6.72 2142.6 55 46 4 36 Divandarreh
Figure 3. Average wind speeds for different months based on the recorded data from 2004 to 2008, for
a) Divandarreh, and b) Bijar
206
Wind Power Calculation
Calculation of annual energy production from a
WPS in a given site requires the considered turbine
power curve with weibul distributions of wind
speed for the site. In this study, Enercon E-53
(800 KW), Nordex 77 (1.3 MW), Gamesa G90
(2 MW), and Vestas V112 (3 MW) wind turbine
technologies are considered. The information
related to these wind turbines can be obtained
from their manufactures web sites (www.vestas.
com, www.nordex-online.com, www.enercon.de,
and www.gamesa.es).
Capacity factor (CF) is one of important in-
dicators for assessing the performance of a wind
turbine. The capacity factor of a WPS at a given
site can be defined as
CF
E
E
p
rated
= (7)
where, E
p
is the produced energy by the system
in the specific period, and E
rated
is the energy that
Figure 4. Wind rose diagrams based on the recorded data from 1992 to 2006, for a) Bijar, and b) Di-
vandarreh
207
could be produced by the system, while the ma-
chine operates at its rated power in the same period.
The CF can be written as (Jangamshetti &
Rau, 1999):
CF
V
V f V dV f V dV
R V
V
V
V
R
F
C
R
= +

1
3
3
( ) ( ) (8)
where, the v
c
is the cut-in wind speed of wind
turbine generator in m/s, the v
R
is the rated wind
speed of wind turbine generator in m/s, and v
F
is
the cut-out wind speed of wind turbine generator
in m/s. The above equation can be calculated as
CF
V
V
e
K
K
V
C
C
R
V
C K
R
C
=

3
3
3
( )
3
3 3
[ ( , ) ( , )]
V
C K
V
C K
e
R
K
C
K V
C
F
K

(9)
where, is the incomplete gamma function (Jan-
gamshetti & Rau, 1999; Suresh, et al., 2001).
Energy Cost Analysis
The LCOE for WPSs can be described as the ratio
of the total annualized cost to the annual electricity
produced by the system. The following expression
can be used to estimate the LCOE delivered by a
WPS (Gokcek & Genc, 2009; Nouni, et al., 2006),
Figure 5. Frequency distributions of wind speed at 60 m height for a) Bijar, and b) Divandarreh
208
LCOE
C R C R C R C R C R C
E
wt wt bb bb ci ci in in misc misc om
P
=
+ + + + +

$/KWh (10)
Here, the E
p
is annual energy production by
delivered WPS, C
wt
is cost of wind turbine, C
bb

is cost of battery bank, C
ci
is the civil work and
installation cost, C
in
is cost of the inverter, C
misc

is miscellaneous costs such as connecting cables,
control panel and other components; and C
om
is
annual operation and maintenance cost. The R
wt
,
R
bb
, R
ci
, R
in
and R
misc
present the capital recovery
factors (R) for wind turbine, battery bank, civil
work and installation, inverter and other miscel-
laneous components, respectively.
For a given discount rate (r) and useful system
lifetime (n), the capital recovery factor can be
defined as follows:
R
r r
r
n
n
=
+
+
( )
( )
1
1 1
(11)
A break-up of relative costs for different com-
ponents of a typical WPS can be easily obtained
(Nouni, et al., 2006). The cost evaluation is made
by means of this cost break-up for all WPSs. A
typical cost table for different wind power tech-
nologies is presented (Sathyajith, 2006). A specific
cost of WPS can be calculated as follows,
C
WPS
=I
WPS
P
R
[$] (12)
Where, the I
WPS
is the specified cost of the WPS.
The estimation of the KWh cost of energy
delivered by the WPS operating at the given sites
has been done under the following assumptions:
1. The lifetime of the WPS (n) is assumed to
be 25 years.
2. The discount rate (r) is taken as 12%.
3. Operation and maintenance cost (C
om
) is
considered to be 2% of initial capital cost
of the WPS project (Nouni, et al., 2006)
4. Useful lifetime for the battery bank and
inverter are assumed to be 7 and 10 years,
respectively (Nouni, et al., 2006).
5. It is assumed that the WPS production is equal
to the amount of energy output in each year
during its useful lifetime (Trksoy, 1995).
The results of cost analysis performed in this
study for the WPS with different size ranges are
presented in Table 3. From this table, it is seen
that the predicted maximum and minimum values
regarding electricity cost per kWh for each WPS
are calculated by taking into account the limit val-
ues of the band interval of WPS specific cost. The
minimum levelised cost of electricity is calculated
that WPS- Vestas V112 (3 MW) is 0.074 $/kWh,
while its maximum value is 0.118 $/kWh. These
values are the predicted lowest values for WPS
in both cases of Divandarreh and Bijar.
According to the all band intervals, the highest
electricity costs are calculated in the case of WPS-
Gamesa G90 (2 MW) in Bijar, as 0.116 $/kWh
for lower-limit and as 0.186 $/kWh for upper-
Table 3. Cost analysis per kwh for WPS in Bijar and Divandarreh
WPS Divandarreh Bijar
CF Cost ($/kwh) CF Cost ($/kwh)
Min Max Min Max
Enercon E-53 (0.8 MW) 0.2746 0.088075 0.140919 0.2677 0.090345 0.144552
Nordex 77 (1.5 MW) 0.2614 0.092522 0.148035 0.2544 0.095068 0.152109
Gamesa G90 (2 MW) 0.2151 0.112437 0.1799 0.2079 0.116331 0.18613
Vestas V112 (3 MW) 0.3286 0.073601 0.117762 0.3286 0.073601 0.117762
209
limit. As seen from the calculation of annual
energy production, the WPS of 3 MW rated
power among the WPS considered in the study is
most attractive in terms of the levelised unit cost.
Dynamic Impacts Analysis
and Stability Improvement
Case Study
In this section, transient stability of Kurdistan
electric network in the presence of two wind
farms in Bijar and Divandarreh are analyzed.
A combination of FSIG and DFIG turbines are
used in the mentioned wind farms. Single line
diagram of Kurdistan network with wind farms
is shown in Figure 6. Two 50-MW wind power
plants are added to 63 KV bus, near to the cities
of Divandarreh and Bijar. It is noteworthy that in
Kurdistan, only there is one conventional power
plant (Sanandaj Power Plant) with above 200 MW.
Detailed system information and power system
parameters are given in (Saleh, 2010).
Simulation Results
In this study, the power system simulation pro-
gram, Power Factory (DIgSILENT), is used as
a suitable tool for power system modeling and
simulation. In the simulation environment, the
conventional power plant exciter is represented
using the standard model EXST1; the power
system stabilizer is represented using a dual-input
power system stabilizer model (PSS2A), and
governor-turbine is represented by the standard
model GAST. Load model is represented using a
static model. The voltage dependencies on active
and reactive powers are considered as 1 and 2,
respectively.
For transient stability investigation, a three
phase short circuit with duration of 0.34 sec is
considered in an important location. The fault
location is shown in Figure 6. Figure 7 depicts
Figure 6. Single-line diagram of the Kurdistan electric network with two wind farms
210
system response including voltages at Divandarreh
and Bijar buses (connected to the wind farms),
rotor angle of Sanandaj power plant, WPSs speed,
and active and reactive powers.
As discussed, the WPSs commonly use the
induction generators to convert the wind energy
into electrical energy. The induction generators
act as reactive power consumers. Therefore, the
system voltage would be affected in the presence
of wind turbines, especially in the case of fixed-
speed type of WPSs. This issue can be also seen
from simulation results shown in Figure 7. A
decrease in related bus voltages with a permanent
oscillation in active power and speed of aggre-
gated generators are indicated. It is shown that in
view point of reactive power compensation, the
Kurdistan grid is much weaker in Bijar than Di-
vandarreh area.
Stability Improvement
For the sake of system frequency and real power
compensation in the presence of WPSs, several
control approaches are already presented (Bev-
Figure 7. System response following a three phase short circuit
211
rani & Hiyama, 2011; Bevrani, Daneshfar, &
Daneshmand, 2010). In order to reactive power
compensation, which is the main control issue in
the present work, traditionally the capacitor banks
are suggested to use. However, it is noteworthy
that capacitor banks cannot provide dynamic
compensation for events such as the sudden drop
of voltage. In response to above challenge, to
improve stability after grid disturbances such as
short circuit faults, the STATCOM technology as
a powerful control tool is examined.
The STATCOM is extensively being used in
power systems because of their ability to provide
flexible power flow control (Muyeen, et al., 2005).
The main motivation for choosing STATCOM
in wind farms is their ability to provide bus bar
system voltage support either by supplying and/
or absorbing reactive power into the system. The
applicability of a STATCOM in wind farms has
been investigated and the results from early studies
indicate that it is able to supply reactive power
requirements of the wind farm under various oper-
ating conditions, for improving transient stability
(Chun, et al., 2000), as well as enhancement of
the steady-state stability margin (Saad-Saoud, et
al., 1998). Regarding the grid codes mentioned
in Section 2, it is also investigated that the me-
dium voltage STATCOM technology which adds
the missing functionality to wind farms in order
to become grid code compliant. Especially, the
voltage control and the fast dynamic behavior
during balanced as well as unbalanced grid faults
(fault ride-through) are highlighted (Maibach, et
al., 2007).
An appropriately sized STATCOM can provide
the necessary reactive power compensation when
connected to a weak grid. Also, a higher rating
STATCOM can be used for efficient voltage con-
trol and improved reliability in the interconnected
grid with wind farms. However, it is noteworthy
that the STATCOM rating is limited by economic
issues. The location of STATCOM is generally
chosen to be a point in the system which needs
reactive power. Simulation results show that
STATCOM provides effective voltage support
at the bus which is connected to the Bijar wind
farm. That is why, for stability improvement of the
example at hand, a 30-Mvar STATCOM is con-
nected to 63 KV bus, near to the Bijar wind farm.
Another reason for choosing the mentioned
place is that the location of the reactive power
support should be as close as possible to a point
at which the support is more needed. Furthermore,
in the present case study, in addition to the losses
reduction and increase of power transfer capabil-
ity, the location of the STATCOM to the center
of averaged load is more appropriate because the
impact of voltage change is more significant at
this point.
But it is notable that the shipping of reactive
power at low voltages in the system running close
to its stability margin is not very efficient. Also,
the total amount of reactive power transfer avail-
able will be influenced by the transmission line
power factor. Hence, the compensation devices
are always kept as close as possible to the center
of equivalent load as the ratio V/Vnominal will
be higher for the load bus under fault conditions
(Prabhakar, 2008).
The system response in the presence of STAT-
COM for the accrued fault is shown in the Figure
8. Results show a considerable improvement in
transient stability. The transient behavior of wind
farms are also improved by injecting large amounts
of reactive power during the fault recovery.
Flexibility in voltage control for power quality
improvement, fast response, and applicability for
use with high power/load fluctuation are the main
advantages of the proposed STATCOM-based
control strategy.
Some important research needs in future can be
summarized as follows:
212
Coordination between STATCOM, energy
storage devices, power system stabilizers,
and excitation controls of conventional
power plants can be considered as an im-
portant topic for further research in the
feld of power systems stability improve-
ment. Determine the proper location and
size optimizing of STATCOM is another
research topic that should be considered.
A more complete dynamic model is need-
ed in order to stability analysis and control
synthesis in interconnected power systems
with a high degree of wind power pen-
etration. Further study is needed to defne
new grid codes for contribution of large
WPSs into the power system stability/per-
formance improvement. Future grid codes
should clearly impose the requirements on
Figure 8. System response with STATCOM support
213
the regulation capabilities of the active/re-
active power of WPSs.
Control performance standards compliance
verifcation remains a major open issue for
wind power units. This concerns specifc
WPS capabilities and will require the de-
velopment of additional standards for test-
ing, from the level of the component up to
the entire WPS.
Advanced computing algorithm and fast
hardware measurement devices are also
needed to realize more effective optimal/
adaptive control schemes for the power
systems with a high penetration of WPSs.
Since, naturally the wind power is stochas-
tic, still it is diffcult to straightly use wind
turbine kinetic energy storage in the regu-
lation tasks such as frequency control. The
contribution of WPSs in active power and
frequency regulation refers to the ability
of these units to regulate their power out-
put (Bevrani, et al., 2010). More effective
practical algorithms and control method-
ologies are needed to perform these issues.
Further studies are needed to coordinate
the timing and the size of the kinetic en-
ergy discharge with the characteristics of
conventional plants.
To allow the increase of wind power pen-
etration, a change in regulation reserve
policy may be required. In this direction,
in addition to the deregulation policies, the
amount and location of wind turbines, gen-
eration technology, and the size and char-
acteristics of the electricity system must be
considered as important technical aspects.
Continuous development of communica-
tions and information technology, as well
as market and regulatory frameworks for
generation and consumption is necessary
for a power system with intelligent elec-
tricity meters and intelligent communica-
tions (Bevrani, et al., 2011).
The wind turbine units must meet technical
requirements with respect to the voltage,
frequency, ability to rapidly isolate faulty
parts from the rest to the network, and have
a reasonable ability to withstand abnormal
system operating conditions. They could
be able to function effectively as part of
the existing electricity industry particular-
ly during abnormal power system operat-
ing conditions when power system security
may be at risk.
High wind power penetration, particularly in
the locations far away from major load centers and
existing conventional generation units increases
the risk of tie-line overloading, and may require
network augmentation, and possibly additional
interconnections to avoid flow constraints. With
increasing wind power penetration, the grid codes
for the connection high wind turbines capacity
should be also updated (Bevrani, et al., 2011).
Furthermore, the updating of existing
emergency frequency control schemes for
N-1 contingency, economic assessment/
analysis of the frequency regulation prices,
further study on frequency and voltage sta-
bility using dynamic demand control and
ratios of wind turbine technologies, and
quantifcation of reserve margin due to in-
creasing wind power penetration (Bevrani,
et al., 2010) can be considered as other im-
portant research needs in future.
CONCLUSION
In this chapter an intensive overview of wind en-
ergy status around the world and Iran is presented.
The dynamic model and the main control loops
of wind turbine technologies are explained. As
a practical case study, the wind power potential,
economic issues, and technical challenges for
a high penetration of wind power in Kurdistan
214
electric network are discussed. The possibility
of connecting a STATCOM to the wind power
system in order to provide an efficient control
method is explored. The STATCOM as a pure
static device with no switched passive components,
which provides outstanding performance for both
steady state and dynamic operation is used as a
suitable control solution to decrease the undesir-
able impact of wind power plants on the transient
stability and to improve the system performance.
It is shown that the proposed STATCOM based
design strategy provides dynamic voltage control
and power oscillation damping, and improves the
Kurdistan network transient stability.
ACKNOWLEDGMENT
This work is supported by Department of Electri-
cal Engineering at University of Kurdistan. The
authors would like to thank Mr. Naji Ghaderne-
zhad from West Regional Electric Co., and Mr.
Chareh-Khah from Kurdistan Meteorological
Organization for their help and useful comments.
REFERENCES
Akhmatov, V., Knudsen, H., Nielsen, A. H.,
Pedersen, J. K., & Poulsen, N. K. (2006). Mod-
eling and transient stability of large wind farms.
Energy Systems, 25, 123144. doi:10.1016/S0142-
0615(02)00017-0
Al-Abbadi, N. M. (2005). Wind energy resource
assessment for five locations in Saudi Arabia.
Renewable Energy, 30, 14891499. doi:10.1016/j.
renene.2004.11.013
Bansal, R., Bhatti, T., & Kothari, D. (2002).
On some of the design aspects of wind energy
conversion systems. Energy Conversion and
Management, 43, 21752187. doi:10.1016/S0196-
8904(01)00166-2
Bevrani, H. (2009). Robust power system
frequency control. New York, NY: Springer.
doi:10.1007/978-0-387-84878-5
Bevrani, H., Daneshfar, F., & Daneshmand,
P. (2010). Intelligent power system frequency
regulation concerning the integration of wind
power units. In Wang, L. F., Singh, C., & Kusiak,
A. (Eds.), Wind power systems: Applications of
computational intelligence (pp. 407437). Hei-
delberg, Germany: Springer-Verlag.
Bevrani, H., Ghosh, A., & Ledwich, G. (2010).
Renewable energy sources and frequency regula-
tion: Survey and new perspectives. IET Renewable
Power Generation, 4(5), 438457. doi:10.1049/
iet-rpg.2009.0049
Bevrani, H., & Hiyama, T. (2011). Intelligent
automatic generation control. CRC Press.
Bevrani, H., & Tikdari, A. (2010). An ANN-based
power system emergency control scheme in the
presence of high wind power penetration. In
Wang, L. F., Singh, C., & Kusiak, A. (Eds.), Wind
power systems: Applications of computational
intelligence (pp. 215254). Heidelberg, Germany:
Springer-Verlag.
Bevrani, H., & Tikdari, A. (2010). On the neces-
sity of considering both voltage and frequency in
effective load shedding schemes. In Proceedings of
IEEJ Technical Meeting, (pp. 7-11). Fukui, Japan.
Bevrani, H., Tikdari, A., & Hiyama, T. (2010).
An intelligent based power system load shedding
design using voltage and frequency information.
Int. Conf on Modeling, Identification and Control,
(pp. 545-549). Okayama, Japan.
Chun, L., Qirong, J., & Jianxin, X. (2000). Inves-
tigation of voltage regulation stability of static
synchronous compensator in power system. IEEE
Power Engineering Society Winter Meeting, (pp.
4, 2642-47).
215
Conroy, J., & Watson, R. (2008). Aggregate
modeling of wind farms containing full-converter
wind turbine generators with permanent magnet
synchronous machines: Transient stability studies.
IET Renewable Power Generation, 3(1), 3952.
doi:10.1049/iet-rpg:20070091
DIgSILENT GmbH. (2003). Dynamical modeling
of doubly-fed induction machine wind-generators.
DIgSILENT Technical Documentation. Retrieved
from www.digsilent.com
Ekanayake, J., Holdsworth, L., & Jenkins, N.
(2003). Comparison of 5th order and 3rd order
machine models for doubly fed induction gen-
erator (DFIG) wind turbines. Electric Power
Systems Research, 67, 207215. doi:10.1016/
S0378-7796(03)00109-3
Elamouri, M., & Ben Amar, F. (2008). Wind
energy potential in Tunisia. Renewable Energy,
33, 758768. doi:10.1016/j.renene.2007.04.005
Erlich, I., Rensch, K., & Shewarega, F. (2006). Im-
pact of large wind power generation on frequency
stability. In Proc. IEEE Power Engineering Society
General Meeting. Montereal, Que., Canada.
Fernandeza, L. M., Saenza, J. R., & Jurado, F.
(2006). Dynamic models of wind farms with fixed
speed wind turbines. Renewable Energy, 31(8),
12031230. doi:10.1016/j.renene.2005.06.011
Gilbert, M. M. (2004). Renewable and efficient
electric power systems. Hoboken, NJ: John Wiley
& Sons.
Gokcek, M., & Genc, M. G. (2009). Evaluation
of electricity generation and energy cost of wind
energy conversion Evaluation of electricity gen-
eration and energy cost of wind energy conversion.
Applied Energy, 86, 27312739.
GWEC. (2010). Global wind 2009 report. Re-
trieved from http://www.gwec.net
Hansen, A., Jauch, C., Srensen, P., Iov, F., &
Blaabjerg, F. (2003). Dynamic wind turbine models
in power system simulation tool DIgSILENT. Risoe
Report R-1400(EN), Ris National Laboratory,
Denmark.
Hansen, A., Srensen, P., Iov, F., & Blaabjerg, F.
(2004). Control of variable speed wind turbines
with doubly-fed induction generators. Wind Engi-
neering, 28(4). doi:10.1260/0309524042886441
Heier, S. (1998). Grid integration of wind energy
conversion systems. Chicester, UK: John Wiley
& Sons.
Jangamshetti, S. H., & Rau, V. G. (1999). Site
matching of wind turbines generator: A case study.
IEEE Transactions on Energy Conversion, 14(4),
15371543. doi:10.1109/60.815102
Jowder, F. (2009). Wind power analysis and site
matching of wind turbine generators in King-
dom of Bahrain. Applied Energy, 86, 538545.
doi:10.1016/j.apenergy.2008.08.006
Lalor, G., Mullane, A., & OMalley, M. (2005).
Frequency control and wind turbine technologies.
19051913. doi:10.1109/TPWRS.2005.857393
Ledesma, P., & Usaola, J. (2005). Doubly fed
induction generator model for transient stability
analysis. IEEE Transactions on Energy Conver-
sion, 20, 388397. doi:10.1109/TEC.2005.845523
Maibach, P., Wernli, J., Jones, P., & Obad, M.
(2007). STATCOM technology for wind parks
to meet grid code requirements. European Wind
Energy Conference-EWEC. Milan.
Mansouri, M. N., Mimouni, M. F., Benghanem,
B., & Annabi, M. (2004). Simulation model
for wind turbine with asynchronous generator
interconnected to the electric network. Renew-
able Energy, 29, 421431. doi:10.1016/S0960-
1481(03)00225-8
216
Muyeen, S., Mannan, M., Ali, M., Takahashi, R.,
Murata, T., & Tamura, J. (2005). Stabilization of
Grid connected wind generator by STATCOM.
IEEE Power Electronics and Drives Systems, 2.
Muyeen, S., Tamura, J., & Murata, T. (2009).
Stability augmentation of a Grid-connected wind
farm. London, UK: Springer-Verlag.
Nouni, M. R., Mullick, S. C., & Kandpal, T. C.
(2006). Techno-economics of small wind elec-
tric generator projects for decentralized power
supply in India. Energy Policy, 35, 24912506.
doi:10.1016/j.enpol.2006.08.011
Nunes, M. V., Lopes, J. A., Zrn, H. H., Bezerra,
U. H., & Almeida, R. G. (2004). Influence of
the variable-speed wind generators in transient
stability margin of the conventional generators
integrated in electrical grids. IEEE Transactions on
Energy Conversion, 19(4), 692701. doi:10.1109/
TEC.2004.832078
Prabhakar, A. (2008). Application of STATCOM for
improved dynamic performance of wind farms in
a power grid. Master Thesis, Missouri University
of Science and Technology.
Radics, K., & Bartholy, J. (2008). Estimating and
modelling the wind resource of Hungary. Renew-
able & Sustainable Energy Reviews, 12, 874882.
doi:10.1016/j.rser.2006.10.009
Saad-Saoud, Z., Lisboa, M., Ekanayake, J.,
Jenkins, N., & Strba, G. (1998). Application of
STATCOMs to wind farms. IEE Proceedings.
Generation, Transmission and Distribution, 145,
15841589. doi:10.1049/ip-gtd:19982178
Saleh, M. (2010). Dynamic analysis of Kurdistan
electric network in the presence of high penetra-
tion of wind power and determine an appropriate
control solution. Master Thesis, University of
Kurdistan, Iran.
Sathyajith, M. (2006). Wind energy fundamentals,
resource analysis and economics. Berlin/Heidel-
berg, Germany: Springer-Verlag.
Senjyu, T., Sakamoto, R., Urasaki, N., Funabashi,
T., Fujita, H., & Sekine, H. (2006). Output power
leveling of wind turbine generator for all operating
regions by pitch angle control. IEEE Transactions
on Energy Conversion, 21, 467457. doi:10.1109/
TEC.2006.874253
Slootweg, J., Polinder, H., & Kling, W. (2003).
Representing wind turbine electrical generating
systems in fundamental frequency simulation.
516524. doi:10.1109/TEC.2003.816593
Slootweg, J. G. (2003). Wind power: Modelling
and impact on power system dynamics. PhD The-
sis, Delft University of Technology, Netherlands.
Slootweg, J. G., Haan, S. W., Polinder, H., &
Kling, W. L. (2003). General model for repre-
senting variable speed wind turbines in power
system dynamics simulations. IEEE Transactions
on Power Systems, 18, 144151. doi:10.1109/
TPWRS.2002.807113
Suresh, H., Janagmshetti, & Guruprasadu, R. V.
(2001). Normalized power curves as a tool for
identification of optimum wind turbine genera-
tor parameters. IEEE Trans Energy Conv., 16(3),
283-8.
Trksoy, F. (1995). Investigation of wind power
potential at Bozcaada, Turkey. Renewable
Energy, 6(8), 917923. doi:10.1016/0960-
1481(95)00091-7
Ucar, A., & Balo, F. (2009). Evaluation of wind
energy potential and electricity generation at
six locations in Turkey. Applied Energy, 86,
18641872. doi:10.1016/j.apenergy.2008.12.016
Wang, Q., & Chang, L. (2004). An intelligent
maximum power extraction algorithm for inverter-
based variale speed wind turbine systems. IEEE
Transactions on Power Electronics, 19, 1242
1249. doi:10.1109/TPEL.2004.833459
217
Weigt, H. (2009). Germanys wind energy: The
potential for fossil capacity replacement and
cost saving. Applied Energy, 86(10), 18571863.
doi:10.1016/j.apenergy.2008.11.031
ADDITIONAL READING
Akhmatov, V., & Eriksen, P. B. (2007). A Large
Wind Power System in Almost Island Operation
A Danish Case Study. IEEE Transactions on
TPWRS.2007.901283
Ali, M. H., & Wu, B. (2010). Comparison of
Stabilization Methods for Fixed-Speed Wind
Generator Systems. IEEE Transactions on
Power Delivery, 25(1), 323331. doi:10.1109/
TPWRD.2009.2035423
AouzellagLahaani, N., D.Aouzellag, & B.Mendil.
(2010). Static compensator for maintaining voltage
stability of wind farm integration to a distribution
network. Renewable Energy, 35, 24762482.
doi:10.1016/j.renene.2010.03.010
Arulampalam, A., Barnes, M., Jenkins, N., &
Ekanayake, J. (2006). Power quality and stability
improvement of a wind farm using STATCOM
supported with hybrid battery energy storage.
IEE Proceedings. Generation, Transmission and
gtd:20045269
Arulampalam, A., Ekanayake, J., & Jenkins, N.
(2003). Application study of a STATCOM with
energy storage. IET Gener. Transm. Distrib.,
150(3), 373384. doi:10.1049/ip-gtd:20030232
Chang-Chien, L.-R., & Yin, Y.-C. (2009). Strate-
gies for Operating Wind Power in a Similar Manner
of Conventional Power Plant. IEEE Transac-
tions on Energy Conversion, 24(4), 926934.
doi:10.1109/TEC.2009.2026609
Chen, Z., Guerrero, J. M., & Blaabjerg, F. (2009). A
Review of the State of the Art of Power Electronics
for Wind Turbines. IEEE Transactions on Power
Electronics, 24(8), 18591875. doi:10.1109/
TPEL.2009.2017082
Cong, L., & Wang, Y. (2002). Co-ordinated control
of generator excitation and STATCOM for rotor
angle stability and voltage regulation enhancement
of power systems. IET Gener. Transm. Distrib.,
149(6), 659666. doi:10.1049/ip-gtd:20020651
EL-Helw, H., & Tennakoon, S. B. (2008). Evalu-
ation of the suitability of a fixed speed wind
turbine for large scale wind farms considering the
new UK grid code. Renewable Energy, 33, 112.
doi:10.1016/j.renene.2007.08.010
Fang, D., Yuan, S., Wang, Y., & Chung, T. (2007).
Coordinated parameter design of STATCOM stabi-
liser and PSS using MSSA algorithm. IET Gener.
Transm. Distrib., 1(4), 670678. doi:10.1049/
iet-gtd:20060205
Gautam, D., Vittal, V., & Harbour, T. (2009).
Impact of Increased Penetration of DFIG-Based
Wind Turbine Generators on Transient and Small
Signal Stability of Power Systems. IEEE Trans-
doi:10.1109/TPWRS.2009.2021234
Gaztaaga, H., Etxeberria-Otadui, I., Ocnasu,
D., & Bacha, S. (2007). Real-Time Analysis of
the Transient Response Improvement of Fixed-
Speed Wind Farms by Using a Reduced-Scale
STATCOM Prototype. IEEE Transactions on
TPWRS.2007.895153
Han, C., Huang, A. Q., Baran, M. E., Bhattacha-
rya, S., Litzenberger, W., & Anderson, L. (2008).
STATCOM Impact Study on the Integration of a
Large Wind Farm into a Weak Loop Power Sys-
tem. IEEE Transactions on Energy Conversion,
23(1), 226233. doi:10.1109/TEC.2006.888031
218
Jangamshetti, S. H., & Rau, V. G. (2001). Normal-
ized Power Curves as a Tool for Identification of
Optimum Wind Turbine Generator Parameters.
283288. doi:10.1109/60.937209
Li, H., & Chen, Z. (2008). Overview of different
wind generator systems and their comparisons. IET
Renew. Power Gener., 2(2), 123138. doi:10.1049/
iet-rpg:20070044
Li, P., Keung, P.-K., & Ooi, B.-T. (2009). Devel-
opment and simulation of dynamic. IET Renew.
Power Gener., 3(2), 180189. doi:10.1049/iet-
rpg:20070093
Lu, M.-S., Chang, C.-L., Lee, W.-J., & Wang, L.
(2009). Combining the Wind Power Generation
System With Energy Storage Equipment. IEEE
Transactions on Industry Applications, 45(6),
21092115. doi:10.1109/TIA.2009.2031937
Mohod, S. W., & Aware, M. V. (2010). A
STATCOM-Control Scheme for Grid Connected
Wind Energy System for Power Quality Improve-
ment. IEEE Systems Journal, 4(3), 346352.
doi:10.1109/JSYST.2010.2052943
Muyeen, S. M., Takahashi, R., Ali, M. H., Mu-
rata, T., & Tamura, J. (2008). Transient Stability
Augmentation of Power System Including Wind
Farms by Using ECS. IEEE Transactions on
TPWRS.2008.920082
Muyeen, S. M., Takahashi, R., Murata, T., & Tamu-
ra, J. (2009). Integration of an Energy Capacitor
System With a Variable-Speed Wind Generator.
740749. doi:10.1109/TEC.2009.2025323
Muyeen, S. M., Takahashi, R., Murata, T., &
Tamura, J. (2010). A Variable Speed Wind
Turbine Control Strategy to Meet Wind Farm
Grid Code Requirements. IEEE Transactions on
TPWRS.2009.2030421
Prasad, R. D., Bansal, R. C., & Sauturaga, M.
(2009). Wind Energy Analysis for Vadravadra
Site in Fiji Islands: A Case Study. IEEE Trans-
actions on Energy Conversion, 24(3), 750757.
doi:10.1109/TEC.2009.2025326
Qiao, W., Harley, R. G., & Venayagamoorthy, G.
K. (2009). --Coordinated Reactive Power Control
of a Large Wind Farm and a STATCOM Using
Heuristic Dynamic Programming. IEEE Trans-
actions on Energy Conversion, 24(2), 493503.
doi:10.1109/TEC.2008.2001456
Rodrguez, J. M., Fernndez, J. L., Beato, D.,
Iturbe, R., Usaola, J., & Ledesma, P. (2002).
Incidence on Power System Dynamics of High
Penetration of Fixed Speed and Doubly Fed Wind
Energy Systems: Study of the Spanish Case.
10851095. doi:10.1109/TPWRS.2002.804971
Tamrakar, I., Shilpakar, L., Fernandes, B., &
Nilsen, R. (2007). Voltage and frequency control
of parallel operated synchronous generator and
induction generator with STATCOM in micro
hydro scheme. IET Gener. Transm. Distrib., 1(5),
743750. doi:10.1049/iet-gtd:20060385
Yang, Z., Shen, C., Zhang, L., Crow, M. L., & Atcit-
ty, S. (2001). Integration of a StatCom and Battery
Energy Storage. IEEE Transactions on Power
Systems, 16(2), 254260. doi:10.1109/59.918295
Power System Stability: The ability of a power
system to regain a state of operating equilibrium
after being subjected to a physical disturbance,
with most systems indices (voltage, angle, and
frequency) bounded. Power system stability can
take three different forms of rotor angle, voltage,
and frequency stabilities.
Power System Control: This term is used to
define the application of control theory and tech-
219
nology, optimization methodologies, and expert/
intelligent systems to improve the performance
and functions of power systems during normal
and abnormal operations.
Capacity Factor: Capacity factor is defined as
the ratio of the average power output to the rated
output power of the wind energy converter system.
Wind Rose Diagram: The wind rose diagram
illustrates the wind direction in a given site.
Fixed Speed Wind Turbine: Wind turbine
that is directly connected to the grid with a small
speed variation of its rotor.
Variable Speed Wind Turbine (VSWT):
This type of wind turbine is decoupled from the
grid through a power electronic converter and the
rotor acts as a flywheel.
Static Synchronous Compensator (STAT-
COM): A technology being extensively used as
dynamic shunt compensator for reactive power
control in transmission and distribution system.
Levelised Cost of Electricity (LCOE): The
LCOE for WPSs can be described as the ratio of
the total annualized cost to the annual electricity
produced by the system.
Weibull Distribution: The probability dis-
tribution, which is widely used to describe the
long-term records of wind speeds.
220
Chapter 7
INTRODUCTION
We examine in this chapter some new Lagrangian
heuristics for an important combinatorial optimi-
zation problem, a generalized assignment problem.
The classical assignment problem involves profit-
maximizing assignment of each task to exactly
one agent with each agent being assigned to at
most one task (a one-to-one assignment). In the
generalized assignment problems capacity limits
for agents and/or tasks are recognized allowing
one-to-many or many-to-many assignment.
To our knowledge, the first generalized assign-
ment problem was studied by De Maio and Roveda
(1971). They consider a transportation problem
where each demand point must be supplied by
exactly one supply point. Here the agents are the
supply points and the tasks are the demand points.
The requirements of the demand points do not de-
pend on the particular supply point that supplies it,
i.e., the requirements are agent-independent. This
model was further developed by Srinivasan and
Thompson (1972). They made the requirements
agent-dependent and are the first ones to propose
the model that is known today as the generalized
assignment problem. The term generalized assign-
Igor Litvinchev
Nuevo Leon State University, Mexico
Socorro Rangel
So Paulo State University, Brazil
Many-to-Many
Assignment Problems:
Lagrangian Bounds and Heuristic
ABSTRACT
Modifed Lagrangian bounds and a greedy heuristic are proposed for many-to-many assignment problems
taking into account capacity limits for tasks and agents. A feasible solution recovered by the heuristic is
used to speed up the subgradient technique to solve the modifed Lagrangian dual. A numerical study is
presented to compare the quality of the bounds and to demonstrate the effciency of the overall approach.
DOI: 10.4018/978-1-61350-138-2.ch007
221
Many-to-Many Assignment Problems
ment problem for this type of problem was first
introduced by Ross and Soland (1975).
Cattrysse and Van Wassenhove (1992), Mo-
rales and Romeijn (2004) and Pentico (2007) in
their survey papers identify a variety of applica-
tions in which GAP has been used either directly
or as a sub-problem within a broader model type.
When the tasks are jobs to be performed, and
the agents are computer networks, we obtain
the problem described by Balachandran (1976).
Another example is the fixed-charge plant loca-
tion problem, where the agents are capacitated
plants and the tasks are customers, where each
of the customer demands must be supplied from
a single plant (Geoffrion and Graves (1974)).
Other applications that have been studied are the
location problem (Ross and Soland (1977)), the
maximal covering location problem (Klastorin
(1979)), various routing problems (Fisher and
Jaikumar (1981), Bookbinder and Reece (1988)),
assignment in parallel and distributed computing
(Pirkul (1986), Bokhari (1987)), R & D planning
problems (Zimokha and Rubinshtein (1988)),
loading problem in flexible manufacturing systems
(Kuhn (1995)), production planning (LeBlanc,
Shtub, Anandalingam (1999)).
The real life applications of the assignment
models frequently involve large number of tasks
and/or agents thus resulting in large-scale opti-
mization problems. However, most large-scale
optimization problems exhibit a structure that can
be exploited to construct efficient solution tech-
niques. In one of the most general and common
forms of a structure the constraints of the problem
can be divided into easy and complicated.
In other words, the problem would be an easy
problem if the complicating constraints could be
removed. For example, removing agent constraints
in the assignment problems results in independent
subproblems corresponding to tasks.
A well-known way to exploit this structure is
to form a Lagrangian relaxation with respect to
complicating constraints. That is, the complicating
constraints are relaxed and a penalty term is added
to the objective function to discourage their viola-
tion. The optimal value of the Lagrangian problem,
considered for fixed multipliers, provides an upper
bound (for maximization problem) for the original
optimal objective. The problem of finding the
best, i.e. bound minimizing Lagrange multipliers,
is called the Lagrangian dual. Lagrangian bounds
are widely used as a core of many numerical tech-
niques for integer and combinatorial problems,
as well as to measure the progress of the main
algorithm and derive stopping criteria. In many
approximate and heuristic approaches Lagrang-
ian solution is used as a starting or a reference
point to construct the algorithm. The literature
on Lagrangian relaxation is quite extensive, see,
e.g.Lemarechal (2007), Frangioni (2005) and the
references therein.
Frequently a complex system can be rep-
resented as a number of coupled subsystems.
Accordingly, all constraints can be divided into
binding and block ones, such that dualizing binding
constraints results in a decomposable Lagrangian
problem. In many cases there are different ways
to specify subsystems thus resulting in different
decomposable Lagrangian problems for the same
original problem. For example, in different varia-
tions on the assignment problem we may consider
either tasks or agents constraints as binding ones.
Similar properties have routing problems, produc-
tion scheduling, location problems, to mention a
few (Lasdon, 2002).
An approach to improve classical Lagrangian
bound was proposed in Litvinchev (2007) and fur-
ther developed in Litvinchev, Rangel & Saucedo
(2010). The main idea of this approach is to get a
tighter estimation of the penalty (complementar-
ity) term arising in the classical Lagrangian func-
tion. It is well known that under certain convexity
and regularity conditions the penalty turns to zero
for the optimal primal-dual pair (complementar-
ity condition). However, for nonconvex (integer)
problems the complementarity condition is not
necessarily fulfilled. An auxiliary optimization
problem is used to estimate the penalty term and to
222
construct the modified Lagrangian bound and the
corresponding modified dual problem. Suppose
the original problem has two interesting subset
of constraints, i.e. both subsets considered sepa-
rately have attractive, while different structures.
Then we may relax one subset of constraints in
the standard Lagrangian fashion, while using the
other subset in the auxiliary problem to estimate
the penalty term.
In this chapter we specify this approach for
the many-to-many assignment problem recog-
nizing capacity limits for tasks and agents. From
Lagrangian point of view it is important to note
that relaxing task constraints gives independent
subproblems corresponding to agents, while relax-
ing agent constraints results in tasks subproblems.
Thus both subproblems arising in the modified
bound can be solved in a decomposable fashion.
The modified Lagrangian bounds are numeri-
cally compared with classical ones and a greedy
heuristic is applied to the (unfeasible) Lagrangian
solution to get an approximate solution to the
original problem. Combining the modified La-
grangian bound with a greedy heuristic provides
high quality feasible solutions typically within
0.5% of relative difference between primal and
dual bounds. If the subgradient scheme is used to
solve the dual problem, we may restore feasibility
by the (computationally cheap) greedy algorithm
in each iteration of the subgradient technique and
then use this feasible solution to update parameters
of the iterative scheme. Incorporating a feasible
greedy solution into the subgradient scheme results
in a significant decrease in the number of itera-
tions without dropping the quality of the bounds.
The reminder of the chapter is organized as
follows. In Section 2 we present the basic construc-
tions to derive the modified Lagrangian bounds
and give an illustrative example. The modified
bounds for many-to-many assignment problem are
specified in Section 3. Benders and the subgradient
techniques to solve the dual problem are presented
in Section 4, together with a greedy algorithm
to restore feasibility of the Lagrangian solution.
Numerical results are given in Section 5, future
research directions are presented in Section 6 and
Section 7 concludes.
BASIC CONSTRUCTIONS
TO DERIVE THE MODIFIED
LAGRANGIAN BOUNDS
Consider the problem:
z cx Dx d Ax b x U
*
max | , , , = { }
(1)
where xR
n
and the set U can be of general
structure and may contain, for example, sign con-
straints on x and integrality constraints on some
or all components of x. The constraints Dxd are
m complicating constraints, while constraints
Axb are considered nice in the sense that the
optimization problem formed with only these
constraints, together with xU, is easier than the
original problem. Denote by x* an optimal solu-
tion of (1) and let X={ xU | Axb}.
Relaxing the complicating constraints and
using u, an m-vector of Lagrange multipliers, we
can define the standard Lagrangian problem as:
z u cx u d Dx x X u ( ) max ( ), , = + { } 0 .
(2)
We assume for simplicity that it has an optimal
solution for all u0and therefore we get the well
known Lagrangian bound:
z z u u
*
( ), for any 0 . (3)
The best Lagrangian bound and the associated
Lagrange multipliers u* are obtained by solving
the Lagrangian dual problem:
z u z u w
u
D
( ) min ( )
*
=
0
. (4)
223
Suppose now that the original problem (1) has
two interesting subproblems, i.e., both constraints
Dxd and Axb considered separately, have attrac-
tive, while different structures. In this case it may
be useful to reformulate the original problem prior
to relaxation. Introducing copy constraints,x=y,
in the original problem it is possible to build a
Lagrangian decomposition (Guignard and Kim,
1987) by dualizing them with n multipliers R
n

thus obtaining an x-problem and an y-problem:
z c x x X y Dy d y U
x y
( ) max ( ) | max | , = { } + { }

(5)
The Lagrangian decomposition bound is then
stated as:
z z z w
LD
* *
( ) min ( ) =
(6)
and can strictly dominate the standard Lagrang-
ian bound obtained by dualizing either set of
constraints, Dxd or Axb.
Instead of creating a copy y of the variable x
and imposing the constraint x=y, it is possible to
transform the original problem by introducing an
aggregated copy constraint Dx=Cy, where Cis
a matrix of suitable dimension. The aggregate
copy constraint is then dualized providing an
aggregated Lagrangian decomposition bound
(Maculan & Reinoso, 1992), which, in general,
is weaker than w
LD
.
In what follows we consider constructions
to improve the standard Lagrangian bound and
provide two interpretations of the approach: one
using a reformulation of the original problem, and
another one in terms of a subproblem to estimate
the complementarity term.
We assume certain information about an op-
timal solution to (P), x*:
Assumption. A set WR
n
is known, such that
x*W.
We will refer to W as the localization of x*, or
simply the localization. The set W can be defined
by manipulating the constraints of the original
problem, by querying a decision maker, etc. We
will distinguish the case of a priori localization,
when the set W is defined before the Lagrangian
problem has been solved for some u0 and the
associated bound z(u) has been calculated, and
a posteriori localization, when W is defined or
corrected after the Lagrangian problem has been
solved.
Consider a further modification of the original
problem (1):
z cx x X Dx Dy y W
M
= { } max | , , .
(7)
Since a pair (x,y)=(x*,x*) is feasible to (7),
it is a relaxation of the original problem (1) and
hence z*=z
M
. Note also that if the localization W
is such that yW implies Dyd (this is the case,
for example, when constraints Dyd are included
explicitly in the definition of W) then an optimal
x-solution to (7) is feasible to (1) and hence z*z
M
.
Thus for such a localization we have z*=z
M
.
Dualizing constraints DxDy with multipliers
u0 we get the modified Lagrangian bound:
z z z u cx uDx uDy
M M
x X y W
*
( ) max max = { } + { }

(8)
The modified bound z
M
(u) obtained for the
localization W={y|Dyd,yU} coincides with
the aggregated Lagrangian decomposition bound
calculated for C=D.
To simplify further notations, we use (u)for the
optimal objective value of the first maximization
problem in (8), and (u) for the optimal objective
of the second. The modified dual problem cor-
responding to the bound (MLB) is then stated as:
w u u
MD
u
= + { }
min ( ) ( )
0
. (9)
224
It is possible to derive (8) using other ar-
guments. The feasibility of x* to the standard
Lagrangian problem implies that cx*+u(d-
Dx*)z(u). Since x* is feasible to (1) and u0,
then u(d-Dx*)0 yields immediately z*z(u) as in
(3). The complementarity condition u*(d-Dx*)=0
is fulfilled for convex problems (1) which fulfill
certain regularity assumptions. But even for the
convex case the term u(d-Dx*) can be strictly
positive for uu*. For the nonconvex case the
complementarity term u(d-Dx*) can be strictly
positive for u=u*. Thus we may try to strengthen
the standard Lagrangian bound z(u) using more
tight estimations of the complementarity term
instead of u(d-Dx*)0.
Since x*W, then we have for any localiza-
tion W
z u cx u d Dx cx u d Dy
y W
( ) ( ) min{ ( )}
* * *
+ +

(10)
such that:
z cx u d Dx u d Dy
x X y W
*
max{ ( )} min{ ( )} +

(11)
This bound, after elementary algebraic trans-
formations, coincides with z
M
(u) defined in (8).
We may expect that for those integer programs,
where the constraints Dxd are not active for all
feasible solutions, a reasonable choice of W may
result in ( ) min ( ) u u d Dy
y W
= { } >
0 , thus
improving the standard Lagrangian bound z(u).
Calculating (u) so far was solely intended
to estimate the complementarity term associated
with Dxd without taking into account the original
objective function. We may balance z(u) and (u)
by introducing some information of the original
objective in the modified problem. Let LB and
UB be (unknown) lower and upper bounds for
the optimal objective of (1). Then we may add to
(1) the constraint LBcxUB and dualize it with
multipliers
LB
UB
0.
Estimating the extended complementarity term
u d Dx UB cx LB cx
UB LB
( ) ( ) ( )
* * *
+ +
(12)
and using the localization W as before, gives the
bound
z z z u cx uDx cy uDy
M M
x X y W
*
( , ) max ( ) max ) . = { } + + { }

1

(13)
The unrestricted dual variable stands
for
LB
UB
. The bound z
M
(u,) in (13) uses
m+1multipliers, an m-vector u and a scalar ,
while the Lagrangian decomposition bound, z(),
uses an n-vector of multipliers, .
The modified dual problem corresponding to
the bound (LMB) is stated as:
w u u
MD
u
= + { }
min ( , ) ( , )
, 0
, (14)
where (u,)is used to denote the optimal objective
value of the first maximization problem in z
M
(u,),
while (u,) stands for the optimal objective of
the second.
It is not hard to verify that aggregating to the
problem (7) the valid constraint cx=cyand dual-
izing it together with constraints DxDy we get
exactly the same expression for the bound as in
(13).
Defining Localizations
and Estimating (u,)
The critical issue in using the modified Lagrangian
bound (13) is constructing a suitable localization
W. From the definition of z
M
(u,) in (13) it follows
that, in general, the tighter the localization W is,
the smaller is (u,) and the better is the modified
upper bound z
M
(u,). From this point of view, it
is worth to retain in the definition of W as many
original constraints as possible. However, the
localization should be simple enough to guarantee
225
that calculating (u,) in (13) is easy (by our
assumption on the original problem, the calcula-
tion of (u,) is easy).
In particular, defining W
all
by all original con-
straints, we obviously get w W z
MD all
( )
*
= with
u*=0, *=1. Since it can be difficult to calculate
(u,) under the localization defined by all (many)
constraints, we may try to use estimations of the
corresponding value of (u,). In particular, this
can be done either by simply relaxing some com-
plicating constraints, or dualizing them to get the
Lagrangian bound, standard or modified. If La-
grangian relaxation is used to estimate (u,), we
will refer to this case as the nested Lagrangian
relaxation.
If a localization W y Dy d y U
0
= { } | , is
used to define ( ) min ( ) u u d Dy
y W
= { }
0
0 for
any u0, then:
w z u z u u
z u z u
D
u u
u
M
u
M
=
{ }
=

min ( ) min ( ) ( )
min ( , ) min (
,
0 0
0 0
0
,, ) ( )
w W
MD 0

(15)
and we may possibly strengthen the standard dual
bound w
D
by considering the modified Lagrang-
ian dual (14). The bound w W
MD
( )
0
calculated
fixing =0 coincides with the aggregated Lagrang-
ian decomposition bound (Maculan and Reinoso,
1992).
If the localization W
0
is decomposable, that
is W
0
=W
01
W
02
W
0L
, the calculation of (u,)
in (14) reduces to L independent subproblems of
smaller dimensions. There are many classes of
problems with both X and W having such structure
and resulting in decomposable calculations of
(u,) and (u,). This is often the case for prob-
lems with x
ij
variables: generalized assignment,
facility location, multiple knapsack, cutting and
packing problems, among others.
Another suitable localization resulting in
( ) u 0 i s a sur r ogat e l ocal i zat i on
W y U uDy ud
s
= { } | , yi el di ng al so
w w W
D MD s

( ).
We may interpret the use of localizations W
0
and
W
s
as follows. Instead of calculating the exact
(u,). using the localization W
all
defined by all
original constraints, we simply drop constraints
Axb and use the associated estimation of the
exact (u,). (possibly combining Dxd in a
unique constraint as in W
s
.
Another way to estimate ( , ) u is to use La-
grangian relaxation, standard or modified, instead
of simply dropping constraints. Suppose that the
localization has the form:
W y Py p y Y U
1
= { } | , ,
where Y R
n
, p R
q
and the matrix Pis di-
mensioned accordingly. We assume that the set Y
has a favorable structure (for example, decompos-
able) and we will handle the constraints yY
explicitly, and will dualize the constraints Pyp
using a q-vector of multipliers v 0. Estimating
( , ) u by the standard Lagrangian bound yields:

( , ) max
max ( )
( , ,
u cy uDy
cy uDy v p Py
u v
y W
y Y U
L
+
{ }
+ +
{ }
)), v 0

(16)
while the modified Lagrangian bound (considered
for simplicity without dualizing the objective
copy constraint) results in the estimation

( , ) max
max )
max
,
u cy uDy
cy uDy vPy
y W
y Y U
w U Pw p
+
{ }
+
{ }
+
vvPw u v v
ML
) ( , , ),
{ }
0

(17)
and
ML L
u v u v ( , , ) ( , , ) .
226
Based on these estimations of ( , ) u we get
two modified dual problems associated with the
nested Lagrangian relaxation for localization W
1

that use m+q+1 multipliers:
w W u u
MD
L
u v
L
( ) min ( , ) ( , )
, ,
1
0
= +
{ }
(18)
and
w W u u
MD
ML
u v
ML
( ) min ( , ) ( , )
, ,
1
0
= +
{ }
.
(19)
If both original constraints, Axb and Dxd,
considered separately are easy, then defining
Y={y| Dxd } and {Pxp}={ Axb } results in an
easy calculation of
L
u v ( , , ) . For this case
we have
w W u u v v
MD
L
u v
L L
( ) min ( , ) ( , , ) ( , ) ( , , )
ma
, ,
1
0
0 1 0 1 = + { } + =
xx ( ) | , cy v b Ay Dy d y U w
D
+ { }
2

(20)
and
w W u u
cx uDx uDy Dy d y U
MD
L L
x X
( ) ( , ) ( , , )
max max | ,
1
0 0 0 +
= { } + {
}}
+ { }
max ( ) .
x X
D
cx u d Dx w
1

(21)
Hence w W
MD
L
( )
1
is at least as good as any
of the two standard Lagrangian relaxations.
Illustrative Example
To give an idea on the behavior of the Lagrangian
bounds (standard and modified) using the localiza-
tions discussed in Section 2.1 consider the follow-
ing binary problem (Freville and Hanafi, 2005):
z x x x x
x
*
{ , }
max = + + +
0 1 1 2 3 4
2 2 (22)
8 16 3 6 18
1 2 3 4
x x x x + + + , (23)
5 10 8 16 19
1 2 3 4
x x x x + + + . (24)
The optimal solution to this problem is z*=2
with three alternative optimal points x*{(0, 1,
0, 0), (0, 0, 0, 1), (1, 0, 1, 0)}. Freville and Hanafi
(2005) present several Lagrangian bounds for this
problem instance; a summary of their results is
as follows. The linear programming relaxation
gives z
LP
=3.04. The optimal multipliers associ-
ated to the copy constraints in the Lagrangian
decomposition bound are
*
, , , = ( )
2
3
1 1 1 2 giving
w
LD
= 2 2 667
2
3
. . The classical Lagrangian
bound obtained by dualizing constraint (23) yields
w
D
1
2 2 923
12
13
= . with the corresponding mul-
tiplier u
1
1
13
*
= . Dualizing constraint (24) gives
w
D
2
2 2 947
18
19
= . for u
2
2
19
*
= . The surrogate
relaxation, obtained by combining the two origi-
nal constraints into a single one using two multi-
pliers =( , . ) 1 0 5 gives the bound w
s
= 3 .
To calculate the modified bounds, let Axb be
defined by constraint (24) while constraint (23)
stands for Dxd and is dualized with the multi-
plier u0. Let U x = { }
{ }
1 0
4
, and suppose
that the localization W
1
is defined by the two
original constraints, such that constraint (23) is
used to define the condition yY and is handled
explicitly, while constraint (24) stands for Pyp
and is dualized with multiplier v0 in the estima-
tion of
L
u v ( , , ) . Using a standard technique,
we present equivalently the dual problem
w W
MD
L
( )
1
as a linear programming problem
(master problem) having constraints associated
with all feasible points of x Ax b U | { } (for
( , ) u ) and of Y U (for
L
u v ( , , ) ). Eliminat-
ing the redundant constraints we get the complete
master problem:
227
w W
s t
u
u
MD
L
u v
( ) min
. .
, ,
1
0
2 2 6
3 3 19
0
= +
+ +
+ +
+
+
8 14
2 16 9
4 17 10
3 14 2
19
u v
u v
u v
u v
v

(25)
where the constraints for and correspond to
the all-nonzero feasible solutions. The optimal
solution to the master problem (25) is =
1
25

u =
2
25
, v = 0 , =1 6 . , =1 2 . giving
w W w w
MD
L
D D
( ) ,
1
2.8 min
1 2
= <
{ }
=2.923.
If we do not use the copy constraint cx cy =
that is setting = 0 in the above master problem,
the corresponding solution is, u =
1
13
0 0769 .
v =
1
247
0 004 . , =
20
13
1 538 . ,
=
313
247
1 267 . giving
w W
MD
L
( )
1
693
247
2.8057 = + = .
Instead of using the Lagrangian relaxation

L
u v ( , , ) to estimate ( , ) u we may use a sur-
rogate relaxation. Let
W x x x x x
s
= { } + + +
{ }
1 0 10 5 21 7 14 27 5
4
1 2 3 4
, | . .

be the surrogate localization obtained by mul-
tiplying constraint (23) by
1
1 = , constraint (24)
by
2
0 5 = . and summing them up. These mul-
tipliers are optimal to the surrogate relaxation of
the original problem and result in the surrogate
bound w
s
= 3 . For W=W
s
the modified dual
bound w W
MD s
( )is calculated using the linear
master problem. Note that the constraints corre-
sponding to the variable in the master problem
remain the same as before. The optimal solution
to the new master problem is =
2
15
, u =
1
15
,
= =
4
3
giving w W
MD s
( ) . = 2 2 667
2
3
,
which coincides with the bound w
LD
obtained by
Lagrangian decomposition. If we use a localiza-
tion W=W
0
, defined in this case by the constraint
(23), and do not use the copy constraint (cx=cy),
the associated bound coincides with the aggre-
gated Lagrangian decomposition bound. To cal-
culate this bound we need to set =v=0 in the
master problem. The associated solution is
u =
1
13
0 0769 . , =
20
13
1 538 . ,
=
17
13
1 308 .
giving the bound + =
37
13
2.846 .
Let now constraint (23) stands for Axb and
constraint (24) be dualized. Moreover, the latter
constraint is handled explicitly in the definition
of yY, while the constraint (23) is dualized in
the estimation of
L
u v ( , , ) . After eliminating
the redundant constraints, the complete master
problem becomes:
w W
s t
u
u
u
MD
L
u v
( ) min
. .
, ,
1
0
2 2 10
4 4 29
3 3 24
0
= +
+ +
+ +
+
,
8 15
2 16 12
3 18
18
u v
u v
u v
v

(26)
Its optimal solution is =
4
15
, u =
2
15
,
v = 0, =1 2 . , =1 6 . giving w W
MD
L
( )
1
2.8 =
Fixing =0 in the master problem (and thus
computing the bound without the copy constraint
(cx=cy), we get the bound + = 2.826 . Setting
=v=0 (i.e. calculating the aggregated Lagrangian
decomposition bound) yields + = 2.842 .
MODIFIED LAGRANGIAN
BOUND FOR MANY-TO-MANY
ASSIGNMENT PROBLEM
The assignment problems (AP) involve optimally
matching the elements of two or more sets. When
there are only two sets, they may be referred as
228
tasks and agents. For example, tasks may
be jobs to be done and agents may be the people
or machines that can do them. In its original ver-
sion, the AP involves assigning each task to a
different agent, with each agent being assigned
to at most one task (a one-to-one assignment). In
the generalized assignment problem (GAP) each
task is assigned to one agent, as in the classic
AP, but it allows for the possibility that an agent
may be assigned to more than one task, while
recognizing that a task may use only part of an
agents capacity rather than all of it. Thus GAP is
a one-to-many assignment problem that recognizes
capacity limits for agents (see Martello and Toth
(1990), Pentico (2007) and the references therein).
A further generalization of AP is a many-to-
many assignment recognizing capacity limits of
both tasks and agents. Such a situation arises,
for example, in a medical center, where doctors
(agents) have to attend their patients (tasks) in a
limited time period, while patients cannot also
spend a lot time in the center. This leads to the
following optimization model:
z c x
MMAP x ij ij
j
n
i
m
=

= =

max
{ , } 0 1
1 1
(27)
a x b i m
ij ij
j
n
i
=
=
1
1 , ... , (28)
d x d j n
ij ij
i
m
j
=
=
1
1 , ... . (29)
In what follows we will refer to the problem
(27)-(29) as (MMAP). Here x
ij
=1 if agent i is
assigned to task j, and 0 otherwise, c
ij
is the
profit (utility) of assigning agent i to task j, a
ij
is
the amount of agent is capacity used to execute
task j, and b
i
is the available capacity of agent i.
It is assumed that each task has its own capacity
(time) limit, such that d
ij
is the amount of task js
capacity used when executed by agent i, and d
j
is
the available capacity of task j. Note that (MMAP)
has a double-decomposable structure: if we dual-
ize the first m restrictions (28), then the relaxed
problem decomposes into n independent subprob-
lems, each having a single knapsack-type restric-
tion d x d
ij ij
i
m
j
=

1
, while relaxing the second
group of restrictions (29) we get m single knapsack
constrained subproblems. To derive the modified
bounds for the MMAP problem let us define the
sets:
X x a x b i m X
ij ij ij
j
n
i
i
m
i
= { } =
=
=
=
1 0 1
1
1
, | , ...

X x a x b
i ij ij ij
j
n
i
= { }
1 0
1
, | (30)
Y x d x d j n Y
ij ij ij
i
m
j
j
n
j
= { } =
=
=
=
1 0 1
1
1
, | , ...

Y x d x d
i ij ij ij
i
m
j
= { }
1 0
1
, | (31)
Py p y a y b i m
ij ij ij
j
n
i
{ } { } =
1 0 1
1
, | , ...

(32)
and the localization:
W y y Y Py p
ij 1
1 0 = { }
{ }
, | , . (33)
The original constraints included in the set X
will be considered as easy, while those in Y will
be treated as complicating. Localization W
1
will
be used in the modified Lagrangian dual to cal-
culate the value w W
MD
L
( )
1
. We will handle con-
straints yY explicitly, while restrictions Pyp
will be dualized in the estimation of ( , , ) u v .
L e t u u j n
j
= =
{ }
, 1 0 a n d
v v i m
i
= = { } , 1 0 be the Lagrangian mul-
tipliers. Then the modified Lagrangian dual for
(MMAP) is:
229
w W u v
MD
L
u v
( ) min ( , , )
, ,
1
0
=
(34)
where:
( , , ) ( , ) ( , , ) u v u u v
L
= + , (35)
( , ) max ( ) u c u d x
x X
ij j ij ij
j
n
i
i
=
l
l
l
'
!
1
1
+
1
1
'
!
1
1
+
1
1
= =
1
1 11
m
,
(36)

L
y Y
ij j ij i ij ij
i
m
u v c u d v a x
i
( , , ) max ( =
l
l
l
'
!
1
1
+
1
1
'
!
1
1
11
+
1
1
= =

j
n
i i
i
m
v b
1 1

(37)
SOLVING THE DUAL PROBLEMS
In this section we describe two main procedures
to solve the dual problems defined in Section 3
for the MMAP. First we apply a constraint gen-
eration scheme (Benders method) transforming
the modified dual problem (18) into a large-scale
linear programming problem (Section 4.1). The
main advantage of using Benders technique is
that it generates upper and lower bounds for w
MD
L
in each iteration thus producing near-optimal

solutions with guaranteed quality. The objective
is to compare the quality of the modified and
classical Lagrangian bounds. A simple greedy
heuristic to obtain feasible solutions to (MMAP)
is presented in Section 4.2. The feasible greedy
solution is then used in a subgradient algorithm
to obtain the modified bounds (Section 4.3).
Solving the Dual Problem
by Benders Technique
In this section we apply a constraint generation
scheme (Benders method) to solve the modified
Lagrangian dual (18) without considering the
copy constraints, cx=cy, that is setting =0. The
main focus of this approach is to compare the
quality of the bounds. Benders technique provides
upper and lower bounds for w W
MD
L
( )
1
in each
iteration thus producing near-optimal values of
the dual bounds with guaranteed quality.
Let x t T
ij
t
, =
{ }
1 and y t L
ij
l
, =
{ }
1 be
all feasible points of X and Y. Then the dual prob-
lem (MD
L
) used to compute w W
MD
L
( )
1
can be
stated as follows:
w W
MD
L
u v R
( ) min ( )
, ; , 1 0
=

(38)
subject to
+ =
= = =

u d c u d x t T
j j
j
n
ij j ij ij
t
j
n
i
m
1 1 1
1 ( ) , ... ,
(39)
+ =
= = = =

u d v b v a u d y l L
j j
j
n
i i
i
m
i ij j ij ij
l
j
n
i
m
1 1 1 1
1 ( ) , ... .

(40)
In what follows we will refer to the (master)
problem (38)-(40) as (MP). The latter is an LP
problem having 2+m+n variables and a large
number of constraints - one for each feasible point
of X and Y. To solve master problem (MP) we
use constraint generation scheme in the form of
Benders algorithm. We omit here the complete de-
scription of this well known iterative method (see
Lasdon (2002), and Martin (1999) for details) and
focus only on the constraint generation scheme.
Consider that on the k-th iteration we have a
restricted master problem (RMP
k
), having fewer
constraints (39) and (40) compared to (MP).
Denote its optimal solution by u v
k
, , , ( ) . To
check the feasibility of this solution to all con-
straints (39) we need to verify if:
k
j
k
j
j
n
ij j
k
ij ij
t
j
n
i
m
u d c u d x t T =
= = =

1 1 1
1 ( ) , , ..., ,
(41)
230
or equivalently:
k
j
k
j
j
n
x X
ij j
k
ij ij
j
n
i
m
x X
ij
u d c u d x c
i

=

= =

1 1 1
max ( ) max ( uu d x
j
k
ij ij
j
n
i
m
) ,
= =

1 1

(42)
where the maximization over X is reduced to
independent maximizations over X
i
due to decom-
posable structure of X. That is to verify the fea-
sibility with respect to constraints (39) we need
to solve m independent subproblems each one
having a single knapsack constraint and n binary
variables. Denote by x
ij
k
{ }
their optimal solution.
Similarly, to check the feasibility with respect
to constraints (40) we need to verify:
k
j
k
j
j
n
i
k
i
i
m
y Y
i
k
ij j
k
ij ij
y Y
u d v b v a u d y
j
+
= =

1 1
min ( ) min (vv a u d y
i
k
ij j
k
ij ij
i
m
j
n
j
n
i
m
= = = =

) ,
1 1 1 1

(43)
which results in solving n independent subprob-
lems with a single knapsack constraint and m
binary variables. Let y
ij
k
{ }
be their optimal solu-
tion.
If (42) and (43) are fulfilled, stop with
u v
k
, , , ( ) optimal to (MP). If (42) fails, add:
+
= = =

u d c u d x
j j
j
n
ij j ij ij
k
j
n
i
m
1 1 1
( ) (44)
to the restricted master problem. If (43) fails, add:
+
= = = =

u d v b v a u d y
j j
j
n
i i
i
m
i ij j ij ij
k
j
n
i
m
1 1 1 1
( )
(45)
to the restricted master problem. So in each itera-
tion we add at most two constraints to (RMP
k
) to
get the next restricted master problem (RMP
k+1
).
On the k-th iteration of Benders technique we
have a lower and an upper bound for the optimal
value w W
MD
L
( )
1
of (MP):
LB w W z u u v UB
k k k
MD
L
s k
s s s k
= { } =
=
( ) ( ) min ( ) ( , ) ,
...

1
1

(46)
where the minimum is taken over all previous
iterations and for the iteration s we have:
z u u d c u d x
s
j
s
j
j
n
ij j
s
ij ij
s
j
n
i
m
( ) ( ) = +
= = =

1 1 1

(47)
and
( , ) ( ) u v u d v b v a u d y
s s
j
s
j
j
n
i
s
i
i
m
i
s
ij j
s
ij ij
s
j
n
i
= +
= = = =

1 1 1 1
mm

(48)
The iterative process terminates if, for ex-
ample,( ) / , UB LB LB
k k k
where >0 is a
given threshold.
Assuming the objective coefficients in
(MMAP) are nonnegative we have z
MMAP
0 .
The optimal solution u v , , ,
*
( ) to the problem
(MP) then satisfies:
0
1
= z w W
MMAP MD
L
( )
* *
,
0 = z z u
MMAP
( )
* *
. (49)
Thus we can add to the master problem (MP)
the trivial restrictions 0 and 0 at
beginning of the constraint generation process in
order to prevent the objective function from de-
creasing unboundly at early iterations.
Restoring Primal Feasibility
by a Greedy Heuristic
The modified Lagrangian dual (18) provided high-
quality bounds for the (MMAP) for the instances
tested in (Litvinchev & Rangel, 2008; Litvinchev,
Rangel & Saucedo, 2010). It turned out that the
corresponding integer Lagrangian solutions (x,y)
had a higher degree of primal feasibility and sub-
231
optimality than the standard Lagrangian solution.
In this section we consider a greedy algorithm to
recover primal feasibility. The feasible solution
is also used in a subgradient algorithm to obtain
the modified bounds (Section 4.3).
To get a feasible Lagrangian based solution
we use a simple greedy approach. First we try to
decrease to zero some components currently equal
to 1 to obtain a feasible solution. The choice of
the candidate component is based on the smallest
decrease of a rounding indicator (e.g. minimal cost
component). After a feasible solution is obtained
we try to increase to 1 some zero components
based on the largest increase of another rounding
indicator (e.g., maximal cost component) while
maintaining feasibility.
Let x
0
be a current binary point not necessary
feasible to (MMAP). Let
1
be a set of all pairs
(i,j) with x
ij
0
0 = and
1
be a set of all pairs (i,j)
wi t h x
ij
0
1 = . Denot e
i i ij
j
ij
b a x =
0
,
j j ij
i
ij
d d x =
0
. If min ,
ij i j

{ }
0 then
x
0
is feasible to (MMAP). Otherwise, we first
decrease, in a greedy manner, some positive x
ij
0

to 0 to get a feasible solution (x
gr
). Then we try
to improve this feasible solution by increasing,
in a greedy fashion, some zero components to 1.
Algorithm 1
1. Letx
0
beaLagrangiansolution(x y or ).Setx
gr
=x
0
.
2. Sett r
ij ij
, astheroundingindicators(e.g.t c r c
ij ij ij ij
= = , seeComment
1).
3. Feasibilitytest.
3.1. Forx
0
compute
i j
, .
3.2. Whilemin ,
ij i j

{ }
0do(roundingdown)
3.2.1. Compute:min
ij ij
t
{ }
1
.Letthisminimumbeattainedfor(i,j).
3.2.2. Set:
3.2.3. x
ij
gr
= 0 for(i,j)=(i,j),
3.2.4.
1 1
= { } \ ( , )' i j ,
0 0
= { } ( , )' i j
3.2.5.
i i i j j j i j
a d = + = +
( , )' ( , )'
, ,
3.3. end_while
4. Letx
gr
beafeasiblesolutionobtainedintheroundingdownstep,
andS
0 0
beasetof( , ) i j
0
withbotha
ij i
andd
ij j

5. whileS
0
0 do(roundingup)
5.1. Computemax
ij S ij
r
{ }
0
.Letthismaximumbeattainedfor(i,j).
5.2. Set
5.2.1. x
ij
gr
=1for(i,j)=(i,j),
5.2.2.
1 1
= { } ( , )' i j ,
0 0
= { } \ ( , )' i j
5.2.3.
i i i j j j i j
a d = =
( , )' ( , )'
, ,
5.2.4. updateS
0
5.3. end_while
6. Returngreedysolutionx
gr
.
7. EndAlgorithm1.
232
Algorithm 1 gives a summary of the Greedy
Heuristic, its flowchart can be seen in Figure 1.
Comment 1. The rounding down part of the
Algorithm 1 may be based on pure cost criterion
t
ij
=c
ij
. It is also possible to use another indicator,
setting for example, t
ij
=c
ij
/max{a
ij
,d
ij
}. In this
way we can take into account the impact of the
component (i,j) in violating the constraints (the
larger the values of a
ij
,d
ij
the faster we get feasibil-
ity). Similarly, we can try t
ij
=c
ij
/max{a
ij
/b
i
,d
ij
/d
j
}
since the relative values a
ij
/b
i
,d
ij
/d
j
also give a
measure of feasibility. In the rounding up part of
the Algorithm 1 we may use r
ij
= t
ij
. Alternatively,
we can use r
ij
= c
ij
/min{ a
ij
, d
ij
}. Small values of
a
ij
,d
ij
help to obtain a small degradation in the
solution feasibility. Another possibility is to set
r
ij
=c
ij
/min{a
ij
/b
i
,d
ij
/d
j
}.
Figure 1. Flowchart of Algorithm 1
233
The Lagrangian solution is always feasible
either to the first or to the second group of con-
straints of the problem (MMAP). So we can
simplify Algorithm 1 by considering only
i j
( )
when rounding down, depending on whether x or
y is used for rounding the modified Lagrangian
solution.
Solving the Dual Problem by
the Subgradient Technique
A popular approach to solve the dual problem is
by subgradient optimization. Here we present the
basic steps of the subgradient technique used in
Litvinchev et al. (2010a) to calculate w W
MD
L
( )
1

and modified to use the greedy solution obtained
by the Algorithm 1 (Litvinchev et al., 2010). A
more detailed discussion of subgradient optimiza-
tion can be found in Martin (1999) and Wolsey
(1998).
Let u v
k
, , ( ) be the values of the Lagrangian
multipliers for the iteration k,
k k k k
u v = ( , , )
and x
ij
k
, y
ij
k
be the associated subproblems solu-
tions:
x c u d x
ij
k
x X
k
ij j
k
ij ij
j
n
i
=
l
l
l
'
!
1
1
+
1
1
'
!
1
1
+
1
=
arg max ( ) 1
1
11
,
(50)
y c u d v a x
ij
k
y Y
k
ij ij
k
ij i
k
ij ij
i
m
i
=
l
l
l
'
!
1
1
+
1
1
'
arg max (
1
!!
1
1
+
1
1
.
(51)
After solving the subproblems, a subgradient
is directly identified as:

k
k
ij ij
k
j
n
i
m
ij ij
k
j
n
i
m
c x c y = [ ] = +
= = = =

/
1 1 1 1

(52)

i
k
i
k
i ij ij
k
j
n
i
m
v b a y = [ ] = +
= =

/
1 1
, (53)

j
k
j
k
ij ij
k
j
n
i
m
ij ij
k
j
n
i
m
u d x d y =
= +
= = = =

/
1 1 1 1

(54)
Denote by s
k
a vector composed of all

k k k
, ,
{ }
, let
k k k k
u v =
{ }
, , and set:

k k
k
k LB
k
k
s
s
+
=
1
2
( ) , (55)
where
k
(0,2],
LB
is a lower bound on

*
( ) = w W
MD
L
1
. Sincez w W
MMAP MD
L

( )
1
, we
may set
LB
equal to the objective function value
of (MMAP) associated to a given feasible solu-
tion. In what follows we will apply the greedy
algorithm in each iteration to get a feasible solu-
tion and update
LB
accordingly.
Since u,v0, the multipliers for the next it-
eration are defined as the projection of u
k
and v
k

onto the nonnegative orthant, while has no sign
restrictions:
u u v v
k k k k k k + + +
=
{ }
=
{ }
=
1 1 1
0 0 max , , max , , .

(56)
A summary of the subgradient algorithm to
compute the modified bound for the (MMAP)
is given by Algorithm 2 and its flowchart can be
seen in Figure 2.
The subgradient method is not monotone, that
is, it is not necessary that
k k
+1
. In practice,
the parameter
k
is varying in (0,2], beginning
with
k
=2. If after K consecutive iterations with
a fixed value for
k
the function does not improve
sufficiently, then a smaller value of
k
is used,
say, a half of
k
. The stopping criteria used in
Algorithm 2 are: a) maximum iteration number
is reached; b)
k
is already small enough; or c) the
relative difference between the best integer fea-
234
sible solution found so far and the Lagrangian
bound is within a given threshold.
COMPUTATIONAL TESTS
The objective of the following numerical study is
to compare the relative quality of the bounds as
well as their proximity to the optimal objective. We
numerically compare the Lagrangian bounds for
two sets of instances of (MMAP): Set1 instances
with sizes mn for m=5,8,10 and n=50 and Set2
instances with m=5,10,20 and n=100. The data
were random integers generated as follows:
c U a U d U b
ij ij ij i
[ , ], [ , ], [ , ], 10 5 0 5 25 3 20
= =

( ), ( ), a d d
ij
i
j ij
j
1 1 0 1
(57)
Figure 2. Flowchart of Algorithm 2
235
and divided in three classes (a, b, and c) with
respect to the values of : (=1), b( =0.9) and
c(=0.8). More details of the data generation can
be found in Litvinchev et al. (2010a).
We divided the computational study for prob-
lem (MMAP) in two parts. In Part 1, the standard
and modified bound without the copy constraints
(=0) were calculated using the Benders technique
described in Section 4.1. These results are pre-
sented in Section 5.1. In Part 2 the Lagrangian-type
bounds, standard and modified, were calculated
by the subgradient method. We used two versions
of Algorithm 2 presented in Section 4.3. First we
compute the Lagrangian bounds without using the
greedy solution given by Algorithm 1. That is, the
steps 2 and 3.2 of Algorithm 2 were not executed
and we modified step 3.3. Since all the objectives
coefficients of problem MMAP are positive, we
set
LB
=0 (this parameter is necessary in step 3.5).
This version of the subgradient algorithm will
be named as Algorithm 2a. Then Algorithm 2 was
applied as it is stated in section 4.3.
Algorithm2a(Subgradientalgorithm
withoutusingagreedysolution)
Algorithm2withthefollowing
modifications:
Replacestep2by:Set

UB LB
0 0
0 = = ,
Replacestep3.3by:Let
LB
k
LB
k +
=
1
and

UB
k
UB
k k +
=
{ }
1
max ,
EndAlgorithm2a.
The Benders and the subgradient algorithm
were coded using the modeling language AMPL
(Fourer, Gay & Kernighan, 1993) and all the
associated optimization subproblems solved by
the system CPLEX 10.0 (ILOG, 2006). The runs
with the Benders algorithm was executed on a
machine Pentium 4, 3.2GHz, 2GB RAM and the
runs with the subgradient algorithm were executed
on a machine AMD Athlon 64X2 Dual Core, 2.8
GHz and 2048MB RAM.
For all problem instances we have calculated:
z
IP
- optimal objective of the problem (MMAP),
z
LP
- optimal objective of the LP relaxation of
the problem (MMAP),
Algorithm 2
1. Giveninitialvaluesforu v
0 0 0
, , .
2. Set
UB LB
0 0
= = , .
3. While(notstop)do
3.1. Computex y
ij
k
ij
k
, by(50),(51)andlet
k k k k
u v = ( , , ) .
3.2. UseAlgorithm1toobtainfeasiblesolutionsx
gr
andy
gr
withobjec-
tivevaluescx
gr
andcy
gr
respectively.
3.3. Let
LB
k
LB
k gr gr
cx cy
+
=
{ }
1
max , , ,
UB
k
UB
k k +
=
{ }
1
max ,
3.4. Computes
k k k k
=
{ }
, , by(52)-(54).
3.5. Update
k+1
by(55)with
LB LB
k
= .
3.6. Projectu
k
andv
k
ontothenonnegativeorthant
u u v v
k k k k + +
=
{ }
=
{ }
1 1
0 0 max , , max , .
3.7. Makestoptests.
4. end_while
5. endAlgorithm2
236
zlag - classical Lagrangian bound w
D
(comput-
ed by Benders and Algorithm 2a),
z
lag
gr
- classical Lagrangian bound w
D
(computed
by Algorithm 2),
Z
MD
- modified Lagrangian bound w
MD
L
(com-
puted by Benders and Algorithm 2a),
z
MD
MD
L

(computed by Algorithm 2a),
z
MD
gr

MD
L
(computed by Algorithm 2).

The relative quality of the bounds was mea-
sured by:
rel
z z
z z
md ip
md ip
0 100 =

%,
rel
z z
z z
md ip
lag ip
1 100 =

%,
rel
z z
z z
md ip
lp ip
2 100 =

%,
rel
z z
z z
lag ip
lp ip
3 100 =

%,
where each indicator compares two subsequent
bounds. Here rel0 indicates improvement obtained
by introducing the objective copy constraint
(cx=cy) in the modified bound, rel1 represents
improvement of the modified bound with =
0 over classical, rel2 shows the strength of the
modified bound over LP relaxation, and rel3
compares the quality of classical bound with
LP-bound. The proximity to the optimal integer
solution was represented by:
gap
z z
z
md ip
ip
0 100 =

%,
gap
z z
z
gr md
gr
ip
ip
0 100 =

%,
gap
z z
z
md ip
ip
1 100 =

%,
gap
z z
z
lag ip
ip
2 100 =

%,
gap
z z
z
gr lag
gr
ip
ip
2 100 =

%
gap
z z
z
lp ip
ip
3 100 =

%.
For the problem instance 20 100c we were
not able to find the optimal solution, CPLEX
aborted due to insufficient memory. The best
integer solution found after examining 316,560
nodes in the branch and cut tree was used then to
calculate the indicators.
Results for MMAP Using
Benders Technique
In this section we describe the computational
results obtained when computing the standard
(w
D
) and the modified bound w
MD
L
using the
Benders technique described in Section 4.1, the
objective was to compute the bounds with a con-
trolled precision. The constraint generation stops
when( ) / . UB LB LB
k k k
0 0001 or the time
limit of two hours was reached. The numerical
results obtained for the small and moderate prob-
lems in Set1 and Set2 are given in Tables 1 and
2, respectively. For each problem, the tables give
the instance dimension (m,n), its class (a,b,c), the
relative quality of the bounds (rel1,rel2,rel3), the
proximity to the original optimal value
(gap1,gap2,gap3), and the respective numbers of
iterations, iter(z
MD
) and iter(z
lag
), required for
computing the standard and modified Lagrangian
bounds by the constraint generation method.
An inspection of the Tables 1 and 2 shows that
the modified bound is better than the classical one
(rel1<rel3) for all the generated problems and is
considerably better for problems of class a. The
improvement in the bound for problems of class
237
c is insignificant. When z
LP
is close to z
IP
(small
values of gap3) for both Lagrangian bounds, few
possibilities are available for improvements. That
is the case for the instance 20100b which has
gap3=0.78 and gap1=0.72 (See Table 2). Overall,
the results suggest that the difference between the
modified and classical bounds is larger for larger
values of gap3. In other words, the worse is the
continuous relaxation, the more possibilities we
have for improving the classical Lagrangian
bound. For all the problems except for the last
two in Table 1, we have rel1<rel3; i.e., assuming
that z
lag
sufficiently improves z
LP
for these prob-
lems, the same can be concluded about the relation
between z
MD
and z
lag
. The number of iterations
depends weakly on the type of the Lagrangian
bound and tends to increase with dimension.
Results for MMAP Using a
Subgradient Algorithm
All Lagrangian-type bounds were calculated by
the subgradient method. First we applied Algo-
rithm 2a to compute the modified bound, with
and without the copy constraints (z
MD,
zM
D)
. Then
we applied Algorithm 2 to study how the greedy
solution given by Algorithm 1 can improve the
bound quality. The latter is also used to speed
up the subgradient scheme to solve the modified
Lagrangian dual problem. For both Algorithms 2
and 2a we used K=5, and if
Table 1. Relative quality of the bounds; results by benders; set1 n=50
class rel1 gap1 iter(z
MD
) rel2 gap2 iter(z
lag
) rel3 gap3
a 61.28 6.63 426 60.22 10.82 545 98.73 11.01
b 77.82 5.58 577 77.71 7.17 516 99.87 7.18
c 67.28 2.20 860 66.67 3.27 755 99.07 3.3
a 77.99 4.36 609 72.66 5.59 576 93.16 6.00
b 88.07 1.92 751 86.88 2.18 748 97.82 2.21
c 97.77 3.08 925 95.65 3.15 915 97.83 3.22
a 82.48 3.25 639 74.03 3.94 543 89.75 4.39
b 95.74 1.35 844 90.6 1.41 854 94.63 1.49
c 99.38 1.61 851 93.06 1.62 811 93.64 1.73
Table 2. Relative quality of the bounds; results by benders; set2 n=100
m class rel1 gap1 iter(z
MD
) rel2 gap2 iter(z
lag
) rel3 gap3
a a 66.66 7.40 1195 66.66 11.10 1938 100 11.1
5 b b 77.81 5.05 1548 77.81 6.49 1258 100 6.49
c c 76.85 3.42 2717 76.85 4.45 2452 100 4.45
a a 79.62 3.40 1941 77.62 4.27 2253 97.49 4.38
10 b b 95.00 1.14 2508 93.44 1.20 2843 98.36 1.22
c c 98.93 1.96 2822 98.50 1.98 3548 99.5 1.99
a a 86.95 1.40 2474 76.50 1.61 2414 87.98 1.83
20 b b 96.00 0.72 3988 92.31 0.75 3898 96.15 0.78
c c 93.22 0.55 3117 90.16 0.59 3931 93.44 0.61
238
(( ) / ) .
k k
UB
k

+ + 1 1
0 005
for 5 consecutive iterations with fixed
k
, this
parameter was updated to
k+1=

k
/2. The stopping
criteria were specified as follows: a) at most 250
iterations were permitted; b) the runs stop if
k+1
0.005; or c) (( ) / ) .
UB
k
LB
k
LB
k + + +

1 1 1
0 0001.
The numerical results obtained for the problems
in Set1 and Set 2 using Algorithm 2a are given in
Tables 3 and 4, respectively. For each problem,
the tables give the instance dimension (m,n), its
class (a,b,c), the relative quality of the bounds
(rel0,rel1,rel2,rel3), the proximity to the original
optimal value (gap0,gap1,gap2,gap3).
The results in Table 3 (Set 1) and Table 4 (Set
2) were obtained within a time limit of 240 s and
360 s respectively. The values of the indicators
were calculated based on the best values of the
bounds obtained up to the final iteration. As can
be seen from Tables 3 and 4, for all problem in-
stances we have rel0 < 100%, that is, the use of
the objective copy constraint (cx = cy) has a
beneficial effect in the computation of the modi-
fied bound.
Moreover, gap0 is significantly smaller than
min{gap1,gap2,gap3} for all instances (except
for 20 100c), which means that z
MD
is superior
to all other bounds within the computational time
limit. In most cases gap0 < 0.5% and for all
problems of the class a the exact bound was ob-
tained, gap0 = 0. Note that for problem instances
with gap0 = 0 the proposed bound zM
D c
annot
Table 3. Relative quality of the bounds; results by Algorithm 2a (the subgradient method); set1 n=50
m class rel0 gap0 rel1 gap1 rel2 gap2 rel3 gap3
a a 0 0 59.93 6.65 60.37 11.09 100.73 11.01
5 b b 0 0 78.39 5.88 81.79 7.5 104.33 7.18
c c 15.12 0.36 66.47 2.4 72.78 3.61 109.5 3.3
a a 0 0 76.06 4.38 72.96 5.76 95.92 6
8 b b 5.89 0.12 78.72 1.96 88.61 2.49 112.56 2.21
c c 9.67 0.32 98.02 3.35 104.06 3.42 106.16 3.22
a a 0 0 80.88 3.3 75.19 4.08 92.96 4.39
10 b b 20.09 0.28 85.89 1.39 93.56 1.62 108.93 1.49
c 24.83 0.51 109.24 2.04 117.48 1.86 107.55 1.73
Table 4. Relative quality of the bounds; results by Algortihm 2a (the subgradient method); set2 n=100
m class rel0 gap0 rel1 gap1 rel2 gap2 rel3 gap3
a a 0 0 66.68 7.44 67.04 11.16 100.54 11.1
5 b b 0 0 73.43 5.02 77.34 6.84 105.32 6.49
c c 0.93 0.03 70.79 3.37 75.8 4.76 107.07 4.45
a a 0 0 77.61 3.42 78.05 4.41 100.57 4.38
10 b b 5.75 0.07 76.97 1.18 96.52 1.53 125.4 1.22
c c 7.26 0.17 107.98 2.37 118.92 2.19 110.14 1.99
a a 0 0 91.22 1.5 82.13 1.65 90.03 1.83
20 b b 8.4 0.08 107.71 0.98 125.48 0.91 116.5 0.78
c 77.41 0.79 139.73 1.02 165.02 0.73 118.1 0.62
239
be outperformed by any other bound. For most
instances of class c we have gap3=min{gap1,gap2},
so the LP bound is better than zlo
g,
zmd
o
btained
within the time limit. In Section 5.1 the results of
t he Be nde r s me t hod s howe d t ha t
gap1<gap2<gap3for all classes. However, for the
class c the difference in these indicators was very
modest: for small values of gap3 there is a little
scope for bound improvements. As a byproduct
of the bound computation ( w
MD
L
) an integer
Lagrangian solution y was obtained having a much
higher degree of primal feasibility and subopti-
mality than the standard Lagrangian solution. (see
Litvinchev, Rangel & Saucedo, 2010) for more
details on the constraint violation and suboptimal-
ity of the Lagrangian solutions (x,y)).
To study the effect of introducing a heuristic
solution on the calculation and quality of the dual
bounds we applied Algorithm 2 to compute the
standard ( z
lag
gr
) and the modified ( z
MD
gr
) Lagrang-
ian bounds. The numerical results obtained for
the problems in Set1 and Set 2 are given in Tables
5 and 6, respectively. For each problem, the tables
give the instance dimension (m,n), its class (a,b,c),
the proximity to the original optimal value
(gap0
gr
,gap0,gap2
gr
,gap2) and the number of it-
erations (iter).
For all instances with gap0
gr
=gap0=0.00 the
stopping criterion (c) was fulfilled. For all other
instances the runs were terminated by the stopping
criterion (b). As can be seen from Tables 5 and 6,
incorporating a greedy solution in the subgradient
algorithm slightly improves the quality (of the
approximate values) of the bounds obtained by
the subgradient method, but the effect is rather
modest. For all problem instances the modified
bound (gap0
gr
) is significantly tighter than the
classical Lagrangian bounds (gap2
g
gap2). More-
over, the number of iterations of the subgradient
method reduces significantly by using a greedy
solution. This takes place for all problem in-
stances. Note that the computational cost for one
iteration of the subgradient technique (solving
integer Lagrangian problems) is much higher than
the one for obtaining a greedy solution (simply
reordering data). Thus combining the subgradient
method with a cheap feasibility recovering
technique is favorable for this class of problem.
The greedy solutions derived by the modified
bound are better than the ones resulting from the
standard Lagrangian bound (see Litvinchev et al.,
2010b for the details).
Table 5. Relative quality of the bounds; results by the subgradient method (Algorithm 2 and 2a); set1 n=50
m class gap0
gr
gap0 gap2
gr
gap2
iter(
z
MD
gr
) iter(
z
MD
) iter(
z
lag
gr
)
iter(Z
lag
)
a 0 0 10.91 11.09 5 165 55 100
5 b 0 0 7.22 7.5 5 80 74 125
c 0.12 0.36 3.29 3.61 55 110 80 85
a 0 0 5.62 5.76 5 80 55 110
8 b 0.09 0.12 2.19 2.49 75 125 95 200
c 0.11 0.32 3.17 3.42 55 75 80 145
a 0 0 3.96 4.08 5 130 55 65
10 b 0.07 0.28 1.42 1.62 65 70 74 115
c 0.2 0.51 1.63 1.86 55 170 67 115
240
Discussion
The results obtained by the Benders technique
and presented in Tables 1, 2 demonstrate that the
modified bounds are tighter than the classical
Lagrangian bounds (rel1<100%). For problem
instances where the continuous relaxation is weak
(gap3 is large enough) the modified bounds are
significantly tighter. It is important to note that
the values of the bounds obtained by the Bend-
ers technique were computed with a high accu-
racy, the guaranteed relative error of the bound
computation was less than 0.0001. That is, the
Benders technique provides real values of the
bounds and in this sense the indicators presented
in Tables 1, 2 are independent on the method used
to get the bounds. So we may conclude that the
superiority of the modified bounds is an inherent
property of the bounds and not the result of the
loose computations for the classical bounds. The
precise computations by the Benders technique
provide the bounds with guaranteed accuracy but
result in a high computational cost. Typically the
number of the main iterations to solve the master
problem (MP) increases significantly with the size
of the problem instance: for most instances with
m=10,20 presented in the Table 2 it was necessary
nearly 2 hours of the CPU-time to get the desired
accuracy in the bounds computations. Thus, the
use of the Benders technique was important to
prove numerically the superiority of the modi-
fied bounds, but this method in its current form
cannot be recommended for practical calculations
of the bounds.
To speed up the bounds calculation the sub-
gradient technique was used to solve the classical
and modified dual problems. In contrast to the
Benders method, the subgradient technique does
not provide the value of the bounds with the
prescribed accuracy. That is, terminating iterations
of the subgradient method using the stopping
criteria a)-c), we can expect only approximate
values of the bounds. As we can see from the
Tables 3 and 4, for some instances the value of
the indicator rel1 is higher than 100%. This does
not mean that the classical bound is tighter, this
only means that at the moment that the subgradi-
ent algorithm stopped, a better approximation was
obtained for the classical bound (see the exact
values of rel1 in Tables 1, 2). Tables 3, 4 also
Table 6. Relative quality of the bounds; results by the subgradient method (Algorithm 2 and 2a); set2
n=100
m class gap0
gr
gap0 gap2
gr
gap2
iter( z
MD
gr
) iter( z
MD
) iter( z
lag
gr
)
iter(Z
lag
)
a 0 0 11.28 11.16 5 190 75 95
5 b 0 0 6.53 6.84 5 100 49 110
c 0.02 0.03 4.48 4.76 70 60 75 110
a 0 0 4.31 4.41 5 175 60 105
10 b 0.04 0.07 1.21 1.53 65 55 70 175
c 0.11 0.17 1.99 2.19 70 85 79 65
a 0 0 1.61 1.65 5 85 49 90
20 b 0.07 0.08 0.76 0.91 70 105 65 135
c 0.11 0.79 0.59 0.73 80 180 49 130
241
present the proximity indicator gap0 for the bound
w
MD
L
, obtained by using the objective copy con-

straint (cx=cy), and the proximity indicator gap2
for the classical Lagrangian bound. As can be
seen from the Tables, gap0 is significantly
smaller than gap2. That is the approximation of
the boundw
MD
L
obtained by the subgradient tech-

nique is much closer to the optimal objective
comparing to the approximation of the classical
bound. Moreover, gap0 is almost 10 times
smaller than gap1 obtained by the Benders tech-
nique and presented in Tables 1, 2. Thus, we may
conclude that introducing the objective copy
constraint has a beneficial effect in the computa-
tion of the modified bound (see the end of the
Section 2.1 for theoretical discussion of this sub-
ject).
Computing the classical Lagrangian bound
results in a solution of Lagrangian problem which
is frequently used as a starting or reference point
in heuristic techniques. In the case of the modified
bound we have two Lagrangian solutions, x and
y, both different from the classical Lagrangian
solution. To recover a feasible solution from an
infeasible Lagrangian one we used a simple greedy
heuristic (Algorithm 1). The results presented in
the Tables 5 and 6 show that: a) combining the
subgradient scheme with the best feasible greedy
solution (Algorithm 2) improves significantly the
convergence of the subgradient method and b) this
improvement was obtained without dropping the
quality of the bounds and feasible solutions. We
note that to get the modified bound the Algorithm
2 either stops after 5-8 iterations or achieves the
main improvement (in terms of the quality of the
modified bound and the corresponding feasible
solution) within the same range of iterations.
From practical point of view, it is sufficient to
execute at most 10 iterations of the Algorithm
2, extra iterations will only slightly change the
overall picture.
Thus we may conclude that the proposed ap-
proach (Algorithm 2) can be useful for practical
solution of the MMAP. It takes a few iterations to
get an approximate feasible solution within 0.1%
of the relative suboptimality. The main compu-
tational cost is associated to solution of the n+m
independent one-dimensional knapsack problems
((50), (51)) in each iteration. Recovering feasibility
of the modified Lagrangian solution by the greedy
algorithm is significantly cheaper. The core of the
Algorithm 2 is the modified Lagrangian bound.
It provides tight approximation of the optimal
objective and results in an integer Lagrangian
solution having a high degree of primal feasibil-
ity and suboptimality. The use of this Lagrangian
solution in a simple greedy algorithm results in a
high quality approximate feasible solution to the
original problem.
A procedure for tightening Lagrangian bounds
was proposed. The main idea of the approach
is to estimate the complementarity term in the
Lagrangian function. For convex problems under
certain regularity conditions the complementarity
term turns to zero for the optimal primal-dual pair.
However for integer problems the complemen-
tarity term can be different from zero even for
optimal Lagrange multipliers since we typically
have positive optimal multipliers for non-active
constraints. We may expect that this approach can
be especially useful for discrete problems having a
large number of nonactive constraints for feasible
solutions. Here we used the approach for assign-
ment problems with knapsack-type inequality
constraints, which frequently are non-active for
feasible solutions. An interesting area for future
research is an application of the proposed ap-
proach to packing and cutting problems where,
by the physical nature of the problem, available
resources can not be utilized completely, e.g.,
packing circles into rectangular or circular areas.
Another direction for future research is the use
of a posteriori localization, possibly adjusted in
242
the course of solving the dual problem. It can be
constructed in a surrogate way, similar to local-
ization W
s
mentioned in Section 2.1 or using valid
inequalities as in Litvinchev and Rangel (1999)
(see also Litvinchev and Tsurkov (2003)).
The proposed method results in two subprob-
lems - one is similar to the classical Lagrangian
problem and the other is used to estimate the
complementarity term. Thus we have two La-
grangian solutions both different from the clas-
sical one. In our computational experiment with
many-to-many assignment problems we notice
that although solutions to modified Lagrangian
dual are unfeasible, constraints are typically much
less violated comparing with the classical Lagrang-
ian solution. This is the reason why our simple
greedy algorithm aimed to restore feasibility of
the modified Lagrangian solution provides very
high quality approximated solutions to the original
problem. It is interesting to see if the same behavior
of the modified Lagrangian solutions takes place
for other classes of combinatorial problems.
It was demonstrated by the computational
experiment, that incorporating the greedy feasible
solution in the subgradient scheme improves
the convergence of the subgradient method. An
interesting area for a future research is to use
the modified bounds in combination with other
greedy approaches, e.g., using different choices
for rounding up/down components, as well as with
more sophisticated heuristic techniques (see, e.g.,
Jeet and Kutanoglu (2007)).
Another direction for future research is to im-
prove the behavior of the subgradient technique
by a more suitable choice of the initial values for
the Lagrangian multipliers, for example, using
optimal duals for the copy constraints Dx Dy in
the LP-relaxation of the problem (7). An alterna-
tive to the subgradient method used in this paper
would be using more stable approaches such as
center-based or bundle algorithms (e.g. Bahience
et al. (2002)).
FINAL REMARKS AND CONCLUSION
In this chapter we presented a Lagrangian heuristic
and applied it to a class of the generalized assign-
ment problems. Our contribution is threefold. First,
we proposed the modified Lagrangian bound based
on a more tight estimation of the complementarity
term in the Lagrangian function. For the original
problem having two interesting sets of constraints
we proved that the new bound is at lest as good as
the best of the corresponding classical Lagrang-
ian bounds. Second, we studied numerically the
properties of the two Lagrangian-like solutions
arising in the modified bound computations and
showed that for the modified Lagrangian solution
the original constraints are much less violated than
for the classical Lagrangian solution. Third, we
merged a greedy technique used to recover the
feasibility of the modified Lagrangian solution
with the subgradient algorithm thus obtaining
an approximate iterative scheme producing high
quality feasible solutions.
The usefulness of the approach was demon-
strated for a class of the generalized assignment
problems, though the basic scheme can be applied
for a much broader class of problems. In this
chapter we focused on the general algorithmic
aspects of the proposed Lagrangian based heu-
ristic. We sincerely hope that this new approach
will be useful for the researches and practitioners
working with specific (and not only assignment)
problems in power optimization.
ACKNOWLEDGMENT
This work was partially supported by RFBR,
Russia (09-01-00592), Mexican foundations
CONACyT (61343), PAICYT (CE008-09), and
PROMEP (103.5/09/3905), and Brazilian agencies
CNPq and FAPESP.
243
REFERENCES
Balachandran, V. (1976). An integer generalized
transportation model for optimal job assignment in
computer networks. Operations Research, 24(4),
742759. doi:10.1287/opre.24.4.742
Bokhari, S. H. (1987). Assignment problems in
parallel and distributed computing. Norwell, MA:
Kluwer Academic Publishers.
Bookbinder, J. H., & Reece, K. E. (1988). Ve-
hicle routing considerations in distribution sys-
tem design. European Journal of Operational
Research, 37(2), 204213. doi:10.1016/0377-
2217(88)90330-X
Campbell, G. M., & Diaby, M. (2002). Develop-
ment and evaluation of an assignment heuristic
for allocating cross-trained workers. European
Journal of Operational Research, 138(1), 920.
doi:10.1016/S0377-2217(01)00107-2
Cattrysse, D. G., & Van Wassenhove, L. N. (1992).
A survey of algorithms for the generalized assign-
ment problem. European Journal of Operational
Research, 60(3), 260272. doi:10.1016/0377-
2217(92)90077-M
De Maio, A., & Roveda, C. (1971). An all zero-one
algorithm for a certain class of transportation prob-
lems. Operations Research, 19(6), 14061418.
doi:10.1287/opre.19.6.1406
Fisher, M. L., & Jaikumar, R. (1981). A generalized
assignment heuristic for vehicle routing. Networks,
11, 109124. doi:10.1002/net.3230110205
Fourer, R., Gay, M. D., & Kernighan, B. W. (1993).
AMPL A modeling language for mathematical
programming. Denver, CO: Scientific Press.
Frangioni, A. (2005). About Lagrangian methods
in integer optimization. Annals of Operations
Research, 139, 163169. doi:10.1007/s10479-
005-3447-9
Freville, A., & Hanafi, S. (2005). The multidi-
mensional 0-1 knapsack problem Bounds and
computational aspects. Annals of Operations
Research, 139, 195227. doi:10.1007/s10479-
005-3448-8
Geoffrion, A. M., & Graves, G. W. (1974). Mul-
ticommodity distribution system design by Bend-
ers decomposition. Management Science, 20(5),
822844. doi:10.1287/mnsc.20.5.822
Guignard, M., & Kim, S. (1987). Lagrangian
decomposition: A model yielding stronger La-
grangian bounds. Mathematical Programming,
39, 215228. doi:10.1007/BF02592954
ILOG. (2006). ILOG CPLEX 10.0 (users manual),
Mathematical programming optimizers, version
10.0.
Klastorin, T. D. (1979). On the maximal covering
location problem and the generalized assignment
problem. Management Science, 25(1), 107113.
doi:10.1287/mnsc.25.1.107
Kuhn, H. (1995). A heuristic algorithm for the load-
ing problem in flexible manufacturing systems.
International Journal of Flexible Manufacturing
Systems, 7, 229254. doi:10.1007/BF01325036
Lasdon, L. S. (2002). Optimization theory for
large systems (2nd ed.). Dover.
LeBlanc, L. J., Shtub, A., & Anandalingam, G.
(1999). Formulating and solving production plan-
ning problems. European Journal of Operational
Research, 112(1), 5480. doi:10.1016/S0377-
2217(97)00394-9
Lemarechal, C. (2007). The omnipresence of
Lagrange. Annals of Operations Research, 153,
927. doi:10.1007/s10479-007-0169-1
Litvinchev, I. (2007). Refinement of Lagrangian
bounds in optimization problems. Computational
Mathematics and Mathematical Physics, 47(7),
11011108. doi:10.1134/S0965542507070032
244
Litvinchev, I., Mata, M., Rangel, S., & Saucedo,
J. (2010). Lagrangian heuristic for a class of
the generalized assignment problems. Comput-
ers & Mathematics with Applications (Oxford,
England), 60(4), 11151123. doi:10.1016/j.
camwa.2010.03.070
Litvinchev, I., & Rangel, S. (1999). Localization
of optimal solution and a posteriori bounds for ag-
gregation. Computers & Operations Research, 26,
967988. doi:10.1016/S0305-0548(99)00027-1
Litvinchev, I., & Rangel, S. (2008). Comparing
Lagrangian bounds for a class of generalized
assignment problems. Computational Mathemat-
ics and Mathematical Physics, 48(5), 739746.
doi:10.1134/S0965542508050047
Litvinchev, I., Rangel, S., & Saucedo, J. (2010). A
Lagrangian bound for many-to-many assignment
problems. Journal of Combinatorial Optimization,
19(3), 241257. doi:10.1007/s10878-008-9196-3
Litvinchev, I., & Tsurkov, V. (2003). Aggregation
in large scale optimization. Boston, MA: Kluwer.
Maculan, N., & Reinoso, H. (1992). Lagrangean
decomposition in integer linear programming: A
new scheme. INFOR, 30(1), 15.
Martello, S., & Toth, P. (1990). Knapsack prob-
lems: Algorithms and computer implementations.
New York, NY: Wiley.
Martin, R. K. (1999). Large scale linear and in-
teger programming: A unified approach. Boston,
MA: Kluwer.
Morales, D. R., & Romeijn, H. E. (2004). The
generalized assignment problem and extensions. In
Du, D.-Z., & Pardalos, P. M. (Eds.), Handbook of
combinatorial optimization (Vol. 5, pp. 259311).
Kluwer Academic Publishers.
Pentico, D. W. (2007). Assignment problems:
A golden anniversary survey. European Jour-
nal of Operational Research, 176, 774793.
doi:10.1016/j.ejor.2005.09.014
Pirkul, H. (1986). An integer programming model
for the allocation of databases in a distributed com-
puter system. European Journal of Operational
Research, 26(3), 401411. doi:10.1016/0377-
2217(86)90142-6
Ross, G. T., & Soland, R. M. (1975). A branch and
bound algorithm for the generalized assignment
problem. Mathematical Programming, 8, 91103.
doi:10.1007/BF01580430
Ross, G. T., & Soland, R. M. (1977). Modeling
facility location problems as generalized assign-
ment problems. Management Science, 24(3),
345357. doi:10.1287/mnsc.24.3.345
Srinivasan, V., & Thompson, G. L. (1972). An
algorithm for assigning uses to sources in a special
class of transportation problems. Operations Re-
search, 21, 284295. doi:10.1287/opre.21.1.284
Wolsey, L. A. (1998). Integer programming. New
York, NY: John Wiley & Sons.
Zimokha, V. A., & Rubinshtein, M. I. (1988). R
& D planning and the generalized assignment
problem. Automation and Remote Control, 49,
484492.
ADDITIONAL READING
Bahiense, L., Maculan, N., & Sagastizabal, C.
(2002). The volume algorithm revisited: relation
with bundle methods. Mathematical Program-
ming, 94, 4169. doi:10.1007/s10107-002-0357-3
Barahona, F., & Anbil, R. (2000). The volume
algorithm: producing primal solutions with a
subgradient method. Mathematical Programming,
87, 385399. doi:10.1007/s101070050002
Beasley, J. E. (1993). Lagrangean relaxation. In
Reeves, C. R. (Ed.), Modern heuristic techniques
for combinatorial problems (pp. 243303). Black-
well Scientific Publications.
245
Boschetti, M., & Maniezzo, V. (2009). Benders
decomposition, Lagrangian relaxation and me-
taheuristic design. Journal of Heuristics, 15(3),
283312. doi:10.1007/s10732-007-9064-9
Burkard, R., DellAmico, M., & Martello, S.
(2009). Assignment problems. Philadelphia:
SIAM. doi:10.1137/1.9780898717754
Everett, H. III. (1963). Generalized Lagrange mul-
tiplier method for solving problems of optimum
allocation of resources. Operations Research, 11,
399417. doi:10.1287/opre.11.3.399
Fisher, M. L. (1985). An application oriented guide
to Lagrangian relaxation. Interfaces, 15, 1021.
doi:10.1287/inte.15.2.10
Geoffrion, A. M. (1974). Lagrangian relaxation
and its uses in integer programming. Mathematical
Programming Study, 2, 82114.
Guignard, M. (2003). Lagrangian relaxation.
Top (Madrid), 11(2), 151228. doi:10.1007/
BF02579036
Haidar, M., Al-Rizzo, H., Chan, Y., & Akl, R.
(2009). User-based channel assignment algo-
rithm in a load-balanced IEEE 802.11 WLAN.
International Journal of Interdisciplinary Tele-
communications and Networking, 1(2), 6681.
doi:10.4018/jitn.2009040105
Held, M., & Karp, R. (1970). The traveling sales-
man problem and minimum spanning trees. Op-
erations Research, 18, 11381162. doi:10.1287/
opre.18.6.1138
Jeet, V., & Kutanoglu, E. (2007). Lagrangian re-
laxation guided problem space search heuristics
for generalized assignment problems. European
Journal of Operational Research, 182, 1039
1056. doi:10.1016/j.ejor.2006.09.060
Jornsten, K., & Nasberg, M. (1986). A new La-
grangian relaxation approach to the generalized
assignment problem. European Journal of Opera-
tional Research, 27, 313323. doi:10.1016/0377-
2217(86)90328-0
Kleeman, M., & Lamont, G. (2008). Evolution-
ary multi-objective optimization for assignment
problems. In Bui, L. T., & Alam, S. (Eds.), Multi-
Objective Optimization in Computational Intel-
ligence: Theory and Practice (pp. 364387). IGI
Global. doi:10.4018/978-1-59904-498-9.ch013
Lemarechal, C. (2001). Lagrangian relaxation.
In Junger, M., & Naddef, D. (Eds.), Computa-
tional combinatorial optimization (pp. 115160).
Springer Verlag. doi:10.1007/3-540-45586-8_4
Li, D., & Sun, X. (2006). Nonlinear integer pro-
gramming. Springer.
Lorena, L. A. N., & Narciso, M. J. (1996). Relax-
ation heuristics for generalized assignment prob-
lem. European Journal of Operational Research,
91, 600610. doi:10.1016/0377-2217(95)00041-0
Narciso, M. J., & Lorena, L. A. N. (1999). La-
grangean/surrogate relaxation for generalized
assignment problems. European Journal of Op-
erational Research, 114(1), 165177. doi:10.1016/
S0377-2217(98)00038-1
Pentico, D. W. (2007). Assignment problems:
A golden anniversary survey. European Jour-
nal of Operational Research, 176, 774793.
doi:10.1016/j.ejor.2005.09.014
Shapiro, J. F. (1974). A survey of Lagrangean
techniques for discrete optimization. Annals of
Discrete Mathematics, 5, 113138. doi:10.1016/
S0167-5060(08)70346-7
246
Generalized Assignment Problem: Involves
profit maximizing assignment of tasks to agents
recognizing capacity limits for agents and/or tasks.
Many-to-Many Assignment Problem: In-
volves profit maximizing assignment of tasks to
agents recognizing capacity limits for both agents
and tasks.
Lagrangian Relaxation: Relaxation of a
number of constraints together with aggregating a
penalty term to the objective function to discour-
age their violation.
Lagrangian Multipliers: Coefficients of the
linear combination of the constraints used to form
a penalty term in the Lagrangian function.
Lagrangian Relaxation Bounds: Bounds to
the original optimal objective value provided by
the Lagrangian problem.
Lagrangian Dual Problem: Aims to find
Lagrangian multipliers corresponding to the best
Lagrangian bound.
Lagrangian Heuristic: uses a solution to
the Lagrangian (Lagrangian dual) problem as a
starting or reference point in a heuristic approach.
247
APPENDIX
The literature on Lagrangian relaxation is quite extensive. Here we refer the reader to a few pioneer
and/or review papers providing a clear exposition of the subject: Everett (1963), Held & Karp (1970),
Geoffrion (1974), Shapiro (1974), Fisher (1985), Beasley (1993), Lemarechal (2001), Guignard (2003).
To our knowledge, the latter paper presents the most comprehensive and complete revision of the basic
ideas used in Lagrangian relaxation and its applications.
Numerical techniques of non-differentiable optimization used to solve Lagrangian dual problem can
be found in the references cited above. More recent approaches are presented, e.g., in Barahona & Anbil
(2000), Bahiense, Maculan & Sagastizabal (2002). Various relaxations in nonlinear integer programming
are considered in Li & Sun (2006).
The ideas to combine Lagrangian and Benders decomposition techniques with heuristic approaches
are presented in Boschetti & Maniezzo (2009).
A very complete source on modeling and solution techniques for assignment problems are provided
in the recent book of Burkard, DellAmico & Martello (2009) and in the survey paper Pentico (2007).
More specific Lagrangian-based solution approaches can be found in Jornsten & Nasberg (1986), Lorena
& Narciso (1996), Narciso & Lorena (1999), Jeet & Kutanoglu (2007) and the references therein. Multi-
objective assignment models and solution techniques are presented in Kleeman & Lamont (2008), while
models and algorithms for channel assignment are considered in Haidar et al. (2009) and the references
therein. See also the additional reading section.
248
Chapter 8
Joo Zambujal-Oliveira
Instituto Superior Tcnico & Technical University of Lisbon, Portugal
Power Systems Investments:
A Real Options Analysis
ABSTRACT
Energy projects with extended life cycles and initial investments can be non-proftable under discount
cash fow methods. Therefore, real options analysis has become relevant as a pricing technique for
these types of projects, with private risks and high investment levels. Following this question, this study
analyses different real options approaches to select the most acceptable for investing decisions in the
energy sector. Combined cycle natural gas-fred plants constitute relevant generation assets that build-
ing decisions can mostly be studied by real options tools. Because traditional pricing approaches fail
to consider the worth of fexibility, conditions for creating a signifcantly large options-based value
can be found. Being unable to capture the value associated with the decision makers ability to react
dynamically to changing market conditions, these assets constitute a fne example of fexibility, which
contributes to increasing its intrinsic value.
The study employs a real options approach that doesnt need to capture all the uncertainty and proposes
a process that directly determines the uncertainty associated with the frst period. The results support
that its use can be considered fair. However, it shows that long periods of operation and poor adhesion
to the geometric Brownian motion by the project returns might call into question its use in the energy
market. The values for option pricing have remained inside acceptable ranges, but some shortfalls could
be found. First, the study employs Monte Carlo simulations, which can be viewed as forward-looking
processes, and option pricing problems need backward recursive solutions. Second, the study shows that
its simplicity produces results as accurate as those gathered from approaches with added complexity
and computational needs.
DOI: 10.4018/978-1-61350-138-2.ch008
249
Power Systems Investments
INTRODUCTION
Risks and uncertainties such as electricity, fuel
prices, and construction costs characterize the
energy market (Thompson et al., 2004). In that
sense, case studies for the energy sector are as
relevant as they are for pharmaceutical research
or hi-tech start-ups. Therefore, the energy market
needs a flexible way to approach plant investments
and decision-making processes. Investment timing
is relevant for energy sector because of the need
to trade off the supply and demand for electricity.
Adjustments in investment timing place pressure
on power prices.
This study will evaluate a models ability to
define the optimal investment rule by considering
uncertainties in the costs and revenues, assuming
decision flexibility. Combined cycle natural gas-
fired plants are relevant generation assets that
building decisions can mostly be studied by Real
Options Analysis (ROA), (Blyth, 2008). Because
traditional pricing approaches fail to take into
account the worth of flexibility, the conditions to
create a significant large options-based value can
be found in these plants. Being unable to capture
the value associated with the plant managers
ability to react dynamically to changing market
conditions, these types of uncertainty origins of-
fer a large options-based value. Because peaking
power plants (as gas-fired) can react quickly to
market prices, they constitute fine examples of
asset flexibility, which contributes to increasing
their intrinsic asset value.
Although decision tree analysis improves
discount cash flow (DCF), the manager still
needs to assign different probabilities to potential
results and find out a suitable discount rate. By
contrast, ROA simplifies these assumptions and
makes the process of choosing discount rates no
longer arbitrary. Possible results are assessed and
underlying risk profile is recognised under an
options-based framework.
Given the nature of the energy market, the
problem in question arises from the need of an
energy production company to know if its power
generating plant project has financial viability.
The company also wants to know if the present
moment is the best time for beginning its invest-
ment, given the potential opportunities. Within
the real options paradigm, this issue sets up as an
evaluation of the deferment or delay option for a
power generating plant investment. The real op-
tions approach applies derivative pricing theory
to the analysis of options opportunities in real
assets (Dixit and Pindyck 1994). It can be used
to calculate plant values and optimal operating
policies while considering plant characteristics. A
power plant produces electricity at variable costs
and if the (uncertain) electricity price exceeds its
costs, it creates a positive contribution margin.
Power plants can be seen as a strip of European
call option or as path-dependent American Op-
tion. Since the expansion of the power plants
variable costs is also uncertain, power plants can
also be appraised as swap options (fuel against
electricity). In another perspective, accounting
for different plant construction lead times in the
face of demand uncertainty can originate differ-
ent optimal capacity planning strategies (Gardner
and Rogers 1999).
Increased uncertainty enlarges the option
value of a project by capturing the benefits from
managerial/operational flexibility. However,
option pricing models still need to bypass some
practical shortfalls. Because the source of uncer-
tainty should be traded, energy markets should
be observed to calculate the volatility in electric-
ity prices and assess their impact on revenues.
Besides, as market data is scarce, it is usual to
estimate expected input parameters with Monte
Carlo simulations (Rode et al., 2001).
Considering all arguments, this chapter intends
to explore the applicability of the approach taken
by Brando et al. (2005) for investments made
in the electricity market. This assessment can be
made through a feasibility study of an investment
in a combined cycle natural gas-fired plant using
an ROA. This chapter begins by introducing the
250
approach framing of ROA in the energy market.
After a suitable framework definition, the study
describes the underlying methodology behind the
research question. It aims to assess the relevance
of calculating the projects volatility using two
different approaches. Then, it describes technol-
ogy used in the investment project. The choice
of this investment project results from supported
reasons of environmental efficiency improvement.
At this point, the chapter provides in detail the
uncertainty modelling.
PROJECT APPRAISAL AND
VOLATILITY FUNDAMENTALS
The dynamics of investment in the power genera-
tion industry is characterized by the existence of
different decisions about assets used in the energy
generation. These decisions can be undertaken
based on the formalization of the investment events
in accordance with the real options approach. The
following paragraphs illustrate the diverse types
of real options and its connection with power
generation industry.
Traditional capital budgeting analysis as-
sumes that projects perform every year of their
lifetime. Power plants should have an option to
cease the projects during their lives. This can be
seen as an abandonment real option. This option
gives the owner the right to sell the cash flows
over the remainder of the projects life. When the
present value of the remaining cash flows (sal-
vage value) falls below the liquidation value, the
power facility should be sold. During the projects
evaluation phase, it is necessary to recognize the
value represented by the opportunity, to recoup
part of the investment, in case the project should
be abandoned for scrap value.
When electricity price falls below the cost
of production, it may be most advantageous to
temporarily shutdown until the electricity price
recovers. Deng et al. (1998) valuate thermal
power plants that operate only when the spark
spread is positive. This temporarily shutdown
option highlights the importance of developing
a dynamic management attitude.
A peaking unit power plant can also be used
to illustrate a type of growth real option (intensity
and scale). This kind of option formerly quantifies
the flexibility to expand or contract the scale of a
project. Such units have high variable costs and
can produce additional power if demand and prices
are high. The plant may have option to change the
output rate per unit of time or to change the total
length of production run time. This option lets
the plant to increase the capacity of an existing
generation line.
If the market conditions are weaker than origi-
nally expected, the plant should have a contract
option to operate below capacity or even lessen
the scale, saving part of the planned investment
outlays. This flexibility allows mitigating losses.
Using Monte Carlo simulation, Tseng and Barz
(2002) assess a power plant capable of switching
between two modes.
When a power generation facility can restart
operations for a previous shut down project in a
dormancy state, there are present switching options
that permit the facility to expand or contract. The
option to switch refers to the feasibility of choosing
among alternative operating modes. Basically, this
happens when demand is changing, the manage-
ment can change the output mix or inputs.
The deferment option is valuable when market
demand is uncertain and interest rate is volatile.
The option to choose when to start a project is a
deferment option. It is an extended concept of
opportunity cost, which emphasizes that the deci-
sion to invest immediately entails renouncing the
undertaking a similar investment and postponing
the implementation of this particular investment.
Fleten and Nasakkala (2010) analyzed a license to
build a power plant, evaluating the optimal timing
in function of stochastic carbon emission costs.
The option to expand allows increase in the
scale of a project by investing more follow-on
capital. When the market conditions turn out to
251
be more favourable than expected, the project can
be accelerated or expand the scale of production.
Some energy projects can be engineered in a way
where the output is contracted in future, forgoing
future expenditures as with projects modularized.
An electric utility may have the option to
switch between various fuel sources to produce
electricity (choice of building a coal-fired plant or
a plant that burns either coal or gas). The option
to use different inputs to produce the same output
is known as an input mix option or process flex-
ibility (Trigeorgis, 1993). Brekke and Schieldrop
(2000) analyzed the fuel switch option and the
optimal timing of investing in a plant that can
burn natural gas or oil.
Traditional decision analysis tools to solve real
option valuation problems have derived from Co-
peland and Antikarov (2001) and are illustrated in
Copeland and Tufano (2004). In Black and Scholes
(1973), all risks are market risks and have market
equivalents. In Copeland and Antikarov (2001),
the opposite is shown; the authors consider that
there is no need for a market equivalent because it
might even not exist in a portfolio equivalent. It is
assumed that the market equivalent is the project
value itself. Additionally, this methodology also
does not make any distinction between market
uncertainties and market equivalents, and private
uncertainties and specific risks (Borison, 2005).
As a consequence of these aspects, the methodol-
ogy of Copeland and Antikarov (2001) is hardly
acceptable from the conceptual perspective.
Brando et al. (2005) used basic principles
put forward by Nau and McCardle (1991), later
improved by Smith and Nau (1995), and employ it
to estimate the volatility of returns overtime. The
following expression summarizes the revenues
and costs in the project function:
V R C
CF R p q g C z e o c c
t t
t t t v f
t
i
T
, ,
, , , , , , ,
( ) =
( ) ( ) ( )
+ ( ) =
1 0

(1)
and
CF R C C R C C
t t t t t d T
, ( ) = + [ ] ( ) + +
0
1
(2)
where V(R
t
,C
t
,) represents the net present value
(or project value) and CFt
(
Rt
,
Ct
)
stands for cash-
flows of each period. C0

is the initial investment
and Rt(,p,q,g)corresponds to the revenue function
with p for electricity price, q for net energy produc-
tion and gfor guaranteed production. Ct
(
z,e,p,cv
,
cf
)

represents the cost function with z for natural
gas price, e for efficiency, o for price for carbon
dioxide emissions, cv

for variable operation and
maintenance costs (O&M costs) and Cf

for fixed
O&M costs. Cd

indicates the depreciation costs
and T

contains the residual value. Lastly, indi-
cates the risk-adjusted discount rate, t symbolizes
the discrete periods of time and T specifies the
project maturity.
In order to estimate the forward returns of
each year, the study calculate the project value
at the end each period with following expression.
V CF
CF
i i
i
i
i
T
,
( ) =
+ ( ) =
1 1
(3)
with V
i
representing the project value, CF
i
designat-
ing the cash flows of each period and representing
the risk-adjusted discount rate. After simulating
cash flows to estimate the project value in each
period, the method estimates the project returns.
Considering a realistic approximation between
the geometric Brownian motion (GBM) and the
project value progression, its volatility should
be similar. With a tree or lattice, there are condi-
tions to estimate the option value and take the
optimal decision. In the sequence of Copeland
and Antikarov (2001), the model assumes a ran-
dom variable
c
that represents the return between
period 0 and 1.
252
C
i n
V
V
CF PV CF CF
V
=

=
+ ( )
+
ln
, ..,
1
0
1 1 1
0

[ ] i n 1 1 ,

(4)
Smith (2005) suggested that CF
1
should be
stochastic and the other cash flows (CF
2
,...,CF
n
)
should be expected values conditional on the re-
sults of the first cash flow. The problem in decision
making is finding out the best moment to invest
(now or the next period). So, there is no need
to capture the uncertainty of all the cash flows
(Brando et al., 2005). This improvement permits
to detain directly the uncertainty associated with
the first period, rationalising the estimate and
providing a fair estimate of the projects volatility.
B
i V
V
CF V E CF CF
V
=

=
+ ( ) ( )
+
ln
1
0
1 1 1 1 1
0

(5)
Brando et al. (2005) compared this GBM ap-
proximation with Smith and Nau (1995) binomial
approximations with two individual uncertain-
ties in a risk-neutral environment, concluding
the results are not deeply different. Considering
that a binomial tree increases its complexity
with the number of periods and uncertainties,
the motivation to make use of a single variable
approximation of the stochastic process becomes
clear. However, the matching with GBM should
be verified, particularly the lognormal distribution
and the standard deviation homogeneity of the
returns. About the adequacy of Brando et al.s
(2005) approach to our study, Smith (2005) stated
that it should only be applicable for scale options
and that deferment options cannot be considered
scale options. Brando et al.s (2005) allowed this
vision but called for some flexibility.
Following the framework of Yang and Blyth
(2007), the study first calculates the present value
of the project without considering any option.
Using traditional methods, the assessment of the
costs and revenues of the project would involve a
DCF analysis, which discounts future cash flows
to estimate the net present value and thereby its
financial viability. The present value assumes a
full knowledge of each variables evolution (in-
cluding future energy prices and carbon trading
prices) and discounts the future cash flows using
a risk-adjusted discount rate.
Using expression (3), the process begins with
the definition of a project value at moment 0 (V
0
)
and moment 1 (V
1
). Afterwards, using Equation
(5) and the market data, it is possible to obtain an
initial value for the random variable
B
To perform
the Monte Carlo simulation, V
0
is fixed for all it-
erations and V
1
is incorporated into the expression
(4). This expression represents the project returns
and follows a GBM with a drift rate that repre-
sents the mean of the project returns distribution
and volatility s obtained from the standard de-
viation of the returns variation. Therefore, the
project volatility at a time interval t corresponds
to s t / and represents the annualised per-
centage of the standard deviation. Because the
period of cash flow generation is annual, the
process directly obtains the annual volatility from
the standard deviation because equals s. Implic-
itly, there is a postulation that project values are
log-normally distributed and follow a GBM with
a constant volatility. The process to estimate
volatilities employs ordinary econometric ap-
proaches, using similar regression analyses to
isolate price trends. It runs the regressions with
the logarithm of values and with its residuals and
observes the first-order auto-regressions. The
volatility estimates result from the standard error
of these auto-regressions.
Baltazar (2009) considered revenues as a
function of the electricity price (see expression
(1)). So, revenues follow a GBM formalised by
the expression R(c
)
=Ke
c w
ith a constant K. The
variable c t
a
kes values between 0 and 1 and only
varies from the initial period to the next, holding
the same value for the following periods. After
generating project revenues, expression (5) per-
253
mits to obtain the random variable. After running
the simulation, its available the standard deviation
that corresponds to the volatility of project returns.
COMBINED CYCLE NATURAL
GAS-FIRED PLANT
Technology Description
The companys problem evaluates the projects
financial viability. Also suggests the best time to
start the project, taking into account the maximiza-
tion of present value of the project. The action of
the electric market is characterized by its uncer-
tainty nature regarding the price of electricity and
the fuel cost associated with energy production.
This section begins with the case study analysis
of a nuclear combined cycle power plant describ-
ing technological advances that have enabled a
new generation of power plants with high income
levels and low costs. In particular, the natural gas
combined cycle plants can combine a gas cycle
with a steam cycle, which offers a higher efficiency
than other power plants. As gas-fired facilities,
these plants contribute to the diversification of
primary energy sources and take advantage of
issuing low emissions compared with other fuels.
This relevant characteristic constitutes a key factor
to meet the allowed levels of pollution.
Making use of an ROA, this study models a
problem of a power-generating plant investment
problem as a deferment option associated with
the spread between electricity prices and variable
costs (e.g., fuel costs). The option value rises when
interest rates and the time to maturity increase.
Ceteris paribus, if the strike price increases, the
option price declines. In our real options model,
the plant manager has the right to defer the plant
building. The power plants investment decision
process can be defined by call options.
The consumption of electricity in Portugal
between 2003 and 2007 grew on average 2.6%
per year, which represented an increase above
economic growth. In 2007, the electricity produc-
tion was 47,253 GWh, with 28% generated from
natural gas resources. The use of natural gas to
produce electricity grew 11% between 2003 and
2007 (Eurostat, 2009). A combined cycle natural
gas facility is characterised by advanced technol-
ogy with a low environmental impact and high
performance compared with coal or fuel oil or
natural gas coupled to an alternator. A high volt-
age transformer connects the plant to the national
distribution network. This combined cycle power
plant has efficiency between 55% and 56%, much
higher than other power plants. Such efficiency
permits consumed energy to be considered a clean
source. Its high efficiency also leads to a lower
fuel consumption and has contributed to further
improvements in environmental impact.
The combined cycle natural gas-fired plant
has an installed capacity of a 400 MW-type gas
turbine combined with a steam cycle. The exhaust
gases expelled from the gas turbine are used to
generate steam that drives a steam condensation
turbine. The two turbines are linked by a single
shaft and the quantity of natural gas must be 1.8
times the energy generated. The low emissions
of SO
2
, NOx, CO
2
and particulates in combined
cycle plants have a positive effect on environ-
mental performance shows the carbon emissions
for different power plants.
Considering the technological superiority of
a combined cycle natural gas-fired plant with
regard to CO
2
emissions, the following section
describes the uncertainties in market production
and how the evaluation model considers them.
The definition of parameters associated with
uncertainties comes from market studies, namely
the electricity and fuel markets.
Uncertainty Modelling
Some uncertain variables, such as electricity price,
natural gas price, the price for carbon emissions
and project value, should be modelled (Reedman
et al., 2006). The modulation process implies
254
studying the electricity market and retrieving some
historical price data. In this context, Frayer and
Uludere (2001) considered that it is appropriate to
use contracts with maturities similar to the plants
estimated life cycle.
The Iberian Electricity Market (MIBEL) re-
sulted from a partnership between Portugal and
Spain. Its aim is to integrate the two electricity
systems. The formation of a regional market in
the Iberian Peninsula constitutes a step further
for the construction of an internal market in the
European Union. As a consequence of the daily
transactions of electricity contracts, the price var-
ies every day. This procedure helps corroborate
that futures markets, for electricity and natural
gas, have inadequate liquidity to provide reliable
market-based forecasts. Because the futures mar-
kets represent a small portion of the MIBEL and
the contracts are not traded very far in advance.
Therefore, projections are made to simulate
how prices will change stochastically over time
using Monte Carlo techniques. Brando et al.
(2005) only considered the uncertainties with a
market equivalent. Although the project contains
three market uncertainties (electricity price, natu-
ral gas price and the price for carbon emissions),
only the first two are modulated as stochastic
variables. Because electricity demand has no
market equivalent, it is analysed in the sensitivity
analysis section. In the current context, the study
chooses to model electricity prices using a GBM,
for which it is necessary to estimate volatility.
Using historical market data since January
2000, which is available from the market opera-
tors (OMEL, 2009; OMIP, 2009), it is possible to
extract a historical volatility of electricity price of
54%. Assuming this level of annual volatility can
provide a mid-to-long-term standard deviation in
electricity prices, 10,000 iterations of Monte Carlo
simulations (are carried out for a time framework
of 25 years (Baltazar, 2009).
Algeria provides a large percentage of the
natural gas consumed by the Iberian Peninsula.
Several port terminals for natural gas in Spain and
one in Portugal receive it through the Maghreb
Europe pipeline. The price of Brent constitutes the
pattern for the price of natural gas (Monaghan,
2009). Thus, the natural gas price is considered
a market uncertainty in this plant project. The
estimation of gas price volatility results primar-
ily from historical market data between 1987 and
2009 (EIA, 2009). From the analysis of daily spot
prices, Baltazar (2009) obtained a Brent histori-
cal price volatility of 26%, which serves as the
initial value of the simulation process. Following a
similar process to the one of the electricity prices,
the study uses Monte Carlo simulations (Spinney
and Watkins, 1996).
Today, carbon emission licenses are daily
traded in the carbon exchange market (European
Climate Exchange). The launch of new licenses is
carried out in stages and depends on an administra-
tive grant. After 2012, the changes in the carbon
exchange market will invalidate using current
prices as references for the future. Uncertainty
in the price of carbon emissions results from a
biased market and depends on the release of new
regulations. Thus, the allowances of carbon dont
Table 1. Pollutant emissions for different types of power plants (g/kWh), Source: Baltazar (2009).
Technology Fuel CO2 Nox SO2 Particles
Combined Cycle N. Gas 360 0.3 0 0
Conventional N. Gas 560 2 0 0
Conventional Fuel Oil (3%) 750 2 14 0.1
Conventional Fuel Oil (1%) 750 2 5 0.1
Conventional Coal 900 3 7 0.15
255
assume a nature of market uncertainty (EUETS,
2009). According to the study of other combined
cycle natural gas-fired plants with similar profiles,
the study considers an average annual decrease
of carbon emissions of 1.1%.
Project Value Modelling
The project is yet to start and because of economic
conditions and adverse financial conjuncture,
the company is considering postponing the in-
vestment. The combined cycle natural gas-fired
plant has two stages: investment and production.
These stages are later used to price outlooks and
forward cost simulations. The investment stage
involves the construction and reliability testing
phases. Table 2 describes the base parameters of
the investment project.
The production phase starts four years after
the beginning of the project. This means that
building phase takes four years and the reliabil-
ity tests are performed during the first two years
of production. The present cost of the building
phase is 205 million (Table 3). The production
phase lasts for 25 years with a 10-year guaranteed
power purchase agreement (Baltazar, 2009). Us-
ing historical market data since January 2000, it
is possible to extract trends from the electricity
and natural gas prices. A DCF analysis showed
the discounted 29-year net present value is a
negative value of 57 million (Baltazar, 2009),
considering a risk-adjusted discount rate of 8%
and a tax rate of 26.5% (Table 5). Under these
conditions, the project should be rejected. This
level of project value seems to confirm Frayer
and Uludere (2001), who considered that DCF
models, based on production cost-based simula-
tions (like this one), attribute short cumulative
values to typical gas-fired peaking facilities.
The DCF analysis includes a deterministic risk
appraisal and applies a risk premium over the
risk-free interest rate. Therefore, the stochastic
nature of price variations is not taken into account.
That assumption cannot be accepted in a market
with fast technological development, characterised
by uncertainties about energy prices, prices for
carbon dioxide emissions, raw materials and O
& M costs. Although the simplicity of integrating
the risk analysis into a unique risk-adjusted dis-
count rate is tempting, the disadvantages resulting
from a poor understanding of new risks outweigh
the benefits.
From here, the analysis evolves from a de-
terministic DCF model into two directions. One
considers the electricity and natural gas prices
per se. The other reflects cash flows composed
by the difference between the functions of the
electricity price and the fuel price. The project
income depends primarily on electricity demand
and its growth over the years, as well as the elec-
tricity price charged in the Iberian market. The
electricity cost of production essentially depends
on fuel prices such as the natural gas bought in
producer countries and distributed through the
national natural gas network. Because no reliable
long-term market-based predictions for forward
fuel prices are available, the method uses the
company production cost-based technique. This
method retrieves the cost according to the market
prices, adding up the other costs associated with
Table 2. Base parameters of the project
Description Value
Initial Investment (M) 205
Useful Life (years) 29
Risk-adjusted discount rate (annual) 8%
Corporate tax rate (annual) 26.50%
Table 3. Historical volatility
Variable Volatility
Electricity Price (EP) 54%
Natural Gas (NG) 26%
Aggregate (EP,NG) 120%
256
carbon dioxide emissions and operations and
maintenance.
The next phase combines two uncertainties
(electricity price and natural gas price) in the
cash flow statement and obtains the initial project
volatility. This initial value for volatility is dis-
similar to individual uncertainties and depends
on the structure of the cash flows. Brando et al.
(2005) attributed much of this difference to effects
related to operational leverage. In our case, as the
difference between costs and revenues is small,
a tiny change in revenues or costs has a relevant
impact on cash flows. Volatility changes its value
according these changes.
Considering this framework, the ROA can
show the project creates value under different
economic and technological conditions. A DCF
analysis of deterministic net present value ignores
the flexibility to start the project at the right time.
When the project value varies stochastically over
time, the possibility of waiting and seeing is
clearly an asset for managers (Trigeorgis, 1996).
Assuming the existing technical conditions, a
power plant operation with carbon capture and
storage would not be cost effective. However,
because technology is improving at a fast pace,
the investment might turn out to be cost effective
in the near future.
The study integrates two uncertainties (elec-
tricity price and natural gas price) to evaluate the
model value over time. It assumes that the two
variables follow a GBM with a volatility of 54%,
for the electricity price process and a volatility
of 26% for the natural gas price process. The
procedure excludes any variable correlation val-
ues. When compared with Frayer and Uluderes
(2001) peak (15%) and off-peak values (22%),
the volatility estimates seem to be overvalued,
which confirms Smith (2005).
After considering a two-variable approach,
the process of Brando et al. (2005) implies a
combination of the volatility forecasts associated
with the two random variables. Although there
are alternative ways of achieving it (Smith, 2005;
Amendola and Storti, 2008), the process produces
a volatility level of 120%. When the uncertainties
Table 5. Parameters sensitivity analysis
(2) (1) (p) (2) (2)/(1) () (1) (p) (2) (2)/(1)
-50.00% -51.67% 0.033 10.00% -23.24% -1.324
-58.33% -62.79% 0.076 20.00% -46.89% -1.344
-66.67% -74.60% 0.119 30.00% -71.72% -1.391
-75.00% -85.56% 0.141
() (1) (p) (2) (2)/(1) () (1) (p) (2) (2)/(1)
0.00% 0.00% NA 10.00% -10.87% -0.087
-33.33% -23.58% -0.293 20.00% -21.62% -0.081
-66.67% -34.34% -0.485 30.00% -32.50% -0.083
Table 4. Option values for separate and aggregate volatilities
Option Value NPV V(ep) V(ng) V(ep,ng) rf
Model(1) 159 102 113% 84% 2.16%
Model(2) 154 97 172% 2.16%
Model(1):volatility of each uncertainty; Model(2):aggregate volatility
V(ep): volatility of electricity price; V(ng): volatility of natural gas;
V(ep,ng): aggregate volatility; rf: risk-free rate; NPV: net present value
257
are somewhat correlated, the project volatility
should be higher than the sum of the individual
volatilities. In the absence of additional elements,
Lima and Suslick (2006) confirmed empirically
these considerations in similar projects. In sum-
mary, Table 3 shows the historical volatility of
uncertainty variables of the model.
Using the DCF method, the plant investment
shows not to be cost effective (negative project
value). Although this method does not take into ac-
count price uncertainties and can produce incorrect
information to support decisions, modelling these
uncertainties allows us to catch the investment
opportunity when it becomes cost effective. The
following section discusses the pricing of the de-
ferment option using the approaches suggested by
Copeland and Antikarov (2001), Smith (2005) and
Brando et al. (2005). This section also contains
a sensitivity analysis of uncertainties parameters
and other variables not modelled stochastically.
Deferment Option Analysis
Given the project deferment option, the goal is
to determine when the investment becomes eco-
nomically viable. The present value is negative
but the flexibility to defer the investment offers
management the possibility to benefit from ran-
dom movements in the project value. However, if
market conditions change, it will affect the asym-
metry in the obligation to invest (Trigeorgis, 1996).
The firm has a three-year investment license
for starting the plant construction. If the option
expires before the end of the license, the invest-
ment opportunity is lost. Although there are three
market uncertainties (electricity price, natural gas
price and the price for carbon dioxide emissions)
only the first two are considered. Sensitivity
analysis will lead to the uncertainties which have
no market-equivalent.
This study uses a binomial tree that models
the stochastic process (GBM) over the time of the
project (Wang and Dyer, 2010). Because Panko
(1998) evidenced errors of between 20% and
40% of the appliances, the study avoids the use
of binomial lattices to make it safer to examine
the results. The approximation of project uncer-
tainty applies a risk-neutral valuation and includes
options in the decision nodes of the tree. As in
Copeland and Antikarov (2001), real option value
results from the binomial tree filled with project
volatility replicated from forward simulations.
The model approach integrates two uncertain-
ties (electricity and natural gas price) and one
option (to invest or abandon the project during
the time of license). The payoff of each period
reflects the cash flows that result from the ran-
dom variables. Baltazar (2009) studied an initial
investment value of 205 million and a risk-free
rate of 2.16% (MTSP, 2009).
Using simulated volatilities of 113% (electric-
ity price) and 84% (natural gas price) obtains,
through a binomial tree, a project value of 102
million and three years to maturity deferment
option of 159 million (Model(1)). The model
considers a binomial tree that matches the returns
and assumes an equal initial investment value and
risk-free rate. The changed project value differs
by only about 5% (97 million) from the previous
approach with an option value of 154 million
(Model(2)). The difference can be considered
irrelevant, taking into account the initial invest-
ment value of 205 million.
The next section examines the robustness of the
critical parameters such as project volatility, op-
tion maturity, electricity demand and the price for
carbon dioxide emissions. Assuming all the other
variables remain constant, sensitivity analyses
show how each parameter individually moves and
compares them with the strategic net present value
that corresponds to baseline parameterisation.
Process Evaluation and
Sensitivity Analysis
Several tests can be performed to assess the quality
of the process simulations. First, it is necessary
to corroborate if simulated project values follow
258
a normal distribution. Second, it should be seen
whether the volatility over the expected project
returns remains constant from period to period.
The results of the x
2
test with a p value of 5%
indicate that the null hypothesis of project returns
following a lognormal distribution cannot be
rejected. To examine volatility changes over the
several periods of the project, the paper calculates
the value of
B
by applying expression (5) to all
the periods. Because the difference between the
values of each period remain in the interval of
0.2% and +0.8%, it can be considered that there
are no significant variations in volatility from
period to period.
For the sensitivity analysis, the study em-
ploys the expected threshold values of the cash
flow variation (
p
) to capture the impact of the
variation in the uncertainties parameters. The
expression =((
p
/
bp
)-1)*100 translates the
described method where
p
is the strategic net
present value after revised parameterisation and
bp
corresponds to a baseline parameterisation.
Table 2 considers deviations in the critical
parameters in a range between +10% and 75%.
Each parameter change has a corresponding varia-
tion in the strategic present value (
p
). About the
direction of variation it is possible to detect differ-
ent behaviours between the various parameters.
The values tested for the project volatility are
the absolute values between 30% and 60%, with
intervals of 10% that correspond to variations
between 50% and 75%.
A negative change in the project values vari-
ance (
2
) reflects a change in the same direction
in the projects value. Because the test values
of volatility deviate from the benchmark, this
increases the magnitude of their effects on the
project value. The change in the option maturity
(time deferral of the project) has the same effects
on the project value as the volatility. However, the
same percentage change in the project volatility
causes an update in the project value twice as
great as the change in the deferment period of
the project ().
The relationship between electricity demand
() and price for carbon dioxide emissions ()
with the project value is reversed from those
shown by the two previous parameters (volatility
and maturity). An increase in the value of these
variables causes a reduction in the project value.
The variation in electricity demand leads to twice
the adjustment in the project value of the same
percentage change in the price for carbon dioxide
emissions. Although there are significant changes
in the project value, none of the sensitivity analy-
sis procedures call into question the project with
regard to its viability.
The following section describes some of the
latent gaps in an ROA, which could be used as
indicators of future research lines. These perspec-
tives result from the connection between the DCF
and real options paradigms on risk assessment in
the decision-making process.
In a certain sense, electricity market is different
from other commodities because its expensive to
store electricity on a large scale. Consequently,
electricity prices depend on the customers needs,
resulting in large fluctuations in demand during
the day and in balance problems between demand
and production. For that reason, minimizing the
production costs and building a flexible adjustment
between demand and supply are relevant factors to
mitigate market risks and get a significant return
on investments. Combined-cycle power plants
have low investment costs and short construction
times when compared to coal-fired and nuclear
plants (Fraser et al., 2000).
This chapter analyze investments in gas-fired
power plants based on stochastic electricity and
natural gas prices. Fleten and Nasakkala (2010)
analyze the same variables and find that when the
decision to build is considered, the abandonment
option does not have significant value. This study
can also consider an abandonment option and ad-
259
ditionally examine the effects of emission costs
on the value of installing CO2 capture technology.
Traditional DCF analysis in the energy sector
can produce mistaken assessments by ignoring the
adaptability and possibility of cash flow change
overtime. Some authors consider ROA imprecise
and that might even complicate the procedure of
identifying the best strategy. This article shows
that an ROA can supply added information on the
economic viability of energy projects, by valuat-
ing the deferment option. The energy market is
uncertain and thereby an ROA is an essential tool
in strategic decision making. However, to use
consistently an ROA, there are some questions
to discuss.
For projects with equal returns, assuming
risk, neutrality and aversion, an ROA chooses
the projects with a higher volatility. It contradicts
some rules from the traditional literature, which
encourage minimising the risk from an equal
profitability. Also, an ROA creates some bias in
project selection. Increasing the approval rate of
unprofitable projects in the short-term and decreas-
ing the acceptance rate of projects that with new
information might offer new opportunities in the
long-term. Future research should try to overcome
these problems in order to spread the use of ROA.
In terms of technology, the ROA can be
extended to choose between electricity generat-
ing technologies. The firm can decide when to
invest either in a Natural Gas Combined Cycle
(NGCC) power plant or in an Integrated Gas-
ification Combined Cycle (IGCC) power plant.
Furthermore, instead of assuming that fuel prices
follow standard geometric Brownian motions can
be assumed mean reversions, as in Abadie and
Chamorro (2008).
CONCLUSION
The article shows there is an investment project
to be carried out (profitability index of 1.50) with
a deferment option that contributes 51.62% (for
the index). The proper timing for its implemen-
tation is not now. Therefore, the decision maker
should take the deferment option and delay its
implementation. Compared with other project
analyses (Frayer and Uludere, 2001), it seems
that project volatility is slightly overvalued. This
feature seems to come from the high volatility of
electricity price (Alvarado and Rajaraman, 2000)
and from a specific pattern of cash flows with
high first investments and long periods of opera-
tion. Furthermore, sensitivity analysis examines
changes in project volatility, period of production,
disruptions in electricity demand and changes in
prices for carbon dioxide emissions and shows
some robustness of the results.
Therefore, even with some conceptual con-
traindications, ROA approach of Brando et al.
(2005) can be considered fair. This study shows
that energy markets with long periods of opera-
tion and a reasonable adhesion to the GBM, by
the project returns, might call into question its
use in these markets. Although the values of op-
tion evaluation have remained inside acceptable
ranges, some shortfalls could be found.
First, the study employs Monte Carlo simula-
tions, which can be viewed as forward-looking
processes, and option pricing problems need
backward recursive solutions. So, these simula-
tions can be considered less appropriate to deal
with these kinds of option problems.
Second, Smith (2005) suggested some alter-
native approaches to estimate volatility, such as
Longstaff and Schwartz (2001), Farias and Van
Roy (2003) and Ibanez and Zapatero (2004).
Since our study produces similar results, it is not
obvious the extra complexity and computational
needs needed by these approaches can produce
significantly superior results. Finally, it is broadly
accepted that simulation models have a limited
ability to model complex options problems and
large possibilities of containing spreadsheet errors.
It is also consensual that lattices have relevant
drawbacks on transparency and intuition.
260
REFERENCES
Abadie, L., & Chamorro, J. (2008). Valuing flex-
ibility: The case of an integrated gasification com-
bined cycle power plant. Energy Economics, 30(4),
18501881. doi:10.1016/j.eneco.2006.10.004
Alvarado, F., & Rajaraman, R. (2000). Under-
standing price volatility in electricity markets.
Proceedings of the 33rd Hawaii International
Conference on System Sciences.
Amendola, A., & Storti, G. (2008). A GMM
procedure for combining volatility forecasts.
Computational Statistics & Data Analysis, 52(6),
30473060. doi:10.1016/j.csda.2007.10.001
Baltazar, D. (2009). Real options in engineer-
ing projects. Master Thesis (L53087). Lisbon,
Portugal: Technical University of Lisbon (IST).
Black, F., & Scholes, M. (1973). The pric-
ing of options and corporate liabilities. The
Journal of Political Economy, 81(3), 637654.
doi:10.1086/260062
Blyth, W. (2008). Use of real options as a policy
analysis tool, in analytical methods for energy
diversity and security (pp. 6983). London, UK:
Elsevier.
Borison, A. (2005). Real options analysis: Where
are the emperors clothes? Journal of Applied Cor-
porate Finance, 17(2), 1731. doi:10.1111/j.1745-
6622.2005.00029.x
Brando, L., Dyer, J., & Hahn, W. (2005). Using
binomial trees to solve real-option valuation prob-
lems. Decision Analysis, 2(2), 6988. doi:10.1287/
deca.1050.0040
Brekke, K., & Schieldrop, B. (2000). Invest-
ment in flexible technologies under uncertainty.
In Brennan, M., & Trigeorgis, L. (Eds.), Project
flexibility, agency and competition: New develop-
ments in the theory of real options (pp. 3449).
Oxford University Press.
Copeland, T., & Antikarov, V. (2005). Real op-
tions: Meeting the Georgetown challenge. Journal
of Applied Corporate Finance, 17(2), 3251.
doi:10.1111/j.1745-6622.2005.00030.x
Deng, S., Johnson, B., & Sogomonian, A. (1998).
Exotic electricity options and the valuation of
electricity generation and transmission. Proceed-
ings of the Chicago Risk Management. Chicago.
EIA. (2009). US Energy Information Administra-
tion. Retrieved from http://www.eia.doe.gov
EUETS. (2009). European Union emission trad-
ing system. Retrieved from http://ec.europa.eu/
environment/climat/emission/index_en.htm
Eurostat. (2009). Eurostat database. Retrieved
from http://epp.eurostat.ec.europe.eu/portal
Eyedeland, A., & Wolyniec, K. (2003). Energy
and power risk management: New developments
in modeling, pricing and hedging. New York, NY:
John Wiley and Sons.
Farias, D., & Van Roy, B. (2003). The linear
programming approach to approximate dynamic
programming. Operations Research, 51, 850865.
doi:10.1287/opre.51.6.850.24925
Fleten, S., & Nasakkala, E. (2010). Gas-fired
power plants: Investment timing, operating flex-
ibility and CO2 capture. Energy Economics, 32(4),
805816. doi:10.1016/j.eneco.2009.08.003
Fraser, N., Bernhardt, J., & Jewkes, E. (2000).
Engineering economics in Canada (2nd ed.).
Toronto, Canada: Prentice Hall.
Frayer, J., & Uludere, N. (2001). What is it worth?
Application of real options theory to the valuation
of generation assets. The Electricity Journal, 14(8).
doi:10.1016/S1040-6190(01)00237-8
Gardner, D., & Rogers, J. (1999). Planning
electric power systems under demand. Manage-
ment Science, 45(10), 12891306. doi:10.1287/
mnsc.45.10.1289
261
Gnansounou, E., Schmutz, A., & Dong, J. (2003).
Investment decisions with flexibility in the liberal-
ized electricity markets: An approach using real
options theory. SAFE Annual Conference. Zurich.
Ibanez, A., & Zapatero, F. (2004). Monte Carlo
valuation of American options through computa-
tion of the optimal exercise frontier. Journal of
Financial and Quantitative Analysis, 39, 253275.
doi:10.1017/S0022109000003069
Kehlhofer, R., Bachmann, R., Nielson, H., &
Warner, J. (1999). Combined-cycle gas & steam
turbine power plants (2nd ed.). Tulsa, OK: Pen-
nwell.
Lima, G., & Suslick, S. (2006). Estimating the
volatility of mining projects considering price and
operating cost uncertainties. Resources Policy,
31, 8694. doi:10.1016/j.resourpol.2006.07.002
Longstaff, F., & Schwartz, E. (2001). Valuing
American options by simulation: A simple least-
squares approach. Review of Financial Studies,
14, 113147. doi:10.1093/rfs/14.1.113
Monaghan, B. (2009). How to trade natural gas.
Retrieved from http://www.paddypowertrader.
com/index.php
MTSP. (2009). MTSPortugal database. Retrieved
from http://www.mtsportugal.com/
Nau, R., & McCardle, K. (1991). Arbitrage, ra-
tionality and equilibrium. Theory and Decision,
33, 199240. doi:10.1007/BF00132993
OMEL. (2009). Operador del Mercado Ibrico
de Energa. Retrieved from http://www.omel.es
Panko, R. (1998). What we know about spread-
sheet errors. Journal of Organizational and End
User Computing, 10, 1521.
Reedman, L., Graham, P., & Coombes, P. (2006).
Using a real-options approach to model technol-
ogy adoption under carbon price uncertainty: An
application to the Australian electricity generation
sector. The Economic Record, 82(S1), S64S73.
doi:10.1111/j.1475-4932.2006.00333.x
Rode, D., Fischbeck, P., & Dean, S. (2001). Monte
Carlo methods for appraisal and valuation: A case
study of a nuclear power plant. Journal of Struc-
tured and Project Finance, 7, 3848. doi:10.3905/
jsf.2001.320257
Smith, J. (2005). Alternative approaches for solv-
ing real options problems. Decision Analysis, 2(2),
89102. doi:10.1287/deca.1050.0041
Smith, J., & Nau, R. (1995). Valuing risky projects:
Option pricing theory and decision analysis. Man-
agement Science, 14(5), 795816. doi:10.1287/
mnsc.41.5.795
Spinney, P., & Watkins, G. (1996). Monte Carlo
simulation techniques and electric utility re-
source decisions. Energy Policy, 24, 155163.
doi:10.1016/0301-4215(95)00094-1
Thompson, M., Davison, M., & Rasmussen,
H. (2004). Valuation and optimal operation of
electric power plants in competitive markets.
Operations Research, 52, 546562. doi:10.1287/
opre.1040.0117
Trigeorgis, L. (1993). Real options and interactions
with financial flexibility. Financial Management,
22(3), 202223. doi:10.2307/3665939
Trigeorgis, L. (1996). Real options: Managerial
flexibility and strategy in resource allocation.
Cambridge, MA: MIT Press.
Tseng, C., & Barz, G. (2002). Short-term gen-
eration asset valuation: A real options approach.
Operations Research, 50(2).
Wang, T., & Dyer, J. (2010). Valuing multifac-
tor real options using an implied binomial tree.
Decision Analysis, 7(2), 185195. doi:10.1287/
deca.1100.0174
262
Yang, M., & Blyth, W. (2007). Modelling invest-
ment risks and uncertainties with real options ap-
proach. In IEA (Ed.), Climate policy uncertainty
and investment risk. Paris, France: IEA.
ADDITIONAL READING
Amramand, M., & Kulatilaka, N. (1999). Real
options: Managing strategic investment in an
uncertain world. Cambridge, MA: Harvard Busi-
ness School Press.
Botterud, A., & Korpas, M. (2007). A stochastic
dynamic model for optimal timing of investment
in new generation capacity in restructured power
systems. Electrical Power and Energy Systems,
29(2), 163174. doi:10.1016/j.ijepes.2006.06.006
Boyle, P. (1977). Options: A Monte Carlo Ap-
proach. Journal of Financial Economics, 4,
323338. doi:10.1016/0304-405X(77)90005-8
Brealey, R., & Myers, S. (2002). Capital invest-
ment and valuation (1st ed.). USA: McGraw-Hill.
Brennan, M., & Schwartz, E. (1985). Evaluat-
ing natural resource investments. The Journal of
Business, 58(2), 135157. doi:10.1086/296288
Castellacci, G. and Siclari, M. (2003). Asian bas-
ket spreads and other exotic averaging options.
Energy Power Risk Management, March (2003).
Castellacci, G., Siclari, M., Ross, S. and Liao, W.
(2004). Real option valuation of power generation
assets and spread-switching options. Proceedings
of Forecasting Financial Markets.
Chen, H., & Baldick, R. (2007). Optimizing short-
term natural gas supply portfolio for electric utility
companies. IEEE Transactions on Power Systems,
22(1). doi:10.1109/TPWRS.2006.889144
Copeland, T., & Antikarov, V. (2001). Real op-
tions: A practitioners guide. USA: Texere Pub-
lishing.
Cortazar, G. (2004). Simulation and numerical
methods in real options valuation, in Real Options
and Investment under Uncertainty: Classical
Readings and Recent Contributions, Schwartz
and Trigeorgis (editors), USA: MIT Press.
Cox, J., Ross, S., & Rubinstein, M. (1979). Option
pricing: a simplified approach. Journal of Finan-
cial Economics, 7, 229263. doi:10.1016/0304-
405X(79)90015-1
Deb, R. (2000). Operating hydroelectric and
pumped storage units in a competitive environ-
ment, Electricity Journal.
Deng, S., Johnson, B., & Sogomonian, A. (1999).
Exotic electricity options and the valuation of
electricity generation and transmission assets.
Proceedings of 32nd Hawaii International Confer-
ence on System Sciences.
Devin, K. (1995). Gas in electricity generation. En-
ergy Exploration and Exploitation, 13, 149157.
Dixit, A., & Pindyck, R. (1994). Investment under
uncertainty. NJ: Princeton University Press.
Fuss, S., & Szolgayova, J. (2010). Fuel price and
technological uncertainty in a real options model
for electricity planning. Applied Energy, 87(9).
doi:10.1016/j.apenergy.2009.05.020
Gardner, D., & Rogers, J. (1999). Planning elec-
tric power systems under demand uncertainty
with different technology lead times. Manage-
ment Science, 45(10), 12891306. doi:10.1287/
mnsc.45.10.1289
Geman, H. (2005). Energy commodity prices:
Is mean-reversion dead? Journal of Alterna-
tive Investments, 8(1), 3145. doi:10.3905/
jai.2005.591576
Griffes, P., Hsu, M., & Kahn, E. (1999). Power
asset valuation: real options, ancillary services
and environmental risks. The New Power Markets.
London, UK: Risk Books.
263
Hsu, M. (1998). Spark spread options are hot!
The Electricity Journal, 11(2), 112. doi:10.1016/
S1040-6190(98)00004-9
Hull, J. (2002). Options, futures and other deriva-
tives (5th ed.). USA: Prentice Hall.
Hull, J., & White, A. (1994a). Numerical proce-
dures for implementing term structure models I:
single-factor models. Journal of Derivatives, 2(1),
716. doi:10.3905/jod.1994.407902
Hull, J., & White, A. (1994b). Numerical proce-
dures for implementing term structure models II:
two-factor models. Journal of Derivatives, 2(2),
3748. doi:10.3905/jod.1994.407908
Ingersoll, J., & Ross, S. (1992). Waiting to invest:
investment under uncertainty. The Journal of Busi-
ness, 65(1), 129. doi:10.1086/296555
Jacobs, J., & Schultz, G. (2002). Opportunities
for stochastic and probabilistic modelling in the
deregulated electricity industry. In Greengard,
C., & Ruszczynski, A. (Eds.), Decision Making
Under Uncertainty Energy and Power (p. 128).
New York: Springer.
Kester, W. (1984). Todays options for tomorrows
growth. Harvard Business Review, 62, 153160.
Kumbaroglu, G., Madlener, R., & Demirel, M.
(2008). A real options evaluation model for the
diffusion prospects of new renewable power gen-
eration technologies. Energy Economics, 30(4),
18821908. doi:10.1016/j.eneco.2006.10.009
Larsen, E., & Bunn, D. (1999). Deregulation in
electricity: Understanding strategic and regulatory
risks. The Journal of the Operational Research
Society, 50(4).
Laurikka, H. (2006). Option value of gasification
technology within an emissions trading scheme.
enpol.2005.09.002
Leppard, S. (2004). Valuation and risk manage-
ment of physical assets. In Kaminski, V. (Ed.),
Managing Energy Price Risk: The New Challenges
and Solutions (3rd ed.). Risk Books.
Leslie, K., & Michales, M. (1997). The real power
of real options. The McKinsey Quarterly, (3): 422.
Li, F., & Chiu, W. (2002). Valuing generation as-
sets using real option competitive price analysis:
A Step-by-Step Valuation Example for a Portfolio
of Generating Assets. In Ronn, E. (Ed.), Real
Options and Energy Management (pp. 597634).
Risk Waters Group Ltd.
Luherman, T. (1998). Investment opportunities as
real options: Getting started in numbers. Harvard
Business Review, July-Aug., 51-67.
Madlener, R., Kumbaroglu, G., & Ediger, V.
(2004). Modeling technology adoption as an
irreversible investment under uncertainty: the
case of the Turkish electricity supply industry.
Energy Economics, 27(1), 139163. doi:10.1016/j.
eneco.2004.10.007
Mcdonald, R., & Siegel, D. (1985). Investments
and the valuation of firms when there is an option
to shutdown. International Economic Review,
26(2), 331349. doi:10.2307/2526587
McDonald, R., & Siegel, D. (1986). The value of
waiting to invest. The Quarterly Journal of Eco-
nomics, 101(4), 707728. doi:10.2307/1884175
Merton, R. (1973). Theory of rational option
pricing. The Bell Journal of Economics and
Management Science, 4(Spring), 141183.
doi:10.2307/3003143
Myers, S. (1977). Determinants of corporate bor-
rowing. Journal of Financial Economics, (5): 147.
doi:10.1016/0304-405X(77)90015-0
Pindyck, R. (1999). The long-run evolution of
energy prices. The Energy Journal (Cambridge,
Mass.), 20(2), 127. doi:10.5547/ISSN0195-6574-
EJ-Vol20-No2-1
264
Pindyck, R. (2001). The dynamics of commod-
ity spot and futures markets: A primer. The En-
ergy Journal (Cambridge, Mass.), 22(3), 129.
doi:10.5547/ISSN0195-6574-EJ-Vol22-No3-1
Ronn, E. (Ed.). (2002). Real options and energy
management. USA: Risk Waters Group Ltd.
Roques, F., Nuttall, W., & Newbery, D. (2006).
Using probabilistic analysis to value power gen-
eration investments under uncertainty. Working
Paper, EPRG.
Rosenberg, M., Bryngelson, J., Sidorenko, N., &
Baron, M. (2002). Price spikes and real options:
Transmission valuation. In Ronn, E. (Ed.), Real
Options and Energy Management (pp. 323369).
Risk Waters Group Ltd.
Schwartz, E., & Trigeorgis, L. (2004). Real options
and investment under uncertainty: classical read-
ings and recent contributions. USA: MIT Press.
Shahidehpour, M., Yami, H., & Zuvi, L. (2002).
Market operations in electric power systems: Fore-
casting, scheduling and risk management. USA:
Wiley-IEEE Press. doi:10.1002/047122412X
Shimko, D. (1994). Options on futures spread:
Hedging, speculation and valuation. Journal of Fu-
tures Markets, 14(2). doi:10.1002/fut.3990140206
Sun, N., Ellersdorfer, I., & Swider, D. (2008).
Model-based long-term electricity generation
system planning under uncertainty. Proc. IEEE In-
ternational Conf. Electric Utility Deregulation and
Restructuring and Power Technologies, Nanjing.
Weron, R. (2006). Modelling and forecasting
electricity load and prices: A statistical approach.
UK: John Wiley and Sons Ltd.
Zambujal-Oliveira, J. (forthcoming). Validity of
the homogeneity property in real assets replace-
ment procedures. International Review of Applied
Financial Issues and Economics.
Zambujal-Oliveira, J., & Duque, J. (2010). Opera-
tional asset replacement strategy: A real options
approach. European Journal of Operational
Research. doi:.doi:10.1016/j.ejor.2010.09.011
Binomial options pricing model: Numerical
method for the valuation of options proposed by
Cox, Ross and Rubinstein (1979). The model uses
a lattice based model of the varying price over
time of the underlying asset.
Combined Cycle: An electrical generating
technology in which electricity is produced from
otherwise lost waste heat, exiting from one or
more gas (combustion) turbines. The exiting heat
is routed to a conventional boiler or to a heat re-
covery steam generator for use by a steam turbine
to generate electricity.
Historical Volatility: The annualized Stan-
dard Deviation of percent changes in prices over
a specific period. It is a hint of past volatility in
market prices.
Monte Carlo Simulation: Risk analysis
techniques in which probable future events are
simulated on a computer generating estimated
rates of return and risk indexes.
Option: The right, but not the obligation, to
purchase, sale or contract an underlying asset at
a specified price during a defined time period.
Real Options Analysis: Applies option valu-
ation methods to investment decisions.
Risk-Adjusted Discount Rate: The discount
rate that applies to a particular risky stream of
income; the riskier the projects income stream,
the higher the discount rate.
Spark Spread: The relationship between the
price of electricity and the price of natural gas or
other fuel used to produce electricity. The spark
spread reflects the costs, or expected costs, of
producing power.
265
APPENDIX
Table 6. Net and guaranteed energy production (Baltazar, 2009)
Year 5 6 7 8 9 10 11 12 13 14 15 16 17
Net energy production (GWh) 1515 1510 1633 1544 1498 1478 1475 1539 1507 1477 1405 1335
Guaranteed production (k) 7798 7798 7798 7798 7798 7798 7798 7798 7798 0 0 0
Electricity Price (/MWh) 83.6 85.3 87 88.8 90.5 92.3 94.2 96.1 98 100 102 104
Year 18 19 20 21 22 23 24 25 26 27 28 29
Net energy production (GWh) 1194 1125 1210 1305 1392 1477 1567 1534 1501 1468 1468
Guaranteed production (k) 0 0 0 0 0 0 0 0 0 0 0
Electricity Price (/MWh) 108 110 113 115 117 119 122 124 127 129 132
266
T
a
b
l
e

7
.

R
e
v
e
n
u
e
s

a
n
d

c
o
s
t
s

o
f

a

c
o
m
b
i
n
e
d
-
c
y
c
l
e

p
l
a
n
t

i
n
v
e
s
t
m
e
n
t

(
B
a
l
t
a
z
a
r
,

2
0
0
9
)
.

Y
e
a
r
5
6
7
8
9
1
0
1
1
1
2
1
3
1
4
1
5
1
6
1
7
N
a
t
u
r
a
l

G
a
s

P
r
i
c
e
s

(
/
M
W
h
t
)
3
0
3
0
2
9
3
1
3
2
3
2
3
3
3
3
3
4
3
5
3
5
3
6
3
6
P
C
I

(
%
)
0
.
5
6
0
.
5
6
0
.
5
6
0
.
5
6
0
.
5
6
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
C
O
2

C
o
s
t

(
/
t
)
2
8
2
8
2
9
2
9
2
9
3
1
3
2
3
4
3
5
3
7
3
9
4
0
4
2
C
O
2

E
m
i
s
-
s
i
o
n

(
t
)
5
6
4
,
8
6
1
5
4
9
,
1
4
5
5
4
8
,
2
9
9
5
9
3
,
1
2
8
5
6
1
,
0
4
5
5
4
4
,
6
4
6
5
3
7
,
8
9
9
5
3
7
,
0
3
1
5
6
0
,
3
2
2
5
4
8
,
6
6
7
5
3
7
,
8
6
2
5
1
1
,
7
7
2
4
8
6
,
2
3
8
O
&
M

F
i
x
e
d

C
o
s
t

(
)
2
,
3
9
6
2
,
4
4
4
2
,
4
9
3
5
,
3
9
2
5
,
4
4
3
5
,
4
9
4
5
,
5
4
7
5
,
6
0
1
4
,
4
5
6
4
,
5
1
2
4
,
5
7
0
4
,
6
2
8
4
,
6
8
8
O
&
M
V
a
r
i
-
a
b
l
e

C
o
s
t

(
)
1
,
6
6
2
1
,
6
5
2
1
,
6
8
3
1
,
8
5
6
1
,
7
9
4
1
,
7
7
7
1
,
7
9
2
1
,
8
2
6
1
,
9
4
1
1
,
9
4
0
1
,
9
4
1
1
,
8
8
5
1
,
8
2
9
Y
e
a
r
1
8
1
9
2
0
2
1
2
2
2
3
2
4
2
5
2
6
2
7
2
8
2
9
3
0
N
a
t
u
r
a
l

G
a
s

P
r
i
c
e
s

(
/
M
W
h
t
)
3
7
3
8
3
8
3
9
3
9
4
0
4
1
4
2
4
2
4
3
4
4
4
4
P
C
I

(
%
)
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
0
.
5
5
C
O
2

C
o
s
t

(
/
t
)
4
4
4
6
4
9
5
1
5
3
5
6
5
8
6
1
6
4
6
7
7
0
7
3
C
O
2

E
m
i
s
-
s
i
o
n

(
t
)
4
6
0
,
0
6
4
4
3
5
,
6
8
2
4
1
0
,
2
8
9
4
4
1
,
5
4
1
4
7
6
,
6
1
2
5
0
8
,
9
2
8
5
4
0
,
0
5
0
5
7
3
,
1
7
0
5
6
1
,
5
5
4
5
4
9
,
9
0
9
5
3
8
,
3
9
1
5
3
8
,
3
9
1
O
&
M

F
i
x
e
d

C
o
s
t

(
)
4
,
7
4
8
5
,
6
0
2
5
,
6
6
5
5
,
7
2
9
5
,
7
9
5
5
,
8
6
2
5
,
9
3
1
1
4
,
1
7
2
7
,
6
2
1
7
,
6
9
3
7
,
7
6
8
6
,
8
8
0
O
&
M
V
a
r
i
-
a
b
l
e

C
o
s
t

(
)
1
,
7
6
8
1
,
7
0
8
1
,
6
4
2
1
,
8
0
0
1
,
9
8
3
2
,
1
6
0
2
,
3
3
6
2
,
5
2
9
2
,
5
2
9
2
,
5
2
8
2
,
5
2
6
1
,
8
8
8
267
T
a
b
l
e

8
.

C
a
s
h

f
l
o
w
s

o
f

t
h
e

c
o
m
b
i
n
e
d

c
y
c
l
e

p
l
a
n
t

i
n
v
e
s
t
m
e
n
t

(
B
a
l
t
a
z
a
r
,

2
0
0
9
)

Y
e
a
r
1
-
4
5
6
7
8
9
1
0
1
1
1
2
1
3
1
4
1
5
1
6
R
e
v
e
n
u
e
s
1
3
6
,
1
9
9
1
2
9
,
1
5
2
1
2
6
,
5
3
3
1
4
5
,
4
2
8
1
4
0
,
8
6
0
1
4
0
,
2
0
9
1
3
9
,
9
4
7
1
4
2
,
6
7
2
1
4
8
,
4
5
6
1
4
9
,
1
4
0
1
4
1
,
7
3
6
1
3
7
,
6
7
0
C
o
s
t
s
1
0
4
,
8
8
6
1
0
0
,
0
4
5
9
9
,
7
1
9
1
1
6
,
6
6
8
1
1
2
,
3
5
3
1
1
1
,
5
4
2
1
1
2
,
6
0
8
1
1
4
,
7
3
1
1
2
0
,
6
9
1
1
2
0
,
7
5
2
1
2
1
,
0
0
0
1
1
7
,
8
4
8
E
B
I
T
D
A
3
1
,
3
1
3
2
9
,
1
0
7
2
6
,
8
1
4
2
8
,
7
6
0
2
8
,
5
0
7
2
8
,
6
6
7
2
7
,
3
3
9
2
7
,
9
4
1
2
7
,
7
6
5
2
8
,
3
8
8
2
0
,
7
3
6
1
9
,
8
2
2
D
e
p
r
e
c
i
a
-
t
i
o
n
9
,
7
8
5
9
,
7
9
2
9
,
7
9
2
9
,
7
9
2
9
,
7
2
2
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
E
B
I
T
2
1
,
5
2
8
1
9
,
3
1
5
1
7
,
0
2
2
1
8
,
9
6
8
1
8
,
7
8
5
1
9
,
1
5
6
1
7
,
8
2
8
1
8
,
4
3
0
1
8
,
2
5
4
1
8
,
8
7
7
1
1
,
2
2
5
1
0
,
3
1
1
T
a
x
e
s
5
,
7
0
5
5
,
1
1
9
4
,
5
1
1
5
,
0
2
7
4
,
9
7
8
5
,
0
7
6
4
,
7
2
4
4
,
8
8
4
4
,
8
3
7
5
,
0
0
2
2
,
9
7
4
2
,
7
3
2
I
n
v
e
s
t
m
e
n
t
2
0
5
,
4
2
7
C
a
s
h
-
f
l
o
w
s
2
5
,
6
0
8
2
3
,
9
8
8
2
2
,
3
0
3
2
3
,
7
3
3
2
3
,
5
2
9
2
3
,
5
9
1
2
2
,
6
1
5
2
3
,
0
5
7
2
2
,
9
2
8
2
3
,
3
8
6
1
7
,
7
6
2
1
7
,
0
9
0
D
i
s
c
o
u
n
t

C
a
s
h
-
f
l
o
w
s
1
7
4
2
9
1
5
1
1
7
1
3
0
1
4
1
2
8
2
3
1
1
7
7
0
1
0
9
2
7
9
6
9
9
9
1
5
6
8
4
3
0
7
9
6
2
5
5
9
9
4
9
8
8
Y
e
a
r
1
7
1
8
1
9
2
0
2
1
2
2
2
3
2
4
2
5
2
6
2
7
2
8
2
9
R
e
v
e
n
u
e
s
1
3
3
,
4
6
3
1
2
8
,
7
7
2
1
2
4
,
9
1
6
1
2
0
,
2
1
5
1
2
9
,
3
2
4
1
3
9
,
3
9
5
1
4
8
,
4
1
9
1
5
8
,
0
3
2
1
6
8
,
0
7
2
1
6
5
,
9
2
0
1
6
3
,
8
4
7
1
6
1
,
4
7
6
1
6
0
0
5
8
C
o
s
t
s
1
1
4
,
7
1
7
1
1
1
,
2
0
7
1
0
8
,
6
4
5
1
0
4
,
9
5
6
1
1
5
,
0
0
0
1
2
6
,
4
4
1
1
3
7
,
6
3
9
1
4
9
,
0
2
9
1
6
9
,
6
2
8
1
6
3
,
4
9
5
1
6
3
,
9
3
8
1
6
4
,
3
9
5
1
6
6
7
3
2
E
B
I
T
D
A
1
8
,
7
4
6
1
7
,
5
6
5
1
6
,
2
7
1
1
5
,
2
5
9
1
4
,
3
2
4
1
2
,
9
5
4
1
0
,
7
8
0
9
,
0
0
3
-
1
,
5
5
6
2
,
4
2
5
-
9
1
-
2
,
9
1
9
-
6
,
6
7
4
D
e
p
r
e
c
i
a
-
t
i
o
n
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
9
,
5
1
1
7
,
2
6
2
3
5
4
E
B
I
T
9
,
2
3
4
8
,
0
5
4
6
,
7
6
0
5
,
7
4
8
4
,
8
1
3
3
,
4
4
3
1
,
2
6
9
-
5
0
8
-
1
1
,
0
6
7
-
7
,
0
8
6
-
9
,
6
0
2
-
1
0
,
1
8
1
-
7
,
0
2
8
T
a
x
e
s
2
,
4
4
7
2
,
1
3
4
1
,
7
9
1
1
,
5
2
3
1
,
2
7
5
9
1
2
3
3
6
0
0
0
0
0
0
I
n
v
e
s
t
m
e
n
t
C
a
s
h
-
f
l
o
w
s
1
6
,
2
9
8
1
5
,
4
3
1
1
4
,
4
8
0
1
3
,
7
3
6
1
3
,
0
4
9
1
2
,
0
4
2
1
0
,
4
4
4
9
,
0
0
3
-
1
,
5
5
6
2
,
4
2
5
-
9
1
-
2
,
9
1
9
-
6
,
6
7
4
D
i
s
c
o
u
n
t

C
a
s
h
-
f
l
o
w
s
4
4
0
5
3
8
6
4
3
3
5
5
2
9
4
7
2
5
9
2
2
2
1
5
1
7
7
9
1
4
2
0
-
2
2
7
3
2
8
-
1
1
-
3
3
8
-
7
1
6
N
e
t
P
r
e
s
e
n
t

V
a
l
u
e
-
5
6
9
0
4
268
Chapter 9
INTRODUCTION
Distribution network reconfiguration has a history
about 35 years
1
and literatures indicate initial ef-
forts by Merlin and Back (1975). Their heuristic
work focused on power loss reduction where
it started with closing all network switches to
perform a mesh configuration. Switches are then
opened successively to restore radial configuration
with less power loss. Thereafter, this method got
some improvements by Shirmohammadi and Hong
(1989) where introduced as Sequence Switch Op-
eration Method
2
. Here also, the reconfiguration
Armin Ebrahimi Milani
Islamic Azad University, Iran
Mahmood Reza Haghifam
Tarbiat Modares University, Iran
Optimal Confguration and
Reconfguration of Electric
Distribution Networks
ABSTRACT
Power loss reduction can be considered as one of the main purposes for a distribution systems design-
ers and operators, especially for recent non-governmental networks. Moreover, the nature of power
loss challenges different methods to solve this problem, while various studies indicate effectiveness
of reconfguration and its high portion for this case. Thus, reconfguration can be introduced as an
optimization procedure to obtain economical high quality operation by changing the status of sectional-
izing switches in these networks. Some major points such as using different switch types, considering
number of switching and time varying loads, which are almost neglected or not applied simultaneously
in most pervious essays, are the main motivation to propose this chapter. A heuristic practical scheme
is proposed to perform optimal reconfguration, and all previous neglected topics are fully discussed.
Proposed method will apply to sample distribution networks, and the effectiveness of this method will
be discussed through several case studies and comparisons.
DOI: 10.4018/978-1-61350-138-2.ch009
269
Optimal Confguration and Reconfguration of Electric Distribution Networks
procedure starts by closing all network switches
which are then opened one after another so as to
establish the optimum power flow in the network.
This is while, many approximations of pervious
method have been overcome in this algorithm and
the computation takes less time.
Civanlar et al. (1988) made use of method
which is known as a Branch Exchange opera-
tion. This method suggested the opening of any
switch was required to correspond to the closure
of another switch, ensuring that the radial con-
figuration of the distribution network would be
preserved. A year after, Baran and Wu (1989)
succeeded to improve the method of Branch Ex-
change by offering two approximated power flow
Equations in 1989. Their power flow Equations
introduced by recursive approximation of P, Q
and V for each branch.
Liu (1989) in the same year, Taylor, Chiang
and Jeon (1990) in next years presented various
heuristic method for distribution feeder recon-
figuration. Wagner (1993), compares different
major reconfiguration methods. This comparison
depicts considerable economic advantages by us-
ing presented heuristic methods especially for real
time operations. He indicated that an important
loss reduction was obtained through simulations
in Canadian networks during a one year period.
Genetic Algorithm applied in distribution sys-
tem reconfiguration for reduction of real power
loss by Nara et al. (1992) for the first time. They
compared obtained results with the results of Simu-
lated Annealing (SA) method and concluded that
although the genetic algorithm has less assurance
but it can be faster that SA. After introducing GA
method and due to its effectiveness, many works
done by refine or suggest heuristic techniques to
this algorithm to compensate long solution time.
Sarfi et al. (1996) presented an algorithm
based on Network Partitioning Method after
their efforts to use refined genetic algorithm for
reconfiguration of distribution networks. Before
that, Chen and Cho (1993) have performed an
analysis of hourly reconfiguration scheme. They
have studied the hourly load patterns over an
interval of a year in order to define the hourly
load conditions for each season. They have used
Branch and Bound technique for obtaining mini-
mum loss configuration considering hourly load
patterns over annual intervals. It was the year that
Kashem et al. (1999) investigated the load rear-
rangement and proposed their famous method for
load balancing in distribution systems.
Huang and Chin (2002) proposed an algorithm
based on fuzzy operation in order to deal with
feeder reconfiguration problem. They suggested
this method to reduce power loss as well as to
acquire the load balance. More Artificial intel-
ligence based methods, such as (Ahuja, 2007;
Mendoza, 2009; Milani 2010) are other useful
procedures proposed for reconfiguration of dis-
tribution networks, recently.
By review the full history of distribution
network reconfiguration, can perceive a vast
bibliography on this optimization problem. As it
mentioned before, it includes general attempts to
find a global optimal solution, new approximate
methods (Civanlar, 1988), (Baran, 1989) and al-
ternative paradigms of approaching the problem,
such as evolutionary computation (Nara, 1992)
and other bio-inspired techniques (Ahuja, 2007).
What is more considerable here is that most
of these approaches consider fixed demands at
each load point (usually the maximum demands
forecast for the planning period). However, loads
vary during any given planning period, with a dif-
ferent pattern for each bus, and these variations
must be considered in the search for a minimum
loss configuration. Also it seems considering
different switch types will affect on optimal
configuration and reconfiguration of electric
distribution networks.
MAIN FOCUS OF THE CHAPTER
After a brief introduction to electric distribution
networks, their configuration and reconfiguration,
270
this chapter will focus on some issues involved
with reconfiguration of electric distribution net-
works and provides practical answers to these
major points. Through this investigation readers
learn complete concepts of an optimal reconfigura-
tion procedure to reduce power loss and total costs,
while they will get some basic information and
knowledge about electric distribution networks
configuration. Fully discussion about this subject
in following sections supplies a perfect insight to
this problem and its issues.
INTRODUCTION TO
DISTRIBUTION NETWORKS
AND THEIR CONFIGURATION
Electricity distribution is the final stage in the
delivery before retail of electricity to end users.
A distribution systems network carries electric-
ity from the transmission system and delivers it
to the consumers. In the other words, distribu-
tion network provides the final links between
the transmission system and demands. Figure 1
gives a basic insight from distribution position in
power systems and depicts the main configuration
of the network.
Typically, a distribution network would include
medium-voltage power lines, electrical substations
and transformers, low-voltage distribution wiring
and sometimes electricity meters. These networks
are generally divided into subsystems of radial or
loop feeders fitted with number of switches that
are normally closed or opened. Customers could
be supplied from different substations or by dis-
tributed power generations through different
routes. These paths are characterized by different
mixtures of commercial, industrial and residential
customers who might impose time-varying load
demands and different service reliability require-
ments (Yin & Lu, 2009).
Radial distribution networks are the most con-
ventional configurations of distribution systems.
In such networks, feeders are extended from
distribution substations to lateral feeders while all
service areas are supplied through feeders. By this
definition radial network leaves the station and
passes through the network area with no normal
connection to any other supply while an intercon-
nected conFigure, such as mesh structure, will have
multiple connections to other points of supply.
In general, the main advantages of radial con-
figuration are its simplicity and its low cost (Ahuja,
Das & Pahwa, 2007). In radial configuration, the
number of disconnecting devices reduces and
design of a protection system is not complicated.
Moreover, this configuration is widely used for
effective coordination of its protective devices
(Kashem, Ganapathy & Jasmon, 1999). Figure 2
illustrates single-line diagram of a typical men-
tioned distribution network.
The loop (or ring) system of distribution starts
at the substation and is connected to or encircles
an area serving one or more distribution transform-
ers or load centers. Unlike the radial network, the
conductor of the ring system returns to the same
substation. The loop system is more expensive to
build than the radial type, but it is more reliable
and may be justified in an area where continuity
of service is of considerable importance. Figure
Figure 1. Distribution position in power system
and its main configuration
271
3 illustrates single-line diagram of a typical men-
tioned distribution network.
INTRODUCTION TO DISTRIBUTION
NETWORKS RECONFIGURATION
In original configuration of a distribution network
there are some sectionalizing switches (normally
closed
3
) as well as some tie switches (normally
opened
4
) for both protection and configuration
management. By changing the status of these
switches, the configuration of distribution sys-
tem is varied and loads are transferred among
the feeders. This could be done while the radial
configuration format of electrical supply is still
maintained at the end of this process and all load
points are not interrupted (Rugthaicharoencheep
& Sirisumrannukul, 2009). The whole mentioned
operation for reaching a new optimized configura-
tion is defined as Reconfiguration of distribution
networks.
Figure 4 illustrates a typical radial distribution
network before and after reconfiguration. For this
sample network, the operator has followed loss
reduction purpose. In this case switches S14 &
S15 are closed after reconfiguration. This is while;
S8 & S9 are opened to preserve the initial radial
structure of network under study.
Generally, reconfiguration methods can be
divided in to two groups: General and Spe-
cific. In the specific methods, one initial answer
is obtained and is used in a specific algorithm to
reach further answers up to an improvement point
Figure 2. Single-line diagram of a typical radial distribution network
Figure 3. Single-line diagram of a typical loop distribution network
272
considering the constraints of the problem. In
general methods, an algorithm is used for solving
the problem and a large range of answer are ob-
tained, among which by performing an operation
the most improved answer is selected as a final
one.
The advantages obtained from feeder recon-
figuration using any proposed methods are, for
example, real power loss reduction, balancing
system load, bus voltage profile improvement,
increasing system security and reliability, and
power quality improvement.
Different Switch Types and Number
of Switching Consideration
To conFigure a distribution network can consider
two different switch types; manual and automatic
ones. Considering the combination of these two
different types of switches simultaneously will
result an optimal flexible planning and operating
of these networks with the most minimization in
costs. Therefore, these benefits can achieve trough
two different cases as follow:
Configuration Mode
Cost minimization can achieve by using differ-
ent switch types to conFigure a new distribution
system. This procedure will follow by different
switch type placement in a statistic manner i.e. an
automatic switch will be the best optimal chose
for a place with high switching probability where
manual ones can be selected for places with less
switch operating.
Note that in this case there are two different
costs which together determine the total cost;
first the cost should spend for change the state
of switches and second the cost of automation.
Considering this differentiations in costs, com-
panies can achieve most benefits through their
investigations by establish an optimal balance
between above maintained costs.
Figure 4. Typical radial distribution network before (above) and after (below) reconfiguration
273
Reconfiguration Mode
Except a huge benefit derives from configura-
tion mode by considering mentioned different
switch types and their costs, operators can get
more benefits through an optimal reconfigura-
tion on their networks by applying this scheme.
In this case one of the simplest proceedings is to
involve the objective function with different costs
for different switch types. The result of applying
this procedure in optimization problem under
study will be a massive reduction in number of
switching which consequently reduces not only
operation but also maintenance costs.
Following simple objective function describes
how to generally consider different switch types
conditions in reconfiguration problems:
F Min C w SSA SSA C w SSM
fitness A a a
a
NAS
M
=
\
)
. . . .
2 1
1
3 mm m
m
NMS
SSM
\
)
l
l
l
l
=
1
1

(1)
Where C C
A M
and,
NAS: Number of Automatic Switching after each
reconfiguration
NMS: Number of Manual Switching after each
reconfiguration
C
A:
Cost of operating for Automatic type switches
C
M:
Cost of operating for Manual type switches
SSA
x
: State of automatic switch in the xth con-
figuration
SSM
x
: State of manual switch in the xth con-
figuration
In this definition, SSA
x
and SSM
x
get only
two values in each configuration; one or zero,
where represent state of close or open respectively.
Moreover, in above objective function w
1
and w
2

represent weighing factors refer to automatic and
manual switches respectively. By this definition
the operator can create an optimal balance between
different costs of switches operation regarding
to the existing conditions and the costs. Note that
by omit these factors from proposed objective
function, two switch types will get same share in
reducing the operating costs. Mentioned objective
function can be used simultaneously with other
purposes such as power loss reduction to form a
multi-objective function.
Distribution Feeder Reconfiguration
Considering Time Varying Loads
Access to load data is one of the most important
steps in planning and operating of electrical dis-
tribution systems (Hadian & Haghifam, 2009).
Different type of loads in this system consisting
industrial, commercial and residential can vary
with time over a wide rang. This stochastic nature
of loads will require a proper procedure to con-
sider load variations during different intervals for
both the long and medium term system planning,
and the day-ahead operation. Thus, considering
a probabilistic load flow will help the planer and
operator to execute a flexible configuration and
reconfiguration corresponding to achieve more
benefits in future.
BASIC DETERMINISTIC LOAD FLOW
The Load Flow (LF) analysis has always been one
of the most important analyses in power systems
(Papaefthymiou, 2003). The exact formulation of
LF problem concerns the determination of real
and reactive power flows in each line (branch)
of the power system. The data used to solve this
problem are the active and reactive loads at load
buses as well as the real power generation and
voltage magnitudes at the generator buses. Since,
in this formulation, single values are used as
inputs for the solution of the problem, the name
Deterministic Load Flow (DLF) has been used
274
in order to distinguish this methodology from
Probabilistic Load Flow (PLF) introduces and
discussed in future sections.
Back to DLF problem, the general DLF for-
mulation can be divided into two aspects:
1. Determination of the State Vector of the
system: the magnitude and phase angle of
the unknown bus voltages in the system are
computed.
2. Determination of the power flows on each
line of the system: the state vector is used in
this step for the determination of the active
and reactive power flows in every branch.
In this analysis, the system is modeled by a set
of N buses (nodes), which are interconnected by
transmission lines (branches). The transmission
links are represented by their nominal -equivalent
circuits. From these equivalent circuits numerical
values for the series impedances Z and conse-
quently the line-charging admittance Y will
calculate to form an nn bus admittance matrix
which includes typical element as:
Y Yi Y j Y G jB
ij j ij ij ij ij ij ij ij
= < = + = + | | | | cos | | sin

(2)
This symmetrical matrix has two other important
characteristics:
1. Y
ii
the ii-th self admittance (diagonal ele-
ment) is equal to the sum of all admittances
connected to the i-th node.
2. Y
ik
the ik-th element of Y (non-diagonal
element) is equal to the negative of the
admittance in between nodes i and k.
The voltage at a typical bus i is given by:
V
i
=|V
i
|<i
=
|Vi
|
(cosi
+
json i )

(3)
Thus, the net current injected into the network
at bus i will be given by:
I Y V Y V Y V Y V
i i i iN N in n
n
N
= + + + =
=
1 1 2 2
1
...
(4)
Let Pi and Qi denote the net real and reactive
power entering the network at bus i. Then,
P jQ V Y V
Y VV
or
V V
i i i in n
in i n in n i
n
N
i n n
= =
< +
=
*
| (
| | . | | [cos(

1
+ +
=
=

i n i in in
n
N
n
N
J G jB ) sin( )[
}
1
1

(5)
Consequently,
P V V G B
i i n in in in in
n
N
= +
=
| | . | | [ cos sin ]
1

(6)
Q V V G B
i i n in in in in
n
N
= +
=
| | . | | [ cos sin ]
1

(7)
where; N: Number of Buses,
i
n:
i

n
and
i
n:
i

n
The above Equations constitute the polar form
of the ac Power-Flow Equations. It can be seen
that there are four potentially unknown quantities
associated with each bus i; Pi, Qi, i and |Vi|. For
the solution of this problem, the general formula-
tion followed in power-flow studies is to identify
three types of buses in the system. At each bus i
two of the above quantities are specified and the
remaining two are calculated. These 3 bus types
are introduced as follow:
275
Load Buses (P-Q Nodes)
In these buses loads are connected, so P
i
, and Q
i

are known from historical record, load forecast
or measurements;
i
and |V
i
| are the unknown
quantities.
Voltage Controlled Buses
(Generator Buses, P-V Nodes)
At each bus where a generator is connected, the
active power generation can be controlled by
adjusting the prime mover, and the voltage mag-
nitude can be controlled by adjusting the generator
excitation. Therefore, P
i
and |V
i
| are known and
i
and Q
i
are the unknown quantities.
Slack Bus
The power injection at this bus is determined by
the power balance Equation of the system. In
the formulation of the power flow problem, the
power injection in this bus is not pre-specified or
scheduled. After the power flow problem has been
solved, the difference between the total power
which is injected into the system at all other buses
and the total output plus the losses, are assigned
to the slack bus. The voltage angle of this bus
serves as a reference for the angles of all other bus
voltages. The other known quantity for this bus is
the voltage magnitude |V
i
|. Obviously, there is no
requirement to include the power flow Equations
for the slack bus in the power flow problem.
With these definitions, the problem is trans-
formed to a system of (2N-N
g
-2) non-linear
Equations with (2N-N
g
-2) state variables to be
calculated, where N
g
is the number of voltage-
controlled buses in the system. Due to the
non-linearity of these Equations, power-flow
calculations usually employ iterative techniques
such as Gauss-Seidel and Newton-Raphson
procedures (Papaefthymiou, 2003).
PROBABILITY DENSITY FUNCTION
A typical daily load profile and concept of its
variations is shown in Figure 5. Considering
such these patterns for different loads of network,
a probabilistic density function (PDF) in each
load node can be obtained. It is obvious that by
increasing the numbers of segments, results are
more accurate while more time consuming and
vice versa.
Bell shaped curve shown in Figure 6 illustrates
a sample PDF plotted by Monte Carlo Method
5

(Rubinstein & Kroese, 2008). MCM is a statisti-
cal simulation uses a random sequence of numbers
to describe the statistical behavior of a variable.
In the other words, for this purpose, MCM uti-
lizes repeated trials of the deterministic load flow
technique to determine the probability distribu-
tions of the nodal powers, line flows and losses.
The hypothesis of the Monte Carlo Method,
as a mathematical technique, infers the search of
an efficient solution instead of an accurate solu-
tion (Opazo Mora, Garcia-Santander & Lopez
Parra, 2008). Having expected demand and its
standard deviation, desired PDF can be explains
as an exponential Equation as follow:
f x e
x
( )
. .
.
/
( )
=
1
2
1 2
2
2

(8)
where,
x: Variable
: Standard deviation of x
: Expected value of x
In this Equation, to form a standard distribu-
tion, expected value of x can be consider as
load value of the node with maximum occurrence
probability through studied time interval.
276
PROBABILISTIC LOAD FLOW
This section introduces a probabilistic load flow
using Point Estimate Method (Usaola, 2008).
Let z be a random variable that is function of
several independent variables with specific PDFs
and x = (x
1
,..., x
n
), z = h(x).where,
n: number of independent variables
Hence, The probability density functions of
each variable xk will be f
k
(x
k
), the joint resultant
probability density function will be f
x
(x
1
..., x
n
) and

=
( )
...
,
kn
x
k k kn
f
x x x
if ki kj
1 2
0
(9)
With these definitions,
k,n
will be the central
moment of order n of the variable x
k
, where mean
and variance are
k
and
2
k=

k,2
respectively.
for i = 1,..., m points and k = 1,..., n variables,
consider the points x
k,i
=
k
+
k,i

k
.
So that, each points will be associated to a
weight p
k,i
such that p
k i
i
m
k
n
,
=
= =

1
1 1
.
Then, z=h(x) can be expanded in multivariate
Taylor series around the point
x
= (
1,,

n
). Us-
Figure 5. A typical daily load profile
Figure 6. Sample probability density function
277
ing this series expansion, the mean of z may be
approximated as:
z x
x x
i
E z E h x h x f x
f x h
j
= = =
= +
[ ] [ ( )] ... ( ) ( )
... ( ) ( )
!
1
1
nn
j
j
i
j i i
j
x
h
x
x d

1
. ( ) .

(10)
Since,
... ( ) ...( ) ( ) x x f x dx
k k
j
k k
q
x 1 1 1 1
0 =

(11)
for every
q, ,j if { } k k m n l
m n
, , ..., 1 .

Then,

z x
j
i
j
i
n
j
j i
h
j
h
x
= +

= =

( )
!
,
1
1 2
(12)
Since ( ). ( ). . x f x dx k
k k k k
=
0
If approximate the mean
z
by

z k i k i n
k
m
i
n
p h x
= =
, ,
( , ..., , ..., )
1
1 1
, (13)
from the series expansion of the terms of (13) and
its approximation to the series (12) in a similar
way to the uni-variant case, equaling terms, will
arrive at the following system of Equations:
p
k i
i
m
k i
j
k j , , ,
=
=
1
, for j=1, , 2m-1 and k=1,,
n (14)
To this system of Equations, it can be added
p
n
k i
i
m
,
,
=
=
1
1
for each variable k (15)
Therefore, there are 2m Equations with 2m
unknowns for each variable k, forming a nonlin-
ear system.
Solving these Equations, the moment of order
j for the variable z, m
z,j
, can be then approximated
by:
m E[z ]
z,j
j
k 1
=
= =

p h x
k i
j
k i n
i
m n
, ,
( , ..., , ..., )
1
1

(16)
PROPOSED PROBABILISTIC
PROCEDURE
As discussed in previous section, the Point Esti-
mate Method can be applied to Probabilistic load
flow. Proposed method can be used for following
procedure in order to find the uncertainty of the
branch power flows and consequently total power
losses in distribution network under study:
1. Evaluate the moments of the power injec-
tions of each node.
2. Solve the system Equations (14) and (15).
3. From these values, run deterministic power
flows for the different values of (
1
,..,x
k,i
,..,
n
) to obtain power flow results for whole
distribution network under study. Note that
in this step the solution provides an ensemble
of values for the branch power flows and
also node voltages.
4. From this ensemble of values, the moments of
these variables are found using the Equation
(16).
5. In the last step, in order to follow proposed
method it is still necessary to obtain the val-
ues of the PDF, for the variables of interest
278
which is power losses. This can be made, for
instance, through Gram Charlier A series
expansion introducing in following section
(Usaola, 2008).
Gram-Charlier A Series Expansion
The Gram-Charlier type A series is a standard
measure frequency function, denoted as f(x), with
the mean
GC
= 0 and variance
2
GC
= 1 expanded
in a series of derivatives of the standard measure
normal (Gaussian) function:
G x e
x
( ) . =

1
2
2
2
(17)
so that the mean, variance, and other raw mo-
ments can be used to obtain an approximation of
the probability density function while the Gram-
Charlier type A series for this purpose yields an
approximate probability density function of power
flow. The derivatives of (17) result in a Taylor
series representation called Gram-Charlier type
A series:
f x c H x G x
j j
j
( ) ( ) ( ) =
=
0
(18)
Where c
j
are constants based on a function of
standardized moments, H
j
(x) are Hermite poly-
nomials, and G(x) is the characteristic Gaussian
function. The first two Hermite polynomials are
H
0
(x) = 1 and H
1
(x) = x. Further more, the Hermite
polynomials obtain from following Equation for
i greater than 1:
H x x x
i
k
a
i i
k i k
a
k
( ) ( ) . .
( )
.
( )
, ,
= +
1
2 1
1 2 1
1 3 5
2 33

(19)
where,
n
m
n
m n m
and
t
i
for t odd
i
for t even
\
)
'
!
!( )!
, :
, :
1
2
2
2
!!
1
1
1
1
1
+
1
1
1
1
1

The constant c
j
in the Gram-Charlier type A
series is found using standard measure as follow
while first two constants will be c
0
=1 and c
1
=0,
c
j
m
m
m
m
j
j
pdf
pdf
j
k
j k
pdf
pdf
j
=

1
1
2 2
1
2 1
2
! ( )
( ) .
( )
( )
( ) /
( ( ))
( )

= =
\
)
l
l
l
l
2 1
2
2 1 3 5
2 3
2 1
( )
, ,
.
( )
k
k
t
b
k
i
k
b
ll
l
l

(20)
Using the Hermite polynomials, H
j
(x), and
Gram-Charlier constants, c
j
, a formal expression
of f(x) is:
f x G x
m
H
m
H
m
pdf pdf
pdf
( ) ( ).[ ( ) ( )
(
( ) ( )
( )
= + +
+
1
1
2
1
1
6
1
24
2
2 2
3
3 3
4

4
2
2 4
6 3 + +
m
H
pdf
( )
) ...]

(21)
Since f(x) is found using a standard measure
random variable
GC
= 0 and
2
GC
= 1, f(x) needs
to be transformed into a function f(y) so that the
probability density function has mean
pdf
=
GC

and variance
2
pdf
=
2
GC
.
The standard measure transformation used for
the Gram-Charlier type A series is:
x
y
pdf
pdf
=

(22)
279
Where x are the standard measure variables
and y are the real variables. The new Gaussian
function now becomes,
G y e
y
pdf
pdf
( ) .
.( )
=
1
2
1
2
2
(23)
The resulting probability density function,
f(y), is the actual distribution of power flow as:
f y c H y G y
pdf
j j
j
( ) . ( ). ( ) =
=
1
0
(24)
GENETIC ALGORITHM
Genetic algorithm is one of the optimization
methods based on heredity and evolution. GA
has become an efficient tool for search and opti-
mization problems after its first introduction in
1960s by J. Holland. In This statistic searching
algorithm a population of strings (chromosomes
or the genotype of the genome), which encode
candidate solutions (called individuals or pheno-
types) to an optimization problem, evolves toward
better solutions.
Comparison of GA with Other
Conventional Optimization Methods
Some differences between GA and other conven-
tional optimization methods are as bellow:
GA works with coded parameters rather
than parameters themselves.
GA always operates on a whole population
of points rather than search from a single
point. This useful characteristic improves
the chance of reaching to global optimum
instead of local one.
GA uses ftness function for evaluation
rather than derivatives. So that, it can be
easily applied to any kind of continuous or
discrete problems.
GA use probabilistic transition operates
rather than deterministic rules.
GA does not have any limitation relate to
continuity, differentiability and, gener-
ally exist in conventional methods.
GAs Advantages and its Limitations
Following list gives some advantages of genetic
algorithm generally find in available researches:
Wide solution space.
Complex ftness landscape.
Multi solutions and multi objective
function.
Easy to discover global optimum and resis-
tant to becoming trapped in local optimum.
Easy modifcation to use in different
problems.
Handles poorly understood search spaces
easily.
Well performance in large scale
optimizations.
Very robust to diffculties in the evaluation
of objective function.
Can be employed for a wide variety of op-
timization problems. Solves multi-dimen-
sional, non-differential, non-continuous,
and even non-parametrical problems.
Very easy to understand and practically no
need advance knowledge of mathematics.
Can be easily transferred to existing simu-
lations and models.
Beside all its advantages, there are some limita-
tions which can be sense using this algorithm in
different optimization problems:
Fitness function identifcation
Premature convergence occurrence
Choosing its various parameters such as
size of population, mutation and cross
280
over rates, selection method andwhich
strongly affect on optimal solution.
Not good to identify the local optimum.
Not effective terminator
Have some troubles to fnd the exact global
optimum. It is very often when the popula-
tions have a lot of subjects.
Requires large number of ftness evalu-
ations which increases the computation
time.
Like other artifcial intelligence tech-
niques, the genetic algorithm cannot assure
constant computation times. Even more,
the difference between the shortest and the
longest optimization computation time is
much larger than with conventional gradi-
ent methods.
Basic GA Concepts
To start GA, many individual solutions are ran-
domly generated to form an initial population.
The population size depends on the nature of the
problem, but typically contains several hundreds or
even thousands of possible solutions. This initial
population and also future reprodused individuals
form crospanding chromosomes in the shape of
diffrent strings.
Chromosome
A chromosome or genotype of the genome as a
concept of GA consists a string of information
which represent characteristic of each individual.
By this definition, each gene in a chromosome
introduces only one specific individual (or status)
where each chromosome represents specific popu-
lation (or solution). Before can use these strings
there should be a way to encode and decode the
initial information and final resulted solutions
respectively.
Binary coding is one of these methods widely
used to encode and decode each gene in a chro-
mosome. While there are many other techniques
such as using real and decimal numbers, binary
coding is the most expanded presentation of a
chromosome, because it is simple and traceable.
Note that in all these methods searching process
will apply on coded strings. Figure 7 represents a
simple binary string including coded information.
In this string each gene defines the color of each
point in Figure 8 where 0 and 1 represent white
and black colors respectively. Moreover, first gene
in each section of this string determines quarter
number for each individual.
Selection
Selection is a stage in genetic algorithm in which
individuals are chosen from a population for later
reproduction. This process uses a fitness func-
tion to evaluate each individual corresponding
to desired objective. Providing fitness values for
each indivitual and sort these individuals by their
coresponding fitness valuse in next step, selection
process ends by selecting induvituals with most
fitness and forms a new population for reproduc-
tion process.
While there are many different types of selec-
tion, Roulette Wheel is the most commen and
simplest method to fulfil the selection process.
Figure 9 depics a sample roulette wheel. In this
wheel, each individul get its own portion by related
fitness persentage. Selection point will show the
Figure 7. Sample binary string
281
selected individul after each rotation. Following
this process, the individual with the most fitness
will have the most probability to select and trans-
fer to the next generation. This type of selection
method can get the form of rout instead of rotation
as shown in Figure 9.
Again, in this form selection can be assign by
choosing a random number (for instance: a value
between 0.0 and 1.0 for each iteration in above
example)
Reproduction
Reproduction is a process to generate latter gen-
eration of solutions from those picked previously
by selection process (crossover) or by changing
some genes in a string (mutation). Although there
are some other GA operators apply to generate
new populations but crossover and mutation
are most commen sulotions as well as almost
insepraTable part of a genetic algorith. Type and
implementation of these operators depends on
encoding and also differs from one problem to
another.
Crossover
For each new solution to be produced by crossover
operation, a pair or even a group of parents are
selected for breeding. While there are many dif-
ferent kinds of crossover, the most common type
is single point crossover.In single point crossover
one crossover position k is selected randomly from
[1, 2, , N-1], where N is number of variables
in each individul. In this step, variables exchange
between individuals about this point to produce
new offsprings. This process ultimately results in
the next generation population of chromosomes
that is different from privious generation. Figure 10
illustrates this process, where k is asumed to be 5.
Note that for this kind of crossover one can
also use more points. See Figure 11 for multi point
cross over at points k=5, 8 and 15.
Mutation
Mutation is another genetic operator that alters one
ore more gene values in a chromosome from its
initial state. The purpose of mutation is to maintain
diversity in the genetic algorith. This operator
allows the algorithm to avoid local optimum by
preventing the indivituals to be all exactly the
same. While Figure 12 illastrate a sample Flip-
Bit mutation, Uniform, Non-Uniform, Boundary
and Gaussian are some other known methods for
mutation process.
Figure 8. Input information to form desired binary
string for Figure 7
Figure 9. Sample rout selection method
282
Termination
Genetic process repeats until a termination con-
dition reaches. Some of common terminating
conditions are listed as follow:
Maximum number of generations; Genetic
algorithm stops after the specifed numbers
of genarations have reached.
Time limit; Genetic algorithm stops after
the specifed time has elapsed.
Fitness limit; Genetic algorithm stops if
there is no change in ftness for specifed
number of genarations.
Figure 10. Example of single point cross over
Figure 11. example of multi point cross over
283
This is while, there are many other condicutions
suported by common softwares such as MATLAB
including Function tolerance, Nonlinear con-
straint tolerance, Stall time limit and Stall
generations.
Basic GAs Flowchart
A flowchart describing main steps in basic genetic
algorithm is illustrated in Figure 13 as follow:
PROPOSED FITNESS FUNCTION
FOR GA
Choosing Genetic Algorithm solver, proceed to
existing optimization problem, a proper fitness
functions most suiTable for proposed method is
needed. Let take a look to two typical PDFs for
power loss resulted from proposed load flow with
different configuration of network (Figure 14):
As it is clear from Figure 14, the curve (a)
illustrates less overall power loss with more total
probability for loss values than the curve (b) due
to the networks configuration diversity. Ap-
proximate these curves to the closest triangles,
can calculate desired fitness function with less
computations. These approximated triangles for
above mentioned PDF curves are depicted in
Figures 15-(a) and 15-(b) respectively.
Considering these different triangle shapes for
various configurations under study, basic proposed
fitness function can be form as follow:
F
S w a
fitness
trg
i
i
i
=

\
)
l
l
l
l
=
1
1 1
1
1
2
. ( ) .
ll
l
l
l
l
(25)
Figure 12. Individual before and after Flip-Bit mutation
Figure 13. Sample of a basic Genetic Algorithm
284
Figure 14. Comparison between power loss PDFs for two different configurations of network
Figure 15. Approximated sample PDF with triangle shapes
285
In this Equation, S
trg
represents the space of
approximated triangles (equal to
a a
1 2
2
.
). Mean-
while, W
1
is considered as penalty factor for a
1

and a
2
where for the configuration with high prob-
ability and low power loss this factor makes a
high fitness and vise versa.
Several experiments and studies chose the
value of 10 for this factor which gives a normal-
ized fitness function with more accurate results.
By this fitness function, genetic algorithm will
search for optimum triangles which just represent
the least power lose values with the most total
probability of occurrence simultaneously. This is
while, in some cases regarding to the condition of
the network and decision of the operator, in order
to consider the cost of switching, the Equation (1)
can be combined easily with above mentioned
Equation to form a single fitness function.
COMPUTATION RESULTS
AND DISCUSSION
To examine the efficiency and improvement, the
suggested method is developed in MATLAB, on
a Pentium-4 PC (1.86GHz & 2GB RAM), and
is performed on a sample three-feeder network.
Original configuration of this network is illustrated
in Figure 4, while the system parameters are acces-
sible by Civanlar et al. (1988). Since it was pro-
posed for a study with fixed demands, to perform
suggested method new different load profiles are
added to each load sections which are remained
with no change all over the examination process.
These profiles are created randomly, with loads
ranging from 90% to 110% of the original loads.
As the first scenario, only the lose reduction
is considered in the objective function. Table 1
designates the numerical results of this experi-
ment. With specified load pattern, resulted power
loss probability density function is depicted in
Figure 16.
In the second scenario, two objectives of
switching reduction and reduction in loss are ap-
plied to optimization problem simultaneously.
Table 2 designates the numerical results of this
experiment. With specified load pattern, resulted
power loss probability density function is de-
picted in Figure 17. Although the Table 2 ex-
presses less reduction in loss for this scenario than
previous one, but implementing the second sce-
nario gives more benefits by reduce the number
of switching. This benefit is calculated as 405 $
includes income from total loss reduction and cost
of switching.
To calculate these values following assump-
tions are used in this study:
The cost of electricity is $6.5625 per kWh
To simplify the presentation, the network
under study includes only one switch type
Each switching cost is set to $203 which is
one-twentieth of a new switch installation
cost (Yin & Lu, 2009)
In order to compare the results of suggested
method with other references, the proposed
method is applied to two other test systems. The
first one is a small system (Baran & Wu, 1989)
widely used as a first reference in case studies.
Table 1. Numerical results for first scenario
Minimum Power
Loss
(p.u)
Maximum Power
Loss
(p.u)
Mean Power Loss
(p.u)
Total Number of
Switching
Total Loss
Reduction
Original 0.07 0.95 0.58 _ 0%
After Reconfiguration 0.01 0.73 0.35 6 35.7%
286
Table 2. Numerical results for second scenario
Minimum Power
Loss
(p.u)
Maximum Power
Loss
(p.u)
Mean Power Loss
(p.u)
Total Number of
Switching
Total Loss
Reduction
Original 0.07 0.95 0.58 - 0%
After Reconfiguration 0.03 0.81 0.41 2 25.3%
Figure 16. Power loss probability density function for first scenario (before & after reconfiguration)
Figure 17. Power loss probability density function for second scenario (before & after reconfiguration)
287
Table 3 represents the loss reductions by proposed
method in comparison to other available method
with several cases for this test network. This is
while, total number of switching and total income
are other results available in this Table to carry
out a better comparison. In this examination the
income values are evaluated based on the energy
cost of $0.50 per kWh and the switching cost of
$8.00 for each pair switching operation as are
considered by Fang, Cai & Guo (2005).
These results depict the relation between loss
reduction and number of switching. It is clear that
by increase the number of switching total power
loss decrease and vise versa. Unlike the compared
method, proposed process found optimal values
for number of switching and the power loss by
apply a balance between them. Again these con-
siderations result in a huge amount of benefit
approximately 260,000 $ annually for the network
under study.
Final comparisons relate to a real large sys-
tem. For this network, Table 4 represents the loss
reductions and total number of switching resulted
by proposed procedure in comparison to Lopez
(2004) famous method with three different cases.
CONCLUSION
In this chapter after review the background of
reconfiguration thorough major presented meth-
ods, electric distribution network, basic concepts
and its configuration are introduced. Furthermore
one becomes familiar with reconfiguration pro-
cedure and its main concepts. Two important
points proposed which are original and main
concerns in recent years. The proposed chapter
fully discussed about these points and proposed
new heuristic solutions to overcome mentioned
issues in optimal reconfiguration process. This
is while, a precise tutorial followed to introduce
basic Genetic Algorithm its parameters and main
concepts including advantages and limitations.
Finally, In order to examine the suggested
solution several experiments have been done.
Table 3. Comparison between the result of proposed method and other method
Other Methods Proposed Method
Network Ref. & Case Number Loss Reduction
(%)
Nu mb e r o f
Switching
(for 24 hours)
Tot al I n-
come
($)
Loss Reduction
(%)
Numbe r of
Switching
(for 24 hours)
Tot al In-
come
($)
Baran (1989) Fang
(2005)
Case #1 18.33 4 489.50 27.43 6 732.24
Case #2 23.21 4 624.00
Case #3 24.41 12 625.00
Case #4 27.78 16 702.00
Table 4. Comparison between the result of proposed method and other method (Real network)
Network/Method Compared Method Proposed Method
Real System With 917
Nodes
Case Loss Reduction
(%)
Number of Switching
(for 24 hours)
Loss Reduction
(%)
Number of Switching
#1 12.52 159 12.26 10
#2 10.86 9
#3 10.28 9
288
Moreover, some proposed scenarios are compared
with other methods applied in reliable articles.
Results depict the relation between loss re-
duction and number of switching. This is while;
the proposed method found optimal values for
number of switching and the power loss by apply
a balance between them.
The obtained results in this essay generally des-
ignates its efficiency, while further improvements
such as solutions to reduce the computation time
will result in more optimal reconfiguration of
electric distribution networks. One can consider
more improved GA or other evolutionary methods
to over come such this problem in future works.
Considering different energy prices for different
time intervals in power markets is another sug-
gestion in order to improve proposed method.
More over, presents of Distributed Generations
(DGs) which are almost new concepts in electric
distribution networks, can affect on proposed
optimization problem.
To follow an optimal configuration or recon-
figuration process of these networks it is necessary
to provide some models which are flexible with
presence of DGs and their conditions. This can
also be achieved by adding proper constraints to
proposed objective function.
Another direction to future researches relates
to implementing reconfiguration process for active
power loss reduction simultaneously by capaci-
tor placement in electrical distribution networks.
Further more; one can follow suggested procedure
with more objectives such as networks reliability
or voltage profile improvement.
REFERENCES
Ahuja, A., Das, S., & Pahwa, A. (2007). An AIS-
ACO hybrid approach for multi-objective distribu-
tion system reconfiguration. IEEE Transactions
TPWRS.2007.901286
Baran, M. E., & Wu, F. F. (1989). Network recon-
figuration in distribution systems for loss reduction
and load balancing. IEEE Transactions on Power
Delivery, 4, 14011407. doi:10.1109/61.25627
Chen, C. S., & Cho, M. Y. (1993). Energy loss
reduction by critical switches. IEEE Trans-
actions on Power Delivery, 8, 12461253.
doi:10.1109/61.252650
Civanlar, S., Grainger, J. J., Yin, H., & Lee, S. S.
H. (1988). Distribution feeder reconfiguration
for loss reduction. IEEE Transactions on Power
Delivery, 3, 12171223. doi:10.1109/61.193906
Fang, X., Cai, Z., & Guo, Z. (2005). Dynamic
network reconfiguration using time interval based
strategy and improved moment method. IEEE
Power and Energy Society General Meeting, 1,
271-277.
Hadian, A., Haghifam, M. R., Zohrevand, J., &
Akhavan-Rezai, E. (2009). Probabilistic approach
for renewable DG placement in distribution sys-
tems with uncertain and time varying loads. IEEE
Power and Energy Society General Meeting: Vol.
1. (pp. 1-8).
Huang, K. Y., & Chin, H. C. (2002). Distribution
feeder energy conservation by using heuristic
fuzzy approach. Electrical Power and Energy
Systems, 24, 439445. doi:10.1016/S0142-
0615(01)00056-4
Kashem, M. A., Ganapathy, V., & Jasmon, G.
B. (1999). Network reconfiguration for load
balancing in distribution networks. IEEE Trans-
mission and Distribution Conference: Vol. 146.
(pp. 563-567).
289
Liu, C. C., Lee, S. J., & Venkata, S. S. (1989). Loss
minimization for distribution feeders: Optimality
and algorithms. IEEE Transactions on Power
Delivery, 4, 12811289. doi:10.1109/61.25615
Lopez, E., Opazo, H., Garcia, L., & Bastard, P.
(2004). Minimal loss reconfiguration based on
dynamic programming approach: Application
to real systems. IEEE Transactions on Power
Systems, 19, 693704.
Mendoza, J. E., Villaleiva, L. A., Castro, N. A., &
Lopez, E. A. (2009). Multi-objective evolutionary
algorithms for decision making in reconfigura-
tion problems applied to the electric distribution
networks. Studies in Informatics and Control,
18, 325336.
Merlin, A., & Back, H. (1975). Search for a
minimal-loss spanning tree configuration in
an urban power distribution system. 5th Power
System Computation Conference (PSCC): Vol.
1. (pp. 1-18).
Milani, A. E., & Haghifam, M. R. (2010). A heuris-
tic approach for multi objective distribution feeder
reconfiguration: Using fuzzy sets in normalization
of objective functions. [IJAEC]. International
Journal of Applied Evolutionary Computation,
1(1), 6073. doi:10.4018/jaec.2010040103
Nara, K., Shiose, A., Kitagawa, M., & Ishihara,
T. (1992). Implementation of a genetic algorithm
for distribution systems loss minimum re-config-
uration. IEEE Transactions on Power Systems, 7,
10441051. doi:10.1109/59.207317
Opazo Mora, H., Garcia-Santander, L., & Lopez
Parra, E. (2008). Minimal loss reconfiguration
considering random load: Applications to real
networks. Ingeniare Journal of Engineering,
1(16), 264272.
Papaefthymiou, G. (2003). Research on the
uncertainty in electrical energy generation in
distribution systems with high penetration from
distributed and renewable energy sources genera-
tion. Netherlands: Delft University of Technology
and Eindhoven University of Technology.
Rubinstein, R. Y., & Kroese, D. P. (2008). Simu-
lation and the Monte Carlo method. Hoboken,
NJ: Wiley.
Rugthaicharoencheep, N., & Sirisumrannukul, S.
(2009). Optimal feeder reconfiguration with dis-
tributed generators in distribution system by fuzzy
multi objective and tabu search. International
Conference on Sustainable Power Generation
and Supply, 1, 1-7.
Sarfi, R., Salama, M., & Chikhani, Y. (1996).
Distribution system reconfiguration for loss reduc-
tion: an algorithm based on network partitioning
theory. IEEE Transactions on Power Systems, 11,
504510. doi:10.1109/59.486140
Shirmohammadi, D., & Hong, H. W. (1989).
Reconfiguration for electric distribution net-
works for resistive line loss reduction. IEEE
Transactions on Power Delivery, 4, 14921498.
doi:10.1109/61.25637
Taylor, T., & Lubkeman, D. (1990). Imple-
mentation of heuristic search strategies for
distribution feeder reconfiguration. IEEE
doi:10.1109/61.107279
Usaola, J. (2008). Probabilistic load flow with
wind production uncertainty using cumulants
and cornish fisher expansion. Power Systems
Computation Conference (pp. 1-8).
Wagner, T., Chikhani, A., & Hackman, R. (1991).
Feeder reconfiguration for loss reduction: An
application of distribution automation. IEEE
doi:10.1109/61.97741
290
Yin, S. A., & Lu, C. N. (2009). Distribution
feeder scheduling considering variable load
profile and outage costs. IEEE Transactions on
Power Systems, 24, 652660. doi:10.1109/TP-
WRS.2009.2016300
ADDITIONAL READING
Ah King, R. T. F., Radha, B., & Rughooputh, H.
C. S. (2003). A real-parameter genetic algorithm
for optimal network reconfiguration. International
Conference on Industrial Technology: Vol. 2. (pp.
237-288).
Allan, R. N., Leite da Silva, P. H., & Burchett,
R. C. (1981). Evaluation methods and accuracy
in probabilistic load flow solutions. IEEE Trans-
actions on Power Systems, 100, 25392546.
doi:10.1109/TPAS.1981.316721
Arrowsmith, D., Bernardo, M., & Sorrentino, F.
(2005). Effects of variations of load distribution
on network performance. IEEE International
Symposium on Circuits and Systems (ISCAS):
Vol. 4. (pp. 3773-3776).
Bueno, E. A., Lyra, C., & Cavellucci, C. (2004).
Distribution network reconfiguration for loss
reduction with variable demands. IEEE Transmis-
sion and Distribution Conference and Exposition:
Vol. 1. (pp. 384-389).
Chen, P., Chen, Z., & Bak-Jensen, B. (2008).
Probabilistic load flow: A review. Third Interna-
tional Conference on Electric Utility Deregula-
tion and Restructuring and Power Technologies
(DRPT): Vol. 1. (pp. 1586-1591).
Dickert, J., Hable, M., & Schegner, P. (2009).
Energy loss estimation in distribution networks
for planning purposes. IEEE PowerTech, 1, 16.
doi:10.1109/PTC.2009.5281997
Golkar, M. A. (2003). A new probabilistic load-
flow method for radial distribution networks.
European Transactions on Electrical Power, 3,
167172. doi:10.1002/etep.4450130305
Hadian, A., Haghifam, M. R., Zohrevand, J., &
Akhavan-Rezai, E. (2009). Probabilistic approach
for renewable DG placement in distribution sys-
tems with uncertain and time varying loads. IEEE
Power & Energy Society General Meeting (PES):
Vol. 1. (pp. 1-8).
Hines, W. W., Montgomery, D. C., Goldsman, D.
M., & Borror, C. M. (2003). Probability and Statis-
tics in Engineering. New York: Wiley Publishing.
Huang, Y. C. (2002). Enhanced genetic algorithm-
based fuzzy multi-objective approach to distribu-
tion network reconfiguration. IEEE Proceedings-
Generation. Transmission and Distribution, 149,
615620. doi:10.1049/ip-gtd:20020512
Huo, L., Yin, J., Yu, Y., & Zhang, L. (2008). Dis-
tribution network reconfiguration based on load
forecasting. International Conference on Intel-
ligent Computation Technology and Automation
(ICICTA): Vol. 1. (pp. 1039-1043).
Jianzhong, W., & Yixin, Y. (2003). Global optimi-
zation algorithm to time varying reconfiguration
for operation cost. Chinese Society of Electrical
Engineering, 11, 1317.
Meliopoulos, A. P. S., Cokkinides, G. J., & Chao,
X. Y. (1990). A new probabilistic power flow
analysis method. IEEE Transactions on Power
Systems, 5, 182190. doi:10.1109/59.49104
Morales, J. M., & Perez-Ruiz, J. (2007). Point
estimate schemes to solve the probabilistic power
flow. IEEE Transactions on Power Systems, 22,
15941601. doi:10.1109/TPWRS.2007.907515
291
Neimane, V. (2001). Distribution Network Plan-
ning Based on Statistical Load Modeling Applying
Genetic Algorithms and Monte-Carlo Simula-
tions. IEEE Porto Power Tech Conference: Vol.
3. (pp. 5-9).
Papaefthymiou, G., Schavemaker, P. H., van der
Sluis, L., Kling, W. L., Kurowicka, D., & Cooke,
R. M. (2005). Integration of stochastic generation
in power systems. 15th Power Systems Computa-
tion Conference: Vol. 1. (pp. 773-781).
Queiroz, L. M. O., & Lyra, C. (2009). Adaptive
hybrid genetic algorithm for technical loss re-
duction in distribution networks under variable
demands. IEEE Transactions on Power Systems,
24, 445453. doi:10.1109/TPWRS.2008.2009488
Savier, J. S., & Das, D. (2007). Impact of Net-
work Reconfiguration on Loss Allocation of
Radial Distribution Systems. IEEE Transactions
on Power Delivery, 22, 24732480. doi:10.1109/
TPWRD.2007.905370
Shouxiang, W., & Chengshan, W. (2007). Mod-
ern distribution system analysis. Beijing: China
Higher Education Press.
Su, C. L. (2005). Probabilistic load-flow com-
putation using point estimate method. IEEE
Transactions on Power Systems, 20, 18431851.
doi:10.1109/TPWRS.2005.857921
Su, C. L. (2005). Cumulant-based probabilistic
optimal power flow (P-OPF) with Gaussian
and gamma distributions. IEEE Transactions
TPWRS.2005.846184
Su, C. L. (2005). Distribution probabilistic load
flow solution considering network reconfiguration
and voltage control devices. 15th Power Systems
Computation Conference: Vol. 1. (pp. 773-781).
Tanabe, T., Funabashi, T., Nara, K., Mishima, Y.,
& Yokoyama, R. (2008). A loss minimum recon-
figuration algorithm of distribution systems under
three-phase unbalanced condition. IEEE Power
and Energy Society General Meeting - Conver-
sion and Delivery of Electrical Energy in the 21st
Century: Vol. 1. (pp. 1-4).
Verbic, G., & Canizares, C. A. (2006). Probabilistic
optimal power flow in electricity markets based on
a two-point estimate method. IEEE Transactions
TPWRS.2006.881146
Verbic, G., Schellenberg, A., Rosehart, W., &
Canizares, C. A. (2006). Probabilistic optimal
power flow applications to electricity markets.
International Conference on Probabilistic Meth-
ods Applied to Power Systems: Vol. 1. (pp. 1-6).
Wanp, S., & Wang, C. (2001). Disbibution network
reconfiguration based on interval algorithm to
maximize confidence of energy loss reduction.
Journal of Automation of Electric Power Systems,
1(25), 2731.
Yu, H., Chung, C. Y., Wong, K. P., Lee, H. W.,
& Zhang, J. H. (2009). Probabilistic load flow
evaluation with hybrid latin hypercube sampling
and cholesky decomposition. IEEE Transactions
TPWRS.2009.2016589
Yu, Y., Qiu, W., & Liu, R. (2001). Distribution
system reconfiguration based on heuristic al-
gorithm and genetic algorithm. Power System
Technology, 11, 1922.
Zhu, J. (2009). Optimal reconfiguration of electri-
cal distribution network using the refined genetic
algorithm. In M.E. El-Hawary, Optimization of
Power System Operation. Hoboken, New Jersey:
Wiley Publishing.
292
Electric Distribution Network: An electric
distribution network is a system of cables and
equipments which deliver electric power to the
end users.
Configuration: The process of design and
implement of an electric distribution system.
Reconfiguration: Switching and changing the
status of network for better operation.
Optimization Problem: An optimization
problem is the problem of finding the best solu-
tion from all feasible solutions.
Switch Types: Maneuver switches in an elec-
tric distribution network which can be manual
type or operated automatically.
Load Flow: An important tool involving
numerical analysis applied to power systems to
calculate its parameters such as voltage angles
and magnitudes.
Probability Density Function (PDF): A
function that describes the relative chance for a
random variable to occur at a given point in the
observation space.
ENDNOTES
1
1975 to present
2
SSOM
3
N. C. switches
4
N. O. switches
5
Monte Carlo Method (MCM), 1940s. Gen-
erally MCMs are a class of computational
algorithms that rely on repeated random
sampling - wikipedia.org.
293
Chapter 10
DOI: 10.4018/978-1-61350-138-2.ch010
A. G. Tikdari
H. Bevrani
G. Ledwich
Queensland University of Technology, Australia
A Descriptive Approach
for Power System Stability
and Security Assessment
ABSTRACT
Power system dynamic analysis and security assessment are becoming more signifcant today due to
increases in size and complexity from restructuring, emerging new uncertainties, integration of renewable
energy sources, distributed generation, and micro grids. Precise modeling of all contributed elements/
devices, understanding interactions in detail, and observing hidden dynamics using existing analysis
tools/theorems are diffcult, and even impossible.
In this chapter, the power system is considered as a continuum and the propagated electromechanical
waves initiated by faults and other random events are studied to provide a new scheme for stability
investigation of a large dimensional system. For this purpose, the measured electrical indices (such as
rotor angle and bus voltage) following a fault in different points among the network are used, and the
behavior of the propagated waves through the lines, nodes, and buses is analyzed. The impact of weak
transmission links on a progressive electromechanical wave using energy function concept is addressed.
It is also emphasized that determining severity of a disturbance/contingency accurately, without consid-
ering the related electromechanical waves, hidden dynamics, and their properties is not secure enough.
Considering these phenomena takes heavy and time consuming calculation, which is not suitable for
online stability assessment problems. However, using a continuum model for a power system reduces
the burden of complex calculations.
294
A Descriptive Approach for Power System Stability and Security Assessment
INTRODUCTION
Power system angle instability following loss of
synchronism of the generators can be considered
as a fast instability phenomena (Kundur, 1994,
and Bevrani 2009). Detecting of this phenomena
and performing adequate emergency actions are
important issues to maintain the power system
stability. When a disturbance takes place in a
power system, the rotor angle of the generators
near the occurred fault deviates from their base
frames. This deviation propagates through the
power system. So, for large disturbances, this
may lead to the catastrophic outages, and finally
blackout.
Because of deregulation and economical
issues in power system management and fast
growing of electrical power consumers, the
transmission lines are usually working close to
their stability limits. In other words, nowadays
power systems are working under stress. One
of problems that a stressed power system is
encountered with, is the prediction of instability
location. Occurring of a disturbance in a point of
large scale power system may initiate instabil-
ity in another point of the grid. In this chapter,
it is shown that the reason of this problem can
be found out from electrical wave propagation
phenomenon and identifying the weak links of
the system.
There are many research reports about wave
traveling and propagation phenomenon (Thorp,
1998, Parashar, 2004, Phadke, 2008, Wagner,
1950, Sluis, 2001, Tsai, 2005 and Bank, 2007).
Electromechanical wave propagation issue is
well-discussed in (Thorp, 1998, Parashar, 2004,
Phadke, 2008). They modeled a power system
as a continuum. An online method for detection
of loss of synchronism using an energy function
criterion is developed in (Padiyar, 2006). Several
methodologies are also available to solve the
same problem, using artificial neural networks,
heuristic algorithms and intelligent schemes.
Emerging Phasor Measurement Units (PMUs)
makes it possible to easily monitor the angle
wave propagation throughout the power systems.
They use Global Positioning Systems (GPS) to
provide a synchronous time throughout the power
systems which may be distributed through the
continents. A methodology for islanding and
identifying the weak links of a complex power
system is proposed in (You, 2006 and Yang,
2006); they used the slow coherency approach.
In this chapter, first as a background, the dif-
ferent types of traveling waves in power systems
and islanding control are addressed in Section 2.
In Section 3, the problem at hand and the system
modeling are described. In Sections 4 and 5, the
islanding problem is re-analyzed concerning the
wave propagation problem, and the proposed
descriptive approach for power system stability
assessment is explained. Some simulation results
on two ring power system (64-bus and 200-bus),
a 400-bus meshed system, and the 24-bus IEEE
Reliability Test System (RTS-79) are presented.
Further research directions are addressed in Sec-
tion 6, and the chapter is concluded in Section 7.
Finally, a new power system emergency control framework based on descriptive study of electrical mea-
surements and electromechanical wave propagation in large electric power systems is introduced. Since,
fast and accurate detection of instability is essential in initiating certain emergency control measures,
the proposed methodology could be also useful to detect the contingency condition and performing the
well-known islanding and load shedding techniques. The chapter is supplemented by some illustrative
nonlinear simulations on large scale test systems.
295
BACKGROUND
Wave Propagation
Wave propagation is a phenomenon appeared in
many studying fields. In civil engineering, for
example, it is used to study on propagation of
earthquake waves through the structures com-
ponents. Propagation of electromagnetic waves
in different environments is an example of this
phenomenon in physics and communication.
When you move one end of a rope which another
end is fixed, you can see a one-dimensional (1-D)
wave that propagates through the rope and then
back propagates. Indeed wherever there is a flex-
ible material affected by a perturbation, the wave
propagation phenomenon will appear.
Observing the propagation of a wave through
an environment makes it so useful for analyzing
the related problems. Also wave propagation
phenomenon is an interesting problem in study-
ing of large scale power grids (Thorp, 1998,
Parashar, 2004, Phadke, 2008, Wagner, 1950). A
large electric power system can be introduced as
a very complex and nonlinear dynamic system
which is always subjected by small/large distur-
bances. In many articles, it has been shown that
the disturbances propagate through the system
like a wave (Thorp, 1998, Parashar, 2004, Phadke,
2008, Wagner, 1950). Propagation of a disturbance
through a power system may threat the stabil-
ity if becomes large or if it passes through the
weak lines/elements. Furthermore, it may cause
undesirable tripping of some protection devices
(Phadke, 2008). To study about electromechanical
transients usually a large detailed model of the
whole power system is needed. Preparing such
system models and solving the heavy equations
are so time-consuming; and often it will not give
us a suitable sense on the global power system
and its phenomenon such as electromechanical
wave propagation.
Studying a system based on wave propagation
analysis is needed to observe an overall view of the
system. Therefore, a descriptive approach could
be considered as a good alternative to analyze a
power system based on wave propagation.
All performed studies on the wave propaga-
tion over the years, can be categorized into two
groups. The first group considers voltage wave
propagation (Wagner, 1950, Sluis, 2001). The
voltage-type wave propagation may be created
following a switching or lighting. However, the
second group is electromechanical wave propa-
gation which is produced when a rotor angle of
a synchronous generator deviates from its base
frame following another disturbance. These phe-
nomena are introduced in the below subsections.
Voltage Wave Propagation
Studying about power system stability can be di-
vided into two general groups: static stability and
transient stability. In static stability study, the goal
is to verify the system stability when encountered
with low-amplitude slow disturbances. However,
in the transient stability problems, a large and
suddenly disturbance is occurred. The dynamic
stability is an improved form of static stability
in the case of low amplitude disturbances with
longer life time (Kundur, 1994).
In static stability study the wave length is
about 6000 Km for 50 HZ (Sluis, 2001). There-
fore, in these studies a lumped model of a line is
adequate. But for the transient stability problems,
that the higher frequencies are exist, the traveling
waves cannot be ignored. In other word, in this
situation when a signal exists at one end of a line,
there is no guarantee to appear it at another end
of line, at the same moment. Indeed, the wave
travels through the lines with a delay. This delay
is actually because of charging the capacitance
and inductance elements of the line (Sluis, 2001).
In the lumped -model, only two capacitors and
one inductor are used; and in the T-model there
are only two inductors and one capacitor. So,
the voltage variations are immediately sensed at
another end of line. But as shown in Figure 1, a
296
distributed model contains a lot of elements that
each element includes one inductor and one ca-
pacitor. When a voltage source is applied at one
end of the line, the first capacitor is immediately
charged; but because of first inductor, the second
capacitor charges when the inductor is charged,
and it makes a delay. This delay also exists for
the next elements.
One of most important parameters in this issue
is the characteristic impedance which is calcu-
lated by Equation 1:
Z
L
C
= (1)
where, L and C are defined as line inductance and
capacitance values per length of line, respectively.
When a voltage wave passes through a point in
which the characteristic impedance is changed, the
magnitude of reflected wave and the magnitude
of the wave that lets trough are dependent on the
value of characteristic impedances of two lines
(for example when an overhead transmission line
is connected to an underground cable).
The wave propagation velocity is another
important parameter that can be calculated with
Equation 2.
v
LC
=
1
(2)
Rotor Angle Wave Propagation
As discussed, the angle wave propagation which
is called electromechanical wave propagation
starts when a generator rotor deviates from its
base frame. Then, it propagates throughout the
power system. The velocity of angle wave propa-
gation is slower than voltage wave propagation
(Thorp, 1998 and Phadke, 2008). As has argued
in (Phadke, 2008), the nature of this phenomenon
is not completely clear. It seems to be the result
of local inertias. As the waves may lead to loss of
synchronism for under stressed power systems,
studying of them is one of the interested issues
in power systems stability analysis and security
assessment.
Formulation of wave motion as nonlinear
partial differential equations in a two-dimensional
surface is introduced in (Thorp, 1998). This for-
mulation is in form of wave equations. The swing
equation (Equation 3) of a generating unit can be
expressed as follows:
2H
D P P P
a m e

.. .
+ = = (3)
where, H, , D, , Pe
,
and Pm

are inertia constant,
angular speed, damping factor, rotor angle, elec-
trical output power, and mechanical input power,
respectively.
The above swing equation is introduced as a
second-order hyperbolic wave equation in (Thorp,
1998). Using of these equations lets us to intro-
duce a power system as a continuum system. The
Figure 1. Distributed model of a transmission line
297
methodology of extracting the continuum model
for a two-dimensional (2-D) power system is
well-established in (Parashar, 2004). For a mesh
network power system, which the generators and
loads are located at distributed discrete points,
introducing a 2-D continuum model needs to in-
troduce the system parameters in form of smooth
functions. To distribute the parameters, a Gaussian
filter which is the most common smoothing tool
can be suggested (Parashar, 2004).
Islanding Control
In fact, the angle instability is the loss of synchro-
nism of the system synchronous generators. Angle
instability is usually started when the synchronism
between two generators which are located at two
sides of a line is loosed. In other words, the angles
of two generators are separated, may because of
overloading a link which connects two generators,
directly or indirectly. This link may be a weak link
or a link which encountered a suddenly overload
due to some connections in the network.
There are many technical reports representing
various methodologies for detection/prediction the
angle instability situation. For example, using of
energy function concept is suggested in (Padiyar,
2006). Based on this concept, for the purpose
of stability, a system should be able to convert
whole of kinetic energy achieved throughout a
disturbance into potential energy. On the other
words, when in a stable swing, the potential energy
reaches its maximum value, the kinetic energy
should be zero.
In complex power systems, there are usu-
ally more than one inter-islands connection
line. Therefore, to validate the system stability,
maintaining only one line is not adequate. Based
on the given idea in (Padiyar, 2006 and Wang,
2004), the change of power in all lines belong-
ing to a cutest called critical cutest determines
the stability or instability situation. As argued
in (Padiyar, 2006), the potential energy, under
certain assumption, can be considered as sum of
energies of the lines belonging to the cutest; and
also, the kinetic energy is a function of voltage
angle gradient of the cutest.
When a power system is becoming unstable at
the location of a cutest, it means that the angle of
a portion of power system that connected to the
system (with the critical cutest) moves in opposite
direction of angle motion of the other parts of
system. In this situation, the islanding control is
the most effective control action. Therefore, the
unstable part of system is separated by tripping
the lines in the critical cutest. The detection of
critical cutest/weak link is the first step, and the
islanding control must be executed. Following the
islanding process, we are in face of some islands
with excess load or generation. Load-generation
imbalance in an isolated island leads the island
into another form of instability, but with a slower
dynamic than angle instability. Therefore, the
island formation may not be the final stabilizing
step. However, it can be considered as a way to
arrest the fast angle instability. Following the
islanding, the other emergency control actions
such as load shedding and generation tripping are
usually needed (Tikdari, 2009 and Bevrani 2010).
Slow coherency is a suitable and usual meth-
odology for islanding formation (You, 2004). Fol-
lowing a fault, there are some groups of generators
based on their angle variations. The angle changes
in a group are similar, but the different groups of
generators behave in opposite directions. This is
the concept of slow coherency theory. One may
introduce every coherent group of generators as
an island. The lines that connect the islands may
also provide the power system weak links (You,
2004, Yang, 2006, Wang, 2004).
Slow coherency is known as a good example
for application of singular perturbation theory in
power systems (You, 2004, Yang, 2006, Wang,
2004, Ourari, 2003 and Sowa, 2004). Using this
method, the groups of generators with coherent
angle behaviors can be determined. In transient
stability issues, the slow coherency can be also
used to find out the equivalent dynamic model of
298
a power system (Ourari, 2003 and Sowa, 2004).
In the slow coherency, it is assumed that state
variables of an n-degree performance can be
divided into two categories: r slowest states and
n-r fast states. The r slowest states represent the
groups of generators with slowest coherency. In
slow coherency-based islanding, two assump-
tions can be considered to simplify the process
of finding coherent groups of generators: i) the
groups of coherent generators are independent to
size of disturbance; and ii) the groups of coherent
generators are independent to the modeling ac-
curacy of the generator unit. These assumptions
are important to perform the continuum model
of power system, which is main interest in the
present chapter.
PROBLEM ILLUSTRATION
AND MODELING
In this section, the problem that we are attempting
to solve is described. Suppose a high stressed large
scale power system which its elements are work-
ing near their stability margins. When a system
is explored near the stability limits, disturbances
may easily force the system to a cascading failure
and even blackout. The goal is to widely monitor
system in order to effective protection against
large disturbances. For this purpose, an online
stability/security assessment program is needed.
On the other hand, following a contingency,
the voltage/angle deviations propagate through the
power system. These traveling waves may pass
through the high stressed elements and trigger an
instability event. By using of wave propagation
phenomenon, a descriptive approach for power
system stability assessment can be proposed. In
a power system, the operators and engineers will
be able to use this descriptive tool to track the
system dynamic behaviors, to validate the system
performance against likelihood events, and to
predict the next stable point.
The methodology will be described in the next
sections. As it can be seen, the main aim is looking
for an approach to help the system operators to
immediate identify the proper emergency action.
Following dangerous events, the system operators
have some choices as emergency control actions.
Sometimes a load shedding scheme or a genera-
tion rescheduling action can maintain the system
stability. However, for very large disturbances, it
may be needed to separate the system into two or
more islands and control them separately.
Solutions and Recommendations
Based on power-voltage curves, there is a
maximum value for transmission line loadability
(Kundur, 1994 and Miller, 1982). A transmis-
sion line may encounter overloading following
a contingency. But, it is important to know that
what happens when an overloading is appeared.
The relation between angle and power across
a line is shown in Equation 4 (Kundur, 1994 and
Miller, 1982).
P
E E
Z Sin
Sin
s r
=
0

(4)
where, is the length of transmission line in
Radian; Es

and Er

are the voltage magnitudes
at two ends of the line; Z0

is the characteristic
impedance; P is the active power passes through
the line; and is the angle difference between Es
a
nd Er.

The is also known as transmission and
load angle (Miller, 1982).
Based on Equation 4, if a line active power
increases, the angle across that link also increased.
In this situation, if two synchronous machines are
connected at two ends of the line, increasing of
which is the difference between the positions
of rotors of the machine may lead to loss of their
synchronism.
Therefore, there is also a maximum value for
load angle of a line. Equation 4 shows that this
299
maximum value is 90. However, based on experi-
ences it is not safe to let to increase more than
30 if the transmission line is not compensated
(Miller, 1982).
In transient stability studies, there are some
approaches for the system stability assessment,
following a contingency. For example Equal area
criterion is one of them which is only used for one
machine connected to an infinite bus or for two-
machines system (Kundur, 1994). (Padiyar, 2006)
used energy function criterion for online stability
assessment of a large power system. Major of
proposed criteria say that a system can stay stable
if the generators can release the complete energy
value that they obtained following a contingency.
Because of oscillations and uncertainty in a real
power system, the proposed algorithms for online
stability assessment that uses instance values
of power, angle and the other indices may need
many considerations. While, in these situations it
is needed to know the behavior of the indices that
propagates through the power system like a wave.
System Modeling
A Ring System
If the number of elements in a power system is large
with a set of distributed generators parameters, the
discrete model simulation results are also close
to the continuum model (Thorp, 1998). Equation
5 and Equation 6 can be used for modeling of an
N-bus ring power system: (see Box 1)
The value of P
k
m
can be achieved from steady
states values of angles (k

= k
=
0). The steady
state values of angles for an N-bus ring system
can be calculated as (Equation 6):

k
eq
k
N
=
2
(6)
Now, consider the 64-bus ring power system
given in (Thorp, 1998). A Gaussian disturbance
around line 15-16 can be implemented as follows
(Equation 7).

k k
eq k
e = +

1
2
0 1 15 5
2
. ( . )
(7)
The simulation results are illustrated in Figure
2. This is a regeneration of the given example in
(Thorp, 1998). In Figure 2, the wave propagation
is illustrated versus time. However in Figure 3,
the angle of wave is plotted versus bus number for
different time slots. Figure 3b shows the normal-
ized version of Figure 3a. As can be seen, when
a deviation appears in rotor angle of a generator,
it propagates and in the traveling path, it may
encounter a weak link and may lead to a cascad-
ing failure.
2-Dimensional System
Now, consider a meshed power system with a
configuration shown in Figure 4, in the Cartesian

k k m
k
k k k k k k
D P b
.. .
[ cos( ) cos( )] [sin( ) + =
+
2
1 1 1
++
=
+ =
+
sin( )];
, , ...,
[ cos(
.. .

k k
m N
k N
D P
1
1 1
1
1
2 3 1
2
for
)) cos( )] [sin( ) sin( )]
[ co
.. .
+
+ =

1 2 1 1 2
2
b
D P
N
N N m
N
ss( ) cos( )] [sin( ) sin( )]
N N N N N N
b +
1 1 1 1
(5)
Box 1.
300
Figure 3. Wave propagation; a) angle versus bus number for different time slots, and b) normalized plot
Figure 2. Electromechanical wave propagation on 64-bus ring system
301
characteristic. Each point represents a bus and
each connection represents a transmission line.
For simplicity, assume that each bus consist of
a generator or a load, or nothing. To obtain the
necessary equations, assume that a generator with a
mechanical power of P
A
m
is connected to the point
A (Figure 4). Considering the connections, P
A
e
is
the sum of electrical powers transferred from bus
A to its neighbor buses 1, 2, 3, and 4. Therefore,
P
A
e
can be calculated as Equation 8:
P P P
e
A
e
Ak
L
A
k
= +
=
1
4
(8)
where (Equation 9),
P
V V
x
k
e
Ak
A k
Ak
Ak
= = sin , , , , 1 2 3 4 (9)
Now, the swing equations can be easily written
for each point (Bevrani and Tikdari, 2010). Us-
ing above descriptions, an example is presented
in Figure 5 which illustrates the propagation of
angle wave through a 2-D power system. At t =
0, a disturbance occurs on the middle of the net-
work then, it propagates throughout the system.
The system situation following the disturbance is
shown in a few time slots.
PROPOSED METHODOLOGY
This section describes the process of using island-
ing formation, performing a continuum model for
system stability assessment, and predicting suit-
able emergency actions following a contingency.
The overall view of the algorithm is demonstrated
in Figure 6.
Angle instability is a fast instability phenom-
enon. Therefore, predicting its situation and
performing suitable actions are very important.
Having a continuum model of a power system
can be helpful for predicting the trajectory of the
disturbances, by using of disturbance conditions
as initial states of the continuum model. Here, the
power system continuum model is used to provide
descriptive tool for stability analysis in emer-
gency conditions.
As introduced in the previous section, the
slow coherency theory can be used to identify
system islands. Determining the islands leads to
identify the weak links or critical cutset. Having
knowledge on system weak links/islands helps one
Figure 4. Network configuration
302
to determine most suitable connections/locations
for performing a more carefully islanding plan,
when the system needs to be separated. Here, a
slow coherency is suggested to find weak links and
islands. The weak links are used to check whether
islanding is needed or not. The weak links can be
considered as the locations should be tripped when
the system operator recommends the islanding.
It can be shown that following a contingency,
if the weak links are not collapsed and the in-
stability is also not observed in the links close
to the contingency, the system remains stable.
Therefore, the overall stability can be validated
by monitoring of just few links (i.e. the weak
links and the links near to the contingency loca-
tion). Furthermore, observing the trajectory of
bus voltage angles helps the operators to choose
a suitable islanding plan. As already mentioned,
following an islanding action, the power system
is divided into some islands with excess load/
generation. Therefore, other emergency control
actions (Bevrani and Tikdari, 2010, and Bevrani,
2009) such as load shedding and generation trip-
ping should be performed as shown in Figure 6.
Indeed, following a certain contingency, the
most important goal of the proposed algorithm is
to determine if islanding is needed or not. If the
contingency is not very dangerous and the angle
across a link does not excess from a certain value
(for example 30 for non-compensated lines),
Figure 5. Wave propagation through a network
303
islanding is not needed. However, for the higher
value of transmission angles, the other emergency
action will not be able to restore the system sta-
bility. For more clarification, assume the angle
across a link is increasing and excess 90. After
that, decreasing of active power by, for example,
load shedding could not restore the system because
the angle may track the power at low side of P
curve; so, increases, thus instability occurs. If
these circumstances to be predicted, the system
will undoubtedly need to be separated. The critical
angle values (thresholds) should be determined
based on desire level of security.
Operators and engineers can validate the power
system stability at control rooms by observing the
wave propagation at human machine interface
(HMI). Three modes may be defined for this tool:
real-time, prediction, and test modes.
In real-time mode, the real-time data which
are gathered from the network by the PMUs are
shown as a surface. The operator can see the real-
time states of whole the network. In prediction
mode, the online data are used as initial values
of a continuum model, and the system states at a
certain time value later will be predicted. In the
test mode, the operator or engineer can validate
a certain contingency based on real states of the
system. In this mode the system real-data are
used as initial values of a continuum model and
the certain test is used as a deviation from initial
state; then the post contingency condition will be
shown in HMI for a certain time interval.
SIMULATION RESULTS
A Ring System and Islanding
To illustrate the concept of a coherent group of
generators following a contingency, a 200-bus
ring system is simulated. For simplicity, it is as-
sumed that there is only one generator or one load
Figure 6. Overall framework of the proposed methodology
304
at each bus. All generators are similar with equal
amount of power, and all loads are also equal. The
number of generators (N
G
) is equal to the number
of loads (N
L
), so N
G
= N
L
= 100. The system data
is determined as follows (Equation 10):
P P H D
m L
= = = = 0 3 0 3 1 0 2 . , . , , . (10)
The system is examined under two different
configurations. In the first configuration (con-
fig-1), all generators and loads are distributed
throughout the power system by a uniform random
function. In the second configuration (config-2),
the generators and loads are distributed in a three
areas ring system, as shown in Figure 7. In this
case, all line impedances are assumed to be fixed
at 0.1 p.u., except three lines that their impedances
are Z
75-76
= 0.2. Z
130-131
= 0.15, Z
200-1
= 0.3. For both
configurations, a large disturbance (i.e. tripping
line 200-1) is applied. The angle deviations are
illustrated for config-1 and config-2 in Figure 8
(8.a and 8.b, respectively). The ring system is
opened due to occurred disturbance.
For both cases, the angle deviation behavior
across link 75-76 is illustrated in Figure 9. As
shown, for the unstable case (config-2), the angle
across link 75-76 (i.e. 7
5-76
= 75

76)
, i
s con-
tinually growing and finally this situation leads
toward separation and instability. However, for
the stable case (config-1); although the angle
across link deviates, the system is remained in a
limited boundary and moves to a constant value.
The kinetic energy, potential energy, and total
(kinetic plus potential) energy across link 75-76
are also depicted in Figure 10. As already men-
tioned, the system to be stable if it can be able to
convert all amount of its kinetic energy achieved
during a contingency into potential energy. This
simulation (Figure 10) also shows that following
mentioned fault, config-2 is going to an unstable
condition, while config-1 holds its stability.
As another example, assume a Gaussian dis-
tribution that affects the angles of a system de-
picted in Figure 7 (config-2). Post-contingency
wave propagation is illustrated in Figure 11. In
this case, the center of disturbance is bus 60;
however, it can be seen that the system is sepa-
rated at line 75-76, which is a weak link. Actu-
ally, when a disturbance reaches a weak link
through its propagation trajectory, may lead to an
unstable operating point.
Figure 11b clearly shows the behavior of wave
propagation when it reaches a weak links. For
plotting this figure, the angle variations versus
bus number for each time slot is calculated. As it
can be seen, angle wave is reflected when it
reaches to a transmission line without enough
stability margin, and the system is separated ex-
actly at the weak point.
Application to 24-Bus Test System
Here, the 24-bus reliability test system (RTS)
is used to investigate the effectiveness of the
proposed strategy. Single line diagram of RTS is
illustrated in Figure 12. The RTS with its full data
is introduced in (IEEE RTS Task Force 1979 and
1999); and the generators data are selected the
same as given typical data in (Anderson, 2003).
Here, the test system is divided into three areas.
While most of the generation is located in area
1, most of the load is located in area 3. In area 2,
load and generation are approximately the same.
Area 1 delivers its over generation into area 2
and area 3 through three tie-lines: line 16-19, line
16-14, and line 24-3.
As a serious fault, the connections between
area 1 and area 2 are loosed. Now, the angle in-
stability on link 24-3 can be considered as a good
example to examine the proposed methodology.
Assume lines 16-19 and 16-14 to be tripped at t
= 2 seconds. Following this large disturbance,
line 3-24 will be encountered with a high over-
loading problem. This over-loading is larger than
its angle stability limit. Therefore, as shown in
Figure 13, the angles of generators G1 and G15
located at two sides of the line, are separated; and
305
Figure 8. Wave propagation for: a) config-1; b) config-2
Figure 7. 200-bus ring system (config-2)
306
the angle instability phenomena is immediately
occurred. To save system in this dangerous condi-
tion, one solution is the use of islanding control
(You, 2003).
The angle deviations reduced by PMUs from
the system buses are illustrated in Figure 14. As
it is shown, following the contingency the con-
nection between area 1 and area 3 (link 3-24) is
separated as well as separation of area 2 and area
1. It demonstrates that the islanding between area
1 and area 3 at link 3-24 is a good idea to protect
the system stability. Figure 15 illustrates how an
islanding improves the voltage behavior that has
been suddenly depressed following the mentioned
event.
Figure 9. Angle across link 75 in a) config-1; b) config-2
Figure 10. Energy across link 75: a) kinetic energy for config-1; b) kinetic energy for config-2; c) po-
tential and total energy for config-1; d) potential and total energy for config-2
307
Therefore, the remaining tie-line may be
tripped when appropriate algorithms are not used
to stabilize the resulted two islands. To save an
island with excess load, an under frequency load
shedding (UFLS) algorithm, like those suggested
by the Florida Reliability Coordinating Council
(FRCC) (FRCC Automatic UFLS Program, 2009)
or (Tikdari, 2009, Bevrani & Tikdari, 2010, and
Bevrani, 2009) should be used. However, for the
islands with excess generation, some loads must
be switched on.
By using the PMUs, one can monitor the thorough
behavior of a power system. The PMUs use Global
Positioning Systems (GPS); so, they can offer a
synchronous time for all of data gathered from the
whole network. Therefore, it is possible to improve
a continuum model to a trainable one, as a future
work. The model uses the data gathered from the
system at times t and t + 1. By comparing the
output of the model and the actual values from the
PMUs at time t + 1, the error can be computed.
Figure 11. Wave propagation following a Gaussian disturbance which its center is bus 60 (the system
is separated at link 75-76): a) 2D plot, and b) 3D plot
308
Figure 13. Angles of generators: G
1
and G
15
Figure 12. 24-bus Reliability Test System (RTS)
309
By using the error value, the continuum model
can be trained and improved.
There are many researches that offer optimal is-
landing. As another future work, these approaches
may be included into the proposed methodology
to prepare a more effective approach.
In order to use wave propagation studies as
an automatic tool instead of descriptive tool by
the power system operators, some additional
algorithms may needed. Pattern recognition
methods may be used in these situations. Some
features can be extracted from the angle surface
variation, following different contingencies. They
can produce special patterns and may be used in
stability assessment problems, more effectively.
Moreover, some important research needs in
future are the updating of existing emergency
frequency control schemes for N-1 contingency,
economic assessment/analysis the frequency regu-
lation prices (considering various control strate-
gies, penetration level, and installation location of
renewable energy sources units), further study on
frequency/voltage stability using dynamic demand
control and ratios of renewable energy sources
technologies, and quantification of reserve margin
due to increasing renewable energy penetration.
Figure 15. Voltage response following disturbance (at 2 sec) and islanding plan (at 2.4 sec)
Figure 14. Bus voltage angles deviations following the contingency
310
CONCLUSION
In a power grid, when a propagation energy wave
caused by a disturbance hits a weak link, a reflec-
tion is appeared and a part of energy is transferred
across the link. Based on this basic fact, the pres-
ent chapter proposes an analytical descriptive
methodology to study the dynamical stability and
security of a large scale power system, following
serious disturbances. The possibility of simulta-
neous remote measurement of bus voltage, rotor
angle and system frequency through synchronized
PMUs and intelligent electronic devices empha-
sizes the significance of the proposed descriptive
scheme for real-world complex power systems.
In this chapter, first an overview on the electro-
mechanical waves in power systems, mathematical
formulation and their interesting properties with
a brief literature review is presented. The ampli-
fication of a propagated wave due to reflections
or in combination with waves initiated from other
disturbances is studied and it is shown how the
occurrence of a large disturbance in a power system
may lead to an uncontrolled tripping of generators
and cascading failures, and may finally result in a
blackout; if proper actions are not taken.
Finally, based on given descriptive study of
electrical measurements and electromechanical
wave propagation in large electric power systems,
a power system emergency control scheme is in-
troduced to detect possible plans. The chapter is
supplemented by several simulations on standard
24-bus, 200-bus, and 400-bus test systems.
ACKNOWLEDGMENT
This work is financially supported by Research
Office at University of Kurdistan.
REFERENCES
Anderson, P. M., & Fouad, A. A. (2003). Power
system control and stability. USA: IEEE Press.
Bank, J. N., Gardner, R. M., Tsai, S. J. S., Kook,
K. S., & Liu, Y. (2007). Visualization of wide-area
frequency measurement information. In Proceed-
ing of IEEE PES General Meeting, Tampa, FL.
Bevrani, H. (2009). Power system control-an
overview. In Bevrani, H. (Ed.), Robust power
system frequency control. New York, NY: Springer.
doi:10.1007/978-0-387-84878-5_1
Bevrani, H., Ledwich, G., Dong, Z. Y., & Ford,
J. J. (2009). Regional frequency response analy-
sis under normal and emergency conditions.
Electric Power Systems Research, 79, 837845.
doi:10.1016/j.epsr.2008.11.002
Bevrani, H., Ledwich, G., Ford, J. J., & Dong, Z.
Y. (2009). On feasibility of regional power system
emergency control plans. Energy Conversion and
Management, 49(2), 193204. doi:10.1016/j.
Bevrani, H., & Tikdari, A. G. (2009). On the ne-
cessity of considering both voltage and frequency
in effective load shedding scheme. IEEJ Technical
Meeting, PSE-10-002, Fukui, Japan.
Bevrani, H., & Tikdari, A. G. (2010). An ANN-
based power system emergency control scheme
in the presence of high wind power penetration.
In L. F. Wang, et al., (Eds.), Wind power systems:
Applications of computational intelligence, (pp.
215-254). Heidelberg, Germany: Springer-Verlag:
Series on Green Energy and Technology.
Bevrani, H., & Tikdari, A. G. (2010). Power sys-
tem stability analysis based on descriptive study
of electrical indices. In ASIJ 5th Conference, 6
March 2010, Tokyo.
311
Bevrani, H., Tikdari, A. G., & Hiyama, T. (2010).
An intelligent based power system load shedding
design using voltage and frequency information.
International Conference on Modelling, Identifi-
cation and Control (ICMIC10), (pp. 545-549).
Okayama, Japan.
FRCC. (2009). Automatic underfrequency load
shedding program. (PRC-006-FRCC-01, 2009).
Retrieved from https://www.frcc.com/
IEEE RTS Task Force of APM Subcommittee.
(1979). IEEE reliability test system. IEEE Trans-
actions on Power Apparatus and Systems, 98(6),
20472056. doi:10.1109/TPAS.1979.319398
IEEE RTS Task Force of APM Subcommittee.
(1999). The IEEE reliability test system-1996.
10101020. doi:10.1109/59.780914
Kundur, P. (1994). Power system stability and
control. New York, NY: McGraw-Hill.
Miller, T. J. E. (1982). Reactive power control in
electric systems. New York, NY: Wiley.
Ourari, M. L., Dessaint, L. A., & Do, V. Q. (2003).
Coherency approach for dynamic equivalents of
large power systems. International Conference on
Power Systems Transients IPST 2003 in New
Orleans, USA.
Padiyar, K. R., & Krishna, S. (2006). Online detec-
tion of loss of synchronism using energy function
criterion. IEEE Transactions on Power Delivery,
21(1), 4655. doi:10.1109/TPWRD.2005.848652
Parashar, M., & Thorp, J. S. (2004, June). Con-
tinuum modeling of electromechanical dynamics
in large-scale power systems. IEEE Transactions
on Circuit and Systems, 18511858.
Phadke, A. G., & Thorp, J. S. (2008). Synchronized
Phasor measurements and their applications.
New York, NY: Springer. doi:10.1007/978-0-
387-76537-2
Sluis, L. V. D. (2001). Travelling waves. In
Greenwood, A. (Ed.), Transients in power
systems (pp. 3156). John Wiley & Sons Ltd.
doi:10.1002/0470846186.ch3
Sowa, P., Azmy, A. M., & Erlich, I. (2004).
Dynamic equivalents for calculation of power
system restoration. Retrieved from http://www.
math.wichita.edu
Thorp, J. S., Seyler, C. E., & Phadke, A. G. (1998,
June). Electromechanical wave propagation in
large electric power systems. IEEE Transactions
on CAS, 45(6), 614622. doi:10.1109/81.678472
Tikdari, A. G. (2009). Load shedding in the pres-
ence of renewable energy sources. Dissertation,
University of Kurdistan, Iran.
Tsai, S. J. S., Zuo, J., Zhang, Y., & Liu, Y. (2005).
Frequency visualization in large electric power
systems. IEEE Power Engineering Society Gen-
eral Meeting.
Wagner, C. F., & McCann, G. D. (1950). Wave
propagation on transmission lines (Ed) electrical
transmission and distribution reference, central
station engineers of the Westinghouse electric &
manufacturing company. Lakeside Press.
Wang, X., & Vittal, V. (2004). System islanding
using minimal cutsets with minimum net flow.
Proceedings of the IEEE PES Power System
Conference and Exposition, New York.
Yang, B., Vittal, V., & Heydt, G. T. (2006).
Slow-coherency-based controlled islandingA
demonstration of the approach on the August
14, 2003 blackout scenario. IEEE Transac-
tions on Power Systems, 21(4). doi:10.1109/
TPWRS.2006.881126
312
You, H., Vittal, V., & Wang, X. (2004). Slow
coherency based islanding. IEEE Transactions
TPWRS.2003.818729
You, H., Vittal, V., & Yang, Z. (2003). Self-healing
in power systems: An approach using islanding
and rate of frequency decline based load shed-
ding. IEEE Transactions on Power Systems, 18,
174181. doi:10.1109/TPWRS.2002.807111
ADDITIONAL READING
Abu-Elnaga, M. M., El-Kady, M. A., & Find-
lay, R. D. (1988, Nov). Stability assessment
of highly stressed power systems using the
sparse formulation of the direct method. IEEE
doi:10.1109/59.192977
Bin Qiu Ling Chen Centeno, V. Xuzhu Dong Yilu
Liu. (2001). Internet based frequency monitor-
ing network (FNET). IEEE Power Engineering
Society Winter Meeting.vol.3 (pp. 1166 1171).
Cresap, R. L., & Hauer, J. F. (1981). Aril). Emer-
gence of a New Swing Mode in the Western Power
System. IEEE Transactions on Power Apparatus
and Systems, PAS-100, 20372045. doi:10.1109/
TPAS.1981.316481
De La Ree, J., Centeno, V., Thorp, J. S., &
Phadke, A. G. (2010, June). Synchronized Phasor
Measurement Applications in Power Systems.
IEEE Transactions on Smart Grid, 1, 2027.
doi:10.1109/TSG.2010.2044815
Delin Wang Xiaoru Wang Yi Fang Wenbin Hao.
(2006, Oct.). Study on Dynamic Characteristics of
Electromechanical Wave in the Continuum Model
for Power System. International Conference on
Power System Technology.
Diduch, J. Y., & Chang, C. P. L. (2008, April).
Islanding Detection Using Proportional Power
Spectral Density. IEEE Transactions on
Power Delivery, 23, 776784. doi:10.1109/TP-
WRD.2007.915907
Grobovoy, A., & Lizalek, N. (2002, April). As-
sessment of power system properties by wave ap-
proach and structure analysis. Fifth International
Conference on Power System Management and
Control (pp. 365 370).
Kimbark, E. (1956). Power System Stability. New
York: Dover.
Kook, K. S., Liu, Y., & Bang, M. J. (2008, Septem-
ber). Global behaviour of power system frequency
in korean power system for the application of
frequency monitoring network. IET Generation.
Transmission & Distribution., 2, 764774.
Makarov, Y. V. Reshetov, V.I. Stroev, A. Voropai,
I. (2005, Nov.). Blackout Prevention in the United
States, Europe, and Russia. Proceedings of the
IEEE. Vol. 93 (pp. 1942 1955).
Murphy, R. J. (1996, Jan). Disturbance record-
ers trigger detection and protection. IEEE
Computer Applications in Power, 9, 2428.
doi:10.1109/67.475956
Pariz, N. Shanechi, H.M. Vaahedi, E. (2003, May).
Explaining and validating stressed power systems
behavior using modal series. IEEE Transactions
TPWRS.2003.811307
Rajamani, K., & Hambarde, U. K. (1999, Jul).
Islanding and load shedding schemes for cap-
tive power plants. IEEE Transactions on Power
Delivery, 14, 805809. doi:10.1109/61.772318
313
Semlyen, A. (1974, March) Analysis of Distur-
bance Propagation in Power Systems Based on
a Homogeneoue Dynamic Model. IEEE Trans-
actions on Power Apparatus and Systems. Vol.
PAS-93 (PP. 676 684).
Senroy, N., Heydt, G. T., & Vittal, V. (2006, Nov.).
Decision Tree Assisted Controlled Islanding. IEEE
doi:10.1109/TPWRS.2006.882470
Shu-Jen Tsai Li Zhang Phadke, A.G. Yilu Liu In-
gram, M.R. Bell, S.C. Grant, I.S. Bradshaw, D.T.
Lubkeman, D. Le Tang. (2007, Aug.). Frequency
Sensitivity and Electromechanical Propagation
Simulation Study in Large Power Systems.
IEEE Transactions on Circuits and Systems. I,
Regular Papers, 54, 18191828. doi:10.1109/
TCSI.2007.902542
Taylor, C. W., & Erickson, D. C. (1997, Jan).
Recording and analyzing the July 2 cascading
outage [Western USA power system]. IEEE
Computer Applications in Power, 10, 2630.
doi:10.1109/67.560830
Thomas, A. J., & Mahajan, S. M. (2009, Oct.).
Electromechanical Wave Analysis Through Tran-
sient Magnetic Modeling. IEEE Transactions on
Power Delivery, 24, 23362343. doi:10.1109/
TPWRD.2009.2027504
Thorp, J. S., Seyler, C. E., Parashar, M., &
Phadke, A. G. (1998, Jan). The large scale elec-
tric power system as a distributed continuum.
IEEE Power Engineering Review, 18, 4950.
doi:10.1109/39.646962
Vanfretti, L. Aliyu, U. Chow, J.H. Momoh, J.A.
(2009, July). System frequency monitoring in the
Nigerian power system. IEEE Power & Energy
Society General Meeting (pp. 1 6).
Vittal, V., Rajagopal, S., Fouad, A. A., El-Kady,
M. A., Vaahedi, E., & Carvalho, V. F. (1988, Feb).
Transient stability analysis of stressed power
systems using the energy function method. IEEE
doi:10.1109/59.43206
Wang Yuguo Zhao Wei Xie Yan. (2009, June).
Dynamic feature extraction of power disturbance
signal based on time-frequency technology. Con-
trol and Decision Conference. CCDC 09. Chinese
(pp. 2300 2303).
Xu, G. Vittal, V. Meklin, A. Thalman, J. E. (2010)
Controlled Islanding Demonstrations on the
WECC System. IEEE Transactions on Power
Systems. Vol. PP.
Yilu, L. (2006). A US-Wide Power Systems Fre-
quency Monitoring Network. IEEE Power Systems
Conference and Exposition. (pp. 159 166).
Yingchen Zhang Markham, P. Tao Xia Lang
Chen Yanzhu Ye Zhongyu Wu Zhiyong Yuan Lei
Wang Bank, J. Burgett, J. Conners, R.W. Yilu Liu.
(2010, Sept.). Wide-Area Frequency Monitoring
Network (FNET) Architecture and Applications.
IEEE Transactions on Smart Grid. Vol. 1 (pp.
159 167).
Zhian Zhong Chunchun Xu Billian, B.J. Li Zhang
Tsai, S.-J.S. Conners, R.W. Centeno, V.A. Phadke,
A.G. Yilu Liu. (2005, Nov.). Power system fre-
quency monitoring network (FNET) implementa-
tion. IEEE Transactions on Power Systems, 20,
19141921. doi:10.1109/TPWRS.2005.857386
Angle Instability: Angle instability is a very
fast instability that leads to loss of synchronism.
Emergency Control: Control of a power sys-
tem in dangerous conditions, where the system is
going to the instability or blackout.
314
Islanding Control: Separate of power system
into some isolated sub-systems following large
disturbances, in order to prevent a global black out.
Load Shedding: An emergency control action
to curtail a part of load, and is useful where the
amount of load is larger than available generation.
PMU: Power measurement unit is a device
which uses GPS to collect data from the various
points of network in a synchronous time.
Slow Coherency: A method which is used in
islanding control problems. Using this method-
ology, the system islands and weak links can be
identified.
Wave Propagation: Propagation of a distur-
bance through the power system network like a
propagated wave.
315
Chapter 11
DOI: 10.4018/978-1-61350-138-2.ch011
Rana A. Jabbar
Rachna College of Engineering and Technology, Pakistan
Muhammad Junaid
M. A. Masood
A. Bashir
M. Mansoor
Analyses and Monitoring
of Power Grid
ABSTRACT
Power system analyses and monitoring of power system engineering are as essential as oxygen for
human beings. This innovative approach deals with a 132 kV grid simulation in electrical transient
analyzer program (ETAP). The existing power distribution system in Pakistan consists of approximately
six thousand 11 kV feeders, which are mainly analyzed by software FDR-ANA (Feeder Analyses). This
software does not have capability to provide comprehensive analyses for integrated power system. The
case under study is 132 kV grid situated in Gujranwala electric power company (GEPCO), one of the
distribution companies (DISCOs) of Pakistan electric power company (PEPCO) which has been selected
for comprehensive analyses using ETAP software. This software performs numerical calculations of large
integrated power system with fabulous speed, besides generating output reports. In a developing country
like Pakistan it is frst time that analyses based Off-line monitoring has been made, which includes load
fow, harmonic, transient, short circuit and ground grid analyses. In load fow analysis, current fowing
in every branch, power factor, active and reactive power fow, line losses, voltage magnitude with angle
etc. have been calculated. During harmonic analysis, distorted current and voltage waveforms along
with their harmonic spectrum caused by non-linear loads have been recorded. Transient analysis has
been performed to record different waveforms like variation in bus frequency, bus real power loading,
bus voltage angle, and bus reactive power loading for short interval of time during transient conditions.
316
Analyses and Monitoring of Power Grid
INTRODUCTION
This book chapter comprises of two interna-
tional conferences research papers published in
ELECO09 and PCO10 in which first time in
the history of Pakistan a practical 132 kV grid
containing a large distribution network has been
simulated for analyses purpose using ETAP soft-
ware (Jabbar Khan et al, 2009; Bashir et al, 2010).
For the last few years electrical engineers have
been focusing on the power system studies using
software tools. Recent advances in engineering
sciences have brought a revolution in the field of
electrical engineering after the development of
powerful computer based software. This research
work high-lights the effective use of ETAP soft-
ware for analyses of large electrical power system
which comprises of large power distribution net-
work emanating from 132 kV power grid (Lei et
al, 2002; Inoue, 2007; Takimoto, 2005; Nagata
& Inoue, 2008; Brown et al, 1990; Zhongxi &
Xiaoxin, 1998; Stagg& El-Abiad, 1968).
Motivation to conduct the research work is
to develop a prototype model which can be used
as bench mark for comprehensive simulation of
integrated network at national grid level to ad-
dress power system stability under normal and
abnormal operating conditions. PEPCO, the only
power sector utility in Pakistan, consists of nine
power distribution companies along with national
transmission and power dispatch company but
unfortunately no ON/OFF line monitoring is
currently being performed. For this purpose a
132 kV grid station has been simulated using
ETAP. PEPCO has been experiencing severe
power shortage for last many years. Resultantly,
the country is facing repeated and astonishing
black outs. Despite the shortage of electricity,
one of the main reasons of this energy crisis is
deficiency in the field of analyses and monitoring
of electrical power network. Keeping in view the
above scenario, Rachna College of Engineering
& Technology (RCET) Pakistan, has analyzed the
complete 132 kV grid network which contains
11 kV feeders and rest of distribution network
to predict the actual effects of load on the entire
power system. These analyses include load flow
analysis, harmonic analysis, transient analysis,
ground grid analysis and short circuit analysis.
The data used for analyses purpose is in the
form of one line diagram of complete power
system network starting from power transformer
at grid up-till the load. The ratings of power/dis-
tribution transformers are taken as they actually
exist. Moreover, the conductors/cables, circuit
breakers, CTs, PTs, and rest of power system
elements are also modelled according to their
actual ratings in ETAP.
This 132 kV Grid located in GEPCO region,
having 6 power transformers, 32 feeders, 48 cir-
cuit breakers, 42 current transformers, 8 potential
transformers and 2 incoming lines. Practical power
system under study is a very antique grid which
was inaugurated in 1952 and a centralized grid
which feeds power supply to the other grids in
this region, all the analyses and monitoring are
concentrated on this grid (WAPDA, 2006).
A powerful computational software ETAP is
used in this research paper for modeling and simu-
lation purpose. The complete power system from
grid to tail end load is modeled in this software.
Although MATLAB is also used for power system
In ground grid modeling, step, and touch potentials have been calculated in comparison with set stan-
dards. While performing short circuit analysis, all the possible short circuit faults like line to ground,
double line to ground, 3-phase faults etc. on cycle, 1.5 to 4 cycle, and 30 cycle networks have been
performed to record the short circuit currents. These analyses have been executed using ETAP software,
based upon historical data obtained from original system that will be very helpful for system security
and reliability.
317
simulation world widely, but ETAP has preferred
here due to strong built-in properties regarding
power system studies (Sybille & Hoang, 2002;
Qinghuaz et al, 2009; Kjlle et al, 2002; Gatta et
al, 2003; Gonen, 1968)
Earlier, considerable work has been done on
transient stability studies, but power system of
developing countries like Pakistan have never been
simulated using software technique like ETAP on
such a large scale. The major purpose for transient
stability studies on this power system is to find
dynamic performance which has great impact
in the design and operation of the system. After
transient stability studies on 132 kV grid, machine
power angles and speed deviations, system electri-
cal frequency, real and reactive power flows of
the machines and buses, power flow of lines and
transformers, as well as the voltage level of the
buses in the power system have been determined
thoroughly (Hongbin et al, 2002; IEEE 1995).
A ground grid study is first time addressed
in this research for a large power system. The
touch and step potential measurement will really
be helpful for safety of both grid staff and equip-
ment. It is worth mentioning that in this analysis
three-dimension view grid grounding scheme
has been used on such a large scale using ETAP.
In short circuit analyses on 132 kV grid mo-
mentary, steady-state, short circuit and other fault
interruption currents on cycle network (sub
transient network), 1.5 to 4 cycle network (tran-
sient network), and 30 cycle network at different
buses have also been determined (Osahenvemwem
& Omorogiuwa, 2008).
BACKGROUND
Introduction to ETAP Software
ETAP software is used for simulation purpose
which provides a fully graphical user interface
(GUI) for constructing one-line diagram. Here
elements can graphically add, delete, relocate,
connect, zoom in or out, display grid off or on,
change element size, change element orientation,
change symbols, hide or show protective devices,
enter properties and set operating status, etc.
ETAPs electrical system diagram is a one-line
representation of a three-phase system. The one-
line diagram is the starting point for all studies. The
electrical system can be constructed graphically by
connecting the buses, branches, isolators, circuit
breakers and protective devices in any order from
the one-line diagram edit toolbar.
Single Line Diagram of the System
Figure 1 shows the single line diagram of the
complete power system which is under study. It
is clear that there are two incoming lines of 132
Figure 1. Single line diagram of 132 kV grid
318
kV supplying power to six power transformers
of different ratings at 132 kV grid station, and
these power transformers are connected with 11
kV power distribution network (11 kV feeders).
Monitoring Points are also marked on the same
single line diagram to discuss different results
while performing simulation. In Figure1, point A
is taken at secondary of power transformer, point
B is taken at primary distribution side (i.e. 11 kV
feeder) and point C is taken on a feeder where
furnace load is connected. This large system is
sub divided into sections according to their load
behavior for analyses and monitoring purpose.
Power transformers installed at the grid are given
in Table 1.
Circuit breakers (CB) of different ratings have
been installed at the grid. Some of them are of
SF6 circuit breakers and remaining are oil, air
and vacuum circuit breakers. These CBs type,
voltage ratings and current ratings are given in
the Table 2.
Simulation of 132 Kv Grid
The system under study is simulated in ETAP as
shown in Figure 2.
Simulation of 11 Kv Feeders
For simplicity of analyses, only one power trans-
former named T1 has been selected for simulation.
Following the practical system, feeders are divided
into two different categories, one for furnace load
and other for usual distributed load which may
include static, inductive, dynamic and dc load etc.
Figure 3 shows that 7 no. feeders are emanat-
ing from power transformer T1. Three furnaces
are being supplied from feeder 7. The simulated
network for general purpose feeder is shown in
Figure 4.
Load Modeling
Following the real time data, load modeling is
performed by considering 70% static and dy-
namic load in addition to 30% dc load. Figure 5
and Figure 6 show the simulated model of load
connected with 200 kVA and 50 kVA distribution
transformers respectively.
For the modeling of furnace load, ETAP pro-
vides the provision for exact modeling of such
large electronic load in which rectification and
inversion phenomenon. Figure 7 shows one of
the three furnaces connected to the feeder.
All the three furnaces are of different ratings,
Furnace-1 which takes a load of 62.2 A from the
Table 1. Power transformers at 132 kV grid
Power Transformers MVA rating
T-1 40
T-2 26
T-3 26
T-4 26
T-5 37
T-6 13
Table 2. Circuit breakers at 132 kV grid
CB Types Rated Voltage (kV) Rated Current (A)
V-15F-31 (Vacuum) 12 630
3AF2721-4B (Vacuum) 12 630
WPV-25-O (Vacuum) 12 2500
V-15V-31 (SF6) 145 3150
JB429FORM (Oil) 66 4380
319
Figure 2. Simulated diagram of 132kV grid using ETAP
Figure 3. Simulated diagram of power transformer T1
320
Figure 4. Distribution network in ETAP for general purpose feeder
Figure 5. Load modeling for 200 kVA distribution
transformer
Figure 6. Load modeling for50 kVA distribution
transformer
321
system. Similarly, Furnace-2 and Furnace-3 take
loads of 161.2 A and 136.5 A respectively (Jabbar
et al, 2008a; Jabbar et al, 2008b).
SOLUTIONS AND
RECOMMENDATIONS
Load Flow Analysis
Load flow studies have been carried out on com-
plete power system using ETAP. However for
detailed monitoring purpose three important points
are selected as shown in Figure 1. Point A is the
secondary side of power transformer T1. Point B
is at 11 kV side of general purpose feeder. Point C
is at the 11 kV side of furnace feeder. Consequent
results after performing load flow analysis are
given in Table 3 and Table 4 accordingly.
Table 4 clearly shows that real power on swing
buses is 16.789 MW while reactive power is
11.970 MVAR. Power factor is 81.4% which is
less than the standard set by the utility (i.e. 92%).
ETAP Alerts during Load Flow Analysis
While performing load flow analysis ETAP
provides various alerts which need immediate
attention for smooth running of the system, some
of the critical alerts regarding under voltages are
detailed in Table 5.
Over-Loaded Transformers
Loading wise power system under study was
normal (not over loaded) but it is interesting
due to involvement of large electronic load (i.e.
furnace) another loss is involved in the system
which is known as distortion power loss due
to which system becomes over loaded in some
branches, some of these over loaded transformers
are indicated in Table 6.
Voltage Drop
Table 7 shows the % voltage drop in different
distribution transformers (only few are given
here). Magnitude of voltage drop standard set by
utility is 5% which is significantly violated here.
Figure 7. Load modeling for furnace feeder
Table 3. Load flow results
Monitoring Points kV MW MVAR % PF
Grid (Bus 3) 132 36.68 23.10 84.6
A 11 5.520 3.793 83.1
B 11 0.379 0.253 83.0
C 11 0.717 0.471 83.6
322
Technical Losses
Technical losses calculated while performing
simulation are given in Table 8 at some branches.
Harmonic Analysis
A detailed harmonic analysis (HA) has been
performed on the entire system using ETAP. The
results include current/voltage waveforms and
their harmonic spectrums at different points of
the power system under consideration. In addition
total and individual harmonic distortion in voltage
and current (i.e. [% THDv], [% IHDv], [% THDi]
and [% IHDi]) has also been obtained. Table 9
shows the different results which are recorded after
harmonic studies on different monitoring points.
These results clearly show that % THDv at Point
Table 6. Some over loaded transformers in the system
TF ID Capability (MVA) Loading (input) MVA Loading (input) %
Loading (output)
MVA
Loading (output) %
T114 0.200 1.251 625.3 0.952 476.2
T797 0.200 0.352 175.8 0.330 165.12
T122 0.200 0.452 225.9 0.416 208.231
T618 0.200 0.352 175.8 0.330 165.1
T803 0.200 0.452 225.9 0.416 208.3
T807 0.200 1.195 597.3 0.926 462.3
T809 0.200 0.352 175.8 0.330 165.3
Table 5. Under voltage alerts during load flow results
Device ID Rating Calculated % Value Alert Condition
Bus 210 0.415 kV 0.314 kV 73.2 Under Voltage
Bus 675 11 kV 10.531 kV 96 Under Voltage
Bus 700 11 kV 10.615 kV 96.5 Under Voltage
Bus 238 11 kV 11 kV 10.604 Under Voltage
Bus 638 11 kV 10.623 kV 96.6 Under Voltage
Table 4. Summary report of total demand and losses
Points for Discussion MW MVAR MVA % PF
Swing Bus 16.789 11.970 20.619 81.4 (lag)
Total Demand 16.789 11.970 20.619 81.4 (lag)
Total Motor Load 15.413 7.068 16.956 90.9 (lag)
Total Static Load 0.555 0.393
Apparent Losses 0.821 4.510
323
C has exceeded the limit set by IEEE std. 519-
1992 (i.e. 5% at the point of common coupling).
The current waveform and its harmonic spec-
trum captured at the monitoring point A are shown
in Figure 8 and Figure 9 respectively.
Voltage waveform and its harmonic spectrum
captured at monitoring point A are shown in
Figure 10 and Figure 11. The below waveshape
is at the 11kV bus bar which shows the significant
presence of 5th and 7th harmonic componets that
also contribute very much in the disturbing the
power quality of the system.
Similarly, the current waveform and har-
monic spectrum captured at point B are shown in
Figure 12 and Figure 13 respectively.
Voltage waveform with harmonic spectrum at
point B is given in Figure 14.
Table 7. % Voltage drop in few sections/branches
CKT/Branch
ID % Bus Voltage from % Bus Voltage to %Voltage drop (Mag.)
T114 73.7 96.8 23.08
T624 89.1 96.12 7.58
T625 90.4 96.2 6.27
T628 75 96.7 21.77
T797 90.8 96.7 5.89
T803 89.1 96.12 7.58
T819 75 96.7 21.77
T821 90.7 96.8 5.88
T831 75 96.7 21.77
T833 90.7 96.8 5.88
T839 89.1 96.12 7.58
Table 8. Losses in kW & kVAR in few sections/branches
CKT/Branch ID Losses kW Losses kVAR
T114 83.6 426.3
T624 10.9 55.7
T625 6.7 19.5
T628 76.2 388.7
T797 6.6 33.7
T803 10.9 55.7
T807 76.2 388.7
T819 76.2 388.7
T821 6.6 33.7
T831 76.2 388.7
Table 9. % THDv & % THDi at monitoring points
Monitoring Points % THDv %THDi
A 2.13 3.41
B 3.0 2.25
C 13.1 15
324
Figure 8. Current waveform at point A
Figure 9. Current spectrum at point A
Figure 10. Voltage waveform at point A
325
Figure 11. Voltage waveform spectrum at point A
Figure 12. Current waveform at point B
Figure 13. Current spectrum at point B
326
Figure 14. Voltage waveform at point B
Figure 15. Voltage waveform spectrum at point B
Figure 16. Current waveform at point C
327
Figure 17. Current spectrum at point C
Figure 18. Votage waveform at point C
Figure 19. Votage spactrum at point C
328
Due to presence of dc load voltage waveform
is also distorted as a result of current waveform
distortion. FFT spectrum is shown in Figure 15,
at the monitoring point B, which shows that 5th,
7th, 11th and 13th harmonic components contrib-
ute very much.
Figure 16 and Figure 17 are the current wave-
form and its harmonic spectrum at point C which
is more distorted as compared to points A and B.
The reason behind this distorted pattern is sig-
nificant presence of 5
th
, 7
th
, 11
th
and 13
th
etc.
harmonic contents in the current waveform drawn
by the furnace load.
Similarly, the voltage waveform and har-
monic spectrum at point C are given in Figure 18
and Figure 19 respectively.
Harmonic analysis has been performed on
complete system but due to space limitation it is
not possible to discuss here any more.
Transient Stability Analysis
During transient stability analysis for the system
under consideration given in Figure 1, the stability
limits of the power system can be inspected before,
during and after system changes. Figure 20 is the
Figure 20. Simulated network in ETAP
Figure 21. Bus frequency variations at selected time interval
329
section of power system modeled in ETAP which
is selected for transient analysis, where a group
of different motors is connected to a distribution
transformer.
Figure 21 shows bus frequency variation at
specific time interval (0.5 0.7 sec) when group
of different motors is connected to bus 232 during
transient stability analysis. It is clear from the
Figure 3 that before 0.5 sec the system frequency
was 50 Hz. (i.e. 100%), after 0.5 sec 0.7 sec
during transient the system frequency decreases
significanly and system becomes normal after
that transient interval.
It is observed, during the transient stability
analysis on the selected bus that real and reactive
power loading of the bus changes abruptly during
transient phase. Figure 22 and Figure 23 show
real and reactive power loading at specific time
interval when group of different motors is con-
nected to the system.
During the switching of the motor load to the
power system it is observed that magnitude in bus
voltage and angle is distorted very firmly which
causes for the instability problems in the power
system. Figure 24 and Figure 25 show bus voltage
magnitude and angle variations at perticular instant
Figure 22. Abrupt changes in bus real power loading
Figure 23. Bus real power loading at particular instant of time
330
Figure 24. Variation in bus voltage
Figure 25. Bus voltage angle variation
Figure 26. Variation in bus voltage per Hz
331
of time when group of different motors con-
nected to the bus are switched on.
Magnitude of bus voltage variation with respect
to frequency of the power system when dynamic
load is switched on also causes the instability
problems in the power system. This voltage mag-
nitude variation versus freequency of the power
system is expresed in Figure 26.
The consequences of instability issues dis-
cussed earlier may result in temporary, permanent
loss to electrical installations, mal-operation of
protective devices and power blackouts etc.
Ground Grid Analysis
For ground grid analysis, power transformer T1
at 132 kV grid has been selected. The exact area
in grid where power transformer and its other
related components like bus bar, circuit break-
ers, isolators etc. are situated is measured. The
exact no. of horizontal and vertical rods used for
mesh grounding is also recorded. The same is
implemented in ETAP for ground grid analysis to
evaluate step and touch potential exactly.
Ground Grid Layout
The ground grid layout is consisted of three types
of different views named as top view, soil view
and 3D view which are shown in Figure 27, Figure
28 and Figure 29 respectively.
Calculation of Touch and Step Potential
The step and touch potential measurements on 132
kV grid provide safety limits for grid equipment as
well as staff. Table 10 shows the different potential
measurements located at x-axis and y-axis of 132
kV Grid against tolerable limits.
It is clear from the Table that calculated value
of touch potential is 1946.6 volts which is beyond
the tolerable limits i.e 427.1 volts. Similarly, step
potential is also out of tolerable limits. The situ-
ation is alarming due to this out of range potential.
It is recommended here, to increase the number
of vertical rods to over come this problem.
These step and touch potentials are also ex-
pressed in 3-D view highlighting the limits and
ranges in Figure 30 and Figure 31 respectively.
Figure 27. Top view of 132 kV ground grid
332
Figure 28. Soil view of 132 kV ground grid
Table 10. Calculated step potential and touch potential on tolerable limits
Touch potential Step Potential
Calculated volts 1949.6 1368.5
Tolerable volts 427.1 1216.4
Location x(axis) 51.9 48.3
Location y(axis) 33.2 4.4
Figure 29. 3-D view of 132 kV ground grid
333
Short Circuit Fault Analysis
Three-Phase Fault Calculations
at Different Buses
For short circuit analysis some of the buses are
selected where an artificial short circuit fault
is created. An equivalent voltage source at the
fault location is replaced by all external voltage
sources and machines internal voltage sources.
Three different impedance networks are formed
for each faulted bus where calculations are made
to calculate for sub transient, transient and steady
state short circuit currents. These networks are
cycle network (sub transient network), 1.5 - 4
cycle network (transient network) and 30 cycle
network (steady state network).
Figure 30. Step potential profile of 132 kV grid
Figure 31. Touch potential profile of 132 kV grid
334
Sub-Transient Network
(Half-Cycle Network)
This type of network is used to calculate sub
transient currents at different buses in the sys-
tem. Detail of short circuit sub transient currents
calculations are given in the Table 11.
It is important to note that bus 232 is rated at
0.415 kV and remaining buses are rated at 11 kV.
Transient Network (1.5 -
4 Cycle Network)
Transient currents at selected buses in the power
system are represented by using 1.5 to 4 cycle
network in EATP. Transient calculations are given
in Table 12 at different buses.
Thirty-Cycle Network
This network is used to calculate the steady state
short circuit current at the faulty buses. Table
Table 11. 3-phase short circuit current calculations in half cycle network
Fault Bus ID Cycle (kA Real)
Cycle
(kA Imag)
Imag/ Real kA symm Magnitude
232 5.661 -112.818 19.9 112.960
240 2.851 -12.084 4.2 12.416
693 0.881 -2.5588 2.9 2.705
695 0.993 -2.963 3.0 3.125
696 0.864 -2.503 2.9 2.648
700 2.045 -17.159 8.4 17.280
Table 12. 3-phase short circuit current calculations in 1.5 to 4 cycles network
Fault Bus ID 1.5-4 Cycle (kA Real) 1.5-4 Cycle (kA Imag) Imag/ Real kA symm Magnitude
240 2.632 -11.828 4.5 12.117
693 0.868 -2.548 2.9 2.692
695 0.977 -2.950 3.0 3.107
696 0.852 -2.494 2.9 2.635
700 1.747 -16.596 9.5 16.688
Table 13. 3-phase short circuit current calculations in 30 cycles network
Fault Bus ID 30 Cycle (kA Real) 30 Cycle (kA Imag) X/R Ratio kA Symm Magnitude
232 3.712 -61.298 16.5 61.140
240 2.491 -11.509 4.6 11.775
693 0.858 -2.535 3.0 2.676
695 0.963 -2.932 3.0 3.086
696 0.842 -2.481 2.9 2.620
700 1.621 -15.964 9.8 16.046
335
13 shows steady state short circuit currents at
selected buses.
Three-Phase LG, LL, and LLG (1/2
Cycle Network) Fault Currents at
Selected Buses
The purpose of this analysis on the faulty buses is
to calculate line-to-ground, line-to-line, double-
line-to-ground, three-phase short-circuit fault cur-
rents and sequence voltages at cycles networks.
Table 14 shows 3-phase fault currents and their
respective sequence voltages at the faulty buses.
Three-phase symmetrical short circuit current
at faulty buses are shown in Table 15 as:
Three-Phase LG, LL, AND LLG (1.5 - 4
Cycle Network) Fault Currents at
Different Buses
Line-to-ground, line-to-line, double-line-to-
ground, three-phase short-circuit fault currents
Table 16. 3-phase short circuit current and sequence voltages at faulty buses in transient networks
Fault Bus ID 3 Phase Fault %V kA Symm Line-Ground Fault (%V from Bus) Va Vb Vc
232 0.0 112.960 0.00 100.75 100.57
240 0.0 12.416 0.00 173.21 173.21
693 0.0 2.705 0.00 173.21 173.21
695 0.0 3.125 0.00 173.21 173.21
696 0.0 2.648 0.00 173.21 173.21
700 0.0 17.280 0.00 173.21 173.21
Table 15. 3-phase symmetrical short circuit current calculations in 1/2 cycles network (sub transient
network)
Fault Bus ID Line-Ground Fault (kA Symm rms) Ia 3Io
232 110.73 110.731
240 0.00 0.00
693 0.00 0.00
695 0.00 0.00
696 0.00 0.00
700 0.00 0.00
Table 14. 3-phase short circuit current & sequence voltages at faulted buses in sub transient networks
232 0.0 112.960 0.00 100.12 100.88
240 0.0 12.416 0.00 173.21 173.21
693 0.0 2.705 0.00 173.21 173.21
695 0.0 3.125 0.00 173.21 173.21
696 0.0 2.3648 0.00 173.21 173.21
700 0.0 17.280 0.00 173.21 173.21
336
and sequence voltage values in 1 - 4 cycles
are represented through this study on the faulty
buses. Table 16 and Table 17 indicates 3-phase
fault currents along with their respective sequence
voltages and 3-phase symmetrical short circuit
current calculations in 1.5 - 4 cycles network
(transient network) at the selected faulted buses
respectively.
Three-Phase LG, LL & LLG (30 Cycle
Network) Fault Currents at Different
Buses
Three-phase line-ground, line-line and double-
line-ground fault is calculated for 30-cycle net-
work are summarized in Table 18 and Table 19
respectively.
Table 17. 3-phase symmetrical short circuit current calculations in 1.5-4 cycles network (transient network)
Fault Bus ID Line-Ground Fault (KA Symm rms) Ia 3Io
232 110.259 110.259
240 0.00 0.00
693 0.00 0.00
695 0.00 0.00
696 0.00 0.00
700 0.00 0.00
Table 18. 3-phase short circuit current & sequence voltages at faulty buses in transient networks
232 0.0 61.410 0.00 79.73 79.64
240 0.0 11.775 0.00 173.21 173.21
693 0.0 2.676 0.00 173.21 173.21
695 0.0 3.086 0.00 173.21 173.21
696 0.0 2.626 0.00 173.21 173.21
700 0.0 16.046 0.00 173.21 173.21
Table 19. 3-phase symmetrical short circuit current calculations in 30-cycles network (transient network)
Fault Bus ID Line-Ground Fault (KA Symm rms) Ia 3Io
232 86.370 86.73
240 0.00 0.00
693 0.00 0.00
695 0.00 0.00
696 0.00 0.00
700 0.00 0.00
337
Research conducted during this work is based
upon the historical data of a practical grid which
has further been implemented in ETAP for off-
line monitoring. Keeping in view the latest trends,
on-line monitoring can also be performed on
same network using same software tool along
with hardware setup. For accurate monitoring
and analyses purpose, real time monitoring has
been recommended in future. Moreover, protec-
tive equipment installed at grid and sensitive
load equipment can also be modelled for relay
coordination analysis in ETAP.
CONCLUSION
ETAP Software based simulation of integrated
electric power system comprising of 132 kV grid
including load flow, harmonics, transient, ground
grid and short circuit analyses reveals innovative
contribution to the power industry stake holders
to access in-advance knowledge for acquiring
the security and quality of supply. Overloading
of power/distribution transformers, line conduc-
tors/cables current carrying ability, power factor,
supply demand gap, voltage drop at the tail end,
technical losses, active and reactive power flow,
voltage and current magnitudes, THD in voltage
and current etc. can be analyzed and monitored
at any desired location using this innovative ap-
proach. Moreover, the novel approach of ground
grid 3D analysis is really interesting and will be
helpful for evaluating the exact value of step and
touch potential for safety purpose. Simulation con-
ducted in this research work can be implemented
successfully at any desired section of the power
system network for above mentioned parameters.
REFERENCES
Bashir, A., Jabbar Khan, R. A., Junaid, M., &
Asghar, M. M. (2010). ETAP software based
transient, ground Grid and short circuit analyses
of 132 kV Grid. 3
rd
Global Conference on Power
Control and optimization (PCO 2010). 2-4 Feb-
ruary, 2010, Gold Coast, Queensland, Australia.
Brown, K., Shokooh, F., Abcede, H., & Don-
ner, G. (1990). Interactive simulation of power
system: ETAP techniques and applications (pp.
19301941). IEEE.
ETAP. (n.d.). Enterprise software solution for
power systems (Electrical Transient Analyzer
Program). Retrieved from http://etap.com/training
/tutorials-training-videos.htm
Gatta, F. M., Iliceto, F., Lauria, S., & Masato, P.
(2003). Modeling and computer simulation of
dispersed generation in distribution networks:
Measures to prevent disconnection during system
disturbances. IEEE PowerTech Conference. June
2003, Italy.
Gonen, T. (1986). Electric power transmission
system engineering (pp. 8793). New York, NY:
Wiley.
Hongbin, Z., Renmu, H., & Jian, Z. (2002). Ap-
plication of different load models for the transient
stability calculation. Proc. 2002 IEEE Power
Engineering Society Transmission and Distribu-
tion Conf., (pp. 2014-2018).
IEEE Task Force on Load Representation for
Dynamic Performance. (1995). Bibliography on
load models for power flow and dynamic perfor-
mance simulation. IEEE Transactions on Power
Systems, 10.
338
Inoue, T. (2007). Dynamic simulations of electric
power systems under long-term change in system
generations and loads. The Seventh IASTED
International Conference on Power and Energy
Systems (EuroPES 2007), (pp. 232-237). August
2007, Spain.
Jabbar, R. A., Akmal, M., Junaid, M., & Masood,
M. A. (2008b). Operational and economic impacts
of distorted current drawn by the modern induc-
tion furnaces. Proceedings of AUPEC08. IEEE.
14-17 December, 2008, UNSW, Sydney, Australia.
ISBN 978-0-7334-2715-2
Jabbar, R. A., Akmal, M., Masood, M. A., Junaid,
M., & Akram, F. (2008a). Voltage waveform
distortion measurement caused by current drawn
by modern induction furnaces. Proceedings of
13th ICHQP, IEEE PES. 07 November 2008,
University of Wollongong, Australia. ISBN 978-
1-4244-1771-1
Jabbar Khan, R. A., Junaid, M., & Asghar, M. M.
(2009). Analyses and monitoring of 132 kV Grid
using ETAP software. 6th International Confer-
ence on Electrical and Electronics Engineering
(ELECO 2009). IEEE. 5-8 November 2009,
Bursa, Turkey.
Kjlle, G. H., Aab, Y., & Hjartsj, B. T. (2002).
Fault statistics as a basis for designing cost-
effective protection and control solutions. Proc.
2002 CIGRE Session, Paris, August 2002.
Lei, X., Buchholz, B., Povh, D., & Retzmann, D.
(2002). Power system analysis-software approach
and real-time simulation, vol 2 (pp. 1011-1016).
Power Engineering Society Winter Meeting, 2002.
IEEE. ISBN 0-7803-7322-7
Nagata, M., & Inoue, T. (2008). An efficient
voltage and reactive power control simulation
using long term dynamics simulation. 16
th
Power
Systems Computation Conference 2008, June
2008, Scotland.
Osahenvemwem, A. O., & Omorogiuwa, O.
(2008). Electric transmission line faults in Nigeria:
A cause study of Benin-Irrua 132kV transmission
line. International Journal of Electrical and Power
Engineering, 2(6), 384-388. ISSN 1990-7958
Pakistan Water and Power Development Author-
ity (WAPDA). (2006-07). Annual report. July
2006- June 2007.
Qinghuaz, N. N., Jian Wu Haupt, T., & Srivastava,
A. K. (2009). Power system decoupled simulation
in MATLAB/Simulink, (pp. 1-8). ISBN: 978-1-
4244-4283-6
Stagg, G. W., & El-Abiad, A. H. (1968). Computer
methods in power system analyses (pp. 110127).
New York, NY: McGraw-Hill.
Sybille, G., & Hoang, L. (2002). Digital simulation
of power systems and power electronics using the
MATLAB/Simulink power system blockset, (pp.
29732981). Power Engineering Society Winter
Meeting, 2000. IEEE. ISBN: 0-7803-5935-6
Takimoto, A. (2005). Time domain simulation
based preventive control method for transient
stability. The International Conference on electri-
cal Engineering 2005, July 2005, China.
Zhongxi, W., & Xiaoxin, Z. (1998). Power system
analyses software package (PSASP)-An integrated
powersystem analyses tool (pp. 7-11). ISBN:
0-7803-4754-4
ADDITIONAL READING
Das, J. C. (2002). Power System Analysis: Short-
Circuit Load Flow and Harmonics (1st ed.). Power
Engineering.
Glover, J. D. (2008). Sarma.M.S (4th ed.). Power
System Analysis and Design.
339
Grainger, J. J., & Stevenson, W. (1994). Power
System Analysis. McGraw-Hill Science/ Engg/
Math.
(2002). Natarajan (1st ed.). Computer Aided
Power System. CRC.
Tleis, N. (2007). Power Systems Modelling and
Fault Analysis: Theory and Practice Book. ISBN:
0750680741
340
APPENDIX 1: KEY TERMS, CONCEPTS, AND EQUATIONS
Load Flow Analysis: Calculates the bus voltages, branch power factors, currents, and power flows
throughout the electrical system. Unlike traditional circuit analysis, a power flow study usually uses
simplified notation such as a one-line diagram and per-unit system, and focuses on various forms of AC
power rather than voltage and current. It analyzes the power systems in normal steady-state operation.
Load Flow Calculation Methods: ETAP provides three load flow calculation methods: Newton-
Raphson, Fast-Decoupled, and Accelerated Gauss-Seidel. Newton-Raphson method has been used
during the simulation.
P
Q
J J
J J V
1 2
3 4
(1)
Fast-Decoupled Method: Derived from the Newton-Raphson method. It takes the fact that a small
change in the magnitude of bus voltage does not vary the real power at the bus appreciably, and likewise,
for a small change in the phase angle of the bus voltage, the reactive power does not change apprecia-
bly. Thus the load flow equation from the Newton-Raphson method can be simplified into two separate
decoupled sets of load flow equations, which can be solved iteratively (Equation 2):

P J
Q J V
1
4
(2)
Gauss-Seidel Method: From the system nodal voltage equation (Equation 3):
l Y V
BUS
(3)
P jQ V Y V
T
BUS
+

(4)
Data Required for Load Flow Analysis: Following data is required during load flow analysis:
Bus Data
Branch Data
Power Grid Data
Synchronous Generator Data
Inverter Data
Synchronous Motor Data
Induction Motor Data
Static Load Data
Capacitor Data
Lumped Load Data
Charger & UPS Data
341
Harmonic Analysis: Due to the wide and ever increasing applications of power electronic devices,
such as variable speed drives, uninterruptible power supplies (UPS), static power converters, etc., power
system voltage and current quality has been severely affected in some areas. In these areas components
other than that of fundamental frequency can be found to exist in the distorted voltage and current
waveforms. These components usually are the integer multipliers of the fundamental frequency, called
harmonics. In addition to electronic devices, some other non-linear loads, or devices including saturated
transformers, arc furnaces, fluorescent lights, and cyclo-converters are also responsible for the deteriora-
tion in power system quality.
Harmonic Analysis Calculation Methods: ETAP harmonic analysis program fully complies with
the latest version of the following standards:
IEEE Standards 519, IEEE Recommended Practices and Requirements for Harmonic Control in
Electrical Power Systems
IEEE Standards 141, IEEE Recommended Practice for Electric Power Distribution for Power
Plants
ANSI/IEEE Standard 399, IEEE Recommended Practice for Power System Analysis
THD
F
F
i
=
2
2
1
(5)
Individual Harmonic Distortion (IHD): Simply calculates the ratio of the harmonic component to
the fundamental component. This value is sometimes used to track the effect of each individual harmonic
and examine its magnitude. IHD is determined by (Equation 6):
IHD
F
F
i
=
1
(6)
Root Mean Square (RMS) Total: This is the square root of the sum of the squares of the magnitudes
of the fundamental plus all harmonics in the system. For a system with no harmonics at all, the total RMS
should be equal to the fundamental component RMS. The total RMS is determined by (Equation 7):
RMS F
i
=

2
1
(7)
Transient Stability Analysis: ETAP transient stability analysis program is designed to investigate
the stability limits of a power system before, during and after system changes or disturbances. The
program models dynamic characteristics of a power system, implements the user-defined events and
actions, solves the system network equation and machine differential equations interactively to find out
system and machine responses in time domain.
342
Power System Stability: The property of a power system which insures that it remains in electro-
mechanical equilibrium throughout any normal and abnormal operating conditions.
Stability Limits: There are two types of stability limit for a power system, namely steady-state
stability limit and transient stability limit.
Steady-State Stability Limit: defined as the stability of a system under conditions of gradual or small
changes in the system. This stability can be either found by the load flow calculation for a steady-state
operation, or determined by a transient stability study if there are system changes or disturbances involved.
Transient Stability Limit: defined as the stability of a system during and after sudden changes or
disturbances in the system, such as short-circuits, loss of generators, sudden changes in load, line trip-
ping, or any other similar impact. The system is said to be transient stable if following a severe distur-
bance, all synchronous machines reach their steady-state operating condition without prolonged loss of
synchronism or going out of step with other machines.
Causes of Instability Problems: The major causes to industrial power system instability problems
include:
Short-circuits
Loss of a tie connection to a utility system
Loss of a portion of in-plant co-generation (generator rejection)
Starting a motor that is large relative to the system generating capacity
Switching operations of lines, capacitors, etc.
Impact loading (motors and static loads)
A sudden large step change of load or generation
Consequences of Instability Problems:
Area-wide blackout
Interruption of loads
Low-voltage conditions
Damage to equipment
Relay and protective device malfunctions
Ground Grid Systems: Since the early days of the electric power industry, the safety of personnel
in and around electric power installations has been a primary concern. With ever increasing fault cur-
rent levels in todays interconnected power systems, there is renewed emphasis on safety. The safety of
personnel is compromised by the rise in the ground potential of grounded structures during unbalanced
electric power faults. At such times, humans touching grounded structures can be subjected to voltages.
However, the magnitude and duration of the electric current conducted through the human body should
not be sufficient to cause ventricular fibrillation
Years of research on the effects of electric current on the human body have lead to the development
of standards of permissible values to avoid electrocution. The ground grid systems program utilizes the
following four methods of computation:
FEM: Finite Element Method
IEEE 80-1986
343
IEEE 80-2000
EEE 665-1995
Calculations during Ground Grid Analysis: Following are the calculations performed during this
analysis:
The Step and Touch potentials for any rectangular/triangular/L-shaped/T-shaped confguration of
a ground grid, with or without ground rods (IEEE Std 80 and IEEE Std 665).
The tolerable Step and Mesh potentials and compares them with actual, calculated Step and Mesh
potentials (IEEE Std 80 and IEEE Std 665).
Graphic profles for the absolute Step and Touch voltages, as well as the tables of the voltages at
various locations (Finite Element Method).
The optimum number of parallel ground conductors and rods for a rectangular/triangular/L-
shaped/T-shaped ground grid. The cost of conductors/rods and the safety of personnel in the vi-
cinity of the substation/generating station during a ground fault are both considered.
The Ground Resistance and Ground Potential rise (GPR).
Short Circuit Analysis: ETAP short-circuit analysis program analyzes the effect of three-phase,
line-to-ground, line-to-line, and line-to-line-to-ground faults on the electrical distribution systems. The
program calculates the total short-circuit currents as well as the contributions of individual motors,
generators, and utility ties in the system. Fault duties are in compliance with the latest editions of the
ANSI/IEEE standards (C37 series) and IEC standards (IEC 909 and others).
3-Phase Faults: Device Duty: This study calculates momentary symmetrical and asymmetrical RMS,
momentary asymmetrical crest, interrupting symmetrical RMS, and interrupting adjusted symmetrical
RMS short-circuit currents at faulted buses.
3-Phase Faults: 30 Cycle Network: This study calculates short-circuit currents in their RMS values
after 30 cycles at faulted buses.
LG, LL, LLG, & 3-Phase Faults: Cycle: This study calculates short-circuit currents in their RMS
values at cycle at faulted buses.
LG, LL, LLG, & 3-Phase Faults: 1.5 to 4 Cycle: This study calculates short-circuit currents in
their RMS values between 1.5 to 4 cycles at faulted buses.
LG, LL, LLG, & 3-Phase Faults: 30 Cycle: This study calculates short-circuit currents in their RMS
values at 30-cycles at faulted buses.
344
Chapter 12
INTRODUCTION
In order to deal with the vagueness of human
thought, Zadeh (1965) first introduced fuzzy set
theory. A fuzzy set is a class of objects with a
continuum of grades of membership. Such a set
is characterized by a membership function which
assigns to each object a grade of membership
ranging between zero and one (Zadeh, 1965). A
fuzzy set is an extension of a crisp set. Crisp sets
only allow full membership or non-membership,
whereas fuzzy sets allow partial membership. In
other words, an element may partially belong
to a fuzzy set (Ertugrul and Karakasoglu, 2006;
Liang, 2008; Peidro et. al, 2010). Fuzzy sets and
fuzzy logic are powerful mathematical tools for
Pandian Vasant
University Technology Petronas, Malaysia
Solving Fuzzy Optimization
Problems of Uncertain
Technological Coeffcients
with Genetic Algorithms and
Hybrid Genetic Algorithms
Pattern Search Approaches
ABSTRACT
In this chapter a solution is proposed to a certain nonlinear programming diffculties related to the pres-
ence of uncertain technological coeffcients represented by vague numbers. Only vague numbers with
modifed s-curve membership functions are considered. The proposed methodology consists of novel
genetic algorithms and a hybrid genetic algorithm pattern search (Vasant, 2008) for nonlinear program-
ming for solving problems that arise in industrial production planning in uncertain environments. Real
life application examples in production planning and their numerical solutions are analyzed in detail.
The new method suggested has produced good results in fnding globally near-optimal solutions for the
objective function under consideration.
DOI: 10.4018/978-1-61350-138-2.ch012
345
Solving Fuzzy Optimization Problems of Uncertain Technological Coeffcients with Genetic Algorithms
modeling: uncertain systems in industry, nature
and humanity; and facilitators for commonsense
reasoning in decision making in the absence of
complete and precise information. Their role is
significant when applied to complex phenomena
not easily described by traditional mathematical
methods, especially when the goal is to find a
good approximate solution (Bojadziev and Bo-
jadziev, 1998; Zamirian et. al, 2009). Modeling
using fuzzy sets has proven to be an effective
way for formulating decision problems where the
information available is subjective and imprecise
(Zimmermann, 1992; Jimenez et. al, 2008).
A linguistic variable is a variable whose values
are words or sentences in a natural or artificial
language (Zadeh, 1975). As an illustration, Age
is a linguistic variable if its values are assumed to
be the fuzzy variables labeled young, not young,
very young, not very young, etc. rather than the
numbers 0, 1, 2, 3. (Bellman and Zadeh, 1977).
The concept of a linguistic variable provides a
means of approximate characterization of phe-
nomena which are too complex or too ill-defined
to be amenable to description in conventional
quantitative terms. The main applications of the
linguistic approach lie in the realm of humanistic
systemsespecially in the fields of artificial in-
telligence, linguistics, human decision processes,
pattern recognition, psychology, law, medical
diagnosis, information retrieval, economics and
related areas (Zadeh, 1975).
In this paper a novel, genetic algorithm (GA)
approach is reported that was successfully used for
solving problems originating from the uncertainty
of the technological coefficients in production
planning in industrial engineering. The main
reason for GA method is adopted in solving this
problem is given below.
Some of the advantages of GA over classical
optimization methods include (Deb, 2001):
1. GA is less susceptible to the complexity of
the problem at hand than non-evolutionary
methods;
2. They deal with multiple solutions in one run.
This is useful to achieve solutions rapidly in
the presence of a large number of parameters;
3. They allow the exploration of multiple local
optima;
4. GAs have successfully been applied to
various optimization problems that involve a
large number of parameters, multiple criteria,
and complex criteria relationships.
It is commonly believed that the main driving
principle behind the natural evolutionary process
is the Darwins survival-of-the-fittest principle
(Eldredge, 1989). In most scenarios, nature ruth-
lessly follows two simple principles:
1. If by genetic processing an above-average
offspring is created, it usually survives longer
than an average individual and thus has more
opportunities to produce offspring having
some of its traits than an average individual.
2. If, on the other hand, a below-average off-
spring is created, it usually does not survive
longer and thus gets eliminated quickly from
the population.
The two important main characteristics of
genetic algorithms are provided:
Exploration: The process of visiting entirely
new regions of a search space, to see if anything
promising may be found there. Unlike exploita-
tion, exploration involves leaps into the unknown.
Problems which have many local maxima can
sometimes only be solved by this sort of random
search.
Exploitation: When traversing a search space,
exploitation is the process using information
gathered from previously visited points in the
search space to determine which places might be
profitable to visit next. An example is hill climbing,
which investigates adjacent points in the search
space, and moves in the direction giving the great-
est increase in fitness. Exploitation techniques
346
are good at finding local maxima (Maric, 2010;
Stanimrovic, 2010).
Employing a GA is a global stochastic method
based on the mechanism of natural selection
and evolutionary genetics and is used in various
fields. For a specific problem such as nonlinear
programming, the combination of GA and other
methods such as the gradient based method can
outperform the GA alone. This can be illustrated
by experimental results (Honggang and Jianchao,
1997; Jeya Mala et. al, 2010). Hybrid genetic al-
gorithms have been adopted in this research work.
The GA approach that has been used in this
research work has some disadvantages regarding
global optimization. The major drawback of the
technique in this research work is that it is un-
able to locate the global optimum for a nonlinear
problem with uncertain technological coefficients
for industrial production planning. Therefore we
are motivated to use a more superior technique of
hybrid evolutionary computation to solve nonlin-
ear industrial production planning problems with
uncertain technological coefficients.
The rest of this paper is structured as follows:
Section 2 describes the membership function for
the technological coefficients for the mathemati-
cal model of the problem. Its followed by a brief
review of previous work on the GA approach for
production planning problems in Section 3. Sec-
tion 4 provides a description of the case study of
chocolate manufacturing problems. Illustration on
the problem formulation and mathematical model
for the case study is given in Section 5. Section 6
explains the computational results and findings
on the solutions of the optimization problems of
production planning in an uncertain environment.
The paper ends with a conclusion in Section 7.
FUZZY MEMBERSHIP FUNCTION
In applications it is often convenient to work
with s-curve membership functions for the fuzzy
numbers because of their computational flex-
ibility, and because they are useful in promoting
representations and information processing in
a fuzzy environment. In this study an s-curve
membership function for the fuzzy numbers in
uncertain technological coefficients of the fuzzy
optimization is adopted (Vasant, 2003; Vasant &
Barsoum, 2006).
Decision-makers face up to the uncertainty
and vagueness with subjective perceptions and
experiences in the decision-making process (Er-
tugrul and Karakasoglu, 2006). By using the fuzzy
optimization approach, uncertainty and vagueness
from subjective perception and the experiences of
a decision maker can be effectively represented
and a more effective decision reached (Sanchez,
Jimenez and Vasant, 2007). A quadratic objective
function for a production planning problem was
considered by Turabieh, Sheta and Vasant (2007).
In order to describe the real world problem of
production planning, a non-linear cubic function
is considered in this research work.
We formulate a non-linear cubic function for
the fuzzy optimization problem of industrial pro-
duction planning. This problem cannot be solved
by the fuzzy linear programming approach. There-
fore in this research work global optimization
techniques using genetic GAs have been adopted.
GENETIC ALGORITHMS
FOR PRODUCTION
PLANNING PROBLEMS
Since the late 1980s there has been a growing
interest in Genetic Gas (stochastic optimization
algorithms based on the principles of natural
[Darwinian] evolution). These have been used
widely for parameter optimization, classification
and learning. More recently, production plan-
ning has emerged as an area to which GAs can
be applied. A detailed introduction to GAs can
be found in Goldberg (1989). One of the earliest
reported applications of GAs to production was
reported by Davis (1985). It is a characteristic of
347
the robust optimization process that once fairly
good solutions have been formed their features
will be carried forward into better solutions and
lead ultimately to optimal solutions. It is in the
nature of GAs that new solutions are formed from
the features of known good solutions. Therefore,
it follows that GAs are particularly attractive for
production planning. Compared with other opti-
mization methods, GAs are suitable for traversing
large search spaces since they can do this relatively
rapidly and because the mutation operator diverts
the method away from local minima, which will
tend to become more common as the search space
increases in size. Suitability for large search spaces
is a useful advantage when dealing with produc-
tion problems of increasing size, since the solution
space grows very rapidly, especially when this is
compounded by such features as alternative profit
production optimization. It is important that these
large search spaces are traversed as rapidly as
possible to enable the practical and useful imple-
mentation of automated production optimization.
If the optimization is done quickly, production
managers can try out what-if scenarios and
detailed optimization analyses besides being able
to react to crises as soon as possible. Traditional
approaches to production planning optimization
such as mathematical programming and branch
and bound are computationally very slow in such
a massive search space. Researchers have found
that a key advantage of a GA is that it provides
a general purpose solution to the optimization
problem, with the peculiarities of any particular
example being accounted for in the fitness func-
tion without disturbing the logic of the standard
optimization (GA) routine. This means that it is a
relatively straightforward and convenient to adapt
the software implementation of the method to meet
the needs of particular applications.
Case Study Chocolate Manufacturing
Due to limitations in resources for manufacturing
a product and the need to satisfy certain conditions
in manufacturing and demand, a problem of fuzzi-
ness occurs in industrial systems. This problem
occurs, for example, in chocolate manufacturing
when deciding a mixed selection of raw materials
to produce varieties of products. This is referred
here to as the product-mix selection problem
(Tabucanon, 1996).
There are a number of products to be manu-
factured by mixing different raw materials and
using several varieties of processing. There are
limitations in raw material resources and facility
usage for the different types of processing. The
raw materials and facilities usage required for
manufacturing each product are expressed by
means of fuzzy coefficients. There are also some
constraints imposed by the marketing department
such as product-mix requirements, main product
line requirements and lower and upper limits of
Table 1. Profit Coefficients c
i
, d
i
and e
i
(Profit function in US $ per 10
3
units)
Product (x
i
) Synonym c
i
d
i
e
i
x
1
= Milk chocolate, 250g MC 250 c
1
= 180 d
1
= 0.18 e
1
= 0.01
x
2
= Milk chocolate, 100g MC 100 c
2
= 83 d
2
= 0.16 e
2
= 0.13
x
3
= Crunchy chocolate, 250g CC 250 c
3
= 153 d
3
= 0.15 e
3
= 0.14
x
4
= Crunchy chocolate, 100g CC 100 c
4
= 72 d
4
= 0.14 e
4
= 0.12
x
5
= Chocolate with nuts, 250g CN 250 c
5
= 130 d
5
= 0.13 e
5
= 0.15
x
6
= Chocolate with nuts, 100g CN 100 c
6
= 70 d
6
= 0.14 e
6
= 0.17
x
7
= Chocolate candy CANDY c
7
= 208 d
7
= 0.21 e
7
= 0.18
x
8
= Chocolate wafer WAFER c
8
= 83 d
8
= 0.17 e
8
= 0.16
348
demand for each product. It is necessary to obtain
maximum profit with a certain degree of satisfac-
tion for the decision-maker.
Problem Formulation
Optimization techniques are primarily used in
production planning problems in order to achieve
optimal profit; a certain objective function is
maximized by satisfying a number of constraints.
The first step in an optimal production planning
problem is to formulate the underlying nonlinear
programming (NLP) problem with uncertain tech-
nological coefficients by writing the mathematical
functions relating to the objectives and constraints.
Given a degree of possibility (), the fuzzy
constrained optimization problem can be formu-
lated (Vasant, 2006) as a nonlinear constrained
optimization problem shown. Tables 1, 2, 3, and
4 provide the input data for Equation 1 in the
problem statement and the values for c
i
, d
i
, e
i
,
a
ij
h
, a
ij
l
, r
i
, u
i
and b
j
.
The mathematical formulation (Equation 1),
of the problem is now described. The eight deci-
sion variables are: 250g of Milk Chocolate (x
1
),
100g of Milk Chocolate (x
2
), 250g of Crunchy
Chocolate (x
3
), 100g of Crunchy Chocolate (x
4
),
250g of Chocolate with Nuts (x
5
), 100g of Choc-
olate with Nuts (x
6
), Chocolate Candy (x
7
), and
Chocolate Wafer (x
7
). The objective is to find an
optimal profit value for the production planning
problem with uncertain technological coefficients.
The production planning problem is subject to a
number of constraints imposed by raw material
availability, facility capacity, and the sales depart-
ment. An upper and lower limit is imposed on
each of the decision variables (refer to Table 3).
In Equation 1 there are 21 other inequality con-
straints, as follows: (i) 17 constraints of facility
capacity (refer to Table 2) and resource variables
(refer to Table 4) (ii) 4 constraints by the sales
department (refer to Table 3).
The optimization formulation is,
Maximize
Subject to:
( ) c
i
x
i
d
i
x
i
e
i
x
i
i
a
ij
l
a
ij
h
a
ij
l

=
2 3
1
8
((
\
)
\
)
l
l
l
l
l
l
l
=
ln
1
1
1
C
B
i

88
0 1 2 17
0 15
1
6
7
8
0
1
0 6
2
0
=
=

x
i
b
j
j
r
i
x
i
r
i
x
i
i i
x x
, , ,...,
.
.
xx x
x x
x
i
u
i
i
C B
3
0 6
4
0
5
0 6
6
0
0 1 2 8
0 00 00 00

=
= =
.
.
, , ,...,
. , , 1 1 1 1 . . , . 0 00 0 1 99 and 1 45 m a

(1)
In the non-linear programming problem
(Equation 1), the variable vector x represents a
set of variables x
i
, i = 1, 2,, 8. The optimization
problem contains eight continuous variables and
21 inequality constraints. A test point x
i
satisfying
constraints is either feasible, or infeasible. The
set satisfying the constraints is called the feasible
domain. The aim of the optimization is to maxi-
mize the total production profit for the industrial
production planning problem. The practical mean-
ing of the constraints is available in Tabucanon
(1996). The vagueness factor represents fuzzy
numbers in uncertain technological parameters.
The objective function appears in work by Lin
(2007) and Chaudhuri (2007). The objective
function represents the profit.
COMPUTATIONAL RESULTS
Experimental studies are carried out for GAs and
hybrid GA pattern search (HGAPS) approaches.
In the following sections, the simulation results
for the GA approach is provided in detail and the
improved results of the HGAPS approach are
provided as well. In this research work MATLAB

toolboxes have been utilized for computation,
simulation, 2D and 3D plots.
349
Results and Discussion for
Genetic Algorithm Approach
The GA approach mimics biological processes
in evolving optimal or near optimal solutions to
problems. In this section a practical application
to challenging industrial problems of production
planning are investigated to exploit the full po-
tential of GA to look for a promising near global
optimal solution.
The results for the industrial production plan-
ning problems were obtained by using the follow-
ing genetic algorithm.
Algorithm
Initialization: Generate an initial population.
Initialize parameters and define fitness function.
Crossover: Perform crossover using arithme-
tic crossover.
Mutation: Perform mutation using adaptive
feasibility.
Selection: Evaluate the fitness of each indi-
vidual in the population. Choose N best chromo-
somes to form the next generation.
Termination Criteria: If stopping criteria are
satisfied, then terminate. Otherwise, select the
next generation and go to Step 2.
Table 2. Raw material and Facility usage required (per 10
3
units) (
ij
= [a
l
ij
, a
h
ij
]) and Availability (b
j
)
Material or
Facility
MC 250 MC 100 CC 250 CC100 CN250 CN100 Candy Wafer
Cocoa (kg) [66, 109] [26, 44] [56,9] [22,37] [37,62] [15,25] [45, 75] [9, 21]
Milk (kg) [47, 78] [19, 31] [37,6] [15,25] [37,62] [15,25] [22, 37] [9, 21]
Nuts (kg) [0, 0] [0, 0] [28,4] [11,19] [56,94] [22,37] [0, 0] [0, 0]
Cons. sugar (kg) [75, 125] [30, 50] [66,109] [26,44] [56,94] [22,37] [157,262] [18,30]
Flour (kg) [0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [54,90]
Alum. foil (ft
2
) [375,625] [0, 0] [375,625] [0, 0] [0, 0] [0, 0] [0, 0] [187,312]
Paper (ft
2
) [337,562] [0, 0] [337,563] [0, 0] [337,562] [0, 0] [0, 0] [0, 0]
Plastic (ft
2
) [45, 75] [95, 150] [45, 75] [90,150] [45,75] [90,
150]
[1200,200] [187,312]
Cooking(ton-
hours)
[0.4, 0.6] [0.1, 0.2] [0.3, 0.5] [0.1, 0.2] [0.3,0.4] [0.1,
0.2]
[0.4, 0.7] [0.1,0.12]
Mixing (ton-
hours)
[0, 0] [0, 0] [0.1, 0.2] [0.04,0.07] [0.2, 0.3] [0.07,
0.12]
[0, 0] [0, 0]
Forming(ton-
hours)
[0.6, 0.9] [0.2, 0.4] [0.6, 0.9] [0.2, 0.4] [0.6, 0.9] [0.2,
0.4]
[0.7, 1.1] [0.3, 0.4]
Grinding(ton-
hours)
[0, 0] [0, 0] [0.2, 0.3] [0.07, 0.12] [0, 0] [0, 0] [0, 0] [0, 0]
Wafer making
(ton-hours)
[0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [0, 0] [0.2, 0.4]
Cutting (hours) [0.07,0.2] [0.07,0.12] [0.07,0.12] [0.07, 0.12] [0.07, 0.12] [0.07,
0.12]
[0.15, 0.25] [0, 0]
Packaging1
(hours)
[0.2, 0.3] [0, 0] [0.2, 0.3] [0, 0] [0.2, 0.3] [0, 0] [0, 0] [0, 0]
Packaging2
(hours)
[0.04,0.6] [0.2, 0.4] [0.04, 0.06] [0.2, 0.4] [0.04, 0.06] [0.2,
0.4]
[1.9, 3.1] [0.1, 0.2]
Labour (hours) [0.2, 0.4] [0.2, 0.4] [0.2, 0.4] [0.2, 0.4] [0.2, 0.4] [0.2,
0.4]
[1.9, 3.1] [1.9, 3.1]
350
Output the best solutions.
Description of the Genetic
Algorithm Parameter Setting
Fitness function (Equation 2):
Maximize ( ) c
i
x
i
d
i
x
i
e
i
x
i
i

=
2 3
1
8

(2)
The fitness function is the objective function
that has to be maximized.
Mutation functions make small random
changes in individuals in the population, which
provide genetic diversity and enable the GA to
search a broader space. A step length is chosen
along each direction so that linear constraints and
bounds are satisfied.
Crossover combines two individuals, or par-
ents, to form a new individual, or child, for the
next generation. The following has been specified
that performs the crossover in the crossover func-
tion field. Arithmetic crossover creates children
that are the weighted arithmetic mean of the two
parents. Children are feasible with respect to linear
constraints and bounds.
Stopping Criteria Options
The stopping criteria determines what causes the
algorithm to terminate. The following options
have been considered in the simulation of the GA.
Generations specifies the maximum number
of iterations the GA performs.
Time Limit specifies the maximum time in sec-
onds the genetic algorithm runs before stopping.
Table 3. Demand (u
k
) and Revenues/Sales (r
k
) in US $ per 10
3
units
Product (x
k
) Synonym Demand (u
k
) Revenues/Sales (r
k
)
x
1
= Milk chocolate, 250g MC 250 u
1
= 500 r
1
= 375
x
2
= Milk chocolate, 100g MC 100 u
2
= 800 r
2
= 150
x
3
= Crunchy chocolate, 250g CC 250 u
3
= 400 r
3
= 400
x
4
= Crunchy chocolate, 100g CC 100 u
4
= 600 r
4
= 160
x
5
= Chocolate with nuts, 250g CN 250 u
5
= 300 r
5
= 420
x
6
= Chocolate with nuts, 100g CN 100 u
6
= 500 r
6
= 175
x
7
= Chocolate candy CANDY u
7
= 200 r
7
= 400
x
8
= Chocolate wafer WAFER u
8
= 400 r
8
= 150
Table 4. Raw material availability (b
j
)
Material or Facility Availability
Cocoa (kg) 100000
Milk (kg) 120000
Nuts (kg) 60000
Cons. sugar (kg) 200000
Flour (kg) 20000
Alum. foil (ft
2
) 500000
Paper (ft
2
) 500000
Plastic (ft
2
) 500000
Cooking (ton-hours) 1000
Mixing (ton-hours) 200
Forming (ton-hours) 1500
Grinding (ton-hours) 200
Wafer making (ton-hours) 100
Cutting (hours) 400
Packaging 1 (hours) 400
Packaging 2 (hours) 1200
Labour (hours) 1000
351
Fitness Limit: If the best value is less than or
equal to the value of Fitness Limit, the algorithm
stops.
Stall Generations: If the weighted average
change in the fitness function value over Stall
Generations is less than Function Tolerance, the
algorithm stops.
Stall Time Limit: If there is no improvement
in the best fitness value for an interval of time
in seconds specified by Stall Time Limit, the
algorithms stops.
Function Tolerance: If the cumulative
change in the fitness function value over Stall
Generations is less than Function Tolerance, the
algorithm stops.
Among the above options, the final results show
that the Fitness Limit option has been utilized
successfully in the simulation results.
Number of Runs: 20, 26, 31 and 95.
Equation 1 was solved by the using the GA
with the following parameters.
= Degree of possibility
Table 5. Optimal value for Objective Function
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
f time (s)
0.001 0.69 2.26 0.00 0.00 0.41 2.79 0.00 0.00 559.8 0.012
0.1 0.51 0.00 0.78 0.00 0.00 0.00 0.00 1.31 320.4 0.058
0.2 1.47 0.00 0.00 0.00 0.00 0.04 0.00 0.00 268.3 0.057
0.3 0.35 0.00 0.23 0.15 0.00 0.00 0.25 0.00 161.5 0.056
0.4 0.22 0.00 0.00 0.00 1.78 0.06 0.75 0.00 430.3 0.055
0.5 0.71 0.34 0.56 0.00 1.03 0.00 0.00 0.00 375.0 0.055
0.6 0.53 0.00 0.00 0.47 0.70 0.00 1.06 0.00 441.0 0.055
0.7 0.00 0.51 0.00 0.00 0.79 0.70 0.68 0.00 336.7 0.055
0.8 0.16 0.00 0.00 0.00 1.91 0.27 0.99 0.95 583.0 0.056
0.9 0.76 0.16 0.00 0.00 0.78 0.00 1.38 0.85 250.2 0.055
0.99 1.08 0.80 0.00 0.00 0.78 0.00 1.38 0.85 719.6 0.056
Table 6. Optimal value for Fitness function
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
f time (s)
0.001 0.18 0.00 0.00 0.00 1.46 1.80 0.00 0.39 379.2 0.420
0.1 0.00 0.11 0.81 0.00 0.00 0.28 0.29 0.00 213.5 0.212
0.2 0.00 0.90 0.00 0.00 0.72 0.30 1.55 0.00 510.6 0.201
0.3 0.00 0.00 0.00 0.93 0.00 0.00 0.00 0.78 131.8 0.200
0.4 0.60 0.94 0.00 0.00 0.08 0.00 0.00 0.00 197.3 0.205
0.5 1.35 2.14 0.17 0.00 0.28 0.39 0.00 0.18 523.7 0.202
0.6 0.00 0.00 0.82 0.00 0.00 0.00 0.00 0.69 182.5 0.233
0.7 0.00 0.32 0.51 0.00 0.10 0.00 0.69 1.88 416.5 0.210
0.8 0.32 0.00 0.03 0.00 0.00 0.48 0.00 0.39 128.9 0.203
0.9 0.05 0.00 0.24 0.72 0.15 0.13 0.00 1.31 236.2 0.207
0.99 1.20 0.00 0.00 0.39 0.00 0.00 2.64 0.00 791.4 0.310
352
= Level of satisfaction
= Vagueness factor
f = Objective function or fitness function
x
i
= Decision variables
Time in seconds = CPU (s)
The best optimal solution for the fitness func-
tion and the best feasible solution for the decision
variables with respect to the level of satisfaction
is reported in Table 5. Figure 1, Figure 2, and
Figure 3 provide solutions for = 13.813.
The Total CPU time for the above results is
0.57 s and the average CPU time is 0.052 for =
1 to = 0.99. Number of runs = 50, population
size = 15 and number of generation = 20.
Figure 1. Fitness value f and level of satisfaction
Figure 2. Fitness value (scores) and final population
353
Figure 3. Final population for = 0.99
Figure 4. Fitness value f and generation
354
Figure 4 and Figure 5 show the simulation
results for number of runs = 20, population size
= 100 and generation = 10000. The computed
results for best optimal value for fitness function,
best feasible solutions for decision variables with
respect to is reported in Table 6.
Experimental simulation results for Figure 6,
total CPU time = 2.603 seconds and average CPU
time = 0.237 seconds. Further experiments were
carried out by increasing the number of runs by
31 with population size = 100 and number of
generation = 10000.
Table 7 reports the best optimal solution for
fitness function f, and the best feasible decision
variables with respect to . The total CPU time =
3.155 seconds and average CPU time = 0.287
seconds for running the simulation.
Figure 7. Current best individual solutions x
i
at = 0.99
355
The best individual solution for the decision
variables is shown in Figure 7 for the experiment
with increased population size = 100 and number
of generation = 10000.
Table 8 reports the best optimal value of fitness
function at = 0.99 is 1222.6 and total CPU time
= 1.981 seconds with average CPU time = 0.180
seconds. There is a tremendous improvement in
the fitness function value when the population
size is increased to 100.
A thorough investigation was carried out with
following parameter settings in the genetic algo-
rithms techniques:
Population size = 100; Generation = 10000;
Elite Count = 20; Selection = Uniform, roulette
and tournament; Fitness scaling = proportional,
top and shift linear; Crossover functions = Inter-
mediate, heuristic, arithmetic, single point, two
point and Mutation function = Adaptive feasible.
Table 9 reports very important information on
the simulation results obtained for = 1 to =
25 by independent runs. The total CPU time (s)
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
f time (s)
0.001 0.63 0.00 1.49 0.00 0.00 0.00 0.00 0.00 340.7 0.220
0.1 0.40 1.44 0.00 0.00 0.00 0.69 0.00 0.00 240.4 0.199
0.2 0.00 0.00 0.31 0.68 0.00 0.00 0.00 0.00 96.20 0.466
0.3 0.56 0.00 0.00 0.00 0.00 0.44 0.88 0.71 375.1 0.208
0.4 0.00 0.00 1.33 0.00 1.20 0.13 0.00 0.53 412.6 0.670
0.5 0.00 1.18 1.76 1.70 0.00 0.00 0.00 0.00 489.2 0.191
0.6 0.00 0.00 0.41 0.60 0.00 0.00 0.02 0.00 110.1 0.195
0.7 0.00 0.58 0.61 0.00 0.95 0.00 0.00 0.00 265.6 0.207
0.8 0.00 0.03 0.00 0.00 1.01 1.63 0.00 0.37 277.6 0.192
0.9 1.36 0.00 0.00 0.62 0.00 0.00 0.00 0.13 299.3 0.404
0.99 1.09 0.47 1.27 1.18 0.21 0.00 0.00 0.46 579.9 0.203
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
f time (s)
0.001 0.62 0.00 0.00 0.00 1.26 0.86 0.00 0.00 336.0 0.194
0.1 0.55 0.00 0.00 0.00 0.49 0.00 0.08 0.00 180.3 0.183
0.2 0.00 0.00 0.00 0.00 0.00 0.00 1.09 1.32 337.1 0.178
0.3 0.00 0.00 0.21 0.00 0.06 0.00 1.69 0.33 418.5 0.179
0.4 0.72 1.60 0.00 0.00 0.18 0.53 0.00 0.30 345.9 0.176
0.5 0.00 0.00 0.00 0.96 0.00 0.00 0.00 0.28 92.00 0.179
0.6 0.00 1.60 0.57 0.32 0.00 0.31 1.34 0.00 545.1 0.180
0.7 1.30 0.00 0.00 0.88 0.58 0.00 0.00 0.57 419.7 0.179
0.8 0.00 0.00 0.00 0.00 0.00 0.00 0.29 0.00 60.40 0.179
0.9 1.31 0.66 0.00 0.00 0.00 0.00 0.37 0.90 444.0 0.177
0.99 1.28 0.00 0.61 1.95 2.26 0.00 2.24 0.00 1222.6 0.177
356
and an average CPU time (s) for each from 1
to 41 with respect to the range = 0.001 to =
0.99 provided in the last two rows of the Table 9.
Unfortunately, there is no solution for the fitness
function for = 45 at = 0.99. This is the draw
back of this GA algorithm for the higher value of
in an independent run.
A 3D plot for the fitness value, vagueness
factor and level of satisfaction for = 1 to = 41
is shown in Figure 8. The result obtained for =
45 at = 0.99 motivates to further investigate this
simulation with a single run for = 1 to = 41.
Figure 9 shows a 3D solution obtained with a
single run of the GA for = 1 to = 41. The
optimal value for the fitness function is 1086 at
= 41 and = 0.99. The solution for the problem
is further investigated for values between 39
and 59. Figure 9 was obtained for the solution of
fitness function f.
Population size of 2000 and number of gen-
eration of 100000 with arithmetic cross over and
adaptive feasible mutation have been used in the
above result. The optimal solution for the fitness
function is 925.53 at = 41 and = 0.2. Compu-
tational time for this solution is 0.148 seconds.
Figure10 shows the solution for a run number
of 26. In this case, the number of generations used
is 1000000 and the population size is 100.
Figure 11 shows the optimal results for
values from 39 to 57. The optimal value occurs
at = 41 and = 0.5 with fitness value 1302.59.
The CPU time for this solution is 0.056 seconds.
The simulation was carried out for 95 runs and
the following result obtained by using a popula-
tion size of 200 and the number of generations
equal to 1,000,000.
The number of runs carried out in this simula-
tion results for the optimal fitness function value
are 20, 26, 31 and 95. This is the largest number
of runs that have been considered in these simu-
lation results.
The optimal value for the fitness function is
1302.59, obtained at = 41 and = 0.5 and the
computational time is 0.055 seconds.
The simulation was run for = 1 to = 41 and
the results for the fitness function is 1302.59 at
= 21 and = 0.9. The computational time for
running this solution is 0.056. Figure 12, Figure
13, Figure 14, and Figure 15 give complete views
of the various dimensions of the solutions.
Table 9. Optimal values of Fitness function for = 1 to = 45
= 1 = 5 = 9 =3 =17 =21 =25 =29 =33 =37 =41 =45
0.01 370.1 352.8 412.7 352.8 412.7 48.70 216.3 352.8 370.1 352.8 412.7 216.3
0.1 287.2 392.7 562.0 392.7 562.0 36.10 547.7 392.7 287.2 392.7 562.0 547.7
0.2 518.0 81.60 925.5 81.60 925.5 433.3 616.6 81.60 518.0 81.60 925.5 616.6
0.3 360.0 348.3 248.1 348.3 248.1 445.4 408.8 348.3 360.0 348.3 248.1 408.8
0.4 605.7 154.8 207.8 154.8 207.8 254.2 801.0 154.8 605.7 154.8 207.8 801.0
0.5 205.7 418.1 485.7 418.1 485.7 492.1 154.9 418.1 205.7 418.1 485.7 154.9
0.6 351.2 431.4 453.3 431.4 453.3 414.6 550.6 431.4 351.2 431.4 453.3 550.6
0.7 295.9 47.70 539.5 47.70 539.5 441.5 707.2 47.70 295.9 47.70 539.5 707.2
0.8 475.5 194.4 442.5 194.4 442.5 420.7 551.1 194.4 475.5 194.4 442.5 551.1
0.9 337.2 198.2 408.2 198.2 408.2 88.60 372.4 198.2 337.2 198.2 408.2 372.4
0.99 366.4 311.6 243.1 311.6 243.1 288.1 703.1 311.6 366.4 311.6 243.1 0.000
TT 6.234 3.338 4.312 4.393 3.761 4.051 4.128 2.058 8.306 3.556 3.487 2.128
AT 0.567 0.303 0.392 0.399 0.342 0.368 0.375 0.187 0.755 0.323 0.317 0.193
TT = Total time and AT = Average time
357
The contour in Figure 16 clearly explains the
situation where by the optimal solution for the
fitness function is stuck at 1221 for = 39 and
= 59. In order to to escape from this solution,
another hybrid method is needed to further improve
the final optimal solution.
Results and Discussion for Hybrid
Genetic Algorithm Pattern Search
Approach
The hybrid GA and general pattern search is a
combination of a GA and a general pattern search.
Below is the algorithm for this approach.
Step 1: Initial population: Generate initial
population randomly.
Step 2: Genetic operators: Selection: sto-
chastic uniform, Crossover: arithmetic
crossover operator Mutation, Adaptive fea-
sible mutation operator
Step 3: Evaluation: Best offspring and ft-
ness function.
Step 4: Stop condition: If a pre-defned
maximum generation number, time limit,
ftness limit reached or an optimal solution
is located during the genetic search pro-
cess, then stop; otherwise, go to Step 2.
Step 5: Continue with general pattern
search techniques (Vasant & Barsoum,
2009).
Step 6: End
Figure 17 provides the best optimal solution
for the fitness function with respect to a level of
satisfaction of = 13.813.
Table 10 provides the solution for the best
fitness function and the decision variables at =
13.813. The CPU time to run the simulation for
= 13.813 is 16 minutes. The best optimal solu-
tion for the fitness function is 180776.1 at = 0.1.
Figure 18. explains the detailed solution for
the decision variables versus level of satisfaction.
The CPU time for = 13.813 at = 0.99 is 78.375
seconds.
The solution for the decision variable x
8
is
unrealistic since for the majority values of , x
8
is
zero. Therefore, this solution needs some improve-
ment. For this purpose, the following hybrid
methods is thoroughly investigated.
Figure 8. Fitness value, vagueness and level of satisfaction
358
359
Fitness values for various values of from 1 to
41 are given in Figure 19. The optimal value for
the fitness function occurs at = 9 and = 0.99 is
197980.93. The total CPU time for the simulation
of Figure 19 is 2 hours 18 minutes.
Figure 20 provides a holistic view of the 3D
mesh plot solution for various values of , and
with respect to the level of satisfaction. The CPU
time for the 3D mesh plot is 2 hours 52 minutes.
The main contribution of the hybrid HGAPS
approach is that it is able to locate the best near
global optimal solution for the fitness function
which is far superior to the GA alone. On the
other hand, it is unable to provide a reasonable
computational CPU time for the fitness function
values.
Figure 12. Fitness value f, vagueness factor and level of satisfaction
Figure 13. Vagueness and level of satisfaction
360
Figure 15. Fitness value f and vagueness factor
361
CONCLUSION
The HGAPS algorithm drives the solution to the
optimal point where all constraints are satisfied.
Moreover, for problem having a moderate type of
constraints such as a linear constraint, the HGAPS
simulation results guarantee an accurate and robust
solution with less computational time and greater
convergence stability than classical optimization
techniques such as line search (Vasant, 2008).
The outcomes of the simulation of the GA
algorithm were not satisfactory for a moderate
non-linear fitness function with 21 inequality
linear constraints and 8 bound constraints for a
production planning problem. The simulation
was carried out for various numbers of runs
and generations. Most of the time, the simula-
tion for the GA algorithm produced very poor
feasible solutions for the decision variables and
local optimal solutions for the fitness function.
Figure 16. Contour plots for vagueness and level of satisfaction
Table 10. Optimal value for Fitness Function (f)
x
1
x
2
x
3
x
4
x
5
x
6
x
7
x
8
f
0.001 0.62 0.80 0.94 0.00 0.21 0.24 0.00 0.00 366.4
0.1 346.7 580.5 276.2 460.4 148.7 247.9 189.2 4.54 180776.1
0.2 0.43 0.90 0.73 0.58 0.04 0.68 0.57 0.00 475.5
0.3 314.8 577.5 291.0 485.7 104.2 173.7 175.9 5.34 173329.7
0.4 0.00 0.00 0.00 0.61 0.51 1.69 0.59 0.00 351.2
0.5 278.2 466.2 282.0 521.1 156.5 260.8 164.2 0.00 170805.1
0.6 280.6 497.6 269.3 493.6 146.5 244.2 161.4 0.00 169598.1
0.7 258.1 602.6 260.3 472.7 128.5 214.2 153.6 0.00 166739.9
0.8 305.2 512.4 238.8 398.0 136.8 228.1 159.4 0.00 166566.4
0.9 257.7 444.3 269.2 450.5 146.1 243.6 148.2 0.00 163208.3
0.99 283.8 566.7 153.1 255.2 106.4 177.3 123.5 0.00 144803.8
362
To overcome these major drawbacks, a hybrid
approach has been adopted in this study for the
global optimal solutions. The major drawback of
the GA algorithms is that is not able to provide a
global optimal solution for the fitness function.
This is possibly due to premature convergence.
Further investigation has been carried out by the
hybridize technique HGAPS in order to find a
global optimal solution for the objective func-
tion. The final optimal solution for the objective
function is very promising and is superior to the
GA approach alone. The major goal of this paper
is finally achieved successfully.
Figure 18. Decision variables x
i
and level of satisfaction
363
Figure 20. Mesh plot for fitness value f, and
364
REFERENCES
Bellman, R. E., & Zadeh, L. A. (1977). Local
and fuzzy logics. In Dunn, J. M., & Epstein, G.
(Eds.), Modern uses of multiple-valued logic (pp.
105151). Kluwer Academic Publishers.
Bojadziev, G., & Bojadziev, M. (1998). Fuzzy sets
and fuzzy logic applications. Singapore: World
Scientific Publishing.
Buzacott, J. A., & Shanthikumar, J. G. (1993).
Stochastic models of manufacturing systems.
Englewood Cliffs, NJ: Prentice-Hall.
Chaudhuri, K. (March 2007). Personal com-
munication.
Chong, T. C., Anderson, D. C., & Mitchell,
O. R. (1989). QTC - And integrated design/
manufacturing/inspection system for poismatic
parts. Proceedings of the ASME Conference on
Computers and Engineering (pp. 417-426), San
Francisco, CA.
Davis, L. (1985). Job shop scheduling with genetic
algorithms. In Proceedings of the 1st International
Conference on Genetic Algorithms. Lawrence
Erlbaum.
Deb, K. (2001). Multi-objective optimization using
evolutionary algorithms. New York, NY: Wiley.
Eldredge, N. (1989). Macro-evolutionary dynam-
ics: Species, niches, and adaptive peaks. New
York, NY: McGraw-Hill.
Erturul, I., & Karakaolu, N. (2006). Fuzzy
TOPSIS method for academic member selection
in engineering faculty. International Joint Con-
ferences on Computer, Information, and Systems
Sciences, and Engineering (CIS2E 06).
Goldberg, D. E. (1989). Genetic algorithms in
search optimization and machine learning. To-
ronto, Canada: Addison Wesley.
Honggang, W., & Jianchao, Z. (1997). The
hybrid genetic algorithm for solving nonlinear
programming. IEEE International Conference on
Intelligent Processing Systems. Beijing, China.
Jeya Mala, D., Ruby, S., & Mohan, V. (2010).
A hybrid test optimization framework coupling
genetic algorithm with local search technique.
Computing and Informatics, 29, 133164.
Jimenez, F., Sanchez, G., & Vasant, P. (2008).
Fuzzy optimization via multi objective evolution-
ary computation for chocolate manufacturing. In
Kahraman, C. (Ed.), Fuzzy multicriteria decision
making and applications with recent developments
(pp. 523538). Springer. doi:10.1007/978-0-387-
76813-7_20
Lapedes, A., & Farber, R. F. (1988). How neural
networks work. In Anderson, D. Z. (Ed.), Neural
information processing systems (pp. 442456).
New York, NY: American Institute Physics.
Liang, T. F. (2008). Interactive multi-objective
transportation planning decisions using fuzzy
linear programming. Asia Pacific Journal of
Operational Research, 25, 1131. doi:10.1142/
S0217595908001602
Lin, F. T. (March 2007). Personal communication.
Maric, M. (2010). An efficient genetic algorithms
for solving the multi level incapacitated facility
location problem. Computing and Informatics,
29, 183201.
MATLAB. (2007). Users guide. The Math Works.
Parmee, I. C. (2003). Poor-definition, uncertainty
and human factorsA case for interactive evolu-
tionary problem reformulation. Proceedings of the
3
rd
IEC Workshop of the Genetic and Evolutionary
Computation Conference (GECCO).
365
Peidro, D., Mula, J., & Poler, R. (2010). Fuzzy
linear programming for supply chain planning
under uncertainty. International Journal of In-
formation Technology & Decision Making, 9,
373392. doi:10.1142/S0219622010003865
Rennard, J. P. (2000). Introduction to genetic algo-
rithms. Retrieved May 12, 2008, from http://www.
rennard.org/alife /english /gavintrgb.html#Evol
Sadeh-Koniecpol, N., Hildum, D., Laliberty,
T. J., Smith, S., McANulty, J., & Kjenstad, D.
(1996). An integrated process-planning/produc-
tion-scheduling shell for agile manufacturing.
(Technical Report CMU-RI-TR-96-10), Robotics
Institute, Carnegie Mellon University.
Sanchez, G., Jimenez, F., & Vasant, P. (2007).
Fuzzy optimization with multi-objective evolu-
tionary algorithms: A case study. Proceedings of
the 2007 IEEE Symposium on Computational
Intelligence in Multi-criteria Decision Making
(pp. 58-64). Honolulu, Hawaii.
Stanimrovic, Z. (2010). A genetic algorithm ap-
proach for the capacitated single allocation P-Hub
median problem. Computing and Informatics,
29, 117132.
Swaminathan, J. M., Smith, S., & Sadeh-Koniecpol,
N. (1998). Modeling the dynamics of supply
chains: A multi-agent approach. Decision Scienc-
es, 29, 607632. doi:10.1111/j.1540-5915.1998.
tb01356.x
Tabucanon, M. T. (1996). Multi objective pro-
gramming for industrial engineers. Mathemati-
cal programming for industrial engineers (pp.
487542). New York, NY: Marcel Dekker, Inc.
Takagi, H. (2001). Interactive evolutionary com-
putation: Fusion of the capabilities of EC compu-
tation and human evaluations. Proceedings of the
IEEE, 89(9), 12751296. doi:10.1109/5.949485
Turabieh, H., Sheta, A., & Vasant, P. (2007).
Hybrid optimization genetic algorithm (HOGA)
with interactive evolution to solve constraint
optimization problems for production systems.
International Journal of Computational Science,
1(4), 395406.
Vasant, P. (2003). Application of fuzzy linear
programming in production planning. Fuzzy
Optimization and Decision Making, 3, 229241.
doi:10.1023/A:1025094504415
Vasant, P. (2006). Fuzzy production planning and
its application to decision making. Journal of Intel-
ligent Manufacturing, 15(1), 512. doi:10.1007/
s10845-005-5509-x
Vasant, P. (2011). Hybrid optimization for deci-
sion making in an uncertain environment. LAP
LAMBERT Academic Publishing, 244 pages,
Germany.
Vasant, P., & Barsoum, N. N. (2006). Fuzzy
optimization of units products in mix-products
selection problem using FLP approach. Soft
Computing Journal, 10(2), 144151. doi:10.1007/
s00500-004-0437-9
Vasant, P., & Barsoum, N. N. (2009). Hybrid
general pattern search and simulated annealing
for industrial production planning problems. 3
rd

Global Conference on Power Control and Optimi-
zation, 2-4
th
February 2010, Gold Coast, Australia.
Zadeh, L. A. (1965). Fuzzy sets. Information
and Control, 8, 338353. doi:10.1016/S0019-
9958(65)90241-X
Zadeh, L. A. (1975). The concept of a linguis-
tic variable and its application to approximate
reasoning. Information Sciences, 8, 199249.
doi:10.1016/0020-0255(75)90036-5
366
Zamirian, M., Kamyad, A. V., & Farahi, M. H.
(2009). A novel algorithm for solving optimal
path planning problems based on parametrization
method and fuzzy aggregation. Physics Letters.
[Part A], 373, 34393449. doi:10.1016/j.phys-
leta.2009.07.018
Zimmermann, H. J. (1992). Fuzzy set theory and
its applications. Boston, MA: Kluwer Academic
Publishers.
ADDITIONAL READING
Abraham, A., Vasant, P., & Bhattacharya, A.
(2008). Neuro-Fuzzy Approximation of Multi-
Criteria Decision-Making QFD Methodology.
In Kahraman, C. (Ed.), Fuzzy Multi-Criteria
Decision Making Theory and Applications with
Recent Developments (pp. 301324). Springer.
doi:10.1007/978-0-387-76813-7_12
Aliev, R. A., Fazlollahi, B., Guirimov, B. G., &
Aliev, R. R. (2007). Fuzzy-genetic approach to
aggregate production-distribution planning in
supply chain management. Information Sciences,
177, 42414255. doi:10.1016/j.ins.2007.04.012
Bhattacharya, A., Abraham, A., & Vasant, P.
(2008). FMS Selection Under Disparate Level-
of-Satisfaction of Decision Making Using an
Intelligent Fuzzy-MCDM- Model. In Kahraman,
C. (Ed.), Fuzzy Multi-Criteria Decision Making
Theory and Applications with Recent Develop-
ments (pp. 263280). Springer. doi:10.1007/978-
0-387-76813-7_10
Bhattacharya, A., Vasant, P., & Susanto, S. (2007).
Simulating theory of constraint problem with a
novel fuzzy compromise linear programming
model. In A. Elsheikh, A. T. Al Ajeeli, & E. M.
Abu-Taieh. Simulation and Modeling: Current
Technologies and Applications (pp. 307-336).
IGI Publisher.
Buckley, J. J., & Feuring, T. (2000). Evolutionary
algorithm solution to fuzzy problems: Fuzzy linear
programming. Fuzzy Sets and Systems, 109, 3553.
doi:10.1016/S0165-0114(98)00022-0
Elamvazuthi, I., Vasant, P., & Ganesan, T. (2010).
Fuzzy linear programming using modified logistic
membership function. International Review of
Automatic Control, 3(4), 370377.
Jen, M., Ida, K., Lee, J., & Kim, J. (1997). Fuzzy
nonlinear goal programming using genetic algo-
rithm. Computers & Industrial Engineering, 33,
3942. doi:10.1016/S0360-8352(97)00036-3
Jimenez, F., Cadenas, J. M., Verdegay, J. L., &
Sanchez, G. (2003). Solving fuzzy optimization
problems by evolutionary algorithms. Information
Sciences, 152, 303311. doi:10.1016/S0020-
0255(03)00074-4
Jimenez, F., Sanchez, G., & Vasant, P. (2008).
Fuzzy Optimization via Multi-Objective Evolu-
tionary Computation for Chocolate Manufactur-
ing. In Kahraman, C. (Ed.), Fuzzy Multi-Criteria
Decision Making Theory and Applications with
Recent Developments (pp. 523538). Springer.
doi:10.1007/978-0-387-76813-7_20
Liang, T. F. (2008). Interactive multi-objective
transportation planning decisions using fuzzy
linear programming. Asia Pacific Journal of
Operational Research, 25, 1131. doi:10.1142/
S0217595908001602
Madronero, M. D., Peidro, D., & Vasant, P.Vendor
selection problem by using an interactive fuzzy
multi-objective approach with modified s-curve
membership functions. Computers & Mathemat-
ics with Applications (Oxford, England), 60,
10381048. doi:10.1016/j.camwa.2010.03.060
Modal, S., & Maiti, M. (2002). Multi-item fuzzy
EOQ models using genetic algorithm. Comput-
ers & Industrial Engineering, 44, 105117.
doi:10.1016/S0360-8352(02)00187-0
367
Peidro, D., Mula, J., & Poler, R. (2010). Fuzzy
linear programming for supply chain planning
under uncertainty. International Journal of In-
formation Technology & Decision Making, 9,
373392. doi:10.1142/S0219622010003865
Sasaki, M., & Gen, M. (2003). Fuzzy multiple
objective optimal system design by hybrid ge-
netic algorithm. Applied Soft Computing, 2(3F),
189196. doi:10.1016/S1568-4946(02)00068-6
Varela, L. R., & Ribeiro, R. A. (2003). Evaluation
of simulated annealing to solve fuzzy optimiza-
tion problems. Journal of Intelligent & Fuzzy
Systems, 14, 5971.
Vasant, P. (2010). Hybrid simulated annealing
and genetic algorithms for industrial production
management problems. International Journal
of Computational Methods, 7(2), 279297.
doi:10.1142/S0219876210002209
Vasant, P. (2010). Innovative hybrid genetic
algorithms and line search method for indus-
trial production management. In Chis, M. (Ed.),
Computation and Optimization Algorithms in
Software Engineering: Application and Tech-
niques (pp. 142160). Hershey, PA: IGI Global.
doi:10.4018/978-1-61520-809-8.ch008
Vasant, P. (2010). Hybrid optimization techniques
for industrial production planning. Journal of
Computer Science & Technology, 10(3), 150151.
Vasant, P., & Barsoum, N. (2009). Hybrid genetic
algorithms and line search method for indus-
trial production planning with non-linear fitness
function. Engineering Applications of Artificial
Intelligence, 22, 767777. doi:10.1016/j.engap-
pai.2009.03.010
Vasant, P., & Barsoum, N. (2010). Hybrid pattern
search and simulated annealing for fuzzy produc-
tion planning problem. Computers & Mathemat-
ics with Applications (Oxford, England), 60,
10581067. doi:10.1016/j.camwa.2010.03.063
Vasant, P., Barsoum, N., Kahraman, C., & Di-
mirovski, G. (2007). Application of fuzzy optimi-
zation in forecasting and planning of construction
industry. In Vrakas, D., & Vlahavas, I. (Eds.),
Artificial Intelligent for Advanced Problem Solv-
ing Technique. IGI publisher.
Vasant, P., Bhattacharya, A., & Abraham, A.
(2008). Measurement of Level-of-Satisfaction
of Decision Maker in Intelligent Fuzzy-MCDM
Theory: A Generalized Approach. In Kahraman,
C. (Ed.), Fuzzy Multi-Criteria Decision Making
Theory and Applications with Recent Develop-
ments (pp. 235262). Springer. doi:10.1007/978-
0-387-76813-7_9
Vasant, P., Elamvazuthi, I., & Webb, J. F. (2010).
Fuzzy technique for optimization of objec-
tive function with uncertain resource variables
and technological coefficients. International
Journal of Modeling, Simulation, and Scien-
tific Computing, 1(3), 349367. doi:10.1142/
S1793962310000225
Vasant, P., & Kale, H. (2007). Introduction to
fuzzy logic and fuzzy linear programming. In A.
Frederick & P. Humphreys (Eds.), Encyclopedia
of Decision Making and Decision Support Tech-
nologies. IGI publisher.
Yang, L., Jones, B. F., & Yang, S. H. (2007).
A fuzzy multi-objective programming for op-
timization of the fire station locations through
genetic algorithms. European Journal of Op-
erational Research, 181, 903915. doi:10.1016/j.
ejor.2006.07.003
Zamirian, M., Kamyad, A. V., & Farahi, M. H.
(2009). A novel algorithm for solving optimal
path planning problems based on parametrization
method and fuzzy aggregation. Physics Letters.
[Part A], 373, 34393449. doi:10.1016/j.phys-
leta.2009.07.018
368
Adaptive Feasible Mutation: A significant
operator in genetic algorithm which provide di-
versification in search for optimal solution.
Arithmetic Crossover: A dynamic back
ground operator which organized by itself.
Degree of Satisfaction: Measure of level of
satisfaction by the decision makers.
Genetic Algorithms: An artificial intelligent
technique using evolutionary ideas of natural
selection and genetic.
Pattern Search: A very useful direct search
algorithms for non-derivative, nosy and discrete
objection function in optimization problems.
Uncertain Coefficients: Imprecise technical
coefficients in fuzzy optimization problems.
Vagueness Factor: Measure of imprecise
variables in uncertain environment.
369
About the Contributors
Pandian Vasant was born in Sungai Petani, Malaysia in 1961. Currently, he is a Senior Lecturer of
Engineering Mathematics for Electrical & Electronics Engineering Program and Fundamental & Applied
Sciences Department at University Technology Petronas in Tronoh, Perak, Mlaysia. He has graduated
in 1986 from University of Malaya (MY) in Kuala Lumpur, obtaining his BSc Degree with Honors (II
Class Upper) in Mathematics, and in 1988 also obtained a Diploma in English for Business from Cam-
bridge Tutorial College, Cambridge, England. In the year 2002 he has obtained his MSc (By Research)
in Engineering Mathematics from the School of Engineering & Information Technology of University
of Malaysia Sabah, Malaysia, and has a Doctoral Degree (2008) from University Putra Malaysia in
Malaysia. After graduation, during 1987-88 he was Tutor in operational research at University Science
Malaysia in Alor Setar, Kedah and during 1989-95 he was Tutor of Engineering Mathematics at the same
university but with Engineering Campus at Tronoh, Perak. There after during 1996-2003 he became a
lecturer in Advanced Calculus and Engineering Mathematics at Mara University of Technology, in Kota
Kinabalu. He became Senior Lecturer of Engineering Mathematics in American Degree Program at
Nilai International College (Malaysia), during 2003-2004 before taking his present position at Univer-
sity Teknologi Petronas in Tronoh. His main research interests are in the areas of Optimization Methods
and Applications to Decision Making and Industrial Engineering, Fuzzy Optimization, Computational
Intelligence, and Hybrid Soft Computing. Vasant has co-authored 200 research papers and articles in
national journals, international journals, conference proceedings, conference paper presentation, and
special issue guest editor, lead guest editor for book chapters project, conference abstract, edited book
and book chapters. In the year 2009, Vasant was awarded top reviewer for the journal Applied Soft
Computing (Elsevier). He has been Co-editor for AIP Conference Proceedings of PCO (Power Control
and Optimization) conferences since 2008 and editorial board member of international journals in the
area of Soft Computing, Optimization and Computer Applications. Currently hes a lead managing
editor for GJTO (Global Journal Technology & Optimization) and organizing committee member (PCO
Global) for PCO Global conferences.
Nader Barsoum obtained his PhD from University of Newcastle upon Tyne, UK, 1989 in Eclectic
Machines and Drives, and his first three degrees from Alexandria University, Egypt in Power Engineer-
ing 1976, Applied Mathematics 1979 and Engineering Mathematics 1983. He has academically jointed
University Sains Malaysia, University Malaysia Sabah, University of Western Sydney, and Curtin Uni-
versity of Technology in the past 20 years. He obtained School Research and Developments Excellence
Award 2010, Highest Research Performance Incentive Award 2010, Best Researcher Award 2009, ANAK
Sarawak Award 2009, and University Encouragement Award 1992. He completed several funded projects
such as: Implementation of an Optimal Energy Saving System with solar tracker 2010, Development
of Smart Drive Build in AC Motors for Optimal Operations 2010, Investigation of Electrical Properties
of Palm Oil Using Kerr-effect Technique 2009, Hardware design of body sensor device 2008, Low cost
hydrogen fuel cell for hybrid renewable energy system 2006, and Simulation of Micro-Controller for
Induction Motor Drive 1995. Besides supervising many Master and PhD degrees, he selected several
times as an external examiner and published over 86 papers in both international scientific journals
and conferences. He is the President of Malaysian Solar Energy Society since 2010, Editor In Chief of
Global Journal Of Technology and Optimization since 2010, General Chairman of an Annual Global
Conference on Power Control and Optimization since 2008, and Guest Editor of International Journal
Of Computers & Mathematics With Applications.
Jeffrey Webb has a Ph.D. degree in photothermal physics from Strathclyde University, Scotland,
U.K. He is now an Associate Professor in the School of Engineering, Computing and Science, Swin-
burne University of Technology, Sarawak (Malaysia). His research experience is in the areas of optics,
electromagnetic waves, ferroelectric materials, and nanoelectronics.
* * *
Muhammad Mansoor Asghar did his BSc Electrical Engineering from University of Engineering
& Technology, Lahore, Pakistan in 2009. He is also student in MSc Electrical Engineering. Currently
he is serving as Lecturer/Lab Engineer in Rachna College of Engineering & Technology, Gujranwala,
Pakistan. He has published more than five international research papers. His research interests include
power system analysis, renewable technologies, power electronics, et cetera.
Adnan Bashir did his BSc and MSc Electrical Engineering from University of Engineering & Tech-
nology, Lahore, Pakistan in 2006 and 2009, respectively. Currently he is serving as Lecturer in Rachna
College of Engineering & Technology, Gujranwala, Pakistan. He has published more than three inter-
national research papers. His research interests include power system analysis, renewable technologies,
power electronics, et cetera.
Roman V. Bershanskiy (M 2002) graduated from Moscow Power Institute as an engineer-electrician
in 2002 and since that time has been working with Joint Stock Company System Operator of Unified
Energy System, JSC, SO UES, Moscow, Russia. Now he is a postgraduate student at Energy Systems
Institute, Russian Academy of Sciences, Irkutsk. His scientific interests are connected with real-time
control problems, using FACTS and WAMS for control of power systems, detection of weak nodes,
and cut sets.
Hassan Bevrani received the M.Eng. (Hons.) degree from K. N. Toosi University of Technology,
Tehran, Iran, in 1997, and the Ph.D. degree from Osaka University, Osaka, Japan, in 2004, both in electri-
cal engineering. From 2004 to 2006, he was a Postdoctoral Fellow at Kumamoto University, Kumamoto,
Japan. From 2007 to 2008, he was a Senior Research Fellow at Queensland University of Technology,
Brisbane, Qld., Australia. From 2000, he has been an academic member of University of Kurdistan. Cur-
rently, he is a Professor in Kumamoto University (Japan). His special fields of interest include intelligent
370
and robust control applications in power system and power electronic industry. Prof. Bevrani is a senior
member of Institute of Electrical and Electronics Engineers (IEEE), and a member of the Institute of
Electrical Engineers of Japan (IEEJ) and the Institution of Engineering and Technology (IET).
Selcuk Cebi received his MSc degree on mechanical engineering from Karadeniz Technical University
in 2004 and he received his PhD degree on industrial engineering from Istanbul Technical University
in 2010. His academic career started in Karadeniz Technical University in 2002 as a research assistant
until 2004. Then, he worked at Istanbul Technical University as a research assistant between 2005 and
2010. Now, he is working at Karadeniz Technical University as an Assistant Professor. Dr. Cebi has
interests in decision support systems, multiple criteria decision making, and design topics, and he has
over 20 publications in SCI indexed journals during his academic career.
Muhammad Ali Masood Cheema was born in Gujranwala, Pakistan, in 1984. He did his BSc and
MSc Electrical (Power) Engineering with Honours from University of Engineering & Technology,
Lahore, Pakistan in 2008 and 2010, respectively. He was awarded two Gold Medals during his BSc in
Electrical Engineering. He served in Pakistans largest transformer manufacturing company, PEL, for
one year as Design Engineer. Currently he is serving as Lecturer in Rachna College of Engineering &
Technology, Gujranwala, Pakistan. He has published more than twelve international research papers
including IEEE PES, Springer Journal, and Elsevier Journal. His research interests include power qual-
ity, electric machine design, control systems, power system analysis, distributed generation and load
modeling, et cetera.
Vo Ngoc Dieu received his B.Eng. and M.Eng. degrees in electrical engineering from Ho Chi Minh
City University of Technology, Ho Chi Minh city, Vietnam, in 1995 and 1999, respectively, and his
D.Eng. degree in energy from Asian Institute of Technology (AIT), Pathumthani, Thailand in 2007. He
is Research Associate at AIT and lecturer at Department of Power Systems, Faculty of Electrical and
Electronic Engineering, Ho Chi Minh City University of Technology, Ho Chi Minh City, Vietnam. His
interest is applications of AI in power system optimization.
Dmitry N. Efimov is a senior researcher in the Energy Systems Institute of the Russian Academy
of Science (ESI SB RAS), Irkutsk, Russia. He was born in 1961 in Irkutsk. He graduated from Irkutsk
State Technical University in 1987. Since then he has been with ESI SB RAS. He received his degree
of Candidate of Technical Sciences from the ESI SB RAS in 1998. From 1998 till 2000, D.N.Efimov
was an Executive Director of the International Research Energy Laboratory, Budapest, Hungary. His
research interests include simulation of transients in electric power systems, development, operation,
and dynamic properties of the large power interconnections.
Alexander Z. Gamm graduated from the Electrotechnical Institute, Novosibirsk, Russia, where he
received the Ph.D and D.Sc. degrees. During 1961-1962, he was a scientific worker in the Transport
Power Institute, Novosibirsk. Since then, he has been with the Energy System Institute (ESI), Russian
Academy of Sciences, Irkutsk. At present, he is a Chief Researcher. His special fields of interest include
state estimation, optimization, and real-time control problems. Dr. Gamm is an Academician of the Rus-
sian Academy of Electrical Sciences and member of International Energy Academy.
371
Anna M. Glazunova graduated from the Irkutsk Polytechnical Institute in 1982. Since 1986 she has
been working as a researcher at Energy Systems Institute. She received her PhD in 2002. Her research
interest is methods of state estimation of electricity use.
Irina I. Golub (M1995, SM1997) graduated from Moscow Power Institute as an engineer-elec-
trician. She has been working at Siberian Energy Institute, Irkutsk since 1972. Her scientific interests
are connected with real-time control problems, especially in the field of observability of electric power
systems and allocation of measurements of operating parameters of electric networks. She is a leading
researcher, University Professor, and Doctor of technical sciences.
U. Hger received his diploma degree in Electrical Engineering in 2006 from the Univ. of Dortmund,
Germany. He is working at Technische Universitt Dortmund, Institute of Power Systems and Power
Economics, on his doctoral degree in the field of wide area congestion management by use of fast power
flow controllers.
Mahmoud-Reza Haghifam received his B.Sc., M.Sc., and Ph.D. degrees in 1988, 1990, and 1995,
respectively, all in electric power engineering. Currently, he is the Full Professor of electrical engineering
at Tarbiat Modares University, Tehran, Iran and worked on Electric Distribution Systems Reconfiguration
during several essays and projects. He is a senior member of IEEE and research fellow of Alexander Van
Humboldt, Germany. His research interest includes electric distribution analysis and planning, electric
distribution system operation and automation, power system analysis and planning, development of soft
computing and bio-inspired computing techniques for power system planning, power system reliability
and reactive power control.
Muhammad Junaid was born in Gujranwala, Pakistan, in 1981. He did his BSc, MSc Electrical
(Power) Engineering with Honours from University of Engineering & Technology, Lahore, Pakistan
in 2008 and 2010 respectively. Currently he is serving as Lecturer in Rachna College of Engineering
& Technology (RCET), Gujranwala. He has published more than fifteen international research papers
including IEEE PES, Springer Journal, and Elsevier Journal. He is also author of International Book
Chapter titled Harmonics Modelling & Simulation. During 4
th
Global Conference on Power Control
& Optimization held in Malaysia, he was declared Best Presenter. Keeping in view his valuable research
contributions, RCET has given him Best Research award. His research interests include power quality,
power system analysis, power system operation & control, power electronics, et cetera.
Cengiz Kahraman received his MSc and PhD degrees in industrial engineering from Istanbul Technical
University. He is in the editorial board of some journals such as Journal of Human and Ecological Risk
Assessment, International Journal of Computational Intelligence Systems, New Mathematics and Natural
Computation, Journal of Enterprise Information Management, Technological and Economic Development
of Economy, Baltic Journal on Sustainability, International Journal of Strategic Property Management,
Journal of Multiple Valued Logic and Soft Computing, and he has made the guest editorships of special
issues of some international journals such as Information Sciences, Journal of Enterprise Information
Management, and International Journal Of Approximate Reasoning. He is the editor of such Springer
books as Fuzzy Applications in Industrial Engineering, Fuzzy Multicriteria Decision Making: Theory
and Applications with Recent Developments, and Fuzzy Engineering Economics with Applications.
372
Fouad Kamel is a Senior Lecturer at the University of Southern Queensland in Toowoomba, Faculty
of Engineering and Surveying, Department of Electrical and Computer Engineering since February 2008.
He graduated Diploma Engineer and PhD in photovoltaic systems from Hanover University in Germany
1984, Dr. Fouad worked as a Lecturer and Associate Professor at the Suez Canal University in Egypt
during 1985-1999. In 1999, he moved to New Zealand and worked there between 2000 and 2007 for
tertiary education and research at Christchurch Polytechnic Institute of Technology and the Southern
Institute of Technology. Dr. Fouad has a history of publications in different renewable energy areas.
Rana Abdul Jabbar Khan graduated from University of Engineering and Technology, Lahore,
Pakistan in 1991. He completed ME from University of Wollongong, Australia in 1995 and PhD from
RMIT University Melbourne, Australia in 2003. He has been serving as SDO and XEN in Water and
Power Development Authority (WAPDA), largest organization of Asia, responsible for generation,
transmission, and distribution of electricity country wide. Currently he is on deputation and serving
Rachna College of Engineering and Technology (RCET) as Principal since 2005. He is principal author
of an open access book chapter, titled Harmonics Modelling and Simulation. He is a senior member
of IEEE, IEAust, IEE, and PEC. He has published several papers in IEEE journals. Honoring his valu-
able research contribution he has been nominated for Presidential Award 2009 by the government of
Pakistan. His research interest includes power quality, power system simulation, renewable technologies
and deregulation of electricity, et cetera.
hsan Kaya is currently an Assistant Professor in Yldz Technical University Department of Industrial
Engineering. He received his MSc and PhD degrees on industrial engineering from Seluk University
and Istanbul Technical University, respectively. His main research areas are fuzzy set theory, process
capability analyses, quality control, decision making, and fuzzy statistics.
Irina N. Kolosok graduated from St. Petersburg Technical University. Since 1972 she has been with
Energy Systems Institute (ESI), Russian Academy of Sciences, currently as a leading researcher. She
received the PhD (1986) and DSc (2004) degrees. Her scientific interests are: real-time control prob-
lems, especially in the field of state estimation of electric power systems (EPS), SCADA systems, and
application of AI-methods for on-line EPS control.
Elena S. Korkina graduated from Irkutsk Polytechnic Institute on speciality of an engineer-economist
in 1978. Since 1987 she has been working at the Siberian Energy Institute of the Siberian Branch of
Russian Academy of Sciences in the laboratory of electric power system operation control problems.
Her scientific interests are: real-time control problems, SCADA systems, and up-to-date Information
Technologies.
Victor G. Kurbatsky (M08) was born on May, 27th, 1949 in Russia. He is PhD, Professor, and
Doctor of Science. He is Leading Researcher at the Energy Systems Institute of the Russian Academy
of Sciences, Irkutsk, Russia. Victor Kurbatsky received his degree of Candidate of Technical Sciences
at SibNIIE (Novosibirsk) in 1984 and Doctor of Technical Sciences at the Energy Systems Institute (Ir-
kutsk) in 1997. His research interests include: electromagnetic compatibility and power quality in electric
networks and application of artificial intelligence techniques in power systems. Professor Kurbatsky is
the author of several monographs and manuals and more than 280 scientific papers.
373
Gerard Ledwich received the Ph.D. degree in Electrical Engineering from the University of Newcastle,
Newcastle, Australia, in 1976. He has been the Chair Professor in Power Engineering at Queensland
University of Technology, Brisbane, Australia, since 2006. Previously, he was the Chair in Electrical
Asset Management from 1998 to 2005 at the same university. He was the Head of Electrical Engineering
at the University of Newcastle from 1997 to 1998. Previously, he was associated with the University of
Queensland from 1976 to 1994. His interests are in the areas of power systems, power electronics, and
controls. Prof. Ledwich is a Fellow of the Institution of Engineers Australia (I.E.Aust.).
Igor Litvinchev received his M.Sc. degree from Moscow Institute of Physics and Technology (Fizteh),
Moscow, Russia, and Ph.D. and Dr.Sci. degrees in systems modelling and optimisation from Computing
Centre, Russian Academy of Sciences, Moscow. He has held visiting positions at universities in Brazil,
Mexico, as well as positions at various universities and research centres in Russia. His research focuses
on large-scale system modelling, optimisation, and control. Dr. Litvinchev is a member of Russian
Academy of Natural Sciences and Mexican Academy of Sciences.
Marwan Marwan received the B.Eng. degree from Hasanuddin University, Makassar, Indonesia,
and the M.Eng. degree from the Queensland University of Technology, Brisbane, Australia, in 2000
and 2006, respectively, all in electrical engineering. He is working at The State Polytechnic of Ujung
Pandang Makassar Indonesia as a Lecturer in Energy Conversion Department from 2001 to the present.
Presently, he is preparing Ph.D. at the University of Southern Queensland in Toowoomba, Australia.
Armin Ebrahimi Milani received his B.Sc. degree (Electric Power Engineering) from Islamic Azad
University, South Branch, Tehran, Iran, in 2006 and M.Sc. degree in Electric Power Engineering from
Islamic Azad University, Science and Research Branch, Tehran, Iran, in 2010. He worked his thesis on
Electric Distribution Systems Reconfiguration. Currently, he is with department of Electrical Engineering,
Azad University, Iran. He is a member of IEEE and senior member of Young Researchers Club, Tehran,
Iran. His research interest includes power system optimization and control, electric distribution systems,
power system restructuring, development of soft computing and bio-inspired computing techniques for
power system planning, operation, and control applications.
J. Zambujal-Oliveira received M.S. degrees in Economics at New University of Lisbon (FE) and in
Management Science at Technical University of Lisbon (ISEG). In 2007, he received a Ph.D. degree in
Management Science from the Technical University of Lisbon (ISEG). Before coming to the Department
of Engineering and Management at Technical University of Lisbon (IST), Prof. J. Zambujal-Oliveira
was employed as a Systems Engineer at banks, insurance companies and consulting firms, and as an
Assistant Professor of management at the University of Madeira in Funchal. Prof. J. Zambujal-Oliveiras
research interests include real options analysis and international taxation. In particular, he enjoys work-
ing with graduate students and colleagues in developing, analyzing, and testing numerical methods for
solving stochastic differential equations, developing and analyzing models in transfer prices systems
and developing, analyzing, and testing computational methods in mathematical finance.
Abdeen Mustafa Omer (BSc, MSc, PhD) is a qualified Mechanical Engineer with a proven track
record within the water industry and renewable energy technologies. He has been graduated from
374
University of El Menoufia, Egypt, BSc in Mechanical Engineering. His previous experience involved
being a member of the research team at the National Council for Research/Energy Research Institute in
Sudan and working director of research and development for National Water Equipment Manufactur-
ing Co. Ltd., Sudan. He has been listed in the WHOS WHO in the World 2005, 2006, and 2007. He
has published over 200 papers in peer-reviewed journals, 50 review articles, and 40 chapters in books.
Weerakorn Ongsakul received his B.Eng. degree in electrical engineering from Chulalongkorn
University, Bangkok, Thailand, in 1988, and his M.S. and Ph.D. degrees in electrical engineering from
Texas A&M University, College Station, in 1991 and 1994, respectively. Currently, he is an Associate
Professor of Energy Field of Study, and Dean of School of Environment, Resources and Development
at Asian Institute of Technology (AIT), Pathumthani, Thailand. His interests are in computer applica-
tions to power systems, parallel processing applications, AI applications in power systems, and power
system deregulation.
Alexei S. Paltsev graduated from Irkutsk State Technical University specializing in computer science
in 2005. Since 2000 he has been with the Energy Systems Institute (ESI), Russian Academy of Sciences,
currently as a leading Engineer. He finished his postgraduate courses at ESI in 2008. His scientific in-
terests are multi-agent systems and EPS state estimation.
Daniil A. Panasetsky (SM07) is a postgraduate student at the Energy Systems Institute m of the
Russian Academy of Sciences, Irkutsk, Russia. He graduated with honors from Irkutsk State Technical
University specializing in electrical engineering in 2006. His research interests include stability analy-
sis of power systems, emergency control, FACTS devices, and application of artificial intelligence to
power systems.
Dmitri B. Popov is a leading engineer of Energy Systems Institute (Siberian Energy Institute until
1997) of the Russian Academy of Sciences, Irkutsk, Russia. He was born in 1967 in Irkutsk. In 1984
he was granted the certificate in mathematical programming at the Institute of System Dynamics and
Control Theory of Russian Academy of Sciences. In 1995 he graduated from Irkutsk State Technical
University specializing in electrical engineering. D.B. Popov joined the Siberian Energy Institute in 1989.
His research interests are transient stability analysis, database application development, users interface
design, program interface development, asynchronous conditions, and prevention by customization of
automatic equipment.
Socorro Rangel received her M.Sc. degree from Campinas State University (UNICAMP), Campinas,
Brazil, and Ph.D. degree from Brunel University, Uxbridge, England. She is currently an Associate Pro-
fessor at the State University of So Paulo (UNESP), So Jos do Rio Preto, Brazil. Her main research
interests lies on large-scale system modelling and optimisation.
Christian Rehtanz received his diploma degree in Electrical Engineering in 1994 and his Ph.D. in
1997 at Technische Universitt Dortmund, Germany. From 2000 he was with ABB Corporate Research,
Switzerland and from 2003, Head of Technology for the global ABB business area Power Systems.
From 2005 he was Director of ABB Corporate Research in China. From 2007 he is Professor and Chair
375
for power systems and power economics at Technische Universitt Dortmund. His research activities
include technologies for network enhancement and congestion relief like stability assessment, wide-area
monitoring, protection, and coordinated FACTS- and HVDC-control.
Mohammad Saleh received B.Sc. and MSc degrees in electrical engineering from Tarbiat Moallem
University (Tabriz), and University of Kurdistan (Sanandaj), Iran, in 2007 and 2011, respectively. His
current research interests include Grid integration of wind energy converters.
Ghasem Tikdari was born on 1985 in Tikdar, Kerman, Iran. He received the B.S. degree from Bahonar
University, Kerman, Iran in 2007 in Electronic Engineering and received the M.S. degree from University
of Kurdistan, Sanandaj, Iran in 2009 in Electrical Power System Engineering. His research interests are
on power system emergency control, renewable energy sources, artificial intelligence techniques, and
electronic and power electronic device designing. He also experienced in some industrial works which
use Distributed Control System (DCS), Supervisory Control and Data Accusation (SCADA), and field-
bus communication protocols.
Nikita V. Tomin, PhD, (M08) was born on December 18th in 1982, in Russia. Dr. Tomin is a Senior
Researcher at the Energy Systems Institute of the Russian Academy of Science, Irkutsk, Russia. In 2007
he defended his PhD thesis at the Energy Systems Institute SB RAS (Irkutsk). Dr. Tomin specializes in
the field of artificial intelligence technologies in electric power systems. He is the author and co-author
of more than 60 scientific papers.
Nikolai I. Voropai (M1996, SM1998; F2009) is Director of the Energy Systems Institute (Siberian
Energy Institute until 1997) of the Russian Academy of Sciences, Irkutsk, Russia. He is also Head of
Department at Irkutsk Technical University. He was born in Belarus in 1943. He graduated from Len-
ingrad (St. Petersburg) Polytechnic Institute in 1966 and has been with the Siberian Energy Institute
since. N.I. Voropai received his degree of Candidate of Technical Sciences at Leningrad Polytechnic
Institute in 1974, and Doctor of Technical Sciences at the Siberian Energy Institute in 1990. His research
interests include: modeling of power systems; operation and dynamic performance of large intercon-
nections; reliability, security and restoration of power systems; development of national, international
and intercontinental electric power grids. N.I. Voropai is a member of CIGRE, a Fellow of IEEE, and a
member of PES. He is the IEEE PES Region 8 Zone East Representative.
376
377
Index
Symbols
30 Cycle Network (Thirty-Cycle Network) 317,
333-334, 336, 343
3-Phase Faults 316, 343
A
acid rain 103, 109, 121
active power fow 23, 25, 36
adaptive feasible mutation 356-357, 368
agricultural zones 108
Analytic Hierarchy Process (AHP) 166, 171, 173,
181-182, 186, 190-192, 194, 197
angle instability 294, 297, 301, 304, 306, 313
Ant Colony Optimization (ACO) 170, 186, 189,
195
Ant Colony Search Algorithm (ACSA) 59, 91
arithmetic crossover 349-350, 357, 368
Artifcial Intelligence (AI) 3-5, 13, 20-21, 54, 56,
59, 67, 80, 189-191, 193, 195-196, 269, 280,
345, 367
Artifcial Neural Networks (ANN) 21-23, 50, 166,
169, 176, 186, 189-190, 192-194, 294
B
bad data 7, 9-10, 12, 22-23
binomial options pricing model 264
biomass energy 127-128, 135
boundary branch 16-18
boundary nodes 7-13
box agents 38
branch data 340
bus data 340
C
calculated scheme 7-9, 12-13, 19, 50
capacitor data 340
capacity factor 138-139, 157, 198, 206, 219
Carbon Dioxide (CO2) 95-98, 100, 103-105, 108,
112-113, 120, 127, 136, 164, 181, 184, 251,
253, 255-260
carbon dioxide content 97
characteristic impedance 296, 298
Circuit Breakers (CB) 316-318, 331
climate change 100-101, 103, 108-110, 112-113,
121, 126, 129, 136, 162, 165, 178, 181
combined cycle 248-249, 253, 255, 259-260, 264
confguration 117, 195, 268-273, 283, 285, 287-
289, 292, 299, 301, 304, 343
consumption nodes 6
control system 2, 32, 34, 37, 43, 46, 50, 55-56,
203-204
Cost Effciency (CE) 96, 138, 242
critical cutest 297
current 5-7, 9, 13, 21, 24-25, 32-33, 35, 39-40, 43-
45, 49, 52, 87, 95-98, 126, 139-140, 147, 158,
163, 165, 185, 203-204, 231, 240, 254, 274,
315-316, 318, 322-328, 334-338, 340-342, 354,
366
current quality 341
D
degree of satisfaction 348, 368
device duty 343
Differential Evolution (DE) 51, 59, 90-91, 169,
188, 206, 220, 243, 261, 312
Direct Load Control (DLC) 141
Discount Cash Flow (DCF) 248-249, 252, 255-259
Distribution Companies (DISCOs) 315-316
Divandarreh 204-210
Double-Line Ground (LLG) 335-336, 343
Doubly-Fed Induction Generator (DFIG) 199-201,
203, 209, 215
drive train model 202
Dual-Input Power System Stabilizer Model (PSS2A)
209
378
Index
E
Earth energy 117
economic load dispatch 57-58, 88, 91, 93-94
electrical circuit 10, 163
electrical component 163
electrical energy 135, 137-138, 148-150, 156-158,
160-161, 187, 201, 210, 289, 291
electrical generation 163
Electrical Transient Analyzer Program (ETAP) 315-
322, 328-329, 331, 337-338, 340-341, 343
electrical vehicle 162-163
electric distribution network 287, 292
electricity network 145, 163, 200
Electricity Price (EP) 59, 90-91, 96, 140, 147, 149,
152-154, 159, 161, 163, 169, 206, 208, 249-
257, 259
electric motors 163
Electric Power Research Institute (EPRI) 140
Elimination and Choice Expressing Reality (ELEC-
TRE) 171, 176, 184, 187-188, 194
emergency control 1-6, 20, 32, 34, 43, 46, 49, 51,
54-56, 214, 294, 297-298, 302, 310, 313-314
Emergency Demand Response Program (EDRP)
141-142, 158
Emerging Energy Technology (EET) 167, 169, 176,
185-186, 197
Enercon E-53 206
energy cost analysis 207
Energy Effciency (EE) 55, 96-97, 110, 112, 118-
120, 122, 124, 126-127, 129-130, 133, 145,
159, 162, 165, 181, 183, 192
energy function 44, 54, 57, 60-62, 65-67, 69-70, 74,
81, 88, 90, 94, 293-294, 297, 299, 311, 313
energy management 137, 154, 162-163, 263-264
Environmental Performance (EP) 59, 90-91, 96,
110, 159, 169, 206, 208, 253, 256
equal area criterion 51, 299
ETAP software 315-317, 337-338
Evolutionary Programming (EP) 59, 90-91, 96, 159,
169, 206, 208, 256
exergy 122-124, 188
F
fast-decoupled method 340
Feeder Analyses (FDR-ANA) 315
Fixed-Speed Induction Generator (FSIG) 199, 209
Fixed Speed Wind Turbine (FSWT) 201, 217, 219
Florida Reliability Coordinating Council (FRCC)
307, 311
forecasting model 22
Fuel Constrained Economic Load Dispatch (FELD)
57, 72-73, 88, 94
functional decomposition 1, 7, 9, 13, 49
fuzzy logic 55, 59, 91, 166, 169, 188, 196, 344,
364, 367
fuzzy optimization 344, 346, 364-368
fuzzy sets 54, 188-189, 192, 196-197, 289, 344-
345, 364-366
fuzzy set theory 344, 366
fuzzy TOPSIS 171, 174, 188, 364
G
Gaussian Filter 297
Gauss-Seidel Method 340
generalized assignment problem 220, 228, 243-246
generator agent 35-38
generator node 26-27, 30
Genetic Algorithms (GA) 21-22, 52, 59, 91, 128,
169-170, 176, 186, 189-190, 193, 269, 279-
281, 283, 288, 291, 344-351, 355-357, 359,
361-362, 364-365, 367-368
geothermal energy 117, 130, 133, 135
geothermal power 127, 129, 133, 135
global warming 95-98, 100, 103-104, 113, 127,
129, 133, 164-165
Graphical User Interface (GUI) 317
greenhouse effect 95, 97, 103, 113, 136
greenhouse gases 96, 108, 118, 127, 136, 165
ground grid analysis 316, 331, 343
ground grid systems 342
Gujranwala Electric Power Company (GEPCO)
315-316
H
harmonic analysis calculation methods 341
Harmonic Analysis (HA) 315-316, 322, 328, 341
Heating, Ventilating and Air-Conditioning (HVAC)
96
historical volatility 254-255, 257, 264
hopfeld lagrange network 57, 61-62, 64, 94
Hopfeld Neural Network (HNN) 57, 59-61, 69, 80,
84, 90-94
Human Machine Interface (HMI) 303
Hybrid GA Pattern Search (HGAPS) 348, 359,
361-362
hybrid genetic algorithms 344, 346, 367
hydropower 135
hydrothermal economic load dispatch 88, 94
379
Index
I
Individual Harmonic Distortion (IHD) 322, 341
Indoor Environmental Quality (IEQ) 96-97
induction motor data 340
information axiom 166, 171, 175-176, 188, 191,
196-197
Integrated Gasifcation Combined Cycle (IGCC)
259-260
interconnection 3, 20, 43-44, 200
inverter data 340
Iran 198-199, 201, 213, 216, 268, 293, 311
islanding control 294, 297, 306, 314
J
Jacobi matrix 24-27, 29-30
JADE platform 38
K
kinetic energy 43, 135, 213, 297, 304, 306
Kurdistan Electric Network 198-199, 209, 216
kyoto protocol 100-101, 125
L
Lagrangian dual 220-222, 225, 228-230, 237, 242,
246-247
Lagrangian dual problem 222, 237, 246-247
Lagrangian function 57, 59, 61-62, 67, 74, 88, 90,
94, 221, 241-242, 246
Lagrangian heuristic 242, 244, 246
Lagrangian multipliers 62, 228, 233, 242, 246
Lagrangian relaxation 59, 61, 94, 221, 225-227,
245-247
Lagrangian relaxation bounds 246
large scale power grids 295
Levelised Cost of Electricity (LCOE) 205, 207-208,
219
lignocellulosics 113
Line-Ground (LG) 335-336, 343
Line-Line (LL) 335-336, 343
load agent 36-37
load fow 13, 15-16, 19, 23, 26, 52, 54, 91, 273-
277, 283, 289-292, 315-316, 321-322, 337-338,
340, 342
load fow analysis 315-316, 321, 340
load fow calculation methods 340
load shedding 34-37, 40, 43, 45-46, 214, 294, 297-
298, 302-303, 307, 310-312, 314
localization 223-228, 241-242, 244
lumped load data 340
M
MACBETH 177-178, 187
many-to-many assignment problem 222, 227, 246
market conditions 109, 137, 163, 248-250, 257
Market Introduction of New Energy Technologies
(EMINENT) 184, 195
Master Problem (MP) 226-227, 229-230, 240
maximal cost component 231
Maximum Power Point Trajectory (MPPT) 188,
202-203
meta-heuristic methods 59
modifed Lagrangian bounds 220, 222, 236
monitoring 1-6, 8, 20-21, 23, 49-54, 56, 121, 124,
127, 302, 312-313, 315-316, 318, 321-323,
328, 337-338
Monte Carlo simulation 250, 252, 261, 264
multi-agent systems 1-2, 51, 53
Multi Attribute Decision Making (MADM) 171,
197
Multi Criteria Decision Making (MCDM) 166, 171,
176-179, 185-186, 188, 190, 194, 197
Multi Objective Decision Making (MODM) 171,
192, 197
N
Natural Gas Combined Cycle (NGCC) 253, 259
Natural Gas (NG) 97-99, 114, 187, 251, 253-259,
261-262, 264, 275, 304
neural network 2, 21-23, 57, 59-65, 69, 75-76, 80,
82-85, 88, 90-94, 187-188, 190-191, 193, 195
Newton-Raphson Method 340
nodal admittance matrix 25-26
nodal voltages 9, 12, 24-26
Nonlinear Programming (NLP) 344, 346, 348, 364
Nordex 77 206
North American Electric Reliability Corporation
(NERC) 6
O
objective copy constraint 236, 238, 241
on-site renewable energy 145, 157-158, 163
operating control 1, 200
operational management 2
optimal ftness function 356
optimal primal-dual pair 221, 241
optimization problem 34, 58, 60, 62, 65, 77, 94,
195, 220-222, 269, 273, 279, 283, 285, 288,
292, 346-348
380
Index
organic sulphur 113, 130
P
Pakistan Electric Power Company (PEPCO) 315-
316
Particle Swarm Optimization (PSO) 59, 90-92, 169
pattern search 169, 344, 348, 357, 365, 367-368
Permanent Magnet Synchronous Generator (PMSG)
199, 201, 203-204
Phasor Measurement Units (PMUs) 2, 7-14, 16-19,
47, 53, 294, 303, 306-307, 310
power factor 203, 211, 315, 321, 337
power grid data 340
power system control 2, 33, 52-53, 199, 218, 310
power system dynamic stability 1
power system stability 44, 51, 198, 212, 218, 293-
295, 298, 303, 310-312, 316, 342
power system voltage 37, 341
Preference Ranking Organisation Method for En-
richment Evaluations (PROMETHEE) 171,
188
Probability Density Function (PDF) 159, 161, 275-
279, 283-286, 292
protection 4, 32-34, 43-44, 53-54, 97, 121, 128-129,
131-133, 270-271, 295, 298, 312, 338
Q
quadratic programming 61, 90-91, 93
R
Rachna College of Engineering & Technology
(RCET) 316
radial schemes 15-16, 18-19
reactive power fow 315, 337
Real Options Analysis (ROA) 248-250, 253, 256,
258-260, 264
Reconfguration 109, 268-273, 286-292
regional electric network 198
Reliability Test System (RTS-79) 294, 304, 308,
311
renewable energies 95-97, 101, 117, 126, 198, 201
renewable energy 3, 95-96, 99-102, 106-107, 117-
118, 122-129, 132-133, 135, 137, 144-145,
150, 157-158, 160, 162-163, 165, 176-178,
184, 187-191, 193-195, 198-199, 214-217, 289,
293, 309, 311
resource management 121, 135
risk-adjusted discount rate 251-252, 255, 264
Root Mean Square (RMS) 341, 343
S
Sanandaj power plant 209-210
short circuit analysis 316, 333, 343
sigmoid function 60-63, 68-69, 82, 90, 94
Simple Additive Weighting Method (SAW) 171
Simulated Annealing (SA) 10, 22, 59, 91-92, 131,
134, 169-170, 269, 365, 367
slow coherency 294, 297-298, 301-302, 312, 314
smart grid 3, 51, 54, 137, 139, 144-146, 160-163,
312-313
Soft Computing (SC) 92, 164, 166-169, 171, 176-
178, 185-188, 191, 194-197, 365, 367
solar energy 97, 117, 135, 158-160, 162, 188-190,
194
spark spread 250, 263-264
stability assessment 293-294, 298-299, 301, 309,
312
stability limits 199, 294, 298, 328, 341-342
STATCOM-based control approach 198
State Estimation (SE) 1, 5-10, 12-14, 16-19, 21, 34,
49-53, 55-56, 255
state variables 2, 6, 13, 21-22, 25, 35, 44, 47, 50,
275, 298
static load data 340
static power converters 341
Static Synchronous Compensator (STATCOM) 198-
199, 211-212, 214-219
steady-state stability limit 26, 342
stratospheric ozone 95, 103
subgradient technique 220, 222, 233, 239-242
surrogate localization 225, 227
sustainable development 95-96, 100, 104, 106, 110,
120, 122-124, 128-129, 131-135
switch 147, 163, 250-251, 268-269, 272-273, 285,
292
switch types 268-269, 272-273, 292
synchronous generator data 340
synchronous motor data 340
T
Tabu Search (TS) 59, 91, 124, 166, 169, 171, 186,
189, 289
Technique for Order Performance by Similarity to
Ideal Solution (TOPSIS) 164, 166, 171-172,
174, 184, 188-189, 194, 196-197, 364
tie-lines 7, 304
Total Harmonic Distortion (THD) 337
trace metals 108
transfer function 63, 68-69, 94, 169
381
Index
transient stability 44-46, 209, 211, 214-216, 218,
295, 297, 299, 313, 317, 328-329, 337-338,
341-342
transient stability analysis 215, 313, 328-329, 341
transient stability limit 342
transmission system 32, 35, 37, 60, 90, 150, 163,
270, 337
Two-Dimensional (2-D) 296-297, 301
U
uncertain coeffcients 368
Under Frequency Load Shedding (UFLS) 307
Uninterruptible Power Supplies (UPS) 43, 51, 340-
341
V
vagueness factor 348, 352, 356, 359-360, 368
Variable Speed Wind Turbine (VSWT) 201-203,
218-219
VlseKriterijumska Optimizacija I Kompromisno
Resenje (VIKOR) 171, 191, 194
voltage magnitudes 7, 10, 13, 16, 273, 298
W
WAMS (Wide-Area Measurement System) 7-8, 21
water privatisation 107
Weibull Distribution 198, 219
Western European Union for the Coordination of
Transmission of Electricity (UCTE) 6, 43, 45
wholesale markets 5
wind energy 130, 135, 189, 198-202, 204, 210,
213-219
wind farms 198-200, 209-211, 214-218
Wind Power Systems (WPSs) 199-200, 207-208,
210, 212-214, 219, 310
wind rose diagram 204, 219
World Energy Council (WEC) 96-98, 130, 132

E 1613501382

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

E 1613501382

Uploaded by

Copyright:

Available Formats

Pandian Vasant

Petronas University of Technology, Malaysia

( / ) , which is called the first

for the reverse

are positive updating step sizes for

output of multiplier neuron representing

input of multiplier neuron corresponding to

multiplier neuron updating step sizes.

updating step sizes for multiplier neurons.

in each iteration thus producing near-optimal

- modified Lagrangian bound w

(computed by Algorithm 2).

, obtained by using the objective copy con-

obtained by the subgradient tech-

You might also like