Generator Maintenance Scheduling (GMS) in a Restructured

Power System

Major project report submitted in partial fulfillment of the

Requirements for the award of the degree of

Bachelor of Technology in Electrical Engineering

By

Burhan Nazir (19207145006)

Faizan Zahoor Khan (19207145012)

Tanzeeb Ahmad (19207145034)

Under the guidance of

Mr. Adil Yousuf

Department of Electrical Engineering

Institute of Technology, University of Kashmir
Srinagar, Jammu and Kashmir - 190024. (India)
 CERTIFICATE

It is to certify that the contents of the report entitled “Generator Maintenance

Scheduling (GMS) in a Restructured Power System” is a Bonafide work carried out by
Mr. Burhan Nazir (19207145006), Mr. Faizan Zahoor Khan (19207145012) and
Mr. Tanzeeb Ahmad (19207145034) under my supervision and guidance in partial
fulfillment of the requirements for the award of the degree of Bachelor of Technology in
Electrical Engineering. The contents of the report have not been submitted earlier for the
award of any other degree or certificates and I hereby commend the work done by them in
this connection.

Mr. Adil Yousuf

Department of Electrical Engineering
Institute of Technology, University of Kashmir
Srinagar, Jammu and Kashmir - 190024. (India)

1
 CERTIFICATE OF APPROVAL

This project titled “Generator Maintenance Scheduling (GMS) in a Restructured

Power System” carried out by Mr. Burhan Nazir (19207145006), Mr. Faizan Zahoor

Khan (19207145012) and Mr. Tanzeeb Ahmad (19207145034) is hereby approved as the

creditable study of technology in Electrical Engineering and is presented in a satisfactory

manner.

It warrants its acceptance as a prerequisite in fulfillment of the requirements for the

award of the degree of Bachelor of Technology in Electrical Engineering at Institute of

Technology, University of Kashmir, Srinagar, J&K.

Internal Examiner External Examiner

Mr. Mohd. Aarish Shaheen

Coordinator
Department of Electrical Engineering
Institute of Technology, University of Kashmir
Srinagar, J&K

2
 ACKNOWLEDGEMENT

We would like to articulate our sincere gratitude towards all those who have contributed
their precious time and helped along in our project work. Without them it would have been
a tough job to complete and understand this project work.

We would especially like to thank Mr. Adil Yousuf, our project supervisor for his firm
support and guidance and invaluable suggestions throughout the project work.

We express our greatest appreciation to Mr. Mohd. Aarish Shaheen Coordinator,

Department of Electrical Engineering for his encouragement, comments and timely
suggestions throughout the course of this project work. We express our indebtedness to all
the faculty members and staff of the department of Electrical Engineering, for their
guidance and effort at appropriate times which has helped us a lot.

Burhan Nazir (19207145006)

Faizan Zahoor Khan (19207145012)
Tanzeeb Ahmad (19207145034)

3
 ABSTRACT

Since the structure has changed from being regulated to being deregulated, the centralized
maintenance system is no longer valid. These days, utility company manager’s major
responsibilities include reducing operations and maintenance (O&M) expenses and
maintaining service reliability. Perhaps the biggest single controllable cost of a utility
operation is preventive maintenance. It is believed that attaining an ideal trade-off between
the expense of maintenance and the reliability of service in a reformed power system
requires careful planning and effective coordination among self-interested entities. Time-
directed maintenance and manufacturer recommendations were highly emphasized in
traditional maintenance programs in vertically integrated utilities.

This project aims to develop a trustworthy and practical model for the generation
maintenance planning in a restructured power system. To illustrate how the developed
model is used, a case example is presented. According to the analysis, the maintenance
strategy, maintenance expenses, and market pricing all have an impact on the maintenance
schedule in the end. In order to portray a more realistic situation for the maintenance
scheduling of generator units in a deregulated environment, the study shows that the cost
components must be carefully studied and simulated since they are key aspects to produce
an effective maintenance plan.

4
 TABLE OF CONTENTS

CertificateI

Certificate of approval II

Acknowledgement III

Abstract IV

Abbreviation and Symbols VIII

List of Tables IX

List of Figures X

Chapter1

Introduction1

Chapter2

Literature Review5

Chapter3

Power System Industry10

3.1 Introduction10

3.2 Vertically Integrated Utilities 11

3.3 Restructured Power System 12

3.3.1 Reasons for restructuring or deregulation of Power Industry13

3.3.2 Various entities involved in deregulation 13

Chapter 4

5
Maintenance Scheduling in restructured Power System16

4.1 Introduction16

4.2 Significance of Maintenance Scheduling17

4.3 Solution approaches in maintenance scheduling18

4.3.1 Single objective approach18

4.3.1.1 Heuristic approach19

4.3.1.2 Mathematical techniques19

4.3.1.3 Artificial Intelligence Approach21

4.3.2 Multi objective approach28

Chapter 5

Generator Maintenance Scheduling – methodology and problem formulation 32

5.1 Introduction32

5.1.1 Optimization objectives32

5.1.2 Problem constraints33

5.2 Methodology for Generator Maintenance Scheduling33

5.2.1 Methodology Ⅰ: Single objective approach by considering the GENCO’s and ISO’s
objective separately34

5.2.1.1 Problem formulation35

5.2.2 Methodology Ⅱ: Multi objective approach by considering the GENCO’s and ISO’s
objective simultaneously38

5.2.2.1 Problem formulation39

Chapter 6

6
Solution Algorithm for Generator Maintenance Scheduling in Restructured Power
System41

6.1 Case Study41

6.1.1 Solution by using single objective approach44

6.1.2 Solution by using multi objective approach48

6.2 Comparison between single objective approach and multi objective approach51

6.3 Conclusion53

Bibliography54

ABBREVIATIONS AND SYMBOLS

7
 GMS Generator Maintenance
Scheduling
ISO Independent Service
Operator
GENCO Generating Companies
TRANSCO Transmission Companies
DISCO Distribution Companies
x¿ Unit Maintenance Status, 1, if unit is on Maintenance
Pimin
Minimum Power output of a generating Unit i (MW)
Pimax
Maximum Power output of a generating Unit i (MW)
Rmin Minimum reserve Margin
Pdt Load at time period t
C (P¿ )
Cost of fuel
P¿
Di Energy produced by unit i in time t (MW)
S Maintenance Duration in weeks
MCE Maximum number of units that can be maintained simultaneously
PMC Market Clearing Expenses
FMC Permanent Maintenance Cost
HW Flexible Maintenance Cost
I All weeks of the hour
Cp Reliability index
u(t) Penalty factor
Unit step function

8
 LIST OF TABLES

Table 6.1: Weekly electricity prices and weekly peak load in percentage of annual
peak42

Table 6.2: Generating unit data and operating and maintenance cost data44

Table 6.3: Starting week of each unit for maintenance using GA46

Table 6.4: Starting week of each unit for maintenance using NSGA-II 49

Table 6.5 Comparison between single objective approach and multi objective
approach52

9
 LIST OF FIGURES

Figure 3.1: Power System Structure under Vertically Integrated Utilities ..12

Figure 3.2: Deregulated Power System 15

Figure 4.1: Flowchart of Genetic Algorithm 25

Figure 4.2: Flowchart of NSGA-II31

Figure 6.1: Algorithm of single objective approach 45

Figure 6.2 : Reliability index of GENCO’s proposal, ISO’s proposal and final maintenance
schedule 47

Figure 6.3: Performance of GENCO’s objective function in single objective

approach47

Figure 6.4: Performance of GENCO’s objective function in single objective

approach48

Figure 6.5: Reliability index of GENCO’s maintenance schedule 50

Figure 6.6: Performance of GENCO’s objective function in multi objective approach50

Figure 6.7: Performance of ISO’s objective function in multi objective approach51

Figure 6.8: Number of iterations versus time graph for both approaches52

10
CHAPTER 1

Introduction
The electrical sector faces significant challenges and is currently transitioning in a number
of ways. Following the deregulation of the industry, the opening of the energy market to
competition has significantly brought about fundamental changes. The government no
longer fundamentally controls electricity prices; instead, the market influences them. It is
meant to encourage innovation, bring down costs, and result in improved services. But the
reality is that leaving behind the initial monopoly (state) system is a drawn-out process. In
fact, this new system introduces new actors and redefines the functions or activities of the
current ones in a way that creates fresh concerns, particularly with regard to the
management and regulation of the market. The industry is expanding consistently in
tandem with this development.

This brief overview of the energy industry highlights the significant issues that the
electrical industry is currently facing in trying to balance the pursuit of profits, which
resulted from the competition-opening policy, with environmental sustainability.

Naturally, but not always, the issue of profitability and the issue of dependability are
intertwined. In fact, electrical companies work to prevent expensive unexpected failures
and reduce the ensuing downtime. Therefore, maintenance management is a significant
economic concern since it reduces investment costs by extending the life of the generating
units, avoids unneeded downtime, and reduces operational costs. Preventive maintenance
scheduling of generating equipment is a challenging task in a large continuously operating
entity such as an electric power system. The removal of generating units for maintenance
can create excessive risk to the system under certain load conditions. Maintenance

1
scheduling should therefore include an appreciation of the risk to the system when
generating units are removed from service for maintenance.

Recently, there has been pressure on the global electric utility industry to be deregulated or
reorganized in order to boost productivity, save costs, or provide customers with more
options. More study is required in this area to improve intelligent knowledge processing.
To handle these circumstances, the current centralized system for power system control,
operation, and planning needs to be modified. Maintenance and scheduling in a
restructured power system are becoming increasingly important with the promotion of the
deregulation of electric power systems. The three primary components of a deregulated
electricity system are generating (GENCOs), transmission (TRANSCOs), and distribution
(DISCOs). The primary responsibilities of these three components will stay the same, but
new forms of unbundling, coordination, and standards must be devised to comply with
FERC requirements, guarantee unrestricted access for all users and competition without
discrimination. Each segment has a specific role to play in ensuring the system's requisite
level of reliability. Therefore, in order to maintain the competitive energy market, it is the
duty of each segment to carry out the required maintenance on their facilities. In this
approach, GENCOS will make the bulk of the decisions about unit maintenance scheduling
rather than just ISO anymore. GENCOS will attempt to plan the maintenance of its units in
accordance with their operational circumstances.

Naturally, GENCOS want to get the most out of their investment. They all want to reduce
their own personal market investment losses. Each GENCO therefore aspires to schedule
maintenance during the weeks when the MCP is lowest in order for MIL to decrease.
Therefore, in accordance with the projection for the market clearing price, the GENCO's
objective function is to sell as much electricity as feasible. Numerous restrictions are being
considered, including generation capacity, maintenance continuity, and maintenance
duration. When the GENCOS units are offline for maintenance, ISO aims to maximize the
system's reserve at every time interval, provided that the purchase price doesn't increase
from a predetermined level.

2
In a centralized electric power system an acceptable generation maintenance schedule is
determined by the system operator and imposed on producers. The goal of creating such a
plan is to strike the right balance between maximum reliability and minimal expense. Keep
in mind that maintenance delays reduce dependability and raise operating expenses. The
currently reorganized electric energy systems, however, render this centralized framework
obsolete. Although independent system operators (ISO), the term used in the new
framework, still have the responsibility of ensuring reliability, producer or consumer
profits or costs are no longer their concern.

In reality, the ISO, which tries to maximize social welfare, looks for an annual plan for
generation maintenance that provides a same level of dependability every week of the year.
However, each producer aims to plan the maintenance of its own units so that the loss of
earnings due to maintenance outages is as small as possible in order to maximize profit.
These goals are obviously at odds with one another: the ISO strategy typically results in
scheduling units for maintenance during times of low demand, whereas the separate
strategies of the manufacturers typically result in scheduling units for maintenance during
times of low price. A generation maintenance plan that ensures an appropriate level of
system security may be negotiated by the ISO with producers. Similar to this, any
generator is required by regulatory agreement to participate in the necessary negotiations
with the ISO to schedule its mandatory maintenance.

Theoretically, it is possible to think of optimizing a system of independent entities by

building a sizable mathematical program and solving it centrally (for example, through the
ISO) utilizing the present computing resources and solution methods. But in reality, this is
frequently not achievable. One needs complete knowledge of local objective functions and
constraints in order to solve an issue centrally. One must look to the coordinating aspects
of decomposition in order to optimize specific goals in restructured power systems. In
particular, to arrive at the best answer with limited information, one must coordinate
entities. Here, resources that the entities compete for are identified, and each entity is
charged based on how much of those resources it uses. Finding equilibrium resource prices

3
while satisfying local and system limitations will be the goal in order to coordinate the
entities

4
CHAPTER 2

Literature Review

Antonio J. Conejo, Raquel Garcia-Bertrand and Manuel Diaz-Salazar proposed

Generator Maintenance Scheduling in Restructured Power System [1]. The scheduling of
generation maintenance in a competitive electric energy environment is the subject of their
work. In a unified setting, the framework administrator determines an upkeep booking plan
that accomplishes an ideal unwavering quality while limiting expense and forces into all
makers. Their work sets up a suitable organizing component that permits accomplishing
age support plan that fulfills maker greatest benefit measures while accomplishing an
adequate degree of unwavering quality in every seven day stretch of the year.

Aijaz Ahmad and D.P. Kothari proposed A Practical Model for Generator Maintenance
Scheduling with Transmission Constraints [2]. A transmission network representation was
used to develop an optimal maintenance scheduling for generating units in a power system
in their research. The maintenance model now includes a DC load flow to account for
network constraints, making the maintenance schedule more practical. For the purpose of
scheduling maintenance for transmission-constrained generators, a mixed interior linear
programming model has been developed.

R. Eshraghnia, M.H. Modir Shanechi and H. Rajabi Mashhadi proposed Generation

Maintenance Scheduling in Power Market Based on Genetic Algorithm [3]. The
generating units of a GENCO in a Restructured Power System can be optimally scheduled
for preventive maintenance using a genetic algorithm presented in their research. Technical

5
constraints like generation capacity, maintenance duration, and maintenance quantity are
being considered. It uses a genetic algorithm to determine the best maintenance schedule
for the anticipated market clearing price.

The proposal modern heuristic techniques for scheduling generator maintenance in power
systems [4] was proposed by K.P. Dahal, J.R. McDonald and G.M. Burt. To get around
some of the drawbacks of traditional solution techniques, it is suggested to use computer-
based modern heuristic solutions like simulated annealing and genetic algorithms. For a
GMS test problem, the applications of simulated annealing and genetic algorithm
approaches using an integer representation have been shown. An integer programming
problem with reliability-based objective function and common problem constraints is used
to formulate a test problem. The performance of the solution technique is discussed, as
well as its implementation.

The proposal A Survey of Generator Maintenance Scheduling Techniques [5] was made
by Al-Arfaj Khalid and Karamitsos Loannis. Using traditional mathematical
programming approaches or heuristic methodologies, numerous maintenance scheduling
strategies have been presented. The simplest solution is offered by heuristic methods,
which rely on trial and error techniques. Techniques based on mathematical optimization
are entirely different from traditional programming and heuristic methods that rely on trial
and error. These methods have been put out to address trivial maintenance scheduling
issues. This focuses on the distinctions between the two approaches to the GMS problem.
This document includes numerous solutions.

Maintenance Scheduling: Description, Literature Survey, and Interface with Overall

Operations Scheduling [6] was the idea put forth by Zia A. Yamayee and Palo Alto. This
paper presents a description of the maintenance scheduling problem that covers all
significant and frequently used terminologies and features of the problem, a thorough
critical review of the literature in this field with references to some materials that are
indirectly related, and interactions between the maintenance scheduling sub problem and
other operations scheduling sub problems like production scheduling and unit
commitment.

6
The concept of maintenance scheduling optimization using a genetic algorithm with a
probabilistic fitness function [7] was proposed by A. Abdul Whab, R. Billinton, A. A.
Eldamaty, and S. O. Faried. This article outlines a novel method for scheduling generator
unit maintenance. In this method, the optimization process is carried out using a genetic
algorithm, and the fitness function is assessed using a well-being analysis. When compared
to schedules produced using the levelization technique, those produced using the
optimization technique had greater overall reliability levels.

The concept of scheduling maintenance for reliable gearbox systems [8] was put forth by
Gokturk Poyra Zoglu and Hyung Seon Oh. In order to keep the system's overall
operation costs to a minimum, this paper offers a transmission switching application for
centralized transmission switching. The suggested model has the following limitations:
power flow restrictions, ramp-up and down limits, unit minimum up and downtime limits,
power balance equalities, power generation limits, voltage magnitude limits, ramp-up and
down limits. It has the advantage of lower operation costs for transmission companies and
central dispatcher. A mixed integer linear programming formulation is used to describe an
optimization procedure.

Particle swarm optimization based approach for GMS [9] was the idea put forth by Dipti
Srinivasan and Chin Aik Koay. In order to resolve a multi-objective, highly constrained
generator maintenance scheduling problem, this work offers a particle swarm
optimization-based approach. It is demonstrated that the strategy based on particle swarm
optimization is efficient in producing workable schedules in a respectable amount of time.
In this study, actual data from a real power system were used, and the outcomes on the
same set of data were compared to those from other evolutionary approaches. In addition, a
unique notion for spawning and selection mechanisms in a hybrid particle swarm method
is introduced in this research. The findings imply that this hybrid model converges to a
better solution more quickly than the conventional PSO technique. It is anticipated that this
hybrid strategy can be easily used to achieve improved convergence for similar
optimization and scheduling challenges.

7
A survey of particle swarm optimization applications in electric power systems [10] was
proposed by M. R. Al Rashidi and M. E. El-Hawary. Recently, numerous research
domains have paid more attention to particle swarm optimization (PSO). This article
provides a thorough analysis of the various PSO applications used to resolve optimization
issues in the field of electric power systems. It focuses on the PSO's essential traits and
benefits over alternative optimization techniques. Additionally, current trends in the
growth of PSOs in this region are investigated. The paper also highlights PSO's potential
theoretical investigations and potential future applications in the field of electric power
systems.

Generator maintenance scheduling of power system using hybrid technique [11] was
suggested by Suraj Kumhar and Mantosh Kumar. This paper discusses how to schedule
generator maintenance for a power system while maintaining crew/manpower, generation
limits, precedence constraints, demanded load, maintenance window, loss of load
probability, and reliability constraints. The scheduling is based on the objective function's
minimization.

Short-term transmission line maintenance scheduling in a deregulated system [12] was

proposed by Marwali and Shahidehpour. In this study, they suggested a decomposition
method for line maintenance scheduling under transmission and voltage limitations based
on duality theory. The given formulation consists of a master program, two independent
programming sub problems, and a master program. Transmission and voltage issues are
independently resolved in the sub problems while the maintenance issue is resolved in the
master problem. The proposed method will optimize the trade-off between maintenance
cost and revenue loss because halting a transaction or buying reactive electricity is related
to revenue loss.

A technique for the automated scheduling of the maintenance of generating facilities [13]
was put out by Christiaanse and Palmer. The effects of maintenance schedules on the
cost and dependability of the power system rise along with the size, number, and
complexity of producing facilities. The approach and computer program that have been
created as an operational and planning tool for scheduling the maintenance of generating

8
facilities are described in this article. The approach is a way to create maintenance plans
that improve dependability and adhere to all real-world scheduling restrictions. The
maintenance scheduling problem, the model upon which the scheduling approach is based,
and the optimization technique are all described in the article.

9
CHAPTER 3

Power System Industry

3.1 Introduction

Market-based strategies for unbundling a variety of services offered by self-interested

organizations such as power generators (GENCO), transmission providers (TRANSCO),
distribution businesses (DISCO), and a host of others have been developed as a result of
the restructuring of the electric power industry. Decision-making instruments must be
developed to assess the impact of competition when these institutions transition towards
restructured market-based functioning. Power firms are facing more difficulties selecting
the right operational planning processes for scheduling equipment maintenance as a result
of market rivalry. The decision must take into account the cost-revenue tradeoffs involved
in assessing the impact of maintenance outages on restructured utilities, as well as
coordination between long-term and short-term maintenance programs.

The restructure is thought to offer a significant venue for the application of decomposition
techniques to coordinate the optimization of different objectives among the self-interested
actors. Earlier, several deconstruction approaches were inspired by the inadequacy of the
time's computing capability to tackle large-scale centralized problems. Since then,
computing technology has dramatically advanced, making it much easier for power
engineers to handle exceedingly complex problems. As a result, interest in decomposition
methods drastically decreased. Decomposition approaches, however, are now becoming
more pertinent for an important class of optimization problems in restructured power
systems.

10
3.2 Vertically Integrated Utilities

A utility that is vertically integrated owns the entire supply chain, including generating,
transmission, and distribution. In the past, all utilities had a monopoly on the generation
and sale of power and were vertically integrated. Electric utilities exercised exclusive
control over the sale of energy within their service areas for many years, functioning as
monopolies. The three aspects of the electrical industry were covered under the control:
generation, transmission, and distribution. Utility companies also had control over billing
and metering operations. In order to safeguard the interests of both consumers and utilities
during the monopoly era, several rules, legislation, and regulatory bodies were established
to regulate the operations of electric utilities. In addition to creating and enforcing
standards for offering utility-related services, the oversight included reviewing and
approving tariffs (or rates) charged for electric energy.

The vertically integrated utilities' organizational structure is depicted in Figure(3.1). Each

utility:

● Controls and owns all or the majority of the generation, transmission, and

distribution infrastructure located within its geographic area.

● Possesses a monopoly over the sale of electricity to subscribers within its

administrative area.

● Is required to satisfy the demands of its consumers by selling them electricity.

● Either has rates (prices) that are established by regulatory bodies or is publicly

owned and not operated for profit.

11
.

..

Figure 3.1 Power System Structure under Vertically Integrated Utilities

3.3 Restructured Power System

Power system restructuring is the process of commercializing the current power system in
a completely different way in order to increase competition and transparency in the power
industry and provide customers with more reliable and affordable power. The basic goal of
restructuring is to improve competition and lower net costs by letting market forces
determine the price of power supply. By allowing the power supply to operate in a
competitive manner and giving customers the option of selecting their preferred electric
energy supplier, restructuring fosters an environment of open competition.

3.3.1 Reasons for restructuring or deregulation of Power Industry

12
Customers and private businesses both can benefit from the competitive climate in a
number of ways. According to various claims, some of the major advantages of
deregulating the power industry would be:

● The cost of electricity will decrease: It is generally accepted that competitive

pricing are lower than monopolist rates. In a highly competitive market, the
producer will aim to sell the power at its marginal cost.

● Customers will have the option of choosing their retailer. Retailers will compete on

the other services supplied to customers as well as the price they are willing to
offer. These can include improved strategies, dependability, quality, etc.

● Customer-focused service: Retailers would offer superior service to the monopolist.

● Regulation and a lack of competition prevented electric companies from innovating

or taking chances on novel concepts that would raise customer value. In a market
with little restrictions, the electric utility will always try to innovate something for
the betterment of service and in turn save costs and maximize the profit.

3.3.2 Various Entities Involved in Deregulation

Deregulation has brought in a number of new companies to the energy market while also
expanding the range of activity for many of the already established players. Different
market structures differ in how each entity is specifically defined and what function it
serves inside the system. However, the following entities can be recognized broadly:

GENCO (Generating Company): GENCO is a generator owner-operator that operates one

or more generators and sells power on the open market. Similar to how a coal mining
business might sell coal in bulk at its mine, GENCO sells electricity at its locations.

TRANSCO (Transmission Company): TRANSCO transports large amounts of power from

its source to its destination. The TRANSCO is responsible for the ownership and upkeep
of the gearbox facilities, as well as many of the administrative and engineering tasks
necessary to guarantee the system's proper operation. The TRANSCO owns and maintains

13
the monopoly's gearbox lines, but does not run them, in some deregulated industries.
Independent System Operator (ISO) handles that. For the use of its lines, TRANSCO is
compensated.

DISCO (Distribution Company): It is the owner-operator of the local power distribution

system that provides electricity to certain businesses and residences. In some locations, the
local distribution role and retail function are merged, meaning that wholesale power is
purchased either through the spot market or through direct contracts with GENCOS and
then supplied to end users. However, the DISCO doesn't sell the power in a lot of other
situations. It exclusively owns and runs the neighborhood distribution network, and it
makes money by moving electricity through its network.

RETAILCO (Retail Energy Service Company):It sells electricity as a retailer. The retail
divisions of the erstwhile vertically integrated utilities will make up a large portion of
them. A RETAILCO purchases electricity from GENCOS and sells it straight to
customers. No physical components of the power network are owned by RETAILCO.

Market Operator: The market operator offers a venue where buyers and sellers can
exchange electricity. It utilizes a computer program that compares offers and bids made by
sellers and purchasers. The market operator is in charge of the settlement procedure. A
day-ahead market is normally administered by the market operator. The system operator is
in charge of managing any near-real-time markets.

Independent System Operator (ISO): The ISO is the organization in charge of making sure
the security and dependability of the entire system. Being an autonomous body, it stays out
of the trades on the power market. With the exception of a small amount of reserve
capacity, it typically does not own any generating resources. The ISO obtains a variety of
services, such as the supply of emergency reserves or reactive power from other system
participants, in order to preserve the security and dependability of the system. In other
nations, ISO also controls the transmission infrastructure. In these systems, the ISO is
commonly referred to as the Transmission System Operator (TSO).In the case of a ISO

14
being completely neutral of every other activity except coordinate, control and monitor the
system.

Customers: A consumer of electricity is referred to as a client. The end user has a variety
of options for purchasing electricity in a fully deregulated market when the retail sector is
likewise open to competition. It can opt to acquire power directly from a GENCO or even
the neighborhood retail service provider, or it can bid on a purchase on the spot market. On
the other hand, only the biggest consumers are allowed to select their supplier in markets
where competition only occurs at the wholesale level.

Figure 3.2 Deregulated Power System

15
CHAPTER 4
Maintenance Scheduling in Restructured
Power System

4.1 Introduction
A list of maintenance tasks that must be completed at regular intervals in order to prevent
breakdowns is known as maintenance scheduling. It has been repeatedly demonstrated that
sporadic continuing repair causes assets to degrade, have a shorter lifespan, and have
higher long-term capital costs. The basic objective of maintenance planning is to stop
equipment failure in its tracks. This covers examinations, modifications, routine
maintenance, and scheduled shutdowns. Regular periodic maintenance inspections can
lengthen the useful life of your assets and lessen the need for component replacement or
repair.

When autonomous entities like GENCOs and TRANSCOS and their self-interested
motivations are taken into account, maintenance scheduling becomes a challenging
discrete unpredictable nonlinear optimization problem. Generating unit maintenance and
gearbox line maintenance are the two basic types of facility maintenance in power
systems. The upkeep of generating and transmission might theoretically be studied
separately. To stimulate practical optimization and workable solutions for the two issues,
system restrictions including network traffic limits, energy demands, and reliability needs
will be intimately tied to the two solutions. For a number of on-line facilities, maintenance
scheduling provides a preventive outage plan for a given time period, such as one year. In

16
order to best achieve a certain target, the solution process often divides the time horizon
into manageable time chunks, such as weeks, and then decides when each facility outage
should start.

The simultaneous resolution of all parts of the operation scheduling problem, of which
maintenance scheduling constitutes a large portion, is necessary for the efficient running of
an electric power system. In order to minimize costs and meet system limits, Generation
Maintenance Scheduling (CMS) entails scheduling units for preventative maintenance
during specific times and at a desired security level. In addition to being highly
dimensional and nonlinear, this problem is also stochastic in nature. Therefore, the
optimization techniques offered are most appropriate for solving this challenging,
enormous problem with limits. Conventional methods that have been used to solve it have
generally consisted of formulating GMS as an appropriate optimization algorithm with
cost or reliability as an objective criterion and the other objectives being included in the
form of constraints. The quantity of constraints to include depends on the model's capacity
to manage them and produce schedules that are acceptable from an operational standpoint.
The maintenance window, generation capacity, peak reserve, hydropower, personnel, etc.
are few of the examples of these constraints. Additionally, using only the simple
generation model and ignoring the transmission network, past methods have primarily
used the simple generation model to determine maintenance plans. This is necessary in
order to prevent analysis complexity brought on by the addition of transmission
constraints.

4.2 Significance of Maintenance Scheduling

An electric power system's primary duty is to supply consumers with electricity at the
lowest possible cost while maintaining levels of reliability that are acceptable. One crucial
component of system planning is the estimation of the cost of generation. A vital tool in
the planning of an electric system is reliability analysis. These two factors, economies and
reliability, sometimes clash, posing a variety of difficult issues for power system
managers, planners, designers, and operators. Engineers and managers of power systems
have been working to attain the highest level of reliability at the most reasonable price.

17
When determining reliability, one must consider the possibility of not being able to meet
demand even in the event of unplanned unit failures. There are other aspects of reliability
assessment besides the ability to transfer the generated energy to the load sites.
Additionally, it evaluates the generation system's surplus capacity, which is needed to
handle unpredictable generator unit failures, including for maintenance on the producing
facilities, in order to limit unpredictable generator unit failure. The ability of the producing
units to work efficiently and dependably over the long term depends on regular
maintenance, which is crucial to lowering the risk of a capacity outage. Regular
maintenance lowers the chance that energy will go unsupplied, increases the availability of
units, and improves the reliability.

On the other side, taking generating units out of service for repair will decrease the
capacity that is available and may raise the chance that the system won't be able to keep up
with rising demand. A properly planned maintenance schedule can increase the system's
dependability, but a poorly managed schedule would immediately increase the likelihood
of energy shortages. Generator maintenance schedule problem decisions also have a
significant and direct impact on the operation and planning of the power system, including
economic dispatch, unit commitment, load dispatch, generation co-ordination, generation
expansion planning, and other planning activities.

4.3 Solution approaches in maintenance scheduling

4.3.1 Single objective approaches

Numerous solutions to this complex, stochastic, nonlinear optimization problem have been
offered in the literature during the course of almost three decades of international research
in GMS. Here are the various problem-solving techniques that have been adopted:

● Heuristic algorithm

● Mathematical techniques

● Artificial intelligence approach

18
4.3.1.1 Heuristic approach

The simplest solution, based on trial-and-error methods, is offered by heuristic approaches.

If a starting point is not carefully selected, the approach tends to slip into a local minimum,
which means that these techniques may not always result in the global optimal solution for
a difficult problem. Because of their simplicity and flexibility, heuristic approaches were
formerly utilized to solve maintenance scheduling issues for centralized power systems.
Maintenance scheduling is carried out consecutively in heuristic approaches. According to
a previously established scheduling order, each unit is taken into consideration separately.
Earlier phases of the research focused more on heuristic techniques. However, these
techniques cannot provide the genuine optimal answer.

4.3.1.2 Mathematical techniques

The basic mathematical methods include branch-and-bound (BaB), linear, integer, and
mixed-integer programming (LP, IP, and MIP), decomposition, and dynamic programming
(DP). The following sections explore various mathematical approaches that have been
utilized to solve maintenance scheduling issues in the literature:

i. Linear, integer and mixed integer programming

Linear programs (LP) are the most fundamental kind of mathematical programming. It has
been successfully used to solve issues ranging from common examples in business,
economics, and transportation to more extreme ones in the behavioral sciences. An
optimizations issue known as the LP model has linear variables for both the objective
function and the constraint function. With the extra restriction that the variables be
integers, integer programming is essentially LP. It becomes mixed integer programming
(MIP) if just some of the variables must be integers while the others are real.
Combinatorial scheduling issues can frequently be expressed as IP or MIP issues.

ii. Dynamic programming

19
In essence, dynamic programming is a full enumeration strategy that uses a divide-and-
conquer strategy to try and reduce the amount of computing required. The strategy works
through a number of smaller issues until it resolves the main issue. Each sub problem’s
contribution to the objective function and its ideal solution are determined. The ideal
solution, which outperforms the sub problems that have already been solved, is found for
each iteration. It uses all the information gathered previously in the resolutions of all
earlier sub problems to arrive at a solution for the current sub problems. Dynamic
programming makes an effort to use a variety of goals, including maximizing the
minimum net reserve and reducing the chance of failing to fulfil the demand for power.
The curse of dimensionality affects dynamic programming when dealing with complex
issues like maintenance scheduling issues.

iii. Branch and Bound

The Branch-and-Bound technique involves applying the following stages repeatedly. First,
the optimal solution's known location in the solution space (i.e., the set of choice variables
under consideration) is divided into subsets. Second, a subset is disregarded from further
analysis if all of its components breach the limitations imposed by the minimization issue
(fathomed). Third, a maximum value for the objective function's minimum is computed.
When the choice variables are required to belong to each remaining subset, lower bounds
are determined on the value of the objective function. The optimal decision variable cannot
lie in a subset if its lower bound exceeds the upper bound of the minimization problem,
hence that subset is fathomed. When just one subset of the decision variables is left and its
upper and lower bounds are equal, convergence has occurred.

iv. Benders decomposition

To take advantage of mixed-integer programs, the Benders decomposition algorithm was

created. In a restructured power system, which is represented by a collection of
autonomous entities, Benders decomposition has a wide range of applications. The original
issue will be divided into a master problem and a sub problem based on the linear
programming duality theory. With only a few (or no) constraints, the master issue is first

20
solved. To determine whether this solution satisfies the remaining requirements, the sub
problem is utilized as a test. If so, the solution is the best one because it minimizes the
objective given all the restrictions. If not, the master issue will be solved again with
additional constraints, starting with the constraint that is least likely to be satisfied. The
decision is made based on the Benders algorithm may require iterations between the
master problem and the sub problem

4.3.1.3 Artificial Intelligence Approach

In the literature, there are numerous Artificial Intelligence (AI) techniques with various
presentations that can be used to resolve the maintenance-scheduling issues of a power
system. Expert systems, Simulated Annealing (SA), fuzzy logic, neural networks,
evolutionary optimization, including evolutionary programming, evolutionary strategy,
and Generic Algorithm (GA), as well as their hybrids are all examples of artificial
intelligence (AI). The following discussion discusses a few AI strategies.

i. Simulated Annealing (SA)

Physically speaking, annealing is the process of first heating a solid to a high temperature,
then slowly cooling it by gradually lowering the temperature of the surrounding air. The
program simulates the metallurgical annealing process, which involves heating a material
and gradually lowering the temperature to reduce flaws while also minimizing system
energy. The starting state of this procedure is comparable to a thermodynamic system at
the beginning. The method generates a new point at random after each iteration. Based on
a probability distribution, the new point's separation from the present point or the size of
the search is determined. The relationship between the distribution's size and temperature
is reciprocal. The algorithm accepts all new points that decrease energy, but it also accepts
points that increase energy with a fixed probability. The algorithm is capable of global
exploration for better solutions and avoids becoming stuck in local minima by accepting
locations that increase the energy. To gradually lower the temperature as the algorithm
runs, an annealing schedule is chosen. The program narrows the scope of its search as the
temperature drops in order to gradually approach a minimum.

21
 ii. Genetic algorithm

Search algorithms called genetic algorithms are rudimentary representations of some of the
natural evolution's processes. In many cases, genetic algorithms offer very efficient search
methods that can be applied to classification and optimization problems. Instead of
working with a single point, genetic algorithms operate on populations of points; each
point is essentially a vector in hyperspace that represents a potential solution to the
optimization problem. As a result, a population is simply an ensemble, or collection, of
hyperspace vectors. In the population, each vector is referred to as an individual or
occasionally as a chromosome. The dimensionality of the problem vector could be
different from the size of the bit string chromosome because real numbers are frequently
expressed using binary integers in genetic algorithms. The number of real parameters in
the optimization problem is equal to the number of elements in each vector (individually).
A vector element often corresponds to one of the numeric vector's parameters or
dimensions. The format of each parameter will determine how many bits can be used to
encode each element. The size of the hyperspace being searched is determined by the total
amount of bits.

From past several years there has been a growing interest in problem-solving systems based
on the principles of evolution and machine learning, where such systems maintain a
population of potential solutions. They have some selection process based on fitness of
individuals and some ―genetic‖ operators . Holland [15] introduced the concept of a
genetic algorithm (GA). Like other AI the basic idea behind GA was to make computers do
what humans do. In order to apply a genetic algorithm to solve an optimization problem,
candidate solutions must be encoded using an appropriate representation, such as a numeric
string, and an evaluation function must be formulated to assign a quality value to every
solution produced.

22
The GA represents an iterative process, where each iteration is called a generation.
Population size specifies how many individuals there are in each generation. With a large
population size, the genetic algorithm searches the solution space more thoroughly, thereby
reducing the possibility that the algorithm will return a local minimum that is not a global
minimum. However, a large population size also causes the algorithm to run more slowly.
Each population contains a number of chromosomes, and a chromosome consists of several
"genes", and each gene is represented by 0 or 1.

There are two basic mechanisms that link a GA to the problem it is solving: encoding and
evaluation. The encoding is carried out using binary strings (chromosomes) of ones and
zeros in each bit, which is the most popular representation. An evaluation function is used
to measure the chromosome's performance, or fitness, on the problem to be solved. The GA
uses a measure of fitness of individual chromosomes to carry out reproduction.

Genetic algorithms apply three genetic operators: selection, crossover and mutation.

I. Selection
There are different types of selection methods which are given below :

▪ Tournament selection: where a small subset of individuals is chosen at random,

then choosing the best individual from that set to be a parent.

▪ The roulette wheel selection: The most common chromosome selection technique

is the roulette wheel selection (Goldberg) [16]. Roulette selection chooses parents
by simulating a roulette wheel, in which the area of the section of the wheel
corresponding to an individual is proportional to the individual's expectation. For
example, for a given population each chromosome is given a slice of a circular
roulette wheel equal to the chromosome fitness ratio. To select a chromosome for
mating, a random number is generated, and the chromosome whose segment spans
the random number is selected. It is like rotating a roulette wheel where each
chromosome has a section on the wheel proportional to its fitness. The roulette

23
 wheel is spun, and when the arrow comes to stop on one of the slices, the
corresponding chromosome is chosen

II. Crossover

Crossover specifies how the genetic algorithm combines two individuals, or parents, to
form a crossover child for the next generation. The crossover operator is applied with a
certain crossover probability, once a pair of parent chromosomes is selected. Generally, a
value of 0.7 for the crossover probability produces good results.

First, the crossover operator randomly chooses a crossover point where two parent
chromosomes "break", and then exchanges the chromosome parts after that point. As a
result, two new offspring are created. There are different types of crossovers for example:

▪ Single point crossover: where a single locus chosen at random then a parent

chromosome break and all bits after that point be swapped.

▪ Two point crossover: this involves choosing two points at random and swapping

the corresponding parts from the two parents defined by the two points

III. Mutation

Mutation specifies how the genetic algorithm makes small random changes in the
individuals in the population to create mutation children. Mutation can occur at any gene in
the chromosome with some probability. Typically, the mutation probability is in the range
between 0.001 and 0.01. Mutation provides genetic diversity and enables the genetic
algorithm to search a broader space. Genetic algorithms guarantee the continuous
improvement of the average fitness of the population, and after a certain number of
generations the population evolves to an optimal or near-optimal solution.

Given a clearly defined problem and a binary string representation for candidate solutions,
a basic GA can be represented in the following steps:

Step 1: Represent the problem variable domain as a chromosome of and define the
population size, the crossover and the mutation probability, and the evaluation function.

24
Step 2: Randomly produce an initial population of chromosomes.

Step 3: Compute the fitness of each one.

Step 4: Choose a couple of chromosomes for mating.

Step 5: Generate a pair of offspring chromosomes by applying genetic operators.

Step 6: Position the formed offspring chromosomes in the new population, then repeat Step
4, until the size of the new population becomes equivalent to the size of the initial
population.

Step 7: Substitute the initial (parent) chromosome population with the new (offspring)
population.

Step 8: Go to Step 3, and repeat the process until the termination criterion is fulfilled.

The flowchart of the Genetic algorithm is shown in Figure 4.1

25
 Figure 4.1 Flowchart of Genetic Algorithm

Genetic algorithms become popular as a powerful optimization tool appropriate for a

diversity of problems. Both GA by itself or a combination of GA and other techniques are
broadly addressed in the literature for solving maintenance scheduling for power systems.
Negnevitsky and Kelareva [20] have used GA is solving maintenance scheduling in power
systems. The objective was to maximize reserve margins subject to maintenance and
system constraints. They have designed a representation which is suitable for a variety of
problems and appropriate chromosome evaluation is suggested. A case study was solved
using GA, and the result shows that chromosome representation plays an important role in
GA where it may reduce problem complexity by including constraints. Abdulwhab et al.
[7] use the genetic algorithm optimization technique to maximize the overall system
reliability for a specified future time period in which a number of generating units are to be
removed from service for preventive maintenance.

With the advancement of research, GA with modified genetic operators, such as string
reversal and reciprocal exchange mutation, is used to solve the generator maintenance
scheduling (GMS) problem. They have used three types of encoding; integer encoding,
binary for integer encoding, and real encoding. The GMS problem is solved to minimize
the expected energy production cost and maximizing the reserve objectives subject to
maintenance windows, consecutive periods of maintenance, crew, demand reserve and
reliability. The result shows that only integer coding GA finds the global optimum
solution, irrespective of the nature of the objective function and system size. Also,
modified genetic operators were shown to be effective in reducing computation time and
improving search efficiency of the GA. It has been reported that the performance of the
GA approach can be improved by combining it with other techniques . The GA/SA hybrid
approach has been employed to solve a maintenance scheduling problem by . The hybrid
approach presented by use the integer encoding for solving the captive power plant
maintenance scheduling problem with levelized reserve reliability objective function. From
the comparison of results obtained from application of only GA and from hybrid GA/SA

26
techniques for scheduling, can be seen that the hybrid GA/SA solution technique yields
better results. Dahal et al.[4] investigated the applications of GA and SA using an integer
representation to encode candidate solutions to GMS test problems with a reliability
criterion. The evaluation function is a weighted sum of the objective function and the
penalty function for violations of the constraint. The authors concluded that the SA and
GA are robust and stable techniques for solving GMS problems. Also, Dahal et al. [4]
propose solving centralized maintenance problems using GA with integer‘s representation
using fuzzy evaluation functions. Since fuzzy logic can be used to deal with multiple
objectives, it was used to combine the objective of maximizing reliability and considering
the flexibility in the manpower constraint. The fuzzy evaluation function is developed as a
combination of a crisp penalty function for inflexible load constraint and a fuzzy penalty
function for the objective and the flexible manpower constraints. The results obtained
using the fuzzy logic evaluation function were compared with those obtained from GAs
with crisp evaluation functions, and the fuzzy logic method was shown to achieve an
effective trade-off between reliability and manpower within the allowed flexibility.

Burke and Smith [21] presented a technique named Memetic approach for solving real
scale maintenance scheduling problems in centralized structures. The objective was to
minimize the sum of the overall fuel and maintenance costs. Memetic is a genetic
algorithm combined with Tabu search. Tabu search is a powerful optimization procedure
that has been successfully applied to a number of combinatorial optimization problems.
Memetic takes the concept of evolution as employed in GA. It has a memory as unit of
information instead of gene in GA . A population of information can be created and a good
one has a better chance of survival than a bad one, and they can be combined to form new
ideas. They investigated the use of Memetic algorithms for solving thermal generator
maintenance scheduling problems. A comparison between Tabu search and Memetic
algorithm shows that Tabu is more affective for small problems and Memetic algorithm
will outperform Tabu search for large problems. Also, they show that the Memetic
algorithm using Tabu search as the local optimizer yields greater benefits than simulated

27
annealing which was used previously in solving thermal generator maintenance scheduling
problems.

iii. Particle Swarm Optimization

The dynamics of swarm behavior seen in nature (such as a flock of geese or a swarm of
bees) are mimicked by the Particle Swarm Optimization (PSO) algorithm. In PSO, a
starting group of people are chosen at random, and each represents a potential solution.
The term "particle swarm" refers to this group of people, which is another name for them.
The population of particles iteratively seeks for (or ideally converges to) the global
optimum, where each particle's motion in the design space is dictated by a velocity update
equation that was inspired by the perception of swarm behavior. The best fitness spots are
either recorded or recalled by each person using this approach. The particle best is a
person's best response or accomplishment, and the neighbors are informed of it.

iv. Ant colony optimization

The GMS problem in regulated power systems has also been solved using Ant colony
optimization (ACO), a constructive metaheuristic inspired by ant colony behavior, with a
cost-based objective. The program chooses a maintenance starting period for each
generating unit and then constructs solutions incrementally. Pheromone level, associated
with how frequently the component has been chosen, and some greedy heuristics are used
to determine the selection process.

4.3.2 Multi-objective approach

The multi-objective optimization problems involve more than one objective function that
are to be maximized or minimized. The answer is the set of solutions that define the best
tradeoff between competing objectives. In the single-objective optimization problem, the
superiority of a solution over other solutions is easily determined by comparing the
objective function values. However in multi-objective optimization problem, the goodness
of a solution is determined by the dominance. The presence of multiple objectives in a

28
problem, in principle, gives rise to a set of optimal solutions (largely known as Pareto-
optimal solutions), instead of a single optimal solution. In the absence of any further
information, one of these Pareto-optimal solutions cannot be said to be better than the
other. This demands a user to find as many Pareto-optimal solutions as possible. Classical
optimization methods (including the multi-criterion decision-making methods) suggest
converting the multi-objective optimization problem to a single-objective optimization
problem by emphasizing one particular Pareto-optimal solution at a time. When such a
method is to be used for finding multiple solutions, it has to be applied many times,
hopefully finding a different solution at each simulation run. One of the most important
multi-objective optimization techniques is discussed below.

Non-dominated Sorting Genetic Algorithm II (NSGA-II)

A reliable multi-objective algorithm with various applications in the real world is NSGA-
II. NSGA-II was proposed by Deb et al.in [19]. NSGA-II still offers a lot of usefulness, if
not just as a reliable benchmark to test against, even though it can currently be viewed as
an outmoded method. The next generation is chosen by NSGA-II employing non
dominated-sorting and crowding distance comparison after offspring are produced using a
certain kind of crossover and mutation. The Non-dominated Sorting Genetic Algorithm II
(NSGA-II) appears as a potent tool for tackling multi-objective optimization problems in
the field of optimization, where competing objectives frequently necessitate sophisticated
decision-making. The evolutionary algorithm known as NSGA-II, created by Kalyanmoy
Deb, expands on the principles of genetic algorithms and provides a novel approach to
solving optimization problems with many conflicting objectives. Following are the steps
involved in NSGA-II optimization technique:

Step 1: Initialize population: Create an initial population of candidate solutions

(chromosomes). Each solution encodes the decision variables for the optimization
problem.

Step 2: Evaluate fitness: Calculate the fitness values of each solution with respect to all
objectives in the optimization problem.

29
Step 3: Fast Non-dominated Sorting: Divide the population into different fronts based on
dominance relationships. Solutions in the first front are non-dominated by any other
solution, solutions in the second front are dominated only by solutions in the first front,
and so on.

Step 4: Crowding Distance Assignment: For solutions in the same front, calculate the
crowding distance for each solution. The crowding distance represents how far a solution
is from its neighboring solutions in the objective space. Let f 1 (x), f 2 (x),……. f n (x) be n
objective functions , f 1 max (x ), f 2 max (x ),……. f nmax (x ) be their maximum values and f 1 min ( x )
, f 2 min (x ),……. f nmin (x ) be their minimum values respectively, then crowding distance if
the solution is given by:

n
f m ( x i +1 )−f m (x i−1)
CD i= ∑
m=1 f mmax ( x )−f mmin (x)

Step 5: Selection for Reproduction: Choose parents for the next generation's offspring.
Solutions from less crowded fronts and solutions with higher crowding distances are
preferred.

Step 6: Variation Operators (Crossover and Mutation)

● Crossover: Perform crossover on selected pairs of parent solutions to generate

offspring. Crossover combines genetic information from parents to create new

solutions.

● Mutation: Apply mutation to some of the offspring solutions. Mutation introduces

small random changes to ensure diversity.

Step 7: Offspring Evaluation: Calculate the fitness values of the newly generated
offspring solutions.

Step 8: Merge Parents and Offspring: Combine the parent solutions with the offspring
solutions to create a combined population.

30
Step 9: Environmental Selection: Select solutions for the next generation's population by
maintaining a diverse set of solutions. Solutions from different fronts and those with
higher crowding distances are preferred.

Step 10: Termination Criteria: Check if a termination condition is met. This could be a
maximum number of generations reached or a convergence criterion based on the change
in the Pareto front.

Step 11: Final Output: Once the termination criteria are met, the algorithm stops, and the
final output is the set of solutions that approximate the Pareto-optimal front.

The flowchart of the NSGA- II is shown in Figure 4.2

Figure 4.2 Flowchart of NSGA- II

31
CHAPTER 5
Generator Maintenance Scheduling –
methodology, problem formulation and
solution

5.1 Introduction
With single, multiple, and a set of constraints, the maintenance scheduling problem in
deregulated power systems is treated as an optimization problem. Various costs
(particularly maintenance costs) functions are included in the objective functions for
deregulated power systems that are based on economic principles. Any real-world problem
that has more than one solution involves maximizing or minimizing an objective function.
In general, an electric power system operates with two goals in mind. The utility's overall
running costs must be kept to a minimum, and system reliability must be increased. The
second objective, which is to maintain reliability over a particular threshold, is sometimes

32
used as a constraint rather than as an objective. One of the following two categories can be
used to classify objective functions for maintenance scheduling that have been reported in
the literature: costs and reliability

5.1.1 Optimization objectives

It is particularly challenging to combine the dependability and economic objectives in a

single mathematical formulation since they frequently conflict with how power systems
operate. Consideration of the cost as an optimization aim and the reliability as a constraint
seems more appropriate in the majority of situations. The following criteria should be met
by a GENCO's ideal maintenance schedule for generating units:

i. Lower the GENCO's operating expenses

ii. Maintain system dependability
iii. Increase the lifespan of producing units
iv. Reduce the investment expenses associated with installing new producing units.

5.1.2 Problem Constraints

Some of the restrictions for maintenance schedule are listed below.

i. Maintenance interval: the amount of time needed for maintenance.

ii. Number of maintenance crews.
iii. Maintenance period during which a facility can be maintained.
iv. Sequence precedence: some facilities should be scheduled for repair before others
v. Repeated maintenance: some facilities may require maintenance more frequently
throughout the course of a specific time
vi. Load demand
vii. The facilities rated capacity.
viii. The maximum number of facilities that can be maintained geographically is eight.
ix. The facility's dependability and the rate of forced outages
x. System dependability

33
5.2 Methodology and problem formulation for Generator Maintenance
Scheduling

The approach to problem-solving for a GMS comprises an iterative process to coordinate

maintenance scheduling between the ISO and the producers so that a suitable level of
reliability is attained throughout the weeks of the year in a way that is suitable for each
producer. The iterative process depends on applying an appropriately calibrated incentive
or disincentive to each week of the year to encourage performing maintenance in it. A
suitable level of reliability can be attained every week of the year thanks to this incentive
or disincentive mechanism. We have opted two different methodologies for solving the
GMS problem which are discussed below:

5.2.1 Methodology Ⅰ: Single-objective approach by considering the

GENCO’s and ISO’s objective separately

In a centralized electric power system, the system operator determines an appropriate

generation maintenance schedule and imposes it on producers. Achieving a suitable
balance between maximum reliability and minimal expense is the goal while putting
together such a plan. Be aware that maintenance interruptions raise operating costs and
reduce reliability. In the recently reorganized electric energy systems, this centralized
paradigm is no longer applicable. Although producer/consumer earnings or costs are no
longer its concern, the operator, known as the independent system operator (ISO) in the
new paradigm, is still responsible for ensuring reliability.

The ISO actually seeks a generation maintenance annual plan that assures comparable
reliability throughout the weeks of the year in order to maximize societal welfare. To
maximize their profits, each producer, however, tries to plan out their own unit
maintenance so that the amount of money they lose due to maintenance outages is kept to a
minimum. These goals are obviously at odds with one another: the ISO strategy typically
results in scheduling units for maintenance during times of low demand, whereas the
separate strategies of the manufacturers typically result in scheduling units for
maintenance during times of low price. The ISO has the right to bargain with

34
manufacturers for a generation maintenance schedule that ensures a sufficient level of
system security. Similar to this, any generator is required by regulation to negotiate with
the ISO in the proper manner to schedule its mandatory maintenance

The methodology presented aims at solving the GMS problems separately by the GENCO
and the ISO. The GENCO and ISO both presents their respective maintenance schedules
on the basis of their objectives. The ISO then imposes the penalty on the GENCO based on
the difference in their reliability indices. The GENCO then compares the ISO reliability
index with its possible schedules and decides the final schedule with maximum profit and
ample amount of reliability

5.2.1.1 Problem formulation

i. GENCO’s objective:

The profit of a GENCO is determined by adding the individual unit profits from the
auctions over the scheduling horizon to determine the total profit. To maximize its own
profit, each GENCO resolves the appropriate maintenance scheduling challenge. As a
result, the objective role of GENCO in maintenance planning might be expressed as
follows:

Objective function of GENCO

[( )
❑ N T N T
C i∗1000
Max ∏ ¿ ∑ ∑ [ ( MCEi∗P¿−C ( P¿ ) )∗( 1−x ¿ ) ] HW −∑ ∑ PM ∗x ¿∗Pimax + ( FM C i∗x ¿∗Pimax
❑ i=1 t =1 i =1 t =1 8760
(1)

The objective function (1) reflects the profit of GENCO, which is determined by
subtracting its total revenues from the corresponding costs, which comprise production
costs, fixed costs, and variable maintenance costs.

Constraints of problem

The continuity of maintenance activities, certain time intervals for maintenance of

particular generating units, maintenance personnel limits, and the maximum and minimum

35
capacities of each generating unit are a few fundamental constraints that should be
established. The constraints of the GENCO's maintenance scheduling problem are listed
horizon.

1. Maintenance outage duration : This constraint ensures that for each unit that is
maintained the required number of time periods

∑ x ¿ =Di ∀ i
i=1
2. Continuous maintenance : This constraint ensures that the maintenance of any
unit must be completed once it begins

x i ( t )−x i ( t−1 ) ≤ x i ( t + Di−1 ) ∀ i, ∀ t

3. Maximum number of units simultaneously in maintenance: This constraint

limits the maximum number of units that can be maintained simultaneously by
a producer.

∑ x¿ ≤ S ∀ t
i=1
4. Capacity and Minimum Power Output: The power generated for each online
unit must be within a certain range represented by its minimum power output
Pimin and its capacity Pimax

Pimin ≤ P ¿ ≤ Pimax
ii. ISO’s objective:

The ISO should arrange the units for maintenance by maximizing the system capacity
reserve in the whole year. The objective role of GENCO in maintenance planning might be
expressed as follows:

Objective function of GENCO

N N
Maximize Rt =∑ Pimax −Pd −∑ Pimax x ¿ ∀ t (2)
i=1 i=1

36
The objective function (2) reflects the reserve of the ISO which is determined by
subtracting sum of demand and maximum capacity of units under maintenance from the
total capacity of system when no unit is under maintenance

Constraints of problem

1. Minimum net reserve: The minimum net reserve is the lowest of all the weekly
net reserves over the maintenance scheduling horizon. This constraint ensures a
net reserve above a specified threshold for all periods.
N

∑ Pimax ( 1−x ¿ )−Pd > R min ∀ t

i=1

2. Maintenance outage duration : This constraint ensures that for each unit that is
maintained the required number of time periods

∑ x ¿ =Di ∀ i
t =1
3. Continuous maintenance : This constraint ensures that the maintenance of any
unit must be completed once it begins

x i ( t )−x i ( t−1 ) ≤ x i ( t + Di−1 ) ∀ i, ∀ t

4. Maximum number of units simultaneously in maintenance: This constraint

limits the maximum number of units that can be maintained simultaneously by
a producer.

∑ x¿ ≤ S ∀ t
i=1
5. Capacity and Minimum Power Output: The power generated for each online
unit must be within a certain range represented by its minimum power output
Pimin and its capacity Pimax

37
 Pimin ≤ P ¿ ≤ Pimax
The reliability index of the GMS at period t is given by

∑ Pimax ( 1−x ¿ )−P d (t) (3)

I ( t )= i=1
❑

iii. Final objective function of GENCO

❑ N T N T
Max ∏ ¿ ∑ ∑ [ ( MCEi∗P¿−C ( P¿ ) )∗( 1−x ¿ ) ] HW −∑ ∑ ¿ ¿ ¿ ¿ ¿ ¿
❑ i=1 t =1 i =1 t =1

(4) The final objective function (4) of GENCO determines the overall
profit of the GENCO by considering the penalty imposed by ISO

5.2.2Methodology ⅠⅠ: Multi-objective approach by considering the

GENCO’s and ISO’s objective simultaneously

To adapt to the scenarios of declining operational costs and to provide customers with
additional options, the current centralized system for the control, operation, and planning
of the electricity system needs to be modified. Maintenance planning in a restructured
power system is crucial given the push for deregulation of electric power networks. The
three primary components of a deregulated electricity system are generating (GENCOs),
transmission (TRANSCOs), and distribution (DISCOs). The primary responsibilities of
these three parts will remain the same, but in order to comply with FERC regulations, new
arrangements for unbundling, coordination, and rules must be made in order to ensure
competition and open access to all users without discrimination.

Naturally, GENCOs want to get the most out of their investments. They all want to limit
the MIL's impact on their personal interests. Therefore, every GENCO aspires to schedule
maintenance for the weeks when the MCP is at its lowest, causing MIL to decrease.
Therefore, in accordance with the projection for the market clearing price, the GENCO's
objective function is to sell as much electricity as feasible. Numerous restrictions are being
considered, including generation capacity, maintenance continuity, and maintenance

38
duration. When GENCO units are offline for maintenance, ISO aims to maximize the
system's reserve at every time interval provided that the purchase price stays at a certain
level.

The methodology presented aims at solving the GMS problem by considering the
objectives of both the GENCO and ISO simultaneously. The objective functions of
GENCO and ISO are solved simultaneously by using multi-objective optimization
technique.

5.2.2.1Problem formulation

GMS objectives:

The profit of a GENCO is determined by adding the individual unit profits from the
auctions over the scheduling horizon to determine the total profit. To maximize its own
profit, each GENCO resolves the appropriate maintenance scheduling challenge. As a
result, the objective role of GENCO in maintenance planning might be expressed as
follows:

[( )
❑ N T N T
C i∗1000
Max ∏ ¿ ∑ ∑ [ ( MCEi∗P¿−C ( P¿ ) )∗( 1−x ¿ ) ] HW −∑ ∑ PM ∗x ¿∗Pimax + ( FM C i∗x ¿∗Pimax
❑ i=1 t =1 i =1 t =1 8760

The objective function reflects the profit of GENCO, which is determined by subtracting
its total revenues from the corresponding costs, which comprise production costs, fixed
costs, and variable maintenance costs.

The ISO should arrange the units for maintenance by maximizing the system capacity
reserve in the whole year. The objective role of GENCO in maintenance planning might be
expressed as follows:

N N
Maximize min Rt =∑ Pimax− Lt−∑ Pimax x ¿ (2)
i=1 i=1

39
The objective function (2) reflects the reserve of the ISO which is determined by
subtracting sum of demand and maximum capacity of units under maintenance from the
total capacity of system when no unit is under maintenance.

Constraints of problem

1. Minimum net reserve: The minimum net reserve is the lowest of all the weekly
net reserves over the maintenance scheduling horizon. This constraint ensures a
net reserve above a specified threshold for all periods.
N

∑ Pimax ( 1−x ¿ )−Pd > R min ∀ t

i=1

2. Maintenance outage duration : This constraint ensure that for each unit that is
maintained the required number of time periods

∑ x ¿ =Di ∀ i
i=1
3. Continuous maintenance : This constraint ensures that the maintenance of any
unit must be completed once it begins

x i ( t )−x i ( t−1 ) ≤ x i ( t + Di−1 ) ∀ i, ∀ t

4. Maximum number of units simultaneously in maintenance: This constraint

limits the maximum number of units s can be maintained simultaneously by a
producer.

∑ x¿ ≤ S ∀ t
i=1
5. Capacity and Minimum Power Output: The power generated for each online
unit must be within a certain range represented by its minimum power output
Pimin and its capacity Pimax

40
 Pimin ≤ P ¿ ≤ Pimax

CHAPTER 6
Solution Algorithm for Generator
Maintenance Scheduling in Restructured
Power System

6.1 Case study

In this case study, the maintenance scheduling of a generating system, containing 32
generating units for the time interval of 52 weeks is considered. The generating units
encompass a single GENCO.

We use the electricity price of nor-pool deregulated power system. Table 6.1 gives the data
on weekly electricity prices and weekly peak loads in percentage of annual peak load and
Table 6.2 shows the list of generating unit data. The annual peak load for the test system is
2850 MW.

Week MCE Peak Week MCE Peak Week MCE Peak Week MCE Peak
load load load load

01 36.7 86.2 14 28.8 75 27 29.2 75.5 40 25.5 72.4

41
02 39.3 90 15 25.2 72.1 28 34.1 81.6 41 27.7 74.3

03 37.2 87.8 16 33.2 80 29 33.2 80.1 42 27.9 74.4

04 35.6 83.4 17 29.1 75.4 30 38.1 88 43 33.1 80

05 37.9 88 18 35.7 83.7 31 25.2 72.2 44 38.1 88.1

06 35.9 84.1 19 37 78 32 30.9 77.6 45 38.2 88.5

07 35.4 83.2 20 37.8 88 33 33.1 80 46 40.3 90.9

08 33.5 80.6 21 36.4 85.6 34 26 72.9 47 42.2 94

09 27.2 74 22 33.7 81.1 35 25.8 72.6 48 38.7 89

10 26.7 73.7 23 39.7 90 36 25 70.5 49 42.8 94.2

11 25.1 71.5 24 38.3 88.7 37 31.1 78 50 45 97

12 25.9 72.7 25 39 89.6 38 23.8 69.5 51 58 100

13 24.9 74 26 36.6 86.1 39 25.6 72.4 52 43.4 95.5

Table 6.1 Weekly electricity prices and weekly peak load in percent of annual peak

Unit Pmax Pmin Ai bi ci PMC FMC Maintenance

Duration

01 12 2.4 0.02533 25.547 24.3891 10 5 1

2

02 12 2.4 0.02649 25.675 24.4110 10 5 1

3

03 12 2.4 0.02801 25.802 24.6382 10 5 1

7

42
04 12 2.4 0.02842 25.931 24.7605 10 5 1
8

05 12 2.4 0.02855 26.061 24.8882 10 5 1

1

06 20 4 0.01561 37.963 118.9083 0.3 5 1

7

07 20 4 0.01359 37.777 118.4576 0.3 5 1

0

08 20 4 0.01161 37.963 118.9083 0.3 5 1

7

09 20 4 0.01059 38.777 119.4576 0.3 5 1

0

10 76 15.2 0.00962 13.507 81.8259 10 0.9 3

3

11 76 15.2 0.00876 13.327 81.1364 10 0.9 3

2

12 76 15,2 0.00895 13.353 81.2980 10 0.9 3

8

13 76 15.2 0.00932 13.407 81.6259 10 0.9 3

3

14 100 25 0.00623 18.000 217.8952 8.5 0.8 4

0

15 100 25 0.00599 18.600 219.7752 8.5 0.8 4

0
16 100 25 0.00612 18.100 218.3350 8.5 0.8 4
0

43
17 100 25 0.00588 18.280 216.7752 8.5 0.8 4
0

18 100 25 0.00598 18.200 218.7752 8.5 0.8 4

0

19 100 25 0.00578 17.280 216.7752 8.5 0.8 4

0

20 155 54.25 0.00481 10.736 142.7348 7 0.8 5

7

21 155 54.25 0.00473 10.715 143.0288 7 0.8 5

4

22 155 54.25 0.00481 10.736 143.3179 7 0.8 5

7

23 155 54.25 0.00487 10.758 143.5972 7 0.8 5

3

24 197 68.95 0.00259 23.000 259.1310 5 0.7 6

0

25 197 68.95 0.00260 23.100 259.6490 5 0.7 6

0

26 197 68.95 0.00263 23.200 260.1760 5 0.7 6

0

27 197 68.95 0.00264 23.400 260.5760 5 0.7 6

0

28 197 68.95 0.00267 23.500 261.1760 5 0.7 6

0
29 197 68.95 0.00261 23.040 260.0760 5 0.7 6
0

44
 30 350 140 0.00150 10.841 176.0575 4.5 0.7 8
6

31 400 100 0.00194 7.4921 310.0021 5 0.3 8

32 400 100 0.00195 7.5031 311.9102 5 0.3 8

Table 6.2 Generating unit data and operating and maintenance cost data

6.1.1 Solution by using single objective approach

The goal of the ISO's solution to the maintenance scheduling problem is to maximize the
reserve over the course of the year's weeks. According to its target function, each GENCO
specifies the ideal maintenance scheduling and gives it to ISO. The ISO contrasts its
reliability index with reliability indices associated with GENCOs' maintenance planning. It
will be allowed if they are sufficiently reliable; otherwise, the units that breach the limits
will be identified, and ISO will request a time rescheduling. And if the schedule isn't
altered or the ISO's reliability isn't fulfilled, impose a penalty in this case, on the GENCO.
Only one particular GENCO is taken into consideration in this research for the problem-
solving. Figure 2 depicts a flowchart of the suggested solution to the maintenance
scheduling problem. Benefit of GENCO reduces after ISO's objectives are achieved. Now,
GENCO must make a final decision as to whether it will accept the scheduling or stick to
the earlier scheduling and make up for its generation loss by purchasing from other
GENCOs. It goes without saying that the choice is based on the pace of benefit loss and the
state of GENCO.

The algorithm of proposed approach for maintenance scheduling is shown Figure 6.1

45
 Figure 6.1 Algorithm of single objective approach

Results:

Table 6.3 shows the optimum annual maintenance schedule for 32 generating units based
on the single objective approach i.e. considering the GENCO’s and ISO’s objectives
separately and finally determining the optimum maintenance schedule by considering the
penalty imposed on GENCO by ISO. The figure 6.2 shows the plots of reliability indices
of GENCO’s original schedule, ISO’s schedule and the GENCO’s final schedule.

The above proposed method uses the Genetic Algorithm to find the optimum schedule for
preventive maintenance concerning the forecasted market clearing expenses. GENCO’s
objective is to sell electricity as much as possible and the goal of ISO is to maximize the
reserve of the system at every time interval. The reserve capacities of all the schedules are
compared with each other in Figure 6.3 It is clear from the figure 6.3 the reliability of final
maintenance schedule is improved after imposing the penalty on GENCO. The figure 6.3

46
shows the performance of single objective approach over 100 generations based on
GENCO’s objective. Figure 6.4 shows the performance of single objective approach over
100 generations based on ISO’s objective.

Unit GENCO’s ISO’s Final Unit GENCO’s ISO’s Final

schedule schedule schedule schedule schedule schedule
01 39 41 28 17 24 02 44
02 34 30 11 18 07 34 48
03 06 04 05 19 22 37 35
04 32 13 23 20 45 11 22
05 14 27 50 21 11 35 36
06 15 02 37 22 36 22 11
07 47 35 50 23 08 04 09
08 46 05 30 24 02 04 41
09 44 27 20 25 01 07 43
10 01 27 35 26 01 17 38
11 04 41 19 27 46 47 18
12 26 17 49 28 17 10 31
13 11 21 14 29 47 19 32
14 17 07 49 30 28 09 02
15 45 15 06 31 39 39 16
16 24 28 10 32 16 31 14
Table 6.3 Starting week of each unit for maintenance using GA

47
Figure 6.2 Reliability index for GENCO’s proposal, ISO’s proposal and final maintenance
schedule

Figure 6.3 Performance of GENCO’s objective function in single objective approach

48
Figure 6.4 Performance of ISO’s objective function in single objective approach

6.1.2 Solution by using multi objective approach

The primary goal of the GENCO is to get the most benefits out of their investments. Every
GENCO will try to schedule the maintenance for the weeks when the market clearing
expenses is at its lowest value. On the other hand when the GENCO’s units are in
maintenance, the ISO aims to maximize the systems reserve at every time interval. These
two objectives are contradictory in nature. In multi-objective approach of solving the
maintenance scheduling problem, we aim consider both the objectives simultaneously and
solve the problem to get the desired optimum schedule. In this approach the scheduling
problem is solved by using the NSGA-II optimization technique. The NSGA-II has proven
to be the one of the efficient tools for solving the optimization problems with multiple
objective.

Results:

Table 6.4 shows the optimum annual maintenance schedule for 32 generating units based
on the multi objective approach i.e. considering the GENCO’s and ISO’s objectives

49
simultaneously. The Figure 6.3 shows the plots of reliability index of maintenance
schedule by using the multi objective approach.

The above proposed method uses the NSGA-II to find the optimum schedule for
preventive maintenance concerning the forecasted market clearing expenses. GENCO’s
objective is to sell electricity as much as possible and the goal of ISO is to maximize the
reserve of the system at every time interval. The figure 6.6 shows the performance of
single objective approach over 100 generations based on GENCO’s objective. Figure 6.7
shows the performance of single objective approach over 100 generations based on ISO’s
objective.

Unit GENCO’s schedule Unit GENCO’s schedule

01 03 17 09
02 43 18 44
03 24 19 18
04 43 20 08
05 24 21 47
06 39 22 18
07 23 23 37
08 33 24 17
09 27 25 26
10 05 26 34
11 44 27 31
12 27 28 40
13 34 29 11
14 40 30 31
15 42 31 01
16 33 32 09
Table 6.4 Starting week of each unit for maintenance using NSGA-II

50
 Figure 6.5 Reliability index for GENCO’s maintenance schedule

Figure 6.6 Performance of GENCO’s objective function in multi objective approach

51
Figure 6.7 Performance of ISO’s objective function in multi objective approach

6.2 Comparison between single objective approach and multi objective approach

S.No Single objective approach Multi objective approach

1 Single objective approach takes Multi objective approach takes

more time to solve the relatively less time for solving the
scheduling problem scheduling problem

2 This approach deals with the This approach deals with the
GENCO’s and ISO’s objective GENCO’s and ISO’s objective
separately simultaneously

3 Penalty factor is needed to be No penalty factor is incorporated in

incorporated in order to preserve this approach
the reliability or reserve of the
system

4 Single objective approach Multi objective approach uses

proposes only one solution for pareto dominance relationship

52
 maintenance scheduling where the partner of a non
problem dominated individual is chosen
among the individuals of the
population that are dominated

Table 6.5 Comparison between single objective approach and multi objective approach

The comparison between the plots of number of iterations vs time taken of the two
approaches is shown in figure 6.8. It is clear from the figure that single objective approach
takes more time than the multi objective approach for same number of generations.
Moreover it can also be seen from the convergence graphs in figure 6.3, figure 6.4, figure
6.6 and figure 6.7 the multi objective approach converges quickly than the single objective
approach. Hence the multi objective approach will take less time to solve the maintenance
problem.

Figure 6.8 : Number of iterations versus time graph for both approaches

53
6.3 Conclusion:

In a deregulated electricity system, the annual maintenance schedule for a GENCO's

producing units is developed and considered from several angles. To determine the ideal
plan for preventive maintenance in relation to the anticipated market clearing price, the
genetic algorithm and NSGA-II are used. This work establishes a suitable framework for
coordination that enables the achievement of a generation maintenance plan that satisfies
producer maximum-profit criteria while attaining a sufficient level of reliability in each
week of the year. In order to find a solution that is agreeable to all parties, the coordination
mechanism takes into account the perspectives of both producers and the ISO. In order to
examine the effectiveness of the two approaches, the solutions produced by the Genetic
algorithm and NSGA-II are compared in this project. It has been discovered that NSGA-II

is more effective than the Genetic algorithm.

54
Bibliography
[1] A.J. Conejo, R,Garcia-Bertrand and M.Diaz-salazar, Generation maintenance
scheduling in restructured power systems, IEEE Transactions on Power System,
20(2):984-992.

[2] Aijaz Ahmad, D.P. Kothari (2000) , A Practical model for Generator Maintenance
scheduling with transmission Constraints, Electric machines and Power Systems ,
28:6,501-513.

[3] R.Eshraghnia, M.H.M Shanechi and H.R. Mashhadi, Generation Maintenance

Scheduling in Power Market Based on Genetic Algorithm, IEEE PES Power systems
Conference and Exposition, (2006) :1814-1819.

[4] Dahal KP, McDonald JR, Burt GM. Modern heuristic techniques for scheduling
generator maintenance in power systems . Transactions of the Institute of Measurement
and Control. 2000;22(2):179-194.

[5] A Khalid, Karamitsos Loannis , A survey of generator maintenance scheduling

techniques, Global Journal of Researches in Engineering 12(1), 10-17, 2012.

[6] Z. A. Yamayee, Maintenance Scheduling: Description, Literature Survey, and

Interface with Overall Operations Scheduling, IEEE Transactions on Power Apparatus and
Systems, 101(8): 2770-2779.

[7] A.Abdulwhab, R.Billinton, A.A. Eladamaty and S.O. Faried (2004), Maintenance
Scheduling Optimization Using a Genetic Algorithm (GA) with a Probalistic Fitness
Function, Electric Power Components and Systems, 32:12, 1239-1254.

[8] G. Poryazoglu and HyungSeon Oh, Scheduling maintenance for reliable transmission
systems, 2016 IEEE/PES Transmission and Distribution Conference and Exposition
(T&D): 1-5.

55
[9] Chin Aik Koay and D. Srinivasan, Particle swarm optimization-based approach for
generator maintenance scheduling, Proceedings of the 2003 IEEE Swarm Intelligence
Symposium. SIS’03 :167-173.

[10] M.R. AlRashidi and M.E. El-Hawary, A Survey of Particle Swarm Optimization
Application in Electric Power Systems, IEEE Transactions on Evolutionary Computation,
13(4): 913-918.

[11] Suraj Kumhar and Mantosh Kumar, Generator maintenance scheduling of power
system using hybrid technique, International Research Journal of Engineering and
Technology (3): 418-423.

[12] Marwali. M. and Shahidehpour. M., Short-term transmission line maintenance

scheduling in a deregulated system, PICA’99, pp. 31-37, California, 1999.

[13] W.R. Christiaanse and A.H. Palmer, A technique for the automated scheduling of the
Maintenance of Generating facilities, IEEE transactions on Power Apparatus and Systems,
91(1): 137-144.

[14] H.H. Zurn and V.H. Quintana, Generator maintenance scheduling via successive
approximations dynamic programming, IEEE transactions on Power Apparatus and
Systems,94(2): 665-671.

[15] Holland, J.H. Genetic Algorithms. Scientific American, 267,66-72.

[16] Goldberg, D.E., Holland. J.H. Genetic Algorithms and Machine Learning 3 ,95-99.

[17] http://www.nordpool.com

[18] M. Shahidehpour and M. Marwali, Maintenance scheduling in restructured power

system, USA-2000.

[19] K. Deb, A. Pratap, S. Agarwal and T.Meyarivan, A fast and elitist multi objective
genetic algorithm: NSGA-Ⅱ, IEEE transactions on Evolutionary Computation, 6(2):182-
197.

56
[20] M. Negnevitsky and G. Kelareva, Applications of genetic algorithms for maintenance
scheduling in power system, IEEE transactions on power systems 9(3): 447-452.

[21] Burke, E.K., Smith, A.J., A memetic algorithm to schedule planned maintenance for
the national grid, Journal of Experimental Algorithms (JEA) 4,1-es

57