prr.hec.gov.pkprr.hec.gov.pk/jspui/bitstream/123456789/1800/1/910S.pdfv ABSTRACT Unit Commitment is...
Transcript of prr.hec.gov.pkprr.hec.gov.pk/jspui/bitstream/123456789/1800/1/910S.pdfv ABSTRACT Unit Commitment is...
UNIT COMMITMENT USING HYBRID
APPROACHES
BY
Engr. Aftab Ahmad 2K1-UET/PhD-EE-04
Supervisor
Prof. Dr. Azzam ul Asar
DEPARTMENT OF ELECTRICAL ENGINEERING UNIVERSITY OF ENGINEERING AND TECHNOLOGY
TAXILA (PAKISTAN)
JUNE 2010
DECLARATION
The substance of this thesis is original work of the author and due reference and
acknowledgement has been made, where necessary, to the work of others. No part of the thesis
has already accepted for any degree, and it is not being currently submitted in candidature of any
degree.
Engr. Aftab Ahmad 01-UET/Ph D-EE-04
Thesis Scholar
DEDICATED TO
My father, mother
Wife, Children
and
Dr. Aftab Ahmad
v
ABSTRACT
Unit Commitment is an important and vital optimization task in a power control centre.
After load forecasting it is the second step in the planning process. It consists of two linked
optimization problems. It comprises unit on/off scheduling problem and the economic dispatch
sub-problem. The on/off scheduling problem is a 0-1 combinatorial problem with equality and
inequality constraints, while the economic dispatch sub-problem is a nonlinear constrained
optimization problem.
Unit commitment is a nonlinear, large scale, combinatorial, constrained optimization
problem. The complete unit commitment optimization problem is to minimize the total
production cost (TPC) of utility in such a way that the constraints such as load demand, spinning
reserve, minimum and maximum power limits of units, minimum up (MUT) and minimum down
times (MDT) are satisfied. Therefore, based on the forecasted load demand, preparing proper
on/off schedule of generators can result in cost saving for utility. It is much more difficult
problem to solve due to its high dimensionality.
The present work is based on the scheduling of thermal units. The generation of initial
feasible UC schedules is much important, for the UCP. When initial feasible schedules
(generation > load + spinning reserve) are generated randomly, it is difficult to get feasible
schedules for the whole daily forecasted load curve.
In this work initial schedule is generated by considering the peak, off peak load of the
forecasted load curve, must run and must out units based on a new priority list method. The
proposed method is very efficient and fast in generating initial unit commitment schedules. In
this work the MUT and MDT constraints are checked and repaired by using bit change operator.
The trial solutions were generated by taking upper four units in the priority list at each hour to
avoid entrapment in local minimum.
vi
In the unit commitment problem, the economic load dispatch (ELD) sub-problem is an
intensive part and its calculations consume a large amount of time. Convex economic dispatch
using load to efficient unit and incremental cost criterion methods have been solved. In this
work, the ELD calculation for non convex problem has been solved using genetic algorithm
(GA) based on real power search method.
In the present work, three hybrid approaches have been developed for the convex and
non-convex cost functions and applied first time to solve the unit commitment problem. To
implement these algorithms a flexible and extensible computational framework has been
developed to run in visual C++ environment. The proposed algorithms are (i) “hybrid of dynamic
programming, particle swarm optimization and artificial neural network algorithms (DP-PSO-
ANN) for convex cost function, (ii) “Neuro-Genetic hybrid approach for non-convex cost
function” and (iii) “hybrid of full load average production cost and maximum power output, for
convex and non convex cost functions”. For comparison the neural network trained with back
propagation learning rule has also been developed. The proposed models have been tested on
IEEE 3 and 10 units standard test systems. The significant improvement in the total production
cost shows the promise of these hybrid models.
National utility system, National Transmission and despatch Company (NTDC) has been
reviewed with reference to its operation problems. Four test systems consisting of 12, 15, 25 and
34 units of NTDC system have been tested.
vii
ACKNOWLEDGEMENTS
My greatest thanks to the ALMIGHTY ALLAH The Most Gracious, the Most
Bountiful, the Omni potent and the Omni present, the MASTER OF THE WORLD, who gave
me the strength and spirit to fulfill the requirements of this thesis. All the respect to our HOLY
PROPHET (May Peace Be Upon Him), who after a lot of hardships and difficulties, made me
able to recognize our ALLAH and to distinguish virtue and evil.
I wish to place on record my deep sense of admiration to Dr. Aftab Ahmad (Late),
Director, NTDC, PAKISTAN for his advice, encouragement and help in my work. May
Almighty ALLAH rest the departed soul in eternal peace!
I would like to express my profound gratitude, most sincere appreciation and special
thanks to my supervisor Prof. Dr. Azzam ul Asar, for his moral support, valuable suggestions,
guidance and encouragement that made me able to complete this research work.
I would like to thank Dr. Habibullah Jamal, Professor, Department of electrical
engineering, Vice Chancellor (Jan. 2001- April 2009) University of Engineering and
Technology, Taxila, PAKISTAN for providing the financial assistance for conducting this
research. Thanks to Prof. Dr. Ahmad Khalil Khan, Dean Faculty of Electrical and Electronics
Engineering, Prof. Dr. Mumtaz Ahmad Kamal, Dean Faculty of Civil and Environmental
Engineering, Prof. Dr. Abdur Razzaq Ghumman, Prof. Dr. Shahab Kushnood, Dean Faculty of
Mechanical and Aeronautical Engineering and Prof. Dr. Adeel Akram, Dean Faculty of
Telecommunication and Information Engineering for encouragement, cooperation and support.
Special thanks to the members of my PhD research monitoring committee for guidance and
suggestions regarding the research work.
Thanks are also to my friends and colleagues especially to Prof. Dr. Tahir Nadeem Malik,
Chairman Department of Electrical Engineering for his helpful discussions and valuable
suggestions throughout this research work.
viii
The assistance in the work by Muhammad Ibrahim, Mian Muhammad Usman, Muhammad
Suleman, Adeel Mukhtar, and Syed Azhar Ali Zaidi.
Finally, I would like to give my special thanks to my wife and children for their patience
and support.
ix
CONTENTS
ABSTRACT
ACKNOWLEDGEMENT
LIST OF FIGURES
LIST OF TABLES
CHAPTER 1
Introduction 1.1 General........................................................................................................................ 1
1.2 Problem statement....................................................................................................... 2
1.3 Objectives ................................................................................................................... 3
1.4 Scope of the work ....................................................................................................... 4
1.5 Thesis Organization .................................................................................................... 5
CHAPTER 2
Unit Commitment Problem --- A Brief Literature Survey 2.1 Introduction................................................................................................................. 7
2.2 Power System Operational Planning .......................................................................... 7
2.3 Unit commitment --- Literature Survey ...................................................................... 9
2.4 Single Classical/Deterministic Approaches................................................................ 9
2.4.1. Priority List ................................................................................................................. 9
2.4.2. Dynamic Programming............................................................................................. 10
2.4.3 Branch and Bound..................................................................................................... 11
2.4.4 Integer and Mixed Integer programming.................................................................. 11
2.4.5 Lagrange Relaxation Method.................................................................................... 12
2.4.6 Straight Forward Method.......................................................................................... 14
2.4.7 Secant Method .......................................................................................................... 15
2.5 Non classical approaches .......................................................................................... 15
2.5.1 Tabu search ............................................................................................................... 15
x
2.5.2 Simulated Annealing (SA)........................................................................................ 15
2.5.3 Expert System ........................................................................................................... 16
2.5.4 Artificial Neural Network (ANN)............................................................................. 17
2.5.5 Evolutionary Programming (EP) .............................................................................. 18
2.5.6 Genetic Algorithm (GA) ........................................................................................... 18
2.5.7 Fuzzy logic................................................................................................................ 20
2.5.8 Particle Swarm Optimization (PSO)......................................................................... 20
2.5.9 Ant Colony Optimization (ACO).............................................................................. 21
2.5.10 Greedy Randomized Adaptive Search Procedure (GRASP) .................................... 21
2.6 Hybrid approaches .................................................................................................... 21
2.7 Unit Commitment --- Issues and Bottlenecks........................................................... 26
2.8 Discussion................................................................................................................. 26
CHAPTER 3
Optimization Tools for Unit Commitment
3.1 Introduction............................................................................................................... 28
3.2 Particle swarm optimization (P.S.O) ........................................................................ 28
3.3 Artificial Neural Networks (ANN) ........................................................................... 30
3.3.1 Feedforward Neural Network ................................................................................... 31
3.4 Dynamic Programming (DP) or Recursive Optimization......................................... 32
3.4.1 Forward Dynamic Programming Approach.............................................................. 34
3.4.2 Mathematical Formulation of the Dynamic Programming for unit commitment
problem ..................................................................................................................... 34
3.5 Genetic Algorithm (GA) ........................................................................................... 36
CHAPTER 4
Unit Commitment --- Problem Formulation and Single Solution Approaches
4.1 Introduction............................................................................................................... 39
4.2 Characteristics of Power Generation Units............................................................... 39
xi
4.2.1 Unit’s Input-Output characteristic (Heat or Cost) .................................................... 40
4.2.2 Non convex fuel cost characteristic due to valve point effect .................................. 41
4.2.3 Incremental heat or cost characteristic...................................................................... 41
4.2.4 Unit Heat rate (HR) characteristic ............................................................................ 42
4.3 Unit Commitment Problem (UCP) ........................................................................... 43
4.3.1 Objective Function.................................................................................................... 44
4.3.1.1 Fuel Cost ................................................................................................................... 44
4.3.1.2 Start up cost............................................................................................................... 44
4.3.1.3 Shut down cost.......................................................................................................... 45
4.4 Constraints ................................................................................................................ 45
4.4.1 System constraints or coupling constraints............................................................... 45
4.4.2 Unit constraints or local constraints.......................................................................... 46
4.5 Unit Commitment mathematical formulation as an optimization problem .............. 47
4.6 Generation of initial feasible unit commitment schedules........................................ 50
4.6.1 Initial unit commitment scheduling by using priority List method and focusing on
peak and off-peak loads of the daily load curve. ...................................................... 50
4.6.2 Generation of trial solutions / neighbors................................................................... 51
4.7 Minimum up and minimum down Time Constraint Handling ................................. 51
4.8 Minimum up and down time constraint repairing by using bit change operator ...... 52
4.9 Algorithm for the construction of initial unit commitment schedule and M.U.T and
M.D.T constraint handling........................................................................................ 52
4.10 Unit commitment schedule and determination of number of units to be operated... 53
4.11 Economic dispatch Problem (Allocation of Generation).......................................... 54
4.11.1 Economic load dispatch (ELD) calculations............................................................. 54
4.11.1.1 Equal Incremental Cost Criterion ............................................................................. 54
4.11.1.2 Loading to most efficient load .................................................................................. 57
4.11.1.3 Using Genetic Algorithm (GA) (Real Power – Search) .......................................... 58
4.12 Economic Dispatch versus Unit Commitment......................................................... 59
4.13 Conventional/Classical Single Approaches for convex fuel cost function............... 60
4.13.1 Single Approach – I --- Complete Enumeration ...................................................... 61
4.13.2 Single Approach – II --- Conventional Priority List ................................................ 61
xii
4.13.3 Single Approach – III --- Proposed Single Approach.............................................. 62
4.14 Case studies --- Convex cost function ..................................................................... 62
4.14.1 Numerical Results of test system – I........................................................................ 62
4.14.2 Numerical results for test system-III --- 10 unit system .......................................... 70
CHAPTER 5
Proposed New Hybrid Models for Unit Commitment problem based on
convex and Non-Convex Cost Functions
5.1 Introduction............................................................................................................... 80
5.2 Hybrid Model – I: A hybrid of particle swarm optimization (PSO), artificial neural
network (ANN) and dynamic programming (DP). ................................................... 81
5.2.1 Generation of test and training data .......................................................................... 83
5.2.2 Artificial Neural Network using SI learning Rule .................................................... 83
5.2.3 Input and Output of the PSO-ANN Model ............................................................... 83
5.2.4 Scaling of the input and output data ......................................................................... 84
5.2.5 Training process........................................................................................................ 84
5.2.6 Parameters Settings................................................................................................... 85
5.3 Case Studies---Convex cost function........................................................................ 85
5.3.1 Numerical results of test system-I: Three unit system: Hybrid Model-I .................. 85
5.3.2 Numerical Results of Test system-II: Three unit system: Hybrid Model-I .............. 95
5.3.3 Numerical Results of Test system-III: Ten unit system: Hybrid Model-I:............... 97
5.4 Hybrid Model-II: Neuro-Genetic Hybrid Approach............................................... 107
5.5 Case Studies --- non convex cost function.............................................................. 109
5.5.1 Numerical Results of Test Systems–IV and V: 3 units systems: Hybrid Model-II 109
5.6 Hybrid Model –III: Scaleable deterministic hybrid approach. ............................... 115
5.7 Case Studies ---Convex fuel cost function.............................................................. 117
5.7.1 Numerical results of test system –I: three unit system: Hybrid Model-III ............. 117
5.7.2 Numerical Results of Test System-II: Three units system: Hybrid Model-III:
(Convex Fuel Cost Curve) ...................................................................................... 123
5.7.3 Numerical Results of Test System-III: Ten units system: Hybrid Model-III: (Convex
fuel cost function) ................................................................................................... 128
xiii
5.8 Case Studies: Hybrid model-III ---Non-Convex fuel cost function........................ 139
5.8.1 Numerical results of test systems – IV and V: 3 units systems: Hybrid Model- III:
(Non-convex fuel cost Curve)................................................................................. 140
CHAPTER 6
Unit Commitment of National Transmission & Despatch Company Limited
(NTDC)
6.1 Introduction............................................................................................................. 143
6.2 WAPDA --- Brief Overview................................................................................... 143
6.3 National Transmission and Despatch Company ..................................................... 144
6.4 Operational Constraints in NTDC System.............................................................. 144
6.4.1 Hydro-Electric Generation Constraints................................................................... 144
6.4.2 Thermal Generation Constraints ............................................................................. 146
6.4.3 Transmission line Constraints................................................................................. 146
6.4.4 Seasonal Variations in Power Demand................................................................... 146
6.4.5 Spinning Reserve Constraint……………………...……………………………….146
6.4.6 Minimum up and down time constraints………………………………………….146
6.4.7 Start up cost consideration…………….…………..………………………………146
6.4.8 Maintenance cost…………………………………………………………………147
6.4.9 Fuel constraint……………………………………………………………………147
6.4.10 Ramping rates…………………………………………………………………….147
6.4.11 Unit deration………………………………………...……………………………147
6.5 Test systems for NTDC system………….......……...……………………………147
6.6 Case Studies ............................................................................................................ 147
6.7 Numerical results .................................................................................................... 148
CHAPTER 7
Conclusions and Suggestions.................................................................................................. 152
References............................................................................................................................... 156
xiv
APPENDIX A
A.1 Three unit standard test systems ............................................................................. 167
A.1.1 Test System-I --- Three Unit Test System --- Convex Fuel Cost Curve ................ 167
A.1.2 Test System-II --- Three Unit Test System --- Convex Fuel Cost Curve ............... 168
A.2 Ten Unit Standard Test system............................................................................... 169
A.2.1 Test System-III --- Convex Fuel Cost Curve.......................................................... 169
A.3 Three Unit Standard Test Systems --- Non-Convex Fuel Cost Curve………….…170
A.3.1 Test Sysem-IV --- Non- Convex Fuel Cost Curve………………………………..170
A.3.2 Test System-V --- Non-Convex Fuel Cost Curve................................................... 170
A.4 Pakistani Utility NTDC Systems ............................................................................ 171
A.4.1 12 Unit NTDC System............................................................................................ 171
A.4.2 15 Unit NTDC System............................................................................................ 171
A.4.3 25 Unit NTDC System............................................................................................ 171
A.4.4 34 Unit NTDC System............................................................................................ 171
APPENDIX B List of Abbreviations .............................................................................................................. 172
APPENDIX C
Notation …………………………………………………………………………………….174
APPENDIX D
Derived Publications............................................................................................................... 175
xv
List of Figures
Figure 3.1 Multilayer feedforward Neural Network Architecture......................................... 31
Figure 3.2 Supervised learning .............................................................................................. 32
Figure 4.1 Boiler-Turbine-Generators Unit ........................................................................... 40
Figure 4.2 Input-output Curve of a Steam Turbine Generator................................................ 41
Figure 4.3 Input-output curve of a multi valve steam turbine generator with four steam
admission valves ................................................................................................... 41
Figure 4.4 Incremental Heat Rate or Cost characteristic ........................................................ 42
Figure 4.5 Incremental Heat Rate Characteristics of a steam turbine with four valves. ........ 42
Figure 4.6 Heat rate and incremental heat rate curves for convex cost function ................... 43
Figure 4.7 Unit Commitment................................................................................................. 44
Figure 4.8 Representation of Unit Commitment Problem (UCP) .......................................... 49
Figure 4.9 Generation of initial solution by priority list method a graphical representation . 51
Figure 4.10 Generation of new schedule by taking upper 4 units ........................................... 51
Figure 4.11 Repairing of minimum up time ............................................................................ 52
Figure 4.12 Repairing of minimum down time ....................................................................... 52
Figure 4.13 Forecasted load curve ........................................................................................... 53
Figure 4.14 Representation of initial unit commitment schedule ............................................ 53
Figure 4.15 Pseudo Code for Genetic Algorithm Real Power-Search Method........................ 59
Figure 5.1 Flowchart for DP-PSO-ANN Hybrid Model-I…………………………………..82
Figure 5.2 Load patterns for training ..................................................................................... 83
Figure 5.3 Steps for SI learning ANN.................................................................................... 84
Figure 5.4 MSE graph for a load of 500 MW........................................................................ 94
Figure 5.5 MSE graph for a load of 800 MW........................................................................ 94
Figure 5.6 MSE graph for a load of 850 MW........................................................................ 94
Figure 5.7 Absolute Percentage Error graph for Operating Fuel Cost .................................. 94
Figure 5.8 Absolute Percentage Error (APE) for a load of 900MW (DP-BP-ANN Model):
Ten unit system: Test system-III......................................................................... 106
Figure 5.9 Flowchart for GA-PSO-ANN, Neuro-Genetic Hybrid Model-II........................ 108
Figure 5.10 Flow chart for Hybrid Model–III (PMAX-FLAPC) ........................................... 116
xvi
List of Tables
Table 4.1 Priority order based on single approach-II and III: test system-I ............................. 62
Table 4.2 Comparison of the operating fuel cost ($) for 3 Unit Systems :( Test System I) ..... 63
Table 4.3 Summary of Unit Commitment Schedules for 3 Unit Systems (Test System I)…...63
Table 4.4 Number of Units in Operation for 3 Unit Systems (Test System I).......................... 65
Table 4.5 Comparison of the Three Single approaches with Genetic Algorithm and Hopfield
Neural Network Methods for 3 unit systems: Test system-I .................................... 66
Table 4.6 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from Single Approach-I (Enumeration): Test system-I ............................................ 67
Table 4.7 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from
the Single Approach-II (Conventional Priority List): Test system-I ........................ 68
Table 4.8 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from
the Proposed Single approach-III (PMAX): Test system-I....................................... 69
Table 4.9 Priority order based on single approach-II and proposed single approach III:......... 70
Test system-III --- ten unit system............................................................................ 70
Table 4.10 Comparison of the operating fuel cost ($) for 10 unit systems considering 10%
spinning reserve and minimum up/down time constraints without transition cost:
Test System III......................................................................................................... 71
Table 4.11 Comparison of the operating fuel cost ($) for 10 unit systems considering 10%
spinning reserve and minimum up/down time constraints with transition cost Test
System III……………………………….…………………………………….……72
Table 4.12 Comparison of the results of the proposed single approach-III with Genetic
Algorithm and conventional priority list.................................................................. 72
Table 4.13 Summary of Unit Commitment schedules for 10 unit systems: .............................. 73
Single approach-I, II and III: Test System III.......................................................... 73
Table 4.14 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from Single approach-I (Enumeration) considering MUT and MDT constraints with
10 % spinning reserve: Test system-III ................................................................... 74
xvii
Table 4.15 Unit Commitment Schedule, fuel cost, start up cost and total production cost of the
best solution obtained from Single approach-I (Enumeration) considering MUT and
MDT constraints with 10 % spinning reserve: Test system-III ............................... 75
Table 4.16 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from Single approach-II (FLAPC) considering MUT and MDT constraints with 10
% spinning reserve: Test system-III ........................................................................ 76
Table 4.17 Unit Commitment Schedule, fuel cost, startup cost and total production cost of the
best solution obtained from Single approach-II (FLAPC) considering MUT and
MDT constraints with 10 % spinning reserve: Test system-III ............................... 77
Table 4.18 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from Proposed Single approach-III (PMAX) considering MUT and MDT
constraints with 10 % spinning reserve: Test system-III ......................................... 78
Table 4.19 Unit Commitment Schedule fuel cost, startup cost and total production cost of the
best solution obtained from Single approach-III (PMAX) considering MUT and
MDT constraints with 10 % spinning reserve: Test system-III ............................... 79
Table 5.1 Comparison of proposed Hybrid Model-I with Genetic Algorithm, Conventional
Priority List and Hopfield Neural Network methods for 3 unit systems: Test
System-I: .................................................................................................................. 88
Table 5.2 Best Results by the proposed Hybrid Model-I (DP-PSO-ANN) among the sixty
samples: Test system-I............................................................................................. 89
Table 5.3 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from the proposed Hybrid Model-I (DP-PSO-ANN): Test system-I........................ 90
Table 5.4 Best output results obtained by the Hybrid Model-I (DP-BP-ANN) amongst sixty
samples: Test system-I.............................................................................................. 91
Table 5.5 Comparison of the outputs (MW) obtained by the proposed SI – ANN learning and
BP - ANN learning for 3 unit system: Test System-I ............................................... 92
Table 5.6 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
by using Hybrid Model-I (DP-BP-ANN): Test system-I.......................................... 93
Table 5.7 Comparison of the best Operating fuel cost ($) obtained amongst the sixty samples
for 3 unit system by using Proposed Hybrid Model-I: Test system-II ..................... 95
xviii
Table 5.8 Comparison of the outputs (MW) obtained by the proposed SI –ANN learning
(Hybrid Model-I, DP-PSO-ANN) and BP-ANN learning (Hybrid Model-I (DP-BP-
ANN) for 3 unit systems: Test System-II ................................................................. 96
Table 5.9 Targets for the 10 unit system: Test system-III ........................................................ 99
Table 5.10 Best results obtained by the proposed Hybrid Model-I (DP-PSO-ANN) amongst the
sixty samples: Test system-III ............................................................................... 100
Table 5.11 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from proposed Hybrid Model-I (DP-PSO-ANN): Test system-III........................ 101
Table 5.12 Load demand and Total Production Cost (TPC) obtained by proposed Hybrid Model-
I (DP-PSO-ANN): Test system-III: ....................................................................... 102
Table 5.13 Comparison of the best results of the Hybrid Model-I (DP- PSO-ANN) with other
approaches available in the literature: Test system-III .......................................... 103
Table 5.14 Comparison between proposed Hybrid model-I and other approaches for daily
saving and Percentage saving in fuel cost: ten unit system ................................... 104
Table 5.15 Comparison of the results of the proposed Hybrid Model-I with Genetic Algorithm,
dynamic programming, simulated annealing and Lagrange relaxation method:... 105
Table 5.16 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
by using (DP-BP-ANN), considering MUT and MDT constraints with 10 %
spinning reserve: Test system-III........................................................................... 105
Table 5.17 Load demand, operating fuel cost, startup cost and Total Production Cost obtained
by Using (DP-BP-ANN), considering M.U.T and M.D.T Constraints with 10 %
spinning reserve: Test system-III:.......................................................................... 106
Table 5.18 Comparison of the best Operating fuel cost ($) obtained amongst the sixty samples
for 3 unit system by using Proposed Hybrid Model-II: (Non-convex).................. 110
Table 5.19 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from the Hybrid Model-II: (GA-PSO-ANN): Test system-IV .............................. 111
Table 5.20 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from
the Hybrid Model-II (GA-BP-ANN): Test system-IV........................................... 112
Table 5.21 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from
the Hybrid Model-II: (GA-PSO-ANN): Test system-V ........................................ 113
xix
Table 5.22 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from
the Hybrid Model-II: (GA-BP-ANN) Test system-V............................................ 114
Table 5.23 Proposed Priority order based on Hybrid Model-III (PMAX-FLAPC)................. 115
Table 5.24 Comparison of proposed Hybrid Model-III with Genetic Algorithm, Conventional
Priority List and Hopfield Neural Network methods for 3 unit systems:Test
System-I: ................................................................................................................ 119
Table 5.25 Comparison of the Summary of unit commitment schedules of the proposed Hybrid
Models I and III for 3 Unit Systems: Test System -I ............................................. 120
Table 5.26 Comparison of Number of Units in Operation for 3 unit systems for hybrid models-I
and III with three single approaches: Test System I .............................................. 121
Table 5.27 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from the Proposed Hybrid Model-III: (PMAX-FLAPC2, ED based on average load
assigned method): Test system-I............................................................................ 122
Table 5.28 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from the proposed Hybrid Model-III (PMAX-FLAPC3) (ED based on lambda
Iteration method): Test system-I............................................................................ 123
Table 5.29 Comparison of the operating fuel cost ($) for proposed Hybrid Models -I and III for
3 unit system: Test System –II (Convex Fuel Cost Curve) .................................. 124
Table 5.30 Summary of Unit Commitment Schedules for 3 unit systems: Test System II: .... 125
Table 5.31 Summary of number of Units in Operation for 3 unit system: Test system II: ...... 126
Table 5.32 Unit commitment schedule and Power Sharing (MW) of the best solution obtained
by Hybrid Model-III (PMAX-FLAPC with ED based on average load) for 3 unit
systems: Test System-II ......................................................................................... 127
Table 5.33 Unit commitment schedule and Power Sharing (MW) of the best solution obtained
by Hybrid Model-III (PMAX-FLAPC with ED) for 3 unit systems Test System-
II…......................................................................................................................... 128
Table 5.34 Comparison of the best results of the Hybrid Model-III with other approaches
available in the literature: Test system-III ............................................................. 131
Table 5.35 Daily Saving and Percentage saving in fuel cost compared with other approaches132
xx
Table 5.36 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from proposed Hybrid Model-III, with MUT and MDT constraints with 10 %
spinning reserve: (ED by average load to economic unit): Test system-III: ......... 133
Table 5.37 Load demand , fuel cost , start up cost and Total Production Cost (TPC) obtained
from proposed Hybrid Model-III, with MUT and MDT constraints with 10 %
spinning reserve. (ED by average load to economic unit): Test system-III........... 134
Table 5.38 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained
from proposed Hybrid Model-III, with MUT and MDT constraints with 10 %
spinning reserve. (ED by lambda iteration method): Test system-III.................... 135
Table 5.39 Load demand , fuel cost , start up cost and Total Production Cost (TPC) of the best
solution obtained from proposed Hybrid Model-III, with MUT and MDT
constraints with 10 % spinning reserve: Test system-III (ED by Lambda iteration
method) .................................................................................................................. 136
Table 5.40 Summary of Unit Commitment schedules for 10 unit systems with 10% spinning
reserve and considering minimum up time and down time constraints:Test System
III............................................................................................................................ 137
Table 5.41 Comparison of Transition Cost for 10 unit system without considering s. r and
MUT and MDT: Test System III (10 unit system) ................................................ 138
Table 5.42 Start up cost comparison………………………………………………………….138
Table 5.43 Comparison of operating fuel cost ($) for 10 unit systems considering 10% spinning
reserve and considering minimum up/down time constraints without transition cost:
Test System III....................................................................................................... 138
Table 5.44 Comparison of the operating fuel cost ($) for 10 unit systems considering 10%
spinning reserve, considering minimum up/down time constraints and transition
cost: Test System III .............................................................................................. 139
Table 5.45 Comparison of the Operating fuel cost ($) for 3 unit system obtained by using
Proposed Hybrid Model-III compared with Hybrid model -II: (Non-convex)... 140
Table 5.46 Unit Commitment Schedule and Power Sharing (MW) of the best solution
obtained from the Hybrid Model-III (PMAX-FLAPC3): (non-convex):Test
system-IV:........................................................................................................... 141
xxi
Table 5.47 Unit Commitment Schedule and Power Sharing (MW) of the best solution
obtained from the Hybrid Model-III (non-convex UC123, ED 321): Test system-
V.......................................................................................................................... 142
Table 6.1 Comparison of the results for NTDC 12, 15, 25 and 34 unit systems .................... 148
Table 6.2 Summary of UC schedule and operating fuel cost for 12 unit NTDC systems ...... 148
Table 6.3 Summary of UC schedule and operating fuel cost for 15 unit NTDC systems ...... 149
Table 6.4 Summary of UC schedule and operating fuel cost for 25 unit NTDC systems ...... 150
Table 6.5 Summary of UC schedule and operating fuel cost for 34 unit NTDC systems ...... 151
1
CHAPTER 1
Introduction
1.1 General
Optimization is the process of maximizing or minimizing a desired objective function
while satisfying the equality and inequality constraints. The majority of the engineering
problems involve constrained minimization. Optimization has vital and dominant role in the field
of engineering.
The Unit Commitment (UC) is an important research challenge and very significant
optimization task in the daily operational planning of modern power systems, because the
improvement of unit commitment schedules results in the reduction of total production cost
(economic benefits). It is a non-linear, high dimensional, mixed-integer combinatorial
optimization problem with both binary/integer (unit status variable) and continuous (unit output
power) variables. The number of combinations grows exponentially for a large scale system.
Therefore UC is one of the most difficult optimization problems in power systems to solve.
The Unit Commitment Problem (UCP) can be considered as two linked optimization
problems: the Unit Scheduling Problem (Allocation of Generators) and the Economic dispatch
problem (Allocation of Generation). The unit schedule problem is the on/off or 0/1 combinatorial
optimization problem. A feasible unit schedule must satisfy the forecasted load demand, system
spinning reserve requirements, and the constraints on the start up and shut down times during
each planning period. The economic dispatch sub problem is the constrained non-linear
optimization problem. This thesis investigates the application of hybrid models based on
Artificial Intelligence (AI) and conventional techniques.
2
1.2 Problem statement
The unit commitment problem (UCP) in power system is defined as determining the
start-up and shut-down schedules of units to meet the forecast load demand and spinning reserve
over a scheduling period so that the total production cost is minimized while satisfying various
system and unit constraints.
The main issues in the UCP are complexity (high dimensions) of search space, generation
of initial feasible schedules, generation of trial solutions, minimum up and down time and
spinning reserve constraint handling, calculations of non linear economic dispatch sub problem,
handling of non convexity in economic dispatch sub problem due to valve point effects. A major
source of infeasibility is the generation of infeasible schedules, which have to be discarded
immediately. Another major source of infeasibility in feasible solutions is the violation of
nonlinear minimum up/down-time constraints, which has to be checked and repaired. By
considering the minimum up and down time constraints the operating fuel cost rises and an
alternate solution needs to found. The unit commitment schedule is based on forecasted load and
spinning reserve requirements. The spinning reserve constraint also changes the on/off schedule
of units; more units are to be operated to satisfy this constraint. These constraints introduce
problems and complicate the unit commitment problem. The units that have the minimum cost
are kept on line. Some of the more expensive units are kept as standby or peaking. Thus the
actual unit commitment problem has high dimensionality, non linear in nature requiring optimal,
robust and fast solution methodology.
It is the standard practice to represent the characteristic of a unit by a quadratic function
which is convex in nature. Presently large units with multi-valve steam turbines exhibit a large
variation in this characteristic, as a result non convexity appears. Non-convexity can not be
handled by conventional single approaches. Derivative based approaches like lambda iteration
method fails to solve the non-convex economic dispatch sub problem. Various mathematical
programming based optimization techniques have been used to solve unit commitment problem.
Most of these are calculus-based optimization techniques that are based on first and second order
differentiations of objective function and its constraint equations as the search direction. They
require input-output characteristics to be of monotonically increasing nature and thus can not
3
solve non-convex ED problem. AI techniques have the ability to take into account the
nonlinearities and discontinuities commonly present in the power systems.
Many classical approaches so far have been developed and successfully implemented.
Some of the most commonly used approaches are Enumeration, priority list, dynamic
programming, Integer and Mixed integer programming, Lagrange relaxation, Benders
Decomposition and Branch and Bound. Enumeration (Brute force) scheme can be used for global
optimal solution. Priority list (PL) methods are fast but they are highly heuristic and produce
near optimal schedules. The dynamic programming method based on a priority list is flexible,
but the computational time depends upon the dimensions of the problem. Integer programming
(IP) and Mixed-integer programming (MIP) require considerable computational efforts when
dealing with large number of units. In LR method it is difficult to obtain initial feasible solutions.
The deterministic approaches are fast and simple. The main problem with these techniques is
numerical convergence and quality of solution.
The non classical single approaches like Tabu Search (TS), Genetic Algorithms (GA),
simulated annealing (SA), Particle Swarm Optimization (PSO), greedy random adaptive search
procedure (GRASP) and Evolutionary Programming (EP) etc. attract the researchers, because of
their ability to search not only for local optimal but also for global optimal. These approaches
provide near optimal solutions, but also suffer from the curse of dimensionality. For large system
of units they consume a lot of computational time to reach near global minimum and quality of
solution is also affected. Thus there is an incentive to explore hybrid methods to get around
above mentioned problems. These methods combine the strength of one approach with weakness
of other approach.
1.3 Objectives
The primary objectives of this thesis are outlined below:
1. To develop flexible and extensible computational framework as general environment for
implementing the various algorithms for unit commitment solution.
2. To develop efficient PSO based Artificial Neural Network (ANN) hybrid models for
convex and non convex UCP.
4
3. To compare the results of the proposed hybrid models with GA and other approaches
available in the literature.
4. To develop and formulate the scaleable unit commitment hybrid model, with reference to
the following:
• Easy minimum up and down time constraint handling
• Consideration of convex and non convex cost function
• Quick economic dispatch Calculations
• Low transition cost
• Generating quality solution (low total production cost)
Its (Scaleable unit commitment hybrid model) implementation in computational framework
and finally testing on IEEE standard test systems
5. To review the operational problems of the PAKISTANI utility system National
Transmission and Despatch Company “NTDC” and to develop its test systems for unit
commitment studies.
1.4 Scope of the work
The contributions made in this thesis are:
1. Implementation of two single approaches, Brute Force technique, and conventional
priority list methods to generate results for comparison and development of a new
proposed single approach.
2. Development of generation of quick initial feasible unit commitment schedules based on
forecasted load curve, which satisfy the spinning reserve requirements of the system and
also satisfy the Minimum up Time (M.U.T) and Minimum down Time (M.D.T)
constraints. The generation of initial feasible solution is much important, for the Unit
Commitment problem. When initial feasible schedules are generated randomly, it is
difficult to get feasible schedules for 24 hours loads considering minimum up and
minimum down time constraints. Initial unit commitment scheduling is generated by
using a new hybrid priority list method by focusing on peak and off-peak loads of the
daily load curve.
3. Development of a methodology to generate trial/neighbor solutions in order to achieve
global minimum which also reduce the dimensionality of the unit commitment problem.
5
5. Development and testing of three new hybrid algorithms for solving the UCP for convex
and non convex fuel cost functions. The proposed models are listed as follows:
• Hybrid Model–I: PSO-ANN Hybrid Approach.
• Hybrid Model–II: Neuro-Genetic Hybrid Approach.
• Hybrid Model–III: Scaleable Unit Commitment Hybrid Approach. A hybrid of
maximum power (PMAX) output of units and full load average production cost
(FLAPC).
6. Development of Swarm Intelligence (SI) and back propagation (BP) learning based
feedforward neural network for hybrid models-I and II.
7. Formulation of unit commitment using Dynamic Programming (DP) approach for the
generation of test and training data for ANN for convex cost function, for hybrid model-I.
8. Formulation of unit commitment using Genetic Algorithm (GA) approach for the
generation of test and training data for ANN for non-convex cost function
9. For the national utility system “NTDC” the tasks achieved includes:
• Operational constraints of NTDC systems
• Four test systems with convex cost characteristics curves close to the original
machines in the NTDC system have been prepared for unit commitment
problem. The test systems consist of 12, 15, 25 and 34 thermal units.
• Unit commitment study and testing of the four test systems of NTDC.
1.5 Thesis Organization
This thesis consists of seven chapters and four appendices organized as follows:
Chapter 2 presents the literature survey on unit commitment problem based on classical,
non-classical and hybrid approaches with the view to develop some observations and potential
avenues for further investigations.
Chapter 3 discusses the optimization tools used in this thesis. These include the Particle
Swarm Optimization, artificial neural network, dynamic programming and genetic algorithm.
Chapter 4 deals with the mathematical modeling of unit commitment problem and
discussion on three single solution approaches, efficient generation of initial feasible and trial
6
solutions, minimum up and down time constraint handling and economic dispatch calculations.
For validation of the three single approaches the algorithms has been implemented on 3 and 10
unit standard test systems.
Chapter 5 presents the proposed new three hybrid models for convex and non convex fuel
cost function.
• Hybrid Model–I: combines the Dynamic Programming (DP), with feed forward neural
network using Swarm Intelligence (SI) learning rule and has been tested on two 3 unit
and one 10 unit standard test systems for convex fuel cost functions. For comparison the
neural network trained with back propagation learning (BP) rule has also been developed.
• Hybrid Model–II: combines the Genetic Algorithm (GA), with feed forward neural
network using Swarm Intelligence (SI) learning rule and has been tested on two 3 unit
standard test systems for non-convex fuel cost functions. For comparison the neural
network trained with back propagation (BP) learning rule has also been developed.
• Hybrid Model–III: is the integration of maximum power output of units and full load
average production cost. The model has been tested on four 3 unit standard test systems
for convex and non-convex fuel cost functions, one 10 unit standard system, and four 12,
15, 25 and 34 unit NTDC systems.
Chapter 6 gives the discussion on operational problems on PAKISTANI utility system
National Transmission and Despatch Company (NTDC) along with its unit commitment studies.
Chapter 7 discusses the conclusions and suggestions for future research.
7
CHAPTER 2
Unit Commitment Problem --- A Brief Literature Survey
2.1 Introduction
The Unit Commitment Problem (UCP) is a large scale, non-linear, 0-1 combinatorial
optimization problem.
This chapter presents an overview and literature survey on UCP. Final section includes
the directions on which the new approaches evolved with time, discussion and potential avenues
for further investigations including hybrid approaches.
2.2 Power System Operational Planning
The objectives of the power system operational planning involves the best utilization of
available energy resources subjected to various constraints and to transfer electrical energy from
generating stations to the consumers with maximum safety of personal/equipment, continuity,
and quality at minimum cost.
The operational planning involves many steps such as short term load forecasting, unit
commitment, economic dispatch, hydrothermal coordination, control of active/reactive power
generation, voltage, and frequency as well as interchanges among the interconnected systems in
power pools etc.
In the early days the power system consisted of isolated stations and their individual
loads. But at present the power systems are highly interconnected in which several generating
stations run in parallel and feed a high voltage network which then supplies a set of consuming
centers. Such system has the advantages of running the number of stations with greater reliability
and economy, but at the same time the complexity in the operational and control procedures has
increased. The power industry therefore requires the services of the group of men who are
specially trained to look after the operation of the system. These men are known as the system
engineers and are responsible for the operation, control and operational planning of the system.
8
Unit Commitment involves the hour-to-hour ordering of the units on/off in the system to
match the anticipated load and to allow a safety margin. Having solved the unit commitment
problem and having ensured through security analysis that present system is in a secure state
then the efforts are made to adjust the loading on the individual generators to achieve minimum
production cost on minute-to-minute basis. This loading of generators subjected to minimum
operation cost is in essence the economic dispatch.
Load forecasting gives an accurate picture of the expected demand over the following
few hours. In an anticipation of the variations in demand and for reasons of economic operation
of the system the unit commitment activity is carried out.
The solution methods being used to solve the UCP can be divided into three categories as:
• Single classical/Deterministic approaches: A variety of classical/deterministic single
techniques in this context have been reported such as: Priority List (PL), Dynamic
Programming (DP), Exhaustive Enumeration (Brute Force Technique), Branch and
Bound (B&B), Integer /Mixed integer programming (IP/MIP), Lagrangian Relaxation
(LR), Straight Forward (SF) and Secant Methods.
• Single non classical approaches: The popular single non classical approaches which got
attention in recent years are such as: Tabu Search (TS), simulated annealing (SA), Expert
System (ES), Artificial Neural Networks (ANN), Evolutionary Programming (EP),
Genetic algorithms (GA), Fuzzy Logic (FL), Particle Swarm Optimization (PSO), Ant
Colony Optimization (A.C.O), and Greedy Randomized Adaptive Search Procedure
(GRASP).
• Hybrid techniques based on classical and non-classical approaches: More recently
hybrid techniques combining two or more of the above mentioned optimization
techniques were proposed to solve UCP such as: Particle Swarm Optimization Based
Simulated Annealing, Enhanced Lagrange Relaxation, Augmented Lagrange Relaxation,
Fuzzy Adaptive Particle Swarm Optimization, Hybrid Particle Swarm Optimization,
Lagrange Relaxation Parallel Particle Swarm Optimization, Lagrange Relaxation Parallel
Relative Particle Swarm Optimization, Unit Characteristics Classification-Genetic
9
Algorithm, Tabu Search based Hybrid Particle Swarm Optimization, Annealing Genetic
Algorithm, Ant Colony Simulated Annealing, Dynamic Programming- Lagrange
Relaxation, Lagrange Relaxation- Genetic Algorithm, Lagrange Relaxation-Particle
Swarm Optimization, Enhanced Merit Order- Augmented Lagrange Hopfield Network,
Priority List based Evolutionary Algorithm, Memetic Algorithm seeded with Lagrange
Relaxation, Dynamic Programming based Hopfield Neural Network etc.
2.3 Unit commitment --- Literature Survey
Unit commitment is the problem to determine the optimal subset of units to be used
during the next 24 to 168 hours [1]. This section presents a survey of the research work based on
techniques both using conventional as well as Artificial Intelligence (AI) approaches.
Traditionally the UC problem is to minimize the total production costs (TPC), (operating
fuel cost, start-up and shut-down costs) and is referred as the cost-based-unit-commitment
(CBUC) problem [2–3]. A 0.5 percent saving of the operating fuel cost gives savings of millions
of dollars per year for large utilities [4]. A number of methodologies to solve the UCP exist and
are under investigation [5-14].
The next section gives the review of several classical approaches which have been reported
in the literature
2.4 Single Classical/Deterministic Approaches
Classical methods give good results. They are heuristic and have dimensionality problem.
2.4.1. Priority List
In 2003, T. Senjyu, et al. [15] introduced extended priority list (EPL) method. The
approach consists of two steps. The initial UC schedules are produced by priority list method and
then modified using the problem specific heuristics to fulfill unit and system constraints. Some
heuristics are also applied. The Economic Dispatch is performed only on the feasible schedules.
10
In 2006, T. Senjyu, et al. [16] proposed Stochastic Priority List (SPL) method. Some
initial feasible UC schedules are generated by Priority List method and priority based stochastic
window system. Some heuristics are used to reduce search space and computational time.
2.4.2. Dynamic Programming
In 1966, P. G. Lowery, [17] proposed DP in solving UCP. The main concern was to
determine the feasibility of using Dynamic Programming to solve the UCP. Results of the study
show that simple, straightforward constraints are adequate to produce a usable optimum
operating policy. The computer time to produce a solution is small.
In 1981, C. K. Pang, et al. [18] presented a study of three different DP algorithms. The
Dynamic Programming-Sequential Combinations (DPSC) and Dynamic Programming-Truncated
Combinations (DP-TC) and Dynamic Programming-Sequential/Truncated Combinations (DP-
STC), is a combination of the DP-SC and DP-TC methods. Four methods were used to establish
the savings and computer resource requirements.
In 1986, S. D. Bond, et al. [19] presented a dynamic programming which is capable of solving
the generation scheduling problem. The solutions are guaranteed to be optimal and are obtained
by using a state definition which includes the length of time a unit has been on or off. This
information is required to assess the effect of present commitment decisions on future flexibility.
When similar combinations of with the lowest accumulated cost is pursued further. The reduced
search effort lowers run times by an order of magnitude compared with a mixed integer-linear
programming approach relying mainly on constraint violations as a truncation mechanism.
Storage requirements are reduced even more significantly. The algorithm has been tested
successfully on data for a small and medium sized thermal power system. In addition to the usual
upper and lower limits on unit outputs, emergency reserve and MUT and MDT down time
constraints were incorporated.
In 1987, W. L. Snyder, et al. [20] proposed an approach to save computational time. This
algorithm incorporates a number of special features and effectively deals with the control of
problem size. To achieve the computational time saving, individual units were assigned status
11
restriction in any given hour. This approach features the classification of units into groups so as
to minimize the number of unit combinations. Programming techniques are described which
maximize efficiency. This approach has been proved on a medium size utility for which sample
results were presented.
In 1988, W. L. Hobbs, et al. [21] developed an enhanced DP approach. This approach
saves predecessor options. The approach was implemented in an on-line energy management
system. A merit order list is formed which excludes all unavailable, fixed output, peaking, and
must run units. Subsequent combinations of units are formed by decommitting one unit at a time.
The method creates several states from each unique combination and links each state to one of
the possible paths to that combination.
In 1991, C. C. Su, et al. [22] developed a technique using fuzzy DP. The errors in the
forecasted load are considered and membership functions are derived for the load demand, the
total cost, and the spinning reserve using fuzzy set notations. With these membership functions at
hand, a recursive algorithm for fuzzy dynamic programming is presented. The developed
algorithm is used to solve the unit UC of Taiwan power. The proposed fuzzy dynamic
programming approach requires more computer time than the DP approach.
In 1991, Z. Ouyang, et al. [23], presented a heuristic improvement of the truncated
window DP and used a variable window size according to forecast load demand increments. The
corresponding experimental results show a considerable saving in the computation time.
2.4.3 Branch and Bound
In 1983, A. I. Cohen, et al. [24] proposed a new approach based on branch-and-bound
techniques. The method incorporates start-up costs, load demand, spinning reserve, MUT and
MDT constraints.
2.4.4 Integer and Mixed Integer programming
In 1978, T. S. Dillon, et al. [25] proposed an extended and modified version of applying
branch and bound technique for Integer Programming and treats the commitments of both hydro
12
and thermal systems. The method is computationally practical for realistic system. The present
method constitutes a basis for the development of unit commitment programs using integer
programming for practical use in electric utilities.
In 2000, S. Takriti, et al. [26] presented a technique for refining the schedules obtained
by Lagrangian method. Given the schedules generated by Lagrangian iterations, and improved
schedule was found by solving the mixed integer program. The model was an integer program
with non linear constraints and solved for optimal solution using branch and bound technique.
The method gives a significant improvement in terms of quality of the solution for large number
of units.
In 2005, Li Tao, et al. [27] formulated the price-based unit commitment problem based
on the mixed integer programming method. The proposed PBUC solution is for a generating
company having cascaded-hydro, thermal, pump storage and combined-cycle, units. The results
are compared with LR method. The major obstacles are more computation time and memory
requirement to solve large UC problems.
In 2007, B. Venkatesh, et al. [28] demonstrated advantages of using the fuzzy
optimization model and presents fuzzy linear optimization formulation of UC using a mixed
integer linear programming (MILP) routine. In this formulation, start up cost is modeled using
linear variables. The fuzzy formulation provides modeling flexibility, relaxation in constraint
enforcement and allows the method to seek a practical solution. The use of MILP technique
makes the proposed solution method rigorous and fast. The method is tested on a 24 h, 104-
generator system demonstrating its speed and robustness gained by using the LP technique. A
five-generator system is additionally used to create a see-through example demonstrating
advantages of using the fuzzy optimization model.
2.4.5 Lagrange Relaxation Method
In 1983, A. Merlin, et al. [29] proposed a new implementation in solving UCP by
Lagrangian relaxation method. Numerous developments were envisaged, to make the algorithm
flexible such as simultaneous management of pumping units, probabilistic determination of the
13
spinning reserve. This decomposition method used is flexible and Lagrange multiplier provides a
new solution to the conventional problem.
In 1987, R. Nieva, et al. [30] proposed an approach to solve very large and complex
UCP. The proposed approach gives an estimate of suboptimality that indicates the closeness of
the solution near to the optimum. In contrast with the technique of Lagrangian Relaxation, this
approach makes no attempt of maximizing the dual function.
In 1988, F. Zhuang, et al. [31] presented an LR method for large scale problem. The
algorithm in divided into three phases. First the Lagrangian dual of the unit commitment is
maximized with standard subgradient techniques, second a reserve-feasible dual solution is find,
and finally ED is performed. On 100 units to be scheduled over 168 hours, gives a reliable
performance and low execution times. Both spinning and time-limited reserve constraints are
treated.
In 1989, S. Virmani et al. [32] presented a paper in which they provide an understanding
of the practical aspects of the Lagrangian Relaxation methodology for solving the thermal UCP.
In 1995, R. Baldick [33] formulated UCP in generalized form and solved using LR
method. The algorithm, presented, approximately solves the dual optimization problem. The
algorithm was slower in solving the special cases of the generalized UCP than algorithms
demonstrated by other authors. The approach has been tested for ten units for a time period of 24
hours.
In 1995, W.L. Peterson, et al. [34] proposed a Lagrange Relaxation to incorporate unit
minimum capacity and unit ramp rate constrains. The proposed method is used in finding a
feasible UC schedule considering a new approach for ramping constraints. The algorithm
incorporates other practical features such as boiler fire-up characteristics and non-linear ramp up
sequences.
14
In 2000, A. G. Bakirtzis, et al. [35] demonstrated the difference between the lambda
values of the economic dispatch and the UCP based on economic interpretation of the
Lagrangian Relaxation solution framework. During the LR solution of the UCP two sets of
lambdas are used. Although both set of lambdas represent marginal cost of electricity. The first
one, is assigned as a Lagrange multiplier (Lambda) to the UC power balance equations and
second one, is the Lagrange multiplier of the power balance equation in the economic dispatch
problem.
In 2004, W. P. Ongsakul, et al. [36] proposed an enhanced adaptive Lagrangian
relaxation (ELR). Enhanced LR approach consists of heuristic search and adaptive LR. ALR is
enhanced by introducing new 0-1 decisions. After the ALR the best feasible schedule is obtained.
The heuristic search is used to fine tune the schedule. The total system production costs are less
for the large scale system. The computational time is much less compared with others
approaches.
In 2005, D. Murtaza, et al. [37] presented an algorithm for the unit commitment schedule
using the Lagrange relaxation method by taking into account the transmission losses. For better
convergence and faster calculation, a two stage Lagrange relaxation was provided. First,
conventional Lagrange relaxation was applied in order to determine the unit commitment
schedule neglecting transmission loss. The results are then input to the proposed method, and the
unit commitment schedule including transmission losses was produced.
2.4.6 Straight Forward Method
In 2007, S. H. Hosseini, et al. [38] presented, a novel fast straightforward method (SF).
This new approach decomposes the UCP into three sub-problems. The quadratic cost functions
of units are linearized and hourly optimum solution of UC is obtained considering all constraints
except the MUT and MDT constraints and then the MUT/MDT constraints are introduced by
modifying the schedule obtained in the first step through a proposed novel optimization
processing. Finally, by using a new de-commitment algorithm the extra spinning reserve is
minimized.
15
2.4.7 Secant Method
In 2008, K. Chandram, et al. [39] proposed an application of Secant method and
Improved Pre-Prepared Power Demand (IPPD) table (based on the units having low minimum
incremental cost) for solving the UCP. The problem is divided into two sub problems, the unit
on/off scheduling and ED sub problem. Initially, IPPD table obtains the unit 0-1 status
information and then the optimal solution is achieved by Secant method. For solving large scale
problems the convergence is in less iteration.
2.5 Non classical approaches
The growing interest is the application of non classical approaches like Artificial
Intelligence (AI) and Swarm Intelligence (SI) in solving the UCP. AI methods like Neural
Networks, Simulated Annealing, Genetic Algorithm, expert system, evolutionary programming,
and fuzzy logic are used to solve the UCP. The SI techniques like PSO and ACO also gained
prominence for solving UCP. In the following section a survey of the AI and SI methods for
UCP are presented.
2.5.1 Tabu search
In 1998, A. H. Mantawy, et al. [40] presented an approach based on the Tabu Search
method. Initial feasible UC schedules are generated randomly using new proposed rules. TSA is
used to solve the combinatorial optimization sub problem while the quadratic programming is
used to solve the EDP subproblem. Numerical results show an improvement in the quality of
solution compared with other approaches.
2.5.2 Simulated Annealing (SA)
In 1990, F. Zhuang, et al. [41] presented a general optimization method, known as
simulated annealing, and is applied to generation unit commitment. SA was used to generate
feasible solutions randomly and moves among these solutions using a strategy leading to a global
minimum with high probabilities. The method assumes no specific problem structures and is
highly flexible in handling unit commitment constraints. Numerical results on test systems of up
to 100 units were reported.
16
In 1998, A. H. Mantawy, et al, [42] presented a Simulated Annealing Algorithm (SAA)
and proposed new rules for randomly generating initial feasible UC schedules. SAA is used to
solve the combinatorial optimization sub problem while the quadratic programming is used to
solve the EDP subproblem. Numerical results show an improvement in the total production cost
compared with other approaches.
In 1998, S. Y. W Wong, [43] developed an enhanced SA-approach for solving the UCP
by adopting mechanisms to ensure that the candidate solutions produced are feasible and satisfy
all the constraints. During the solution process, the solutions are generated in the neighbor of the
current one and the extent of perturbation of the solutions decreased with decreasing
temperature.
In 2006, D. N. Simopoulos, et al. [44] developed a new enhanced SA combined with a
dynamic ED method. SA is used for generator scheduling. The dynamic ED method is used to
incorporate the ramp rate constraints in the UCP. New rules for the tuning of the control
parameters of the SA algorithm are also presented.
In 2006, A. Y. Saber, et al. [45] presented fuzzy UCP using the absolutely stochastic
simulated annealing method.
2.5.3 Expert System
In 1988, S. Mokhtari, et al. [46] presented in setting up an expert system which combines
the knowledge of the unit commitment programmer and an experienced operator. In scheduling
units an expert system based on consultant has been formulated. This expert system will lead an
inexperienced operator to a better unit schedule. The basic expert system used 56 rules for the
experiments. The authors estimate that 300 rules will be required to satisfy all operational
requirements.
In 1991, S. K. Tong, et al. [47] demonstrated PL based heuristic to form initial UC
schedules based on the given forecasted load. A new expert system approach was used to handle
short term UC problem. In the proposed approach one of the previous schedules as the staring
17
point is used to find the new schedule that will satisfy the present load requirements. A rule
based approach is applied to implement PL scheme for modifying the previous schedule so that a
sub-optimal feasible schedule can be obtained quickly.
In 1993, S. Li, et al. [48] presented a graphics package and a new heuristic method for
unit commitment. The principles of this method can be expanded to consider more complicated
cases with additional constraints. The units are divided into three categories the base, medium
and peak. The computational time is less than two seconds.
2.5.4 Artificial Neural Network (ANN)
In 1992, Sasaki, et al. [49] explored the feasibility of using the Hopfield neural network
to unit commitment in which a large number of inequality constraints are handled by the
dedicated neural network instead of including them in the energy function. Once the states of
generators are determined, their outputs are adjusted according to the priority order in fuel cost
per unit output.
In 1999, T. Yalcinoz, et al. [50] presented an improved Hopfield neural networks method.
A new mapping process was used and a computational method for obtaining the weights and
biases using a slack variable technique for handling inequality constraints. Transmission
capacity, transmission losses, start-up and shutdown costs and MUT/MDT constraints have been
taken into account. The HFNN approach has been tested on a 3 unit and a 10 unit systems.
In 2000, M. H. Wong, et al. [51] used GA to evolve the weight and the interconnection of
the neural network to solve the UC problem. The back-propagation was used to train the weights.
Three selection methods Roulette Wheel, Tournament and Ranking was used as well as two
options for Weight and Connections are combined for running the GA. Roulette Wheel has the
best performance.
18
2.5.5 Evolutionary Programming (EP)
In 1999, K. A. Juste, et al. [52] proposed an algorithm that uses the EP technique in
which populations of initial population is generated randomly and then the solutions are evolved
through selection, competition, and random changes.
2.5.6 Genetic Algorithm (GA)
In 1993, D. Dasgupta et al. [53] presented a genetic approach for determining the priority
order in the commitment of thermal units in power generation. The paper examined the
feasibility of using genetic algorithms and reports some simulation results in near optimal
commitment of thermal units. The genetic-based UC system evaluates the priority of the units
dynamically considering the system parameters, operating constraints and load profile at each
time period in the scheduling horizon.
In 1995, X. Ma, et al. [54] developed a forced mutation operator and the efficiency of the
GA was improved significantly using this operator. Two different coding schemes were devised
and tested. It was observed that the two-point crossover operation is considerably more efficient
than the single-point crossover commonly used in GAs In addition, the effects of GA’s control
variables on convergence were extensively studied. The approach was tested on a 10 unit system.
Test results clearly reveal the robustness and promise of the proposed approach.
In 1996, S. O. Orero, et al. [55] presented an enhanced genetic algorithm incorporating
sequential decomposition logic for faster search mechanism Unit commitment constraints
including ramp rates are considered. The method relies on the selection and grading of the
penalty functions to allow the fitness function to differentiate between good and bad solutions.
The method guarantees the production of solutions that do not violate system or unit constraints,
so long as there are enough generators available in the selection pool to meet the required load
demand. The algorithm has been tested on 26 generators.
In 1996, S. A. Kazarlis, et al. [56] presented Genetic Algorithm by using Varying Quality
Function technique and adding problem specific operators. The coding was implemented in a
binary form. With the technique of varying quality function, the GA finally manages to locate
19
the exact global solution. A nonlinear transformation was used for fitness scaling. New operators
swap-mutation and swap-window hill-climb was implemented. The algorithm is applied to 100
units.
In 2002, T. Senjyu, et al. [57] presented new genetic operator based on unit characteristic
classification and intelligent technique for generating initial populations. The initial population is
generated base on load curve. To handle MUT/MDT constraints new mutation operators were
introduced. New cross over operator, shift operator, and intelligent mutation operators were
proposed. Units are classified in several groups depending upon their MUT/MDT constraints.
For every violated constraint, a penalty term is added to the total cost.
In 2003, E. Gil, et al. [58] proposed a new method for hydrothermal systems. The
proposed GA, using new specialized operators, has demonstrated excellent performance in
dealing with this kind of problem, obtaining near-optimal solutions in reasonable times.
In 2004, G. Loannis, et al. [59] presented a new solution based on an integer-coded
genetic algorithm (ICGA), in which the chromosome size is reduction compared to the binary
coding. The non linear MUT and MDT constraints are directly coded in the chromosome. The
use of penalty functions is avoided because they distort the search space. The ICGA is robust and
execution time is less than other approaches.
In 2006, C. Dang, et al. [60] proposed a floating-point genetic algorithm (FPGA). In
which a floating-point chromosome representation is used based on the forecasted load curve. To
handle MUT and MDT constraints encoding and decoding schemes are used. The fitness
function, constraints, population size, selection, crossover and mutation probabilities are
characterized in detail. The FPGA is also applicable for non-convex cost function.
In 2006, L. Sun, et al. [61] introduced a matrix real-coded genetic algorithm (MRCGA).
A real number matrix representation of chromosome is used that can solve the UC problem
through genetic operations. The search performance is improved through a window mutation.
20
The proposed new mechanism of chromosome repair guarantees that the UC schedule satisfies
unit and system constraints.
2.5.7 Fuzzy logic
In 1997, S. Saneifard, et al. [62] formulated the fuzzy logic to the UCP. A comparison of
results presented in the paper indicates that the use of fuzzy logic provides outcomes comparable
to those of conventional dynamic programming. It is claimed that this approach gives
economical cost of operation.
In 2004, S. C. Pandian, et al. [63] presented a fuzzy logic approach that is very useful to
consider the uncertainty in the forecasted load curve, derating and line losses. Numerical results
are compared w.r.t the operating cost and computation time obtained by using fuzzy dynamic
programming and other conventional methods like dynamic programming, Lagrangian relaxation
methods. For validation of the approach in respect of total production cost and computational
time, case studies on 10, 26 and 34 units have been performed.
2.5.8 Particle Swarm Optimization (PSO)
In 2003, Z. L. Gaing, [64] proposed binary particle swarm optimization (BPSO). The
BPSO is used to solve the combinatorial unit on/off scheduling problem for operating fuel and
transition costs. The ED subproblem is solved using the lambda iteration method for obtaining
the total production cost.
In 2006, B. Zhao, et al. [65] presented an improved particle swarm optimization
algorithm (IPSO) for UC which utilizes more particles information to control the process of
mutation operation. For proper selection of parameters some new rules are also proposed. The
proposed method combines LR technique to 0-1 variable.
In 2007, T. Y. Lee, et al. [66] presented a new approach for UCP named the iteration
particle swarm optimization (IPSO). The proposed method improves the quality of solution in
terms of total production cost and also improves the computation efficiency. A standard 48 unit
system has been tested for validation.
21
In 2009, X. Yuan, et al. [67] proposed a new improved binary PSO (IBPSO). The
standard PSO is improved the using the priority list and heuristic search to improve the MUT and
MDT constraints. The 10-100 units have been tested to validate the proposed approach.
Numerical performance shows that the proposed approach is superior in terms of low total
production cost and short computational time compared with other published results.
2.5.9 Ant Colony Optimization (ACO)
In 2003, T. Sum-im, et al. [68] proposed, ant colony search algorithm (ACSA), which is
inspired by the observation of the behaviors of real ant colonies, and is a new cooperative agent’s
approach based on parallel search. In the proposed approach, a set of cooperating agents called
“ants” cooperates to find good unit schedules the ED sub-problem is solved by the λ-iteration
method.
In 2008, A.Y. Saber, et al. [69] proposed memory-bounded ant colony optimization
(MACO). The proposed approach is applicable for large number of units and solves the
computer memory limit requirements. A heuristic is also incorporated to enhance local search.
2.5.10 Greedy Randomized Adaptive Search Procedure (GRASP)
In 2003, A Viana, et al. [70] presented, an adaptive algorithmic framework based on
another meta-heuristic principle (GRASP – Greedy Randomized Adaptive Search Procedure).
The philosophy applied is slightly different from standard meta-heuristics, the decisions taken by
the method, when building a solution, are somehow adapted according to decisions previously
taken. This dynamic learning-process often leads to very good solutions.
2.6 Hybrid approaches
Hybrid approaches are also used to solve many difficult engineering problems. The aim
of the hybrid methods is to improve the performance of single approaches. The objective of
hybrid of two or more methods is to speed up the convergence and to get better quality of
solution compared with single approaches. A brief review of different hybrid approaches which
have been reported in the literature is presented in this section.
22
In 1992, Z. Ouyang et al. [71] utilized neural networks to generate a pre-schedule
according to the forecasted load curve. The proposed approach significantly reduced the
computational time. Case studies are performed on 26 unit system. A 35 training pattern are used
in the study. The training of each load takes approximately eight to ten minutes.
In 1992, Z. Ouyang, et al. [72] proposed a multi-stage Neural Network-expert system
approach. Through inference the feasible UC schedule is obtained. A load pattern matching
scheme is performed at the pre-processor stage. The trained network performs adjustments in
the schedule to achieve the optimal solution at the post processor stages.
In 1995, D. P. Kothari, et al [73] described a hybrid expert system dynamic programming
approach. The output scheduling of the usual dynamic programming is enhanced by
supplementing it with the rule based expert system. The proposed system limits the number of
constraints and also checks the possible constraint violations in the generated schedule. The
expert system communicates with the operator in a friendly manner and hence the various
program parameters can be adjusted to have an optima1, operationally acceptable schedule.
In 1997, H. Shyh-Jier, et al. [74] proposed genetic algorithm based neural network and
dynamic programming approach for UCP. At the initial stage a set of feasible UC schedules are
generated by genetic-enhanced neural networks. In the second stage these schedules are
optimized by using the DP approach. The computational efficiency is more compared with other
methods.
In 2000, M. H. Wong, et al. [75] presented a technique in which genetic algorithm is
evolved to intelligently decide the initial weights and the connections in the ANN. This approach
prevents the stagnation during training. The approach converges into global minimum for a given
range of space. The evolving neural network has lower training error compared to neural
network with random initial weights.
In 2000, R. Nayak, et al. [76] proposed a hybrid of feed forward neural network and the
simulated annealing. The ANN is used to solve the unit scheduling sub problem and the SA is
23
used to solve the ED sub problem. A set of inputs based on the forecasted load curve and
corresponding UC schedules as outputs satisfying the system and unit constraints are used to
train the network. A reduction in computational time is achieved by this approach.
In 2000, C. P. Cheng, et al. [77] presented an application of Genetic Algorithms and
Lagrangian Relaxation (LRGA) method. The proposed approach incorporates GA into LR
method to improve the performance of LR and to update the Lagrangian multipliers. The method
is easy to implement, better in convergence.
In 2002, C. P. Cheng, et al. [78] proposed the application of the annealing–genetic (AG)
algorithm. The AG is a hybrid of GA into the SA to improve the performance of the SA
approach. The method improves the computational time of the Simulated Annealing and the
quality of solution of Genetic Algorithm and gives near optimal solution of a large scale system.
In 2002, J. Valenzuela, et al. [79] presented memetic algorithm, a hybrid of GA, and LR
is efficient and effective for solving large UC problems. The implementations of standard GA or
MA are not competitive compared with the traditional methods of DP and LR. However, an MA
incorporated with LR proves to be superior to other approaches on large scale problems.
In 2003, T. O. Ting, et al. [80] proposed a Hybrid Particle Swarm Optimization (HPSO).
Problem formulation, representation and the numerical results for a 10 unit are presented. Results
shown are acceptable at this early stage.
In 2003, C. C. A. Rajan, et al. [81] presented a neural based tabu search (NBTS) method.
The algorithm is based on the short term memory procedure of the tabu search method. Systems
consisting of 10, 26, and 34 units have been tested and the results are compared with other
approaches. The results in terms of total production cost and computational time are better than
single approaches like DP and LR.
24
In 2004, D. Srinivasan, at al. [82] proposed an efficient algorithm for aiding unit
commitment decisions. To solve the UCP an evolutionary algorithm with problem specific
heuristic and genetic operators has been employed
In 2004, L. Shi, et al. [83] developed and demonstrated a novel ant colony optimization
algorithm with random perturbation behavior (RPACO). The approach is based on the
combination of colony optimization and stochastic mechanism is developed for the solution of
optimal UC with probabilistic spinning reserve.
In 2004, H. H. Balci, et al. [84] presented a hybrid of PSO and LR. UCP is divided into
sub problems and each sub problem is solved using DP. PSO is used to update the Lagrangian
multipliers. The comparison of results shows that the proposed approach uses less computational
time and gives good quality solutions.
In 2005, T.A.A Victoire, [85] introduced an application of hybrid-PSO and sequential-
quadratic programming technique (SQP) guiding the tabu search (TS). The unit scheduling
problem is solved using an improved random-perturbation scheme. A simple procedure for
generating initial feasible UC schedules is proposed for the TS method. The nonlinear ED
subproblem is solved using the hybrid PSO-SQP technique.
In 2005, S. Chusanapiputt, et al. [86] presented Parallel Relative Particle Swarm
Optimization (PRPSO) and LR for a large-scale system. To reduce the dimensionality problem
and to improve the UC schedules the neighborhood solutions are divided into sub-
neighborhoods.
In 2005, P. Sriyanyong, et al. [87] proposed PSO based LR method for optimal setting of
Lagrange multipliers. In the proposed work, the PSO was used to adjust the lagrange multipliers
in order to improve the performance of lagrange relaxation method.
In 2005, T. Aruldoss, et al. [88] presented a solution model using fuzzy logic. Hybrid of
Simulated annealing, particle swarm optimization and sequential quadratic programming
25
technique (hybrid SA-PSO-SQP) is used to schedule the generating units based on the fuzzy
logic decisions.
In 2006, T. O. Ting, et al. [89] introduced a hybrid particle swarm optimization (HPSO)
which is a combination of binary and real coded particle swarm optimization (BPSO and
RCPSO). The term “hybrid particle swarm optimization” was first mentioned by S. Naka, et al.
where by the term hybrid meant the combination of PSO and GA. The BPSO is used to solve
unit scheduling problem and RPSO is used to solve the ED subproblem.
In 2006, V.N. Dieu et al. [90] proposed an enhanced merit order (EMO) and augmented
Lagrange Hopfield network (ALHN) for solving hydrothermal scheduling (HTS) problem with
pumped-storage units. The proposed approach is based on merit order approach enhanced by
heuristic search based algorithms. The ALHN is a continuous Hopfield network and its energy
function is based on augmented Lagrangian function. EMO is efficient in unit scheduling,
whereas ALHN can properly handle generation ramp rate limits, and time coupling constraints.
In 2007, A. Y. Saber, et al. [91] presented a twofold simulated annealing (twofold-SA)
method. A hybrid of SA and fuzzy logic is used to obtain SA probabilities from fuzzy
membership function. The initial feasible UC schedules are generated by a priority list method
and are modified by de-composed SA using a bit flipping operator. Results indicate a low total
production cost and low execution time compared with other approaches.
In 2007, S. Nasser, et al. [92] presented hybrid particle swarm optimization based
simulated annealing (PSO-B-SA) approach. The unit scheduling sub problem is solved by using
binary PSO and ED sub problem is solved by using real valued PSO. Numerical results
demonstrated show that the PSO-B-SA approach can perform well compared with the other
solutions.
In 2007, A. Y. Saber, et al. [93] proposed a fuzzy adaptive Particle Swarm Optimization
(FAPSO) for UCP. FAPSO precisely tracks a changing schedule. Based on the diversity of
26
fitness the fuzzy adaptive criterion is used for the PSO inertia weight. Using fuzzy IF/THEN
rules the weights are dynamically adjusted.
In 2007, S. S. Kumara, et al. [94] developed DP based direct Hopfield computation method. The
proposed approach solves the UCP in two steps. The generator scheduling problem is solved
using DP and generation scheduling problem is solved using Hopfield neural network.
2.7 Unit Commitment --- Issues and Bottlenecks
The issues and bottlenecks in the UCP may be listed as:
1. High dimensionality
2. Handling of cost base and profit base unit commitment
3. Handling of non convex fuel cost function
4. Generation of infeasible solutions
5. Handling of constraints such as:
i. Minimum up and down time
ii. Transmission
iii. Emission
iv. Security
2.8 Discussion
The global optimal solution of the UCP can be obtained by Brute Force (complete
enumeration) technique, which is not applicable for a power system having large number of units
due to its long computational time. Priority list (PL) methods are highly heuristic but very fast
and give UC solutions with high total production cost. The DP methods are based on
enumeration and PL, but suffer from the curse of dimensionality. Integer programming (IP) and
Mixed-integer programming (MIP) require considerable computational efforts when dealing with
large-scale problems. The main problem with Lagrangian relaxation (LR) method is the
difficulty in obtaining the feasible UC schedules.
The non classical approaches such as Evolutionary Computation, Genetic Algorithm, and
Particle Swarm Optimization etc. attract much attention, because they are able to solve convex
and non-convex fuel cost functions, have the ability to search for near global and can deal easily
with non linear constraints. In case of large-scale problem these single approaches consume long
27
computational time. The main difficulty is their sensitivity on the choice of parameters. Hence,
there is an incentive to explore hybrid algorithms. From the selected above mentioned literature
review, it is observed that the hybrid techniques reduce the search space are more efficient and
have better quality of solutions for small and large scale problems, gives solution in an
acceptable computation time and can accommodate more constraints. Thus enhancing existing
classical and non classical optimization approaches and exploring new single and hybrid
approached to solve unit commitment problem has great importance. Among the hybrid
approaches the Swarm Intelligence techniques are new to apply to the UC problem. PSO is new,
flexible and efficient tool for UCP. The potential avenues for further exploration may be listed
as:
1. How to generate initial feasible schedules considering spinning reserve requirements.
2. How to satisfy MUT and MDT constraints.
3. Exploration of new operators for MUT and MDT constraint handling.
4. To reduce the high dimensionality of the UCP.
5. Hybrid methodology is the useful tool for efficient solution by exploiting the strength of
single classical approaches, and non classical approaches.
6. Exploration of fast and efficient method for utility system.
7. Exploration of PSO based approaches.
8. Hybrid models based on the integration of classical and non classical approaches for
enhancing the computational efficiency and handling of non-convex cost function for the
UC problem.
28
CHAPTER 3
Optimization Tools for Unit Commitment
3.1 Introduction
Optimization is the process of making some thing better. Up till now several heuristic
tools have evolved that facilitates solving optimization problems that were previously difficult or
impossible to solve. These tools include particle swarm optimization, tabu search, artificial
neural networks, genetic algorithm, simulated annealing, etc.
This chapter presents the general overview of the optimization techniques on particle swarm
optimization (PSO), artificial neural networks (ANN), dynamic programming (DP) and genetic
algorithm (GA), with the view to their use in the subsequent chapters.
3.2 Particle swarm optimization (P.S.O)
Particle swarm optimization (PSO) first presented by Dr. Kennedy and Dr. Eberhart
[95,96] in 1995 is one of the evolutionary computational techniques based on the social
behaviours of bird flocking and fish schooling. This is a population based stochastic global
optimization technique. PSO has a population with random search solution. Each potential
solution is represented as a particle in population called swarm. Since its introduction it has
attracted lot of attentions from the researchers around the world. PSO models problem as a set of
n particles each representing a dimension of solution space. These particles move in solution
space in search of optimal solution. The particles follow three principles as described by
Kennedy [96] including evaluating, comparing: and Imitating. PSO been used to optimize real
and discrete functions which otherwise are difficult to solve. It can be easily implemented and
high quality solution with stable convergence.
The velocity and position update equations are given by
( ) ( ) ( ) ( )[ ] ( ) ( )[ ]txtytrctxtytrctwvtv ijjijijijij −+−+=+ ˆ)()(1 22,11 (3.1)
( ) ( ) ( )11 ++=+ tvtxtx ijijij i = 1, 2, 3 …s j= 1, 2, 3…n (3.2)
29
Where, xi,j : is the current position of the particle at iteration j, w : inertia weight, vi,j : is the
current velocity of the particle at iteration j, vi(t+1) : is the updated velocity of the particle, yi : is
the personal best position of the particle (every particle tries to adjust its velocity according to
best positions ever visited that is stored in its memory), and y^i is an instance of xi that is visited
by the particle and yielded best positions currently found. c1, c2 are positive numbers and
represent cognitive and social components respectively, c1, c2 controls the movement of the
particle, r1, r2 are the uniform distribution numbers in the range [0, 1], n be the dimension of the
optimization problem, t the current instant (iteration) and s is the swarm size.
The original formula developed by Kennedy and Eberhart was improved by Shi and
Eberhart [97, 98] with the introduction of an inertia parameter w to prevent premature
convergence and provides balance between local and global exploration. Although
experimentation with the inertia weight is still in progress, it appears that a good general
approach is to decrease the inertia weight linearly from 0.9 to 0.4 over 1,000 iterations. The
equation for w is as given below:
w = wmax – (wmax - wmin) x iter
itermax (3.3)
Where
wmax is the initial weight = 0.9
wmin is the final weight = 0.4
itermax is the maximum number of iterations
iter is the current number of iteration
Maurine Clerc [99] introduced a constriction factor χ that improves the ability of PSO and
developed the following update rule.
The complete PSO formula is:
( ) ( ) ( ) ( )( ) ( ) ( )( )( )txtytrctxtytrctvtv ijijijijijij −+−+=+ ˆ)()(1 2211χ (3.4)
Where
ϕϕϕχ
42
22 −−−
= (3.5)
30
4,21 >+= ϕϕ cc (3.6)
Bergh showed that a dangerous condition arises when a particle position approaches closer to
the global best position i.e. yyx ii ˆ== , the particle now will only depends on the term ( )twvij .
Bergh modified the equation of standard PSO by the guaranteed convergence PSO (GCPSO).
The new equation used is represented by the global best particle as follows:
( ) ( ) ( ) ( ) ( ) ( )( )trttwvtytxtv jjjj 21ˆ1 ,,, −+++−=+ ρτττ (3.7)
Where τ is the index of global best particle so that yy ˆ=τ . The term ( )1, +− tx jτ reset the
particle position so that the particle only depends on jy . The term ( ) ( )( )trt 21−ρ generates
random sample around jy for searching better value of jy . The position update equation for the
global best particle is given by
( ) ( ) ( ) ( ) ( )( )trttwvtytx jjj 21ˆ1 ,, −++=+ ρττ (3.8)
The value of ( )tρ is changed after each iteration using the following rule
( )( )( )
( )⎪⎩
⎪⎨
⎧>>
=+otherwiseif5.0if2
1t
fNtsNt
t cF
cS
ρρ
ρρ (3.9)
Where NS number of successes denotes the number of improvements in the y and NF denotes
the number of failures to improve the y . A single failure is given by ( ) ( )1ˆˆ −= tyty cs and cf are
upper threshold values. The following two rules are required for implementation: ( ) ( ) ( ) 011 =+⇒>+ tNtNtN FFS ( ) ( ) ( ) 011 =+⇒>+ tNtNtN SFF
The optimal choice for the values of cs and cf depends upon the objective function. In high
dimension search spaces it is difficult to obtain better values using random search in only a few
iterations, so it is recommended to set sc = 15, fc = 5.
3.3 Artificial Neural Networks (ANN)
Artificial Neural Networks (ANN) in general is massively interconnected network of a large
number of processing elements called neurons in an architecture, inspired by the brain. ANN
exhibits characteristics such as mapping or pattern recognition, generalization, and fault
tolerance. Neural network learn from examples and various learning laws exist of which
31
supervised and unsupervised are popular. For a particular application the neural network is
defined by its architecture and learning rule.
3.3.1 Feedforward Neural Network
Back propagation is a systematic method of training multilayer feedforward artificial
neural networks. It has been built on high mathematical foundation and has very good
application potential. However, it has limitations. BP algorithm works well on small set of data
and simple networks with few neurons. But when the problem under consideration has several
parameters and many hidden layers, it becomes very difficult for backward propagation to
minimize error. The ANN models using back propagation algorithm for training does not ensure
convergence and hangs in local optima and requires much longer training time. The neural
network can be trained based on swarm intelligence learning rule [100].
A neural network consists mainly of three layers: input layer, hidden layer and output
layer. These layers are arranged in some way to have a multilayer feed-forward structure. A
general network model consists of simple processing elements called neurons with adjustable
parameters called weights. These neurons are arranged in a distinct layered topology and perform
a biased weighted sum of their inputs and pass this activation level through a transfer relation
(sigmoid function) to produce their output. Thus, the parameters of data flow from the input
neurons, forwards through many hidden neurons, eventually reaching the output neurons. All of
the layers are fully interconnected with each other by weights as shown in Fig. 3.1.
Figure 3.1 Multilayer feedforward Neural Network Architecture
Neural network learn by examples without using any programmed rules. The ability to learn
through training is the most important feature of ANN. A supervised learning is given in Fig. 3.2.
32
A neural network uses a learning function to modify the variable connection weights at the
inputs of each processing element according to some neural based algorithm. Multiple layers of
neurons with nonlinear activation functions allow the network to learn nonlinear and linear
relationships between the input and output of the network. The training process requires a set of
examples of proper network behavior. The back propagation searches on the error surface by
means of gradient decent technique in order to minimize error. The performance function of the
neural network is normally chosen to be the mean squared error for each pattern on the training
set,
MSE = 1/ P ∑=
P
i 1(tPi – opi) 2 (3.10)
Where tpi is the target at ith pattern, opi is a network’s output at ith pattern and P is number of
neural network pattern.
Figure.3.2 supervised learning
3.4 Dynamic Programming (DP) or Recursive Optimization
Dynamic programming is a mathematical technique that is applicable to many types of
problem. Dynamic programming is a recursive optimization approach to solving sequential
decision problems. By recursive optimization procedure, we mean one that optimizes on a
step-by-step basis using information from the preceding steps. In short, we “optimize as we go.”
The dynamic programming approach involves the optimization of multistage decision processes.
That is, it basically divides a given problem into stages or sub-problems and then solves the sub-
problems sequentially until the invited problem is finally solved. In some mathematical
programming algorithms, optimization was also achieved on a step-by-step basis, but it was
33
iterative rather than recursive; that is, each step represented a unique solution that was non-
optimal. In dynamic programming, a single step is sequentially related to preceding steps and is
not itself a solution to the problem. A single step contains information that identifies a segment
of the optimal solution.
Because of these features, dynamic programming is most often applied to problems
requiring a sequence of interrelated decisions. Many time-dependent (dynamic) processes are
characterized by sequential decision problems that need to be solved; hence the term dynamic
programming (DP). Other applications particularly well-suited to dynamic programming
involve interrelationships rather than time dependencies, although time dependencies are a
common basis for expressing interrelationships among variables. A more appropriate term for
dynamic programming, therefore, might be recursive optimization.
The fundamental approach of dynamic programming involves (1) the breaking down of a
multistage problem into its subparts, steps, or single stages, a process called decomposition; (2)
making decisions one at a time, or recursively, at each stage, according to a specific optimization
objective; and (3) combining the results at each stage to solve the entire problem, a process
called composition. The act of composition results in a set of sequential decision rules called a
policy. For example, dynamic programming would optimize an n-decision variable problem by
decomposing it into a series of n stages (each decision variable a stage), assigning an optimal
value to each variable, and combining the results from each stage to generate the overall solution
to the problem.
At each stage, the decision rule is determined by evaluating a criterion or objective
function, called the recursive equation or functional equation (functional because it is a
function yielding a single real number).
3.4.1 Forward Dynamic Programming Approach
The dynamic programming algorithm can be run back ward in time starting from the final
hour to be studied back to the initial hour. Conversely, the algorithm can be run forward in time
from the initial hour to the final hour. Dynamic programming sub-divide the 24-h day into
34
discrete intervals [101]. The unit commitment procedure then searches for the most economic
feasible combination of generating units to serve the forecast load and spinning reserve
requirement of the system at each interval of the load cycle. In unit commitment problem the DP
is based on enumeration scheme and priority list methods.
For example, if we have four units, N = 4, then 15 possible combinations for each interval
are
Combinations or states (xi) Units x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14 x15
1 1 1 1 1 0 0 1 0 1 1 0 1 0 0 0 2 1 1 1 0 1 1 0 0 1 0 1 0 1 0 0 3 1 1 0 1 1 0 0 1 0 1 1 0 0 1 0 4 1 0 1 1 1 1 1 1 0 0 0 0 0 0 1
Where xi denotes combinations or states i of the four units. Of course, not all combinations
are feasible because of the constraints imposed by the load level and other practical operating
requirements of the system.
3.4.2 Mathematical Formulation of the Dynamic Programming for unit commitment
problem
Let us set
xi (h) = combination xi of interval h (3.11)
and then
xj (h + 1) represents combination xj of interval (h + 1) (3.12)
The production cost incurred in supplying power over any interval of the daily load cycle
depends on which combination of units is on-line during that interval. For a given combination
xi, the minimum production cost Pi equals the sum of the economic dispatch costs of the
individual units. Accordingly, we designate
Pi (h) = minimum production cost of combination xi (h) (3.13)
and then
Pj (h+ 1) is the minimum production cost of combination xj (h + 1) (3.14)
35
Besides production cost, in the unit commitment problem the transition cost, is also
considered which is the cost associated with changing from one combination of power-producing
units to another combination. Thus, transition cost associated with changing from one
combination of operating units to another will be denoted by
Tjj (h) = cost of transition from combination xi (h) to combination xj (h + 1) between
intervals h and (h + 1) (3.15)
Accordingly, the problem of minimizing costs at one stage is tied to the combinations of
units chosen for all the other stages, thus the unit commitment is a multistage or dynamic cost-
minimization problem.
The multistage decision process of the UCP can be dimensionally reduced by practical
constraints of system operations and by a search procedure based on the following observations:
• The daily schedule has h discrete time intervals or stages, the durations of which are not
necessarily equal. Stage 1 precedes stage 2, and so on to the final stage H.
• A decision must be made for each stage h regarding which particular combination of
units to operate during that stage. This is the stage h sub problem.
• To solve for the H decisions, h sub problems are solved sequentially in such a way that
the combined best decisions for the h sub problems yield the best overall solution for the
original problem.
This strategy greatly reduces the amount of computation to solve the original unit
commitment problem.
The cost fij (h) associated with any stage h has two components given by
fij (h) = Pi (h) + Tij (h) (3.16)
Which are the combined transition and production cost incurred by combination xi during
interval h plus the transition cost to combination xj of the next interval? The cost fij(h) is tied by
Tij (h) to the decision of the next hour (h + 1). The best policy or set of decisions (in the sense of
minimum cost) over the first (h - 1) hours of the daily load cycle; this is equivalent to assuming
36
that we know how to choose the best combination of units for each of the first (h — 1) intervals.
If we agree that combination xi* is the best combination for stage (h - 1), then by searching
among all the feasible combinations xi of the final stage h, we can find
Fi* (h – 1) = )}({
minhx j
{Pi*(h – 1) + Ti*j(h – 1) + Fi(h)} (3.17)
Where Fi* (h — 1) is the minimum cumulative cost of the final two stages starting with
combination xi* (h - 1) and ending with combination xj ((h); the cumulative cost Fj(h) of stage h
equals the production cost Pj (h) since there is no further transition cost involved.
Similarly, starting with the combination xi* (h — 2) at interval (h - 2), the minimum
cumulative cost of the final three stages of the study period is given by
Fi* (h – 2) = )}1({
min−hx j
{Pi* (h - 2) + Ti* (h – 2) + Fj (h – 1} (3.18)
Where the search is now made among the feasible combinations xj of stage (h — 1),
continuing the above logic, we find the recursive formula
Fi* (h) = )}1({
min+hx j
{Pi* (h) + Ti*j (h) + Fj (h + 1)} (3.19)
for the minimum cumulative cost at stage h, where h ranges from 1 to H.
At each stage in the dynamic programming solution of the unit commitment problem the
economic dispatch outputs of the available generating units must be calculated before the
evaluation of the production costs Pi*(h).
3.5 Genetic Algorithm (GA)
Genetic algorithms are optimization and search algorithms based on the mechanics of
natural selection, genetics and evolution. A GA is a random search procedure, which is based on
the survival of the fittest theory. Genetic Algorithms were first proposed by John Holland and
more recently reviewed and enhanced by one of his student David Gold berg [102] and many
37
others. Their basic principle is the maintenance of a population of solutions to a problem in the
form of encoded individuals that evolve in time.
GAs manipulates strings of binary digits (‘1’ and ‘0’) called chromosomes, which represents
multiple points in the search space. Each bit in a string is called an allele. They carry out
simulated evolution on populations of chromosomes. Genetic algorithms, using simple
manipulations of chromosomes such as simple encoding and reproduction mechanisms, can
display complicated behavior and solve some extremely difficult problems without knowledge of
the decoded world.
The mechanics of a simple genetic algorithm is based on three operators’ reproduction,
crossover and mutation.
a. Reproduction: In this process individual strings are copied according to their fitness
value. This process is conducted by spinning a simulated biased roulette
wheel, which is called roulette wheel parent selection.
b. Crossover: The main operator working on the parents is crossover, which happens for
a selected pair with a crossover probability. At first, two strings from the
reproduced population are mated at random and a crossover point is
selected randomly. At the crossover point, the partial strings are
interchanged between the selected pair to produce two new strings.
c. Mutation: Although first two operators produce many new strings, they do not
introduce any new information in to the population at the bit level. As a
source of new bits, mutation is introduced and is applied with a low
probability. It inverts a randomly chosen bit on a string.
These three operators are applied repeatedly until the offspring’s take over the entire
population. When new solution of strings is produced, they are considered as a new generation
and they totally replace the ‘parents’ in order for the evolution to proceed. It is necessary to
produce many generations for the population converging to the near optimum or an optimum
solution, the number increasing according to the problem complexity.
38
This approach consists of the following steps:
Step1: Read system data and select GA parameters (population size, probability of cross
over, probability of mutation).
Step 2: Randomly generate a population of chromosomes, each consisting of bits ‘0’ and
‘1’. Which is a two dimensional array (H x N) Where H = Load duration in hours,
N = Number of units.
Step 3: Obtain a population of P strings of feasible solutions or chromosomes of
population that could satisfy the load demand and spinning reserve.
Step 4: Get the row value and check for the feasibility of the solution corresponding to
the satisfaction of the equality constraint (Generation ≥ Load + spinning reserve)
for each hour of the load duration ‘H’.
Step 5: Decode chromosomes and perform economic dispatch prior to genetic operation.
Step 6: Evaluate the fitness function (objective function).
Step 7: If terminating criteria satisfied GO TO Step 10.
Step 8: else
{Perform Genetic Operators on the population of the strings i.e. Selection,
Crossover and Mutation. Make a new population}
Step 9: GO TO Step 4.
Step 10: Exit
Step 11: Print the best array, the economic loading and the operating fuel cost for each
load.
39
CHAPTER 4
Unit Commitment --- Problem Formulation and Single
Solution Approaches
4.1 Introduction
The Unit Commitment (UC) is an important research challenge and vital optimization
task in the daily operational planning of modern power systems due to its combinatorial nature.
Because the total load of the power system varies throughout the day and reaches a different
peak value from one day to another, the electric utility has to decide in advance which generators
to start up and when to connect them to the network and the sequence in which the operating
units should be shut down and for how long. The computational procedure for making such
decisions is called unit commitment, and a unit when scheduled for connection to the system is
said to be committed. In this work the commitment of fossil-fuel units has been considered which
have different production costs because of their dissimilar efficiencies, designs, and fuel types.
Unit commitment plans for the best set of units to be available to supply the predict forecast load
of the system over a future time period.
In general, the UC problem may be formulated as a non-linear, large scale, mixed-integer
combinatorial optimization problem with both binary (unit status variable) and continuous (unit
output power) variables. This chapter presents the characteristics of power generation unit, unit
commitment problem formulation, modeling aspects of single approaches to solve UCP for
convex and non convex fuel cost function. The remaining discussions in this chapter focus on
algorithm development and their implementation, and case studies.
4.2 Characteristics of Power Generation Units
In analyzing the problems associated with the operation of power system, there are many
possible parameters of interest. Fundamental is the basic cost data and set of input-output
40
characteristics of generation units. Different types of fuel are used in thermal power plants.
Depending on the types of turbine such as single value or multi value, the characteristic differs.
Although the operating cost of these units consists of both fuel and maintenance costs,
only the fuel cost varies directly with the units and also with the level of generation. The fuel
cost is incurred during the running (either at no-load or at any load), start-up and sometimes
shutdown conditions of the steam units.
4.2.1 Unit’s Input-Output characteristic (Heat or Cost)
Unit (Boiler, turbine and generator) input-output curve establishes the relationship
between energy input to the driving system and the net energy output from the generator. A
typical boiler-turbine-generator unit is represented in Figure 4.1.
Figure 4.1 Boiler -Turbine -Generators Unit
In this characteristic the gross input (Rs. / h or tons of coal/h or millions of cubic feet of
gas/h or any other unit) being measured in millions of B.T.U. per hour (MBTU/h) is plotted
against the output in MW of the unit. The input is taken along y-axis. The output is normally the
net electrical output of the unit and is taken along x-axis. Z-axis represent the time axis, on
which usually one hour is taken to convert the output power P in MW to energy in MWh in order
to evaluate the per unit cost of input i.e., Plant is loaded at P (MW) for one hour, then input is
measured in Rs. / h or MBTU/h. (Z-axis can be omitted as each point loading pertains one hour).
B T
G
A/P
Boiler fuel input Steam Turbine
Generator
(Gross)
(Net)
Auxiliary Power supply
41
The input-output characteristics of a steam unit in idealized form are represented in Figure
4.2.
Figure 4.2 Input-output Curve of a Steam Turbine Generator
For a single value turbine the governing is done by throttling of steam and for such units,
the input-output curve is substantially a straight line within its operating range.
4.2.2 Non convex fuel cost characteristic due to valve point effect
Non convex characteristic results due to valve point effect, multiple fuels and prohibited
operating zones. The valve point effects produce a ripple, which is highly non-smooth and
discontinuous as represented in Figure 4.3.
Figure 4.3 Input-output curve of a multi valve steam turbine generator with four steam
admission valves
A= primary valve B= secondary valve C= Tertiary valve D=Quaternary valve
E= Quinary valve
4.2.3 Incremental heat or cost characteristic
The incremental heat rate characteristic is the derivative of the input–output characteristic
(∆H/∆P or ∆F/∆P). This characteristic is widely used in economic dispatching of units. It is
converted to an incremental fuel cost characteristic by multiplying the incremental heat rate. The
42
incremental heat rate characteristics for single and multi value units are represented in Figures
4.3 and 4.4 respectively. The incremental heat rate characteristic of multi valve steam turbine is
discontinuous type.
Figure 4.4 Incremental Heat Rate or Cost characteristic.
Figure 4.5 Incremental Heat Rate Characteristics of a steam turbine with four valves.
4.2.4 Unit Heat rate (HR) characteristic
The heat rate curve is obtained from the unit input-output curve by dividing the input by the
corresponding output (H/P) at any loading condition versus the megawatt output of the unit. This
characteristic is plotted between H/P versus P. While incremental heat rate is given by the ratio
of the change in input (∆H) to the corresponding change in output (∆P) at any operating point.
Heat rate (HR) = MW in put
MBTU/hin input Out
(4.1)
The units of Hear rate are KWhBtuorMWhorMBtuMW
hRs /,,//.
and
43
Incremental heat rate = ∆Input = ∆H or ∆F ∆Output ∆P ∆P (4.2)
This is an important characteristic and defines the average heat rate per KWH of output. The
incremental efficiency, which is the ratio of the change in output to the corresponding change in
input at any loading condition, is clearly the reciprocal of the incremental heat rate. Thus lower is
the incremental heat rate, higher is the incremental efficiency. Since H/P = 1/η, therefore this
characteristic if the reciprocal of the usual efficiency characteristic developed for a machine.
Both these quantities have the same unit which is B.T.U. per KWh. The heat rate and
incremental heat rate curves are represented in Figure 4.6.
Figure 4.6 Heat rate and incremental heat rate curves for convex cost function
The heat rate and incremental heat rate can be converted into fuel cost function and incremental
fuel cost by multiplying them with the cost of the fuel (Dollars per million of B.T.U.).
Fi(Pi) = Hi(Pi) x fuel cost $/h (4.3)
Incremental cost = λdPdF
i
T = $/MWh (4.4)
4.3 Unit Commitment Problem (UCP)
UCP was defined as preparing on/off schedule of generating units in order to minimize the
total production cost of utility and constraints such as system power balance, system spinning
reserve, and unit’s minimum up and down times are satisfied. Figure 4.7 represents the
configuration that represent UC problem with on/off switches. The unit commitment problem is
discussed as follows:
44
Figure 4.7 Unit Commitment
4.3.1 Objective Function
The principal objective is to prepare on/off schedule of the generating units in every sub-
period (typically 1h) of the given planning period (typically 1 day or 1 week) in order to serve
the load demand and spinning reserve at minimum total production cost (fuel cost, start up cost,
shut down cost), while meeting all unit, and system constraints. The following costs are
considered.
4.3.1.1 Fuel Cost
The quadratic approximation is the most widely used by the researchers, which is
basically a convex shaped function. The curve shown in Fig. 4.1 is the operating fuel cost
equation for unit i and is mathematically represented as:
Fi(Pih) = ∑=
N
i 1
[ ai + bi Pih + ci Pih
2 ] (4.5)
(Units without valve point effects)
To take the effects of valve points as shown in Fig. 4.6 a sinusoidal function is added to the
convex cost function and represented as:
Fi(Pih) = ∑=
N
i 1[ ai + bi Pih
+ ci Pih2 + | ei sin fi ( Pih min - Pih )|] (4.6)
(Units with valve point effects)
4.3.1.2 Start up cost
Start up cost is warmth-dependent. Mathematically it is represented as a step function:
45
STih = h-costi: Ti down ≤Xi off (h) ≤ Ti down + c-s-houri $/h (4.7)
c-costi: Xi off (h) > Ti down + c-s-houri $/h (4.8)
4.3.1.3 Shut down cost
The typical value of the shut down cost is zero in the standard systems. This cost is
considered as a fixed cost.
SDih = KPih $/h (4.9)
Where K is the incremental shut-down cost
4.4 Constraints
The UCP is subjected to many constraints that include:
• The total power generated must meet the load demand.
• There must be enough spinning reserve to cover any shortfalls in generation.
• The loading of each unit must be within its minimum and maximum allowable rating
(limits).
• The minimum up and down times of each unit must be observed.
• Unit availability constraint is either unit is available / not available, out aged/Must out,
Must run, and Fixed Output Power (F.O.P).
• Unit initial status +/- either already up or already down.
The constraints which are taken into consideration in this work may be classified into two
groups: system constraints and unit constraints.
4.4.1 System constraints or coupling constraints
Constraints that concern all the units of the system are called system or coupling constraints.
These constraints have two categories: the system power balance and system spinning reserve
constraints.
(i) System Power balance or load constraint:
The system power balance constraint is the most important constraint in the UCP. The
generated power from all committed units must be equal to the load demand. This is formulated
in the balance equation as:
46
∑=
N
i 1Uih(Pih) = Dh h = 1,2,…….H (4.10)
(ii) System Spinning reserve requirements
In this work spinning reserve is computed as an amount which is a percentage of the
forecasted load demand and is represented as:
∑=
N
i 1
Uih(Pimax) ≥ Dh + SRh h = 1,2,…H (4.11)
4.4.2 Unit constraints or local constraints
Constraints that concern individual units are called unit constraints or local constraints are
described as follow:
(i) Units minimum and maximum generation limits
The generation limits represent the minimum loading limit below which it is not economical
to load the unit, and the maximum loading limit above which the unit should not be loaded. Each
unit has generation range, which is represented as:
Uih Pimin ≤ Pih ≤ Pimax Uih (4.12)
for i = 1,2,…..N, h = 1,2,….,H
(ii) Minimum up and down time limits
Once the unit is running, it should not be turned off immediately. Once the unit is de-committed, there is a minimum time before it can be recommitted. These constraints can be represented as:
Xi on (h) ≥ Tiup
Xi off (h) ≥ Tidown (4.13)
For i = 1, 2, …., N. h = 1, 2, … , H.
(iii) Unit availability constraints.
The availability constraint specifies the unit to be in one of the following different
situations: Available/ not available, out aged/Must out, Must run, Fixed Output (F.O.P).
Must run
47
Some units are given a must run status during certain times of the year for reasons of voltage
support on the transmission network or for such purposes as supply of steam for uses outside the
steam plant itself, and to increase the reliability or stability of the system.
Must out units:
Units which are on forced outages and maintenance are unavailable for commitment and these
are the must out units.
Units on fixed generation: (F.O.P)
These are the units which have been prescheduled and have their generation specified for certain
time period. A unit of fixed generation is automatically a must run unit for the designated time
period.
(iv) Unit initial status
The initial status value, if it is positive indicates the number of hours the unit is already up,
and if it is negative indicates the number of hours the unit has been already down. The status of
the unit +/- before the first hour in the schedule is an important factor to determine whether its
new state violates the MUT/MDT constraints. The initial status also affects the start up cost
calculations.
(v) Unit derating constraint
During the life time of a unit its performance could be change due to aging and cause
derating of the unit. Therefore, the unit minimum and maximum limits are changed.
4.5 Unit Commitment mathematical formulation as an optimization problem
The objective function of the unit commitment problem is to minimize the total
production cost and is mathematically formulated as:
Objective function
Minimize TPC = ∑=
H
h 1∑=
N
i 1[ Fi (Pih) + STih + SDih ] $/h (4.14)
Subject to:
• The system constraints (4.10, 4.11) and unit constraints (4.12, 4.13).
The UCP can be considered as two linked optimization problems: the Unit Scheduling Problem
(Allocation of Generators) and the Economic dispatch problem (Allocation of Generation) and
is represented in Figure 4.8. The unit schedule problem is the on/off or 0/1 combinatorial
optimization problem. A feasible unit schedule must satisfy the forecasted load demand, system
48
spinning reserve requirements, and the constraints on the start up and shut down times during
each planning period. The economic dispatch problem is the constrained non-linear optimization
problem. The economic load dispatch is to allocate the generation requirement among the
available units so that the total cost of energy supplied to meet the load demand within
recognized constraints is minimized on minute to minute basis.
49
Figure 4.8 Representation of Unit Commitment Problem (UCP)
50
4.6 Generation of initial feasible unit commitment schedules
The generation of initial feasible solution is much important, for the Unit Commitment
problem. When initial feasible schedules (generation > load + spinning reserve) are generated
randomly, it is difficult to get feasible schedules for 24 hours loads considering MUT and MDT
constraints It takes a very long time. These randomly generated solutions are also far from the
optimal solution. The convergence is slow and likely to get trapped in the local minimum during
the exploration of the solution space.
4.6.1 Initial unit commitment scheduling by using priority List method and focusing on
peak and off-peak loads of the daily load curve.
In this work initial solution is generated using the priority list method. The priority list
method is very fast and efficient method but the solution obtained cannot fulfill all the
constraints particularly the MUT and MDT constraints. Graphical representation is given in
Figure 4.9. Generally more generators are started up at around the peak load, and few units are
started up at light loads based on full load average production cost. To satisfy minimum up time
constraint the units are set continuously ON.
The full load average production cost (FLAPC) is the Heat Rate (HR) multiplied by the fuel
cost Fi(Pi).
Mathematically it is represented as:
(FLAPC) = HRi x Fi(Pi) = Fi(Pimax)/ (Pimax) (4.15)
(MBTU / MWh x $/ MBTU) = ($/MWh)
= ( ai + bi Pimax + ci Pimax
2 ) / (Pimax) (4.16)
Start up of the base units
In power system some units have “must run” status. These are base units having large output
power. Two units located at the bottom of the priority list as base load units, the units have “ON”
fixed status.
51
Unit10 ▓
Unit9 ▓ ▓
Unit8 ▓ ▓ ▓ ▓ ▓
Unit7 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit6 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit3 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit4 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit5 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit2 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit1 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Priority List ↑
1 2 3 4 5 6 7 8 9 10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 4.9 Generation of initial solution by priority list method a graphical representation
4.6.2 Generation of trial solutions / neighbors
At first step in solving the combinatorial optimization problems is to have good
neighbors/trial solutions from an existing solution. More solutions were obtained taking upper
four units in the priority list at every time interval as shown in Figure 4.10. Introduction of these
feasible solutions makes the search closer to the optimum, leading to a faster convergence and
better results.
Unit10 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit9 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit8 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit7 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ � Unit6 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit3 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit4 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit5 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ Unit2 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Unit1 ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓ ▓
Priority List ↑
1 2 3 4 5 6 7 8 9 10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
▓ either on/off sate ▓ On state Off state Figure 4.10 Generation of new schedule by taking upper 4 units
4.7 Minimum up and minimum down Time Constraint Handling
During the optimization process of unit commitment schedules the MUT and MDT
constraints may be violated. They will be checked and repaired if violation occurs.
52
4.8 Minimum up and down time constraint repairing by using bit change operator
A bit change operator is used to modify the bit positions. This operator overcomes the
problems of the minimum up/down time constraint violation. The operator looks at the past
states, the future states and the present state itself of all units to decide whether or not the unit’s
status for the present hour should be flipped or not. The units with small minimum up and down
times have more changes in their status, while the units with large minimum up and down times
will require less change. A simple way to achieve this is by the categorization of units as base
load, sub-base load, peaking units, must run, and can run. An example of repairing the minimum
up and down time is represented in Figures 4.11 and 4.12 respectively.
Units/Hr t-1 t t+1 i 0 0 1 1 0 0 0 j 0 0 1 1 1 0 0
Figure 4.11 Repairing of minimum up time
Units/Hr
t-1 t t+1 i 1 1 0 0 1 1 1 j 1 1 1 1 1 1 1
Figure 4.12 Repairing of minimum down time
4.9 Algorithm for the construction of initial unit commitment schedule and M.U.T and
M.D.T constraint handling
Step 1: Generate a feasible unit commitment schedule satisfying load demand and spinning
reserve using priority list method. Which is a matrix (H x N).
Step 2: Get the row values of the matrix and calculate the total generation in each scheduling
hour.
Step 3: Check that the generated power is greater than load demand plus spinning reserve?
Step 4: Get the column values of the matrix and calculate units start up and shut down times.
(Xi on h, Xi off h)
Step 5: Xi on h ≥ Tiup and Xi off h ≥ Ti
down if NO repair minimum up and down time violations
using bit change operator and modify start up cost else go to step 6.
53
Step 6: Get the row values of the matrix and calculate power output of each unit, operating fuel
cost of each unit, total operating fuel cost of each row, start up cost and total production
cost of each row for each hour.
Step 7: The final UC solution is one having the lowest total production cost.
4.10 Unit commitment schedule and determination of number of units to be operated
Consider a system having a forecasted load as shown in Fig. 4.9 Assuming that 10 units
are available to carry the load.
.
Figure 4.13 Forecasted load curve
An initial unit commitment schedule may be as represented in Figure 4.14.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 units Priority 10 Unit10 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 9 Unit9 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 8 Unit8 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 1 0 0 0 0 7 Unit7 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 1 1 0 0 0 0 6 Unit6 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 1 1 0 0 0 0 5 Unit3 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 4 Unit4 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 3 Unit5 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 2 Unit2 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 Unit1 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 on state 0 off state Figure 4.14 Representation of initial unit commitment schedule
Withdrawal of the unit contributes to the saving of its running cost during the reduced load
period of the system, but either start up cost or shut down cost is incurred when the unit is
restored to service. If these costs are less than the spinning cost, the withdrawal is economically
justified. In fact, other schedules can also be considered, but before the shut down of any unit,
the cost of its continued operation should be weighed against the shut down or start up cost. The
54
problem is thus based on the evaluation of the total production cost in supplying the forecasted
load with a specified spinning reserve with various schedules of commitment. 4.11 Economic dispatch Problem (EDP, Allocation of Generation)
The economic dispatch problem can be classified as convex and non convex. Most of the
researches used convex economic dispatch the non-convex economic dispatch may result due to
valve point effect.
4.11.1 Economic load dispatch (ELD) calculations
In the unit commitment problem, the economic dispatch calculations consume a large
amount of calculation time. In this work, the ELD calculation is performed only for feasible
solutions by gradient method, merit order method, load assigned by operational engineer and
genetic algorithm.
4.11.1.1 Equal Incremental Cost Criterion
Let there be N thermal generating units connected to a bus-bar supplying a load demand of
Pload. It is required to load the units so that the cost of operating fuel in minimum. This problem
may be solved by using “Equal Incremental Cost Criterion”. Working philosophy of their
criterion is as: “When the incremental costs of all the machines are equal, and then cost of
generation would be minimum subject to equality constraints”.
The economic dispatch problem mathematically may be defined as:
Minimize: ( )∑=
=N
1iiiT PFF (4.17)
Subject to: equality constraint ∑=
−=N
1iiD PPΦ (4.18)
And inequality constraint ( )
( ) ⎪⎭
⎪⎬⎫
≤−
≤−
0PP0PP
imin.i
max.ii (4.19)
for i =1,2,…,N.
Where
FT = F1 + F2 + ……FN is the total fuel input to the system
Fi = Fuel input to ith unit
55
Pi = the real power generation of ith unit
However, in this analysis method of Lagrange multiplier will be used. The working
philosophy of this method is that constrained problem can be converted into an unconstrained
problem by forming the Lagrange, or augmented function. Optimum is obtained by using
necessary conditions.
⎟⎠
⎞⎜⎝
⎛−+=
+=
∑=
N
1iiDT
T
PPλ.F
λ.ΦFL (4.20)
The necessary conditions for constrained local minima of L are the following:
0PL
i
=∂∂ (4.21)
0λL=
∂∂
(4.22)
Condition-I
First condition gives
( ) 010λ.PF
PL
i
T
i
=−+∂∂
=∂∂
or
λPF
0λPF
i
T
i
T =∂∂
⇒=−∂∂
Q N21T FFFF −−−−−−−++=
Then
λdPdF
PF
i
T
i
T ==∂∂
And therefore the condition for optimum dispatch is
56
λdpdF
i
T = (4.23)
or
λP2cb iii =+ (4.24)
Where 2iiiiiT PcPbaF ++=
Condition-II
Second condition results in
∑=
=−=∂∂ N
1iiD 0PP
λL
or
∑=
=N
1iDi PP (4.25)
Condition for economic operation
“For most economical operation, all plants must operate at equal incremental production cost
while satisfying the equality constraint given by equation (4.25).”
i
ii 2c
bλP
−= (4.26)
The relations given by equations (4.26) are known as the co-ordination equations. They are
function of λ. An analytical solution for λ is given by substituting the value of Pi in equation
(4.25), i.e.
D
N
1i i
i P2c
bλ=
−∑=
(4.27)
∑
∑
=
=
+= N
1i i
N
1i i
iD
2c1
2cbP
λ (4.28)
Optimal schedule of generation is obtained by substituting the value of λ from equation (4.28)
into equation (4.26).
57
Iterative Method of economic dispatch
The equation (4.27) is a function of λ and can be expressed as: ( ) DPλf = (4.29)
Expanding by Taylor’s series about an operating point λ (k) and neglecting higher order terms
results in
( )( ) ( ) ( )D
kk
k P∆λdλλdfλf =⎟⎠⎞
⎜⎝⎛+ (4.30)
( ) ( )( )
( ) ( )k
kDk
dλλdfλfP∆λ
⎟⎠⎞
⎜⎝⎛
−=
( )( )
( ) ( )k
kk
dλλdf
∆P∆λ
⎟⎠⎞
⎜⎝⎛
=
( )( )
( )k
kk
dλdPi∆P∆λ
⎟⎠⎞
⎜⎝⎛
= (4.31)
or
( )k∆λ( )
∑=
i
k
2c1
∆P (4.32)
and therefore ( ) ( ) ( )kk1k ∆λλλ +=+ (4.33)
Where
( ) ( )∑=
−=N
1i
kiD
k PP∆P (4.34)
The process is continued until ∆P (k) is less than a pre-specified accuracy.
4.11.1.2 Loading to most efficient load
Although the criterion of equal incremental production costs will result in the optimum
economic scheduling of generation. The above method is still in use by utilities. In this method
units are loaded in ascending order of their heat rates, to their most efficient loads, based on the
58
forecasted load at that hour. The most economical unit is loaded first and the remaining load is
shifted on the next unit in the priority list. It the remaining load is less the minimum limit of that
unit than the unit is loaded up to its minimum power output and the remaining loaded is shifted
on the previous unit. This is very quick method and gives near optimal solution in very short
time.
4.11.1.3 Economic Dispatch using Genetic Algorithm (Real Power – Search)
GA works better on non convex fuel cost function. The pseudo code for Real Power-
search method is shown in Figure 4.15.
59
Figure 4.15 Pseudo Code for Genetic Algorithm Real Power-Search Method
4.12 Economic Dispatch versus Unit Commitment
The unit commitment assumes that there are N units available to meet the forecasted load
demand, satisfying spinning reserve and MUT and MDT constraints. The economic dispatch
load the units economically within their limits satisfying system and unit constraints
60
4.13 Conventional/Classical Single Approaches for convex fuel cost function
The single approaches used to solve the UCP are, Complete Enumeration, Merit Order based
on Full Load Average Production Cost, and Merit order based on maximum power output of
each unit.
The algorithm of the classical single approaches is as follows:
Step 1: Read in system data ai, bi, ci, load demand (PD), Pmin and Pmax of each unit.
Step 2: Calculate the full load average production cost (FLAPC) cost of each unit.
Step 3-a: Generate all combinations (2N - 1). Check for the feasible combination
corresponding to the satisfaction of the equality constraint (Generation ≥ Load
demand plus spinning reserve) for each hour of the forecasted load.
b. Generate initial schedule based on Full Load Average Production Cost (FLAPC).
c. Generate initial schedule based on Maximum Power Output of each unit (PMAX).
Step 4: Calculate systems lambda using equation 4.28.
Step 5: Calculate economic loading of machines for each feasible combination (based on
equal incremental cost criteria).
Step 6: Check that the Power output of each unit is within minimum and maximum
generating limits of machines.
Step 7: Is there any violation of minimum or maximum power limit (yes-clamp at
minimum or maximum limit), (No - go to step 9).
Step 8: Recalculate the system lambda and output power of each machine using equations
4.32, 4.33 and 4.34.
Step 9: Satisfying the power balance equation (Generation = Load).
Step 10: Print the unit commitment schedule, power output of each machine, operating fuel
cost of each machine, and total production cost and system lambda for each
feasible schedule.
Step 11: Print the best schedule.
61
4.13.1 Single Approach – I --- Complete Enumeration
This method takes all the combinations (2N – 1) H and than calculated the economic
dispatch of each unit. Where N = number of units, H = number of hours (24). For a total period
of H interval, the maximum number of possible combinations is (2N-1) H.
For example, take a 24-h period (e.g. 24 one hour intervals) and consider with 5, 10, 20
and 40 units. The value of (2N-1) H becomes the following.
N (2N-1)H
5 6.25 x (10)35
10 1.73 x (10)72
20 3.12 x (10)144
40 Too Big
There very large numbers are the upper bounds for the numbers of enumerations
required. The constraints on the units and the loading capacity of the units limit the search space.
Never the less, the real practical barrier in the UCP is the high dimensionality of the possible
search space.
4.13.2 Single Approach – II --- Conventional Priority List
This approach is based on conventional priority list method. A priority order is created
based on the Full Load Average Production Cost (FLAPC). The UC schedule is based on
FLAPC and ED is based on Lambda Iteration Method.
In this method, units are committed to service by observing their heat rate values. Units with
the lowest heat rate are put into operation first. For shutting down the reserve order is followed,
i.e. the units with the highest heat rate is withdrawn first. The load dispatcher takes into account
the hourly forecasted load and spinning reserve requirement, and then schedules the units to
match the load and spinning reserve. The industry, however, still mostly used the simple "merit
order" method.
62
4.13.3 Single Approach – III --- Proposed Single Approach
This approach is also based on priority list method. But the priority list in this case is
based on the Maximum Power (PMAX) limit of each unit. The UC schedule is based on PMAX
and ED is based on Lambda Iteration Method.
4.14 Case studies --- Convex cost function
Following two standard test systems have been selected for the validation of the single
approaches.
a. Test System I --- 3 units with 24 hours load.
b. Test System III --- 10 units system with 24 hours load
Input Data: The description and input data of test systems used for investigation in the case
studies are given in Tables A.1 and A.3 placed in Appendix-A.
Computer Implementation: All the three single approaches have been implemented in C++ on
P-IV Personal Computer.
Output Results: The summary and comparison of results is given in this chapter.
4.14.1 Numerical Results of test system – I
The priority order based on conventional and proposed method is given in Table 4.1.
Table 4.1 Priority order based on single approach-II and III: test system-I
Unit No.
PMAX FLAPC (S/MWH)
Single approach-II Conventional Priority
order (FLAPC)
Single approach-III Proposed Priority order
(PMAX)
1 600 9.7922 2 1
2 400 9.4010 1 2
3 200 11.1888 3 3
The output results of test system – I are shown in the following tables:
Table 4.2 presents the comparison of operating fuel cost ($),
Table 4.3 presents the comparison of Summary of Unit Commitment Schedules for single
approaches,
Table 4.4 presents the Number of Units in Operation for single approaches,
63
Table 4.5 presents the Comparison of single approaches with Genetic Algorithm, and Hopfield
Neural Network methods.
Tables 4.6 to 4.8 presents the best solutions obtained by the single approach-I, II and III.
The Salient features of the conventional approaches in the light of the observations from the
results are as follows:
1. Enumeration method gives good results. But the number of transitions is 10. In test
system III the number of transitions is 4.
2. The operating fuel cost obtained by all the three single approaches remains low compared
to genetic algorithm and Hopfield neural network methods.
3. Single approach-I gives $606.28 saving per day compared with GA and $ 691.28 saving
per day compared with HFNN.
4. Single approach-II gives $381.15 saving per day compared with GA and $ 466.15 saving
per day compared with HFNN.
5. Single approach-III gives $389.28 saving per day compared with GA and $ 474.28 saving
per day compared with HFNN.
All the three approaches are simple, fast and give fair amount of reduction in operating cost as
compared to GA and HFNN.
Table 4.2 Comparison of the operating fuel cost ($) for 3 Unit Systems :( Test System I)
200800201000201200201400201600201800202000202200202400202600
Enu
mer
atio
n
Con
vent
iona
lpr
iorit
y lis
t
G.A
.
H.F
.N.N
Pro
pose
dap
proa
ch
Series1
64
Table 4.3 Summary of Unit Commitment Schedules for 3 Unit Systems (Test System I)
Single approach- I
Single approach- II
Proposed Single approach-III
Load (MW)
Complete Enumeration
Merit Order (FLAPC) with ED
UC(2,1,3)
Merit Order (PMAX) with ED
UC(1,2,3) 1200 1200 1150 1100 1000 900 800 600 550 500 500 500 500 500 600 800 850 900 950 1000 1050 1100 1200 1200
111 111 111 111 110 110 110 100 100 011 011 011 011 011 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 110 110 110 110 110 110 110 110 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
201415.789 202393.346 201632.089 No. of Transitions No. of Transitions No. of Transitions 10 2 4
65
Table 4.4 Number of Units in Operation for 3 Unit Systems (Test System I)
Single approach-I Single approach-II Proposed Single
approach-III
Complete Enumeration
Merit Order (FLAPC) with ED
UC(2,1,3)
Merit Order (PMAX) with ED
UC(1,2,3)
Hr. Load (MW)
No. of units No. of units No. of units
1 1200 3 3 3 2 1200 3 3 3 3 1150 3 3 3 4 1100 3 3 3 5 1000 2 2 2 6 900 2 2 2 7 800 2 2 2 8 600 1 2 1 9 550 1 2 1 10 500 2 2 1 11 500 2 2 1 12 500 2 2 1 13 500 2 2 1 14 500 2 2 1 15 600 1 2 1 16 800 2 2 2 17 850 2 2 2 18 900 2 2 2 19 950 2 2 2 20 1000 2 2 2 21 1050 3 3 3 22 1100 3 3 3 23 1200 3 3 3 24 1200 3 3 3
66
Table 4.5 Comparison of the Three Single approaches with Genetic Algorithm and Hopfield Neural Network Methods for 3 unit systems: Test system-I
Algorithm Daily
Operating Fuel Cost
($)
Amount of Daily Saving
(compared with GA)
% saving in fuel cost compared
with GA
Amount of Daily
Saving (compared
with Hopfield Neural
Network
% saving in fuel
cost compared
with Hopfield Neural
Network(HFNN)
Genetic Algorithm[109]
202021.360 - - - -
Hopfield Neural
Network Method[110]
202106.360
-
-
-
-
Single Approach-I
(Enumeration)
201415.089 606.28 0.300 691.28 0.342
Single Approach-II
(Conventional Priority List)
202393.346 -371.98 -0.184 -0.286.98 -0.142
Proposed Single
Approach-III
201632.089 389.28 0.1926 474.28 0.2346
67
Table 4.6 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from Single Approach-I (Enumeration): Test system-I
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 3 111 600.00 400.00 150.00 1150 5875.320 3760.400 1658.340 11294.060 4 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 100 600.00 0.00 0 600 5875.320 0.000 0 5875.320 9 100 550.00 0.00 0 550 5389.505 0.000 0 5389.505 10 011 0 400.00 100.00 500 0 3760.400 1107.840 4868.240 11 011 0 400.00 100.00 500 0 3760.400 1107.840 4868.240 12 011 0 400.00 100.00 500 0 3760.400 1107.840 4868.240 13 011 0 400.00 100.00 500 0 3760.400 1107.840 4868.240 14 011 0 400.00 100.00 500 0 3760.400 1107.840 4868.240 15 100 600.00 0 0 600 5875.320 0.000 0 5875.320 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 600.00 400.00 50.00 1050 5875.320 3760.400 586.260 10221.980 22 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480
SUM 201415.789
68
Table 4.7 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from the Single Approach-II (Conventional Priority List): Test system-I
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 3 111 600.00 400.00 150.00 1150 5875.320 3760.400 1658.340 11294.060 4 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 110 322.39 277.61 0 600 3276.676 2638.749 0 5915.425 9 110 294.69 255.31 0 550 3030.592 2440.639 0 5471.231 10 110 266.99 233.01 0 500 2786.906 2244.458 0 5031.364 11 110 266.99 233.01 0 500 2786.906 2244.458 0 5031.364 12 110 266.99 233.01 0 500 2786.906 2244.458 0 5031.364 13 110 266.99 233.01 0 500 2786.906 2244.458 0 5031.364 14 110 266.99 233.01 0 500 2786.906 2244.458 0 5031.364 15 110 322.39 277.61 0 600 3276.676 2638.749 0 5915.425 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 600.00 400.00 50.00 1050 5875.320 3760.400 586.260 10221.980 22 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480
SUM 202393.346
69
Table 4.8 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from the Proposed Single approach-III (PMAX): Test system-I
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 3 111 600.00 400.00 150.00 1150 5875.320 3760.400 1658.340 11294.060 4 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 100 600.00 0 0 600 5875.320 0 0 5875.320 9 100 550.00 0 0 550 5389.505 0 0 5389.505 10 100 500.00 0 0 500 4911.500 0 0 4911.500 11 100 500.00 0 0 500 4911.500 0 0 4911.500 12 100 500.00 0 0 500 4911.500 0 0 4911.500 13 100 500.00 0 0 500 4911.500 0 0 4911.500 14 100 500.00 0 0 500 4911.500 0 0 4911.500 15 100 600.00 0 0 600 5875.320 0 0 5875.320 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 600.00 400.00 50.00 1050 5875.320 3760.400 586.260 10221.980 22 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480
SUM 201632.089
70
4.14.2 Numerical results: Test system-III --- (10 unit system)
The case studies are conducted based on the following considerations:
• UC schedules without considering spinning reserve and without considering MUT and
MDT constraints.
• Without s. r. but with considering MUT and MDT constraint
• 10% s. r. without MUT and MDT constraint
• With 10 % s. r. without ED considering MUT and MDT constraint
• With 10 % s. r. with ED considering MUT and MDT constraint without transition cost.
• With 10% spinning reserve, minimum up/down time constraints, transition cost and
Economic Dispatch.
The priority order for ten unit system is given in Table 4.9.
Table 4.9 Priority order based on single approach-II and proposed single approach III: Test system-III --- ten unit system
Unit No.
PMAX FLAPC (S/MWH)
Single approach-II Conventional Priority order
(FLAPC)
Proposed Single approach-III
(PMAX)
Proposed hybrid approach
(PMAX-FLAPC)
1 455 18.6062 1 1 1
2 455 19.5329 2 2 2
3 130 22.2446 4 5 5
4 130 22.0051 3 3 4
5 162 23.1225 5 4 3
6 80 27.4546 6 7 7
7 85 33.4542 7 6 6
8 55 38.1472 8 8 8
9 55 39.4830 9 9 9
10 55 40.0670 10 10 10
The output results of test system–III are shown in the following tables:
Table 4.10 presents the comparison of the results for single approaches based upon above
considerations.
71
Table 4.11 presents the comparison of operating fuel cost ($) considering 10% spinning reserve
and minimum up/down time constraints with transition cost.
Table 4.12 presents the comparison of the results with Genetic Algorithm.
Table 4.13 presents the summary of Unit Commitment schedules for Single approaches-I, II and
III.
Tables 4.14-4.19 presents the Unit Commitment Schedules and Power Sharing (MW) , operating
fuel cost, start up cost and total production cost of the best solutions obtained from Single
approaches-I, II, and III, with MUT and MDT constraints with 10 % spinning reserve.
The Salient features of the conventional single approach-III in the light of the observations
from the results are as follows:
1. Proposed Single approach III give better results than GA. The proposed approach gives
cost saving of $949.39 per day equivalent to 0.167% compared with GA.
2. Proposed single approach III give better results the conventional priority list. The
proposed approach gives a cost saving of $1247.37 per day equivalent to 0.220%
compared with conventional priority list.
Table 4.10 Comparison of the operating fuel cost ($) for 10 unit systems considering 10% spinning reserve and minimum up/down time constraints without transition cost: Test System III
560200560400560600560800561000561200561400561600561800
Enu
mer
atio
n
Con
vent
iona
lpr
iorit
y lis
t
Pro
pose
dap
proa
ch
Series1
72
Table 4.11 Comparison of the operating fuel cost ($) for 10 unit systems considering 10% spinning reserve and minimum up/down time constraints with transition cost:
Test System III
Single Approaches
Single approach - I
(Enumeration) With ED, 10 % s. r
MUT, MDT
Single approach - II (Conventional
Priority List FLAPC) With ED, 10 % s. r
MUT, MDT
Proposed Single approach - III
(Priority List PMAX) With ED, 10 % s. r.
MUT, MDT
UC Schedule
UC12543768910
UC12435678910 UC12534768910
Operating fuel cost ($)
560744.47 561682.98 560775.6
Transition cost ($)
4090 4440 4100
Total Production
Cost ($)
564834.47
566122.98 564875.61
Table 4.12 Comparison of the results of the proposed single approach-III with Genetic Algorithm and conventional priority list
Approach/Model Total
Production Cost ($)
Amount of Daily
Saving compared with GA
($)
% Cost Saving
compared with GA
Amount of daily saving
compared with CPL
($)
% Cost Saving
compared with CPL
Genetic Algorithm 565825.00 - - - - Single Approach-I (Complete Enumeration)
564834.47 990.53 0.175 - -
Single Approach-II (Conventional Priority list, CPL)
566122.98 -297.98 -0.052 - -
Proposed Single Approach-III
564875.61 949.39 0.167 1247.37 0.220
73
Table 4.13 Summary of Unit Commitment schedules for 10 unit systems: Single approach-I, II and III: Test System III
Single
approach-I Single
approach-II Proposed Single
approach-III Hour Load
(MW) Enumeration with 10 % s.r with ED, MUT
and MDT constraint
12543768910
(FLAPC) with 10% s.r.
with ED, MUT and MDT constraint
12435678910
(PMAX) with 10% s.r. with ED, MUT and
MDT constraint
12534768910 UC Schedule UC Schedule UC Schedule 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
700 750 850 950
1000 1100 1150 1200 1300 1400 1450 1500 1400 1300 1200 1050 1000 1100 1200 1400 1300 1100 900 800
1100000000 1100000000 1100100000 1100100000 1101100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100100000 1100000000
1100000000 1100000000 1101000000 1111000000 1111000000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1101000000 1100000000
1100000000 1100000000 1100100000 1100100000 1110100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100100000 1100000000
Operating fuel cost ($)
564834.47 566122.98 564875.61
74
Table 4.14 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from Single approach-I (Enumeration) considering MUT and MDT constraints with 10
% spinning reserve: Test system-III
Power output of each unit (MW) Hour UC schedule 12543768910
Load (MW)
1 2 3 4 5 6 7 8 9 10 1 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 295 0 0 0 0 0 0 0 0 3 1100100000 850 455 370 0 0 25 0 0 0 0 0 4 1100100000 950 455 455 0 0 40 0 0 0 0 0 5 1101100000 1000 455 390 0 130 25 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 20 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 130 130 162 80 25 43 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 20 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 20 25 0 0 0 22 1111111000 1100 455 315 130 130 25 20 25 0 0 0 23 1100100000 900 455 420 0 0 25 0 0 0 0 0 24 1100000000 800 455 345 0 0 0 0 0 0 0 0
75
Table 4.15 Unit Commitment Schedule, fuel cost, start up cost and total production cost of the best solution obtained from Single approach-I (Enumeration) considering MUT and MDT
constraints with 10 % spinning reserve: Test system-III
Hour UC schedule 12543768910
Load (MW)
Fuel cost ($/h)
Startup Cost ($/h)
Total Production
Cost ($)
1 1100000000 700 13683.13 0 13683.13 2 1100000000 750 14554.50 0 14554.50 3 1100100000 850 16809.45 900 17709.45 4 1100100000 950 18597.67 0 18597.67 5 1101100000 1000 20020.02 560 20580.02 6 1111100000 1100 22387.04 1100 23487.04 7 1111100000 1150 23261.98 0 23261.98 8 1111100000 1200 24150.34 0 24150.34 9 1111111000 1300 27251.06 860 28111.06 10 1111111100 1400 30057.55 60 30117.55 11 1111111110 1450 31916.06 60 31976.06 12 1111111111 1500 33890.16 60 33950.16 13 1111111100 1400 30057.55 0 30057.55 14 1111111000 1300 27251.06 0 27251.06 15 1111100000 1200 24150.34 0 24150.34 16 1111100000 1050 21513.66 0 21513.66 17 1111100000 1000 20641.82 0 20641.82 18 1111100000 1100 22387.04 0 22387.04 19 1111100000 1200 24150.34 0 24150.34 20 1111111100 1400 30057.55 490 30547.55 21 1111111000 1300 27251.06 0 27251.06 22 1111111000 1100 23592.97 0 23592.97 23 1100100000 900 17684.69 0 17684.69 24 1100000000 800 15427.42 0 15427.42
SUM 21700 560744.47 4090.00 564834.47
76
Table 4.16 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from Single approach-II (FLAPC) considering MUT and MDT constraints with 10 %
spinning reserve: Test system-III
Power output of each unit (MW) Hour UC schedule 12435678910
Load (MW)
1 2 3 4 5 6 7 8 9 10 1 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 295 0 0 0 0 0 0 0 0 3 1101000000 850 455 265 0 130 0 0 0 0 0 0 4 1111000000 950 455 235 130 130 0 0 0 0 0 0 5 1111000000 1000 455 285 130 130 0 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 20 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 130 130 162 80 25 43 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 20 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 20 25 0 0 0 22 1111111000 1100 455 315 130 130 25 20 25 0 0 0 23 1101000000 900 455 315 0 130 0 0 0 0 0 0 24 1100000000 800 455 345 0 0 0 0 0 0 0 0
77
Table 4.17 Unit Commitment Schedule, fuel cost, startup cost and total production cost of the best solution obtained from Single approach-II (FLAPC) considering MUT and MDT constraints
with 10 % spinning reserve: Test system-III
Hour UC schedule Load (MW)
Fuel cost ($/h)
Startup Cost ($/h)
Total Production
Cost ($)
1 1100000000 700 13683.13 0.00 13683.13 2 1100000000 750 14554.50 0.00 14554.50 3 1101000000 850 16892.15 560.00 17452.15 4 1111000000 950 19261.50 550.00 19811.50 5 1111000000 1000 20132.56 0.00 20132.56 6 1111100000 1100 22387.04 1800.00 24187.04 7 1111100000 1150 23261.98 0.00 23261.98 8 1111100000 1200 24150.34 0.00 24150.34 9 1111111000 1300 27251.06 860.00 28111.06 10 1111111100 1400 30057.55 60.00 30117.55 11 1111111110 1450 31916.06 60.00 31976.06 12 1111111111 1500 33890.16 60.00 33950.16 13 1111111100 1400 30057.55 0.00 30057.55 14 1111111000 1300 27251.06 0.00 27251.06 15 1111100000 1200 24150.34 0.00 24150.34 16 1111100000 1050 21513.66 0.00 21513.66 17 1111100000 1000 20641.82 0.00 20641.82 18 1111100000 1100 22387.04 0.00 22387.04 19 1111100000 1200 24150.34 0.00 24150.34 20 1111111100 1400 30057.55 490.00 30547.55 21 1111111000 1300 27251.06 0.00 27251.06 22 1111111000 1100 23592.97 0.00 23592.97 23 1101000000 900 17764.14 0.00 17764.14 24 1100000000 800 15427.42 0.00 15427.42
SUM 561682.99 4440.00 566122.99
78
Table 4.18 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from Proposed Single approach-III (PMAX) considering MUT and MDT constraints
with 10 % spinning reserve: Test system-III
Power output of each unit (MW) Hour UC schedule 12534768910
Load (MW)
1 2 3 4 5 6 7 8 9 10 1 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 295 0 0 0 0 0 0 0 0 3 1100100000 850 455 370 0 0 25 0 0 0 0 0 4 1100100000 950 455 455 0 0 40 0 0 0 0 0 5 1110100000 1000 455 390 130 0 25 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 20 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 130 130 162 80 25 43 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 20 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 20 25 0 0 0 22 1111111000 1100 455 315 130 130 25 20 25 0 0 0 23 1100100000 900 455 420 0 0 25 0 0 0 0 0 24 1100000000 800 455 345 0 0 0 0 0 0 0 0
79
Table 4.19 Unit Commitment Schedule fuel cost, startup cost and total production cost of the best solution obtained from Single approach-III (PMAX) considering MUT and MDT constraints
with 10 % spinning reserve: Test system-III
Hour UC schedule Load (MW)
Fuel cost ($/h)
Startup Cost ($/h)
Total Production
Cost ($)
1 1100000000 700 13683.13 0 13683.13 2 1100000000 750 14554.50 0 14554.50 3 1100100000 850 16809.45 900 17709.45 4 1100100000 950 18597.67 0 18597.67 5 1110100000 1000 20051.16 550 20601.16 6 1111100000 1100 22387.04 1120 23507.04 7 1111100000 1150 23261.98 0 23261.98 8 1111100000 1200 24150.34 0 24150.34 9 1111111000 1300 27251.06 860 28111.06 10 1111111100 1400 30057.55 60 30117.55 11 1111111110 1450 31916.06 60 31976.06 12 1111111111 1500 33890.16 60 33950.16 13 1111111100 1400 30057.55 0 30057.55 14 1111111000 1300 27251.06 0 27251.06 15 1111100000 1200 24150.34 0 24150.34 16 1111100000 1050 21513.66 0 21513.66 17 1111100000 1000 20641.82 0 20641.82 18 1111100000 1100 22387.04 0 22387.04 19 1111100000 1200 24150.34 0 24150.34 20 1111111100 1400 30057.55 490 30547.55 21 1111111000 1300 27251.06 0 27251.06 22 1111111000 1100 23592.97 0 23592.97 23 1100100000 900 17684.69 0 17684.69 24 1100000000 800 15427.42 0 15427.42
SUM 21700 560775.61 4100 564875.61
80
CHAPTER 5
Proposed New Hybrid Models for Unit Commitment
problem based on convex and Non-Convex Cost Functions 5.1 Introduction
Hybrid models deal with the integration of two or more approaches. Recently, hybrid
techniques combining different classical and non classical optimization techniques have been
proposed to solve UCP. This chapter presents three hybrid models for the solution of convex and
non-convex fuel cost functions.
For the past several years, ANNs methods received a great deal of attention. H. Sasaki, et al.
[49] presented the application of Hopfield neural network to unit commitment. T. Yalcinoz, et al.
[50] presents a new mapping process and a computational method for obtaining the weights
using a slack variable technique for handling inequality constraints. M. H. Wong, et al. [51] used
GA to evolve the weight and the interconnection of the neural network to solve the UC problem.
Ouyang et al. [71] utilizes neural networks to generate a initial feasible schedule according to the
input load curve and then refines the schedule, using a dynamic search. Z. Ouyang, et al. [72]
proposes a multi-stage NN-expert system. H. Shyh-Jier et al. [74] proposes genetic algorithm
based NN and DP to solve the UC problem. H. Sasaki et al. [103] utilizes the Hopfield neural
network in which a large number of inequality constraints are handled by the dedicated neural
network instead of including them in the energy function.
One of the first uses of PSO was for evolving neural network weights and, indirectly, to
evolve the structure. R. C. Eberhart, et al. [104] replaced the standard back propagation learning
algorithm with the swarm intelligence learning rule. Y. D. Valle et al. [105] presented a detailed
overview of PSO including basic concepts, its variants and applications to power system. S.
Mohagheghi et al. [106] investigates the efficiency of swarm intelligence learning rule. The
81
applications of PSO as a training algorithm of ANN have been reported in [107-108]. The
advantage of using Particle Swarm Optimization (PSO) algorithm over other techniques is that it
is computationally inexpensive, easy to implement.
5.2 Hybrid Model – I: A hybrid of particle swarm optimization (PSO), artificial neural
network (ANN) and dynamic programming (DP).
Unit commitment problem has been addressed independently by Dynamic Programming
[20], Artificial Neural Network [49] and Particle Swarm Optimization [64]. In this thesis, a new
approach of PSO based artificial neural network (PSO-ANN) has been proposed to solve UC
problem. The proposed Hybrid Model-I combines the Dynamic Programming (DP) with
Artificial Neural Networks (ANN) using Swarm Intelligence (SI) learning rule. In this model
dynamic programming produces near optimal solution based on training data for neural network
model. The neural network fine tunes the data subject to the target values of power output of
units. The best tuned solution is considered the required solution.
The swarm intelligence learning rule based feed forward network has been used for fine
tuning the near optimal dynamic programming results. The standard back propagation learning
rule neural network has also been used in this hybrid approach for the comparison of results.
Flow chart is represented in Figure 5.1.
82
Figure 5.1 Flowchart for DP-PSO-ANN Hybrid Model-I
The PSO algorithm is applied to the neural network to obtain a set of weights that will minimize
the error function. Weights are progressively updated until the convergence criterion is satisfied.
The UC problem has been decomposed in to discrete load level and formed small ANN models
based on hourly load. Three and ten unit standard test systems have been tested for validation of
the proposed approach. The discussion to follow in next section of this approach is with
reference to context given in chapter 3.
83
5.2.1 Generation of test and training data
In this model the test data and training data are generated by using Dynamic
Programming (DP). Artificial Neural Network (ANN) is trained using Swarm Intelligence (S.I.)
learning rule. For proper training of ANN, a pair of load as input and their corresponding
generation schedules as output are prepared off-line by using DP. Each pair is referred as
input/output database pair. A multilayer feed forward neural network with swarm intelligence
learning rule has been programmed for tuning the power generation of the units.
5.2.2 Artificial Neural Network using SI learning Rule
PSO trains the ANN by changing its weight such that mean square error (MSE) for the
training set of an ANN model is reduced. It is a feedback process which runs until the batch error
falls under an acceptable range or iteration cutoff threshold is reached. In the process of training
ANN, particle position corresponds to the weights in the network. The fitness function
corresponds to the mean square error (MSE) of the network [107].
5.2.3 Input and Output of the PSO-ANN Model
The input to the network is the forecasted load demand as shown in Fig.5.2. For a three
unit system, in the ANN model three inputs (P1, P2, P3) the power generation of the machines and
outputs in the range of P1-P3 has been taken according to the load demand. The networks is
trained with input/ output data pattern with 3 neurons as input, 2 neurons for hidden layers and 3
neurons for output layer. Total we have 24 networks for 24 hour duration each representing an
hourly schedule h. Each network has been trained separately over hourly load data.
0
200
400
600
800
1000
1200
1400
1 3 5 7 9 11 13 15 17 19 21 23
Time (hours)
Load
Dem
and
(MW
)
Figure 5.2 Load patterns for training
84
The target has been taken as the best solution attained so far. The percentage error is calculated
as follows: ((target – network output) / target)) x 100
5.2.4 Scaling of the input and output data
The network input and output data will have different ranges if actual hourly load data is
used so it was normalized to fall within the range [0, 1], to avoid convergence problems during
the training process.
5.2.5 Training process
The training process of an ANN model by PSO approach has the following steps as given
in Figure 5.3.
Initialize swarm with random velocity and position Repeat for each particle i in swarm do if MSE( current network ) < MSE (personal best network) then Personal best network = current network end if if MSE ( current network )< the MSE (global best network) then Global best network = current network end if end for Update velocity and position of each particle according to equations (3.1) and (3.2) Until stop criteria (epoch < 30000 or MSE (global best network)< 1.0e-12) being satisfied
Figure 5.3 Steps for SI learning ANN
The perception training algorithm is a form of supervised learning algorithm where the
weights and biases are updated to reduce errors whenever the network output does not match the
desired values. The load patterns and its corresponding feasible schedules are obtained using
dynamic programming technique in a training set for the ANN. Most of the time is spend on off-
line training of the network. With trained network the on-line operation time is very short. The
training set covers 24 hours load data. We used 24 networks for 3 units system each having
architecture of 3:2:3. The training set contained 20 patterns for each hour. Total number of
training patterns is 480.
The terminating criteria are the maximum number of epochs and the maximum mean square
error.
Swarm intelligence learning rule is used for network training.
85
5.2.6 Parameters Settings
a. The PSO model parameters:
The parameter settings are: Number of particles = 60, scaling constants, c1 = c2 =1.4962,
Constriction factor χ = 0.7298, the range for individual particle position 21≤x≤-21, the range
for individual dimension of particle velocity, is v_max = k*x_max, x_max=21, k=0.1 and is
equal to 2.1≤x≤-2.1, Maximum Epoch = 30000, Maximum Error = 1.0e-12, max_memory =
30, success_counter sc =15, fail_counter fc =5.
b. The BP model parameters:
The parameters selected are: No. of samples = 60, No. of inputs = No. of units, No. of
outputs = No. of committed units according to load demand at each hour, momentum rate α =
0.85, learning rate η = 0.8, No. of hidden layers = 2.
5.3 Case Studies---Convex cost function
Following three standard test systems have been selected for the validation of the proposed
approaches.
a. Test System I --- 3 units with 24 hours load.
b. Test System II --- 3 units with 24 hours load
c. Test System III --- 10 units system with 24 hours load
Input data: The description and input data of test systems used for investigation in the case
studies are given in Table A. 1, Table A. 2 and Table A. 3 placed in Appendix A.
Computer Implementation: The algorithms have been implemented in C++ on P-IV Personal
Computer.
Output Results: The summary and comparison of results is given in this chapter.
5.3.1 Numerical results of test system-I: Three unit system: Hybrid Model-I
Sixty samples consisting of three inputs for three unit system to the neural network
produce fine tuned sixty outputs. The outputs consist of one to three power generations
depending upon forecasted load.
The best fine tuned results generated from both the ANN models amongst the sixty samples
and corresponding near optimal output generated by DP have been tabulated. The mean square
86
error and absolute percentage error has been calculated with reference to the target generation
schedule. The plot of mean square error for three different cases is shown in Figures 5.4, 5.5 and
5.6. The plot of absolute percentage error is shown in Figure 5.7.
The output results of test system–I are shown in the following tables:
Table (5.1) gives the comparison of proposed Hybrid Model-I with Genetic Algorithm,
Conventional Priority List and Hopfield Neural Network methods.
Table (5.2) presents the best results by the proposed Hybrid Model-I (DP-PSO-ANN) among the
sixty samples.
Table (5.3) presents the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained from the proposed Hybrid Model-I amongst sixty samples for SI learning.
Table (5.4) shows the best output results obtained by the Hybrid Model-I (DP-BP-ANN)
amongst sixty samples.
Table (5.5) shows the comparison of the outputs (MW) obtained by the proposed SI –ANN land
BP-ANN learning for 3 unit system.
Table (5.6) presents the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained from the proposed Hybrid Model-I amongst sixty samples for BP learning.
The Salient features of the proposed approach in the light of the observations from the results are
as follows:
1. PSO trains the ANN up to the target values as shown in Table 5.2.
2. In Hybrid Model-I (DP-PSO-ANN) the dynamic programming is used to generate
training and test data. S.I. learning rule is used to train the neural network. This model
gives accurate results.
3. This approach hits the target due to the use of S.I. learning. This model is able to explore
more solution space; therefore it takes long training time, but gives high quality solution.
The Absolute Percentage Error (APE) in this method is almost zero.
4. Figures 5.4-5.6 presents the MSE graph of SI-learning algorithm and BP learning
algorithm. SI learning graph decays slowly and finally hits the target. BP learning graph
decays rapidly and then slows down and unable to hit the target.
87
5. Figure 5.7 indicates the APE graph of both the SI and BP learning algorithms. The APE
in SI learning is zero, but in BP the APE is fluctuating.
6. The Absolute Percentage Error in BP trained model is more than PSO trained model.
The BP learning stuck in local minima and unable to reach to the target value. BP-ANN
model gives results near to target quickly due to the use of B.P. learning algorithm.
7. For three unit test system-I the operating fuel cost for committed units in the 24 hours
for PSO-ANN and BP-ANN models are $201632.089 and $201630.923 respectively.
The cost of BP-ANN learning approach is slightly lower than PSO-ANN learning as BP-
learning approach does not reach the target.
8. The operating fuel cost obtained by all both trained models remains low compared to
genetic algorithm, Hopfield neural network and Conventional Priority List methods.
9. SI-learning gives $389.28 saving per day equivalent to 0.1926% compared with GA and
$474.28 saving per day equivalent to 0.2346% compared with HFNN and $761.26
saving per day equivalent to 0.37612% compared with conventional priority list
methods.
10. All the two learning models give fair amount of reduction in operating cost as compared
to GA, HFNN and conventional priority list methods.
11. The PSO trained model gives accurate and good quality results.
12. Hybrid Model-I, incorporates dynamic programming as the backbone.
13. Hybrid Model-I using SI-learning generates better solutions than the other methods,
mainly because of its intrinsic nature of updates of positions and velocities. The second
reason is due to the hourly basis solution. This is somehow similar to the “divide and
conquer” strategy of solving a problem. Owning to this hourly solution, the complexity
of the search is greatly reduced.
14. The result obtained from the simulation is most encouraging in comparison to the best
known solution so far.
88
Table 5.1 Comparison of proposed Hybrid Model-I with Genetic Algorithm, Conventional Priority List and Hopfield Neural Network methods for 3 unit systems: Test system-I:
Algorithm Daily
Operating Cost ($)
Amount of Daily Saving (comparedwith GA)
% Cost saving in fuel cost compared with GA
Amount of Daily Saving (compared with conventional PL)
% Cost saving in fuel cost compared with PL
Amount of Daily Saving (compared with Hopfield Neural Network
% Cost saving in fuel cost compared with Hopfield Neural Network
Genetic Algorithm
202021.360 - - - - - -
Conventional Priority List (FLAPC)
202393.346 - - - - - -
Hopfield Neural Network
202106.36 - - - - - -
Hybrid Model–I (DP-PSO-ANN) SI- Learning
201632.089 389.28 0.1926 761.26 0.37612 474.28 0.2346
Hybrid Model–I (DP-BP-ANN) BP-Learning
201630.923 390.44 0.1932 762.42 0.37670 475.44 0.2352
89
Table 5.2 Best Results by the proposed Hybrid Model-I (DP-PSO-ANN) among the sixty samples: Test system-I
Targets
Outputs obtained by the proposed
DP-PSO-ANN model Hours P-1 P-2 P-3 P-1 P-2 P-3
1 600.00 400.000 200.000 600.000 400.000 200.000 2 600.00 400.000 200.000 600.000 400.000 200.000 3 600.00 400.000 150.000 600.000 400.000 150.000 4 600.00 400.000 100.000 600.000 400.000 100.000 5 600.00 400.000 0 600.000 400.000 0 6 500.00 400.000 0 500.000 400.000 0 7 433.18 366.820 0 433.180 366.820 0 8 600.00 0 0 600.000 0 0 9 550.00 0 0 550.000 0 0 10 500.00 0 0 500.000 0 0 11 500.00 0 0 500.000 0 0 12 500.00 0 0 500.000 0 0 13 500.00 0 0 500.000 0 0 14 500.00 0 0 500.000 0 0 15 600.00 0 0 600.000 0 0 16 433.18 366.820 0 433.180 366.820 0 17 460.88 389.120 0 460.880 389.120 0 18 500.00 400.000 0 500.000 400.000 0 19 550.00 400.000 0 550.000 400.000 0 20 600.00 400.000 0 600.000 400.000 0 21 600.00 400.000 50.000 600.000 400.000 50.000 22 600.00 400.000 100.000 600.000 400.000 100.000 23 600.00 400.000 200.000 600.000 400.000 200.000 24 600.000 400.000 200.000 600.000 400.000 200.000
90
Table 5.3 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the proposed Hybrid Model-I (DP-PSO-ANN): Test system-I
Power output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600 400 200 1200 5875.320 3760.400 2237.760 11873.480 2 111 600 400 200 1200 5875.320 3760.400 2237.760 11873.480 3 111 600 400 150 1150 5875.320 3760.400 1658.340 11294.060 4 111 600 400 100 1100 5875.320 3760.400 1107.840 10743.560 5 110 600 400 0 1000 5875.320 3760.400 0 9635.720 6 110 500 400 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 100 600 0 0 600 5875.320 0 0 5875.320 9 100 550 0 0 550 5389.505 0 0 5389.505 10 100 500 0 0 500 4911.500 0 0 4911.500 11 100 500 0 0 500 4911.500 0 0 4911.500 12 100 500 0 0 500 4911.500 0 0 4911.500 13 100 500 0 0 500 4911.500 0 0 4911.500 14 100 500 0 0 500 4911.500 0 0 4911.500 15 100 600 0 0 600 5875.320 0 0 5875.320 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500 400 0 900 4911.500 3760.400 0 8671.900 19 110 550 400 0 950 5389.505 3760.400 0 9149.905 20 110 600 400 0 1000 5875.320 3760.400 0 9635.720 21 111 600 400 50 1050 5875.320 3760.400 586.260 10221.980 22 111 600 400 100 1100 5875.320 3760.400 1107.840 10743.560 23 111 600 400 200 1200 5875.320 3760.400 2237.760 11873.480 24 111 600 400 200 1200 5875.320 3760.400 2237.760 11873.480
SUM 20650 201632.089
91
Table 5.4 Best output results obtained by the Hybrid Model-I (DP-BP-ANN) amongst sixty samples: Test system-I
Targets
Outputs obtained by the proposed
DP-BP-ANN model Hours P-1 P-2 P-3 P-1 P-2 P-3
1 600.00 400.000 200.000 599.999 399.999 199.999 2 600.00 400.000 200.000 599.999 399.999 199.999 3 600.00 400.000 150.000 599.999 399.999 149.999 4 600.00 400.000 100.000 599.999 399.999 99.999 5 600.00 400.000 0 599.984 400.000 0 6 500.00 400.000 0 499.998 399.950 0 7 433.18 366.820 0 433.192 366.849 0 8 600.00 0 0 599.999 0 0 9 550.00 0 0 549.999 0 0 10 500.00 0 0 499.999 0 0 11 500.00 0 0 499.999 0 0 12 500.00 0 0 499.999 0 0 13 500.00 0 0 499.999 0 0 14 500.00 0 0 499.999 0 0 15 600.00 0 0 599.999 0 0 16 433.18 366.820 0 433.192 366.849 0 17 460.88 389.120 0 460.880 389.118 0 18 500.00 400.000 0 499.998 399.950 0 19 550.00 400.000 0 550.009 399.969 0 20 600.00 400.000 0 599.984 400.000 0 21 600.00 400.000 50.000 599.999 399.999 49.99 22 600.00 400.000 100.000 599.999 399.999 99.999 23 600.00 400.000 200.000 599.999 399.999 199.999 24 600.000 400.000 200.000 599.999 399.999 199.999
92
Table 5.5 Comparison of the outputs (MW) obtained by the proposed SI – ANN learning and BP - ANN learning for 3 unit system: Test System-I
Targets
Output obtained by the proposed SI-ANN learning
rule
Output obtained by proposed BP-ANN learning rule
P-1 (MW)
P-2 (MW)
P-3 (MW)
P-1 (MW)
P-2 (MW)
P-3 (MW)
P-1 (MW)
P-2 (MW)
P-3 (MW)
600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 150.000 600.000 400.000 150.000 599.999 399.999 149.999 600.00 400.000 100.000 600.000 400.000 100.000 599.999 399.999 99.999 600.00 400.000 0 600.000 400.000 0 599.984 400.000 0 500.00 400.000 0 500.000 400.000 0 499.998 399.950 0 433.18 366.820 0 433.180 366.820 0 433.192 366.849 0 600.00 0 0 600.000 0 0 599.999 0 0 550.00 0 0 550.000 0 0 549.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 600.00 0 0 600.000 0 0 599.999 0 0 433.18 366.820 0 433.180 366.820 0 433.192 366.849 0 460.88 389.120 0 460.880 389.120 0 460.880 389.118 0 500.00 400.000 0 500.000 400.000 0 499.998 399.950 0 550.00 400.000 0 550.000 400.000 0 550.009 399.969 0 600.00 400.000 0 600.000 400.000 0 599.984 400.000 0 600.00 400.000 50.000 600.000 400.000 50.000 599.999 399.999 49.99 600.00 400.000 100.000 600.000 400.000 100.000 599.999 399.999 99.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999
93
Table 5.6 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained by using Hybrid Model-I (DP-BP-ANN): Test system-I
Power output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 599.999 399.999 199.999 1200 5875.310 3760.391 2237.748 11873.449 2 111 599.999 399.999 199.999 1200 5875.310 3760.391 2237.748 11873.449 3 111 599.999 399.999 149.999 1150 5875.310 3760.391 1658.329 11294.030 4 111 599.999 399.999 99.999 1100 5875.310 3760.391 1107.829 10743.530 5 110 599.984 400 0 1000 5875.163 3760.400 0 9635.563 6 110 499.998 399.95 0 900 4911.481 3759.930 0 8671.411 7 110 433.192 366.849 0 800 4284.998 3450.846 0 7735.845 8 100 599.999 0 0 600 5875.310 0 0 5875.310 9 100 549.999 0 0 550 5389.495 0 0 5389.495 10 100 499.999 0 0 500 4911.491 0 0 4911.491 11 100 499.999 0 0 500 4911.491 0 0 4911.491 12 100 499.999 0 0 500 4911.491 0 0 4911.491 13 100 499.999 0 0 500 4911.491 0 0 4911.491 14 100 499.999 0 0 500 4911.491 0 0 4911.491 15 100 599.999 0 0 600 5875.310 0 0 5875.310 16 110 433.192 366.849 0 800 4284.998 3450.846 0 7735.845 17 110 460.88 389.118 0 850 4542.955 3658.317 0 8201.272 18 110 499.998 399.95 0 900 4911.481 3759.930 0 8671.411 19 110 550.009 399.969 0 950 5389.592 3760.109 0 9149.700 20 110 599.984 400 0 1000 5875.163 3760.400 0 9635.563 21 111 599.999 399.999 49.99 1050 5875.310 3760.391 586.159 10221.859 22 111 599.999 399.999 99.999 1100 5875.310 3760.391 1107.829 10743.530 23 111 599.999 399.999 199.999 1200 5875.310 3760.391 2237.748 11873.449 24 111 599.999 399.999 199.999 1200 5875.310 3760.391 2237.748 11873.449
SUM 20650 201630.923
94
Case - I
-0.02
0.00
0.02
0.04
0.06
0.08
0.10
1 11 21 31 41 51 61 71 81 91 101
No. of Epochs
MSE
BP SI
Figure. 5.4 MSE graph for a load of 500 MW.
Case - II
-0.06
0.04
0.14
0.24
0.34
0.44
0.54
0.64
0.74
0.84
1 13 25 37 49 61 73 85 97 109
No. of Epochs
MSE
BP SI
Figure. 5.5 MSE graph for a load of 800 MW.
Case - III
-0.02
0.03
0.08
0.13
0.18
0.23
1 11 21 31 41 51 61 71 81 91 101No. of Epochs
MSE
BP SI
Figure. 5.6 MSE graph for a load of 850 MW
-0.0010
-0.0005
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57
No. of Samples
Abs
olut
e Pe
rcen
tage
Err
or
Figure. 5.7 Absolute Percentage Error graph for Operating Fuel Cost
95
5.3.2 Numerical Results of Test system-II: Three unit system: Hybrid Model-I:
The output results are shown in the following tables:
Table (5.7) shows the comparison of the best operating fuel cost obtained amongst the sixty
samples.
Table (5.8) gives the comparison of the outputs (MW) obtained by the proposed SI–ANN
learning and BP-ANN learning.
The Salient features of the proposed approach in the light of the observations from the
results are as follows:
1. For three unit test system-II the operating fuel cost for committed units in the 24 hours
for PSO-ANN and BP-ANN models are $1992262.671 and $1992262.310 respectively.
The cost of BP-ANN learning approach is slightly lower than PSO-ANN learning as BP-
learning approach does not reach the target.
2. The PSO trained model gives accurate and good quality results.
Table 5.7 Comparison of the best Operating fuel cost ($) obtained amongst the sixty samples for 3 unit system by using Proposed Hybrid Model-I: Test system-II
(GA-PSO-ANN) (GA-BP-ANN) Operating fuel cost Test system-II $199262.671 $199262.310
96
Table 5.8 Comparison of the outputs (MW) obtained by the proposed SI –ANN learning (Hybrid Model-I, DP-PSO-ANN) and BP-ANN learning (Hybrid Model-I (DP-BP-ANN) for 3
unit systems: Test System-II
Targets
Output obtained by the proposed SI-ANN learning
rule
Output obtained by proposed BP-ANN learning rule
Unit-1 (MW)
Unit-2 (MW)
Unit-3 MW
Unit-1 (MW)
Unit-2 (MW)
Unit-3 MW
Unit-1 (MW)
Unit-2 (MW)
Unit -3 MW
600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 150.000 600.000 400.000 150.000 599.999 399.999 149.999 600.00 400.000 100.000 600.000 400.000 100.000 599.999 399.999 99.999 600.00 400.000 0 600.000 400.000 0 599.984 400.000 0 500.00 400.000 0 500.000 400.000 0 499.998 399.950 0 433.18 366.820 0 433.180 366.820 0 433.192 366.849 0 600.00 0 0 600.000 0 0 599.999 0 0 550.00 0 0 550.000 0 0 549.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 500.00 0 0 500.000 0 0 499.999 0 0 600.00 0 0 600.000 0 0 599.999 0 0 433.18 366.820 0 433.180 366.820 0 433.192 366.849 0 460.88 389.120 0 460.880 389.120 0 460.880 389.118 0 500.00 400.000 0 500.000 400.000 0 499.998 399.950 0 550.00 400.000 0 550.000 400.000 0 550.009 399.969 0 600.00 400.000 0 600.000 400.000 0 599.984 400.000 0 600.00 400.000 50.000 600.000 400.000 50.000 599.999 399.999 49.99 600.00 400.000 100.000 600.000 400.000 100.000 599.999 399.999 99.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999 600.00 400.000 200.000 600.000 400.000 200.000 599.999 399.999 199.999
97
5.3.3 Numerical Results of Test system-III: Ten unit system: Hybrid Model-I:
For a ten machine system, in the ANN model ten inputs (P1, P2, P3…P10) the power
generation of the machines and two to ten outputs has been taken Training data with different
samples consisting of various combinations have been generated by dynamic programming. The
networks are trained with 10 neurons as input, 2 neurons for hidden layers and 10 neurons for
output layer for ten unit systems. Total we have 24 networks each representing an hour h. Each
network was trained separately over hourly load data.
The PSO-ANN trainer program is written in C++. A training set covers 24 hours load data.
We use 24 networks for 10 units system each having architecture of 10:2:10. The training set
contained 60 patterns for each hour. Total number of training patterns is 1440 for 10 unit system.
Sixty samples consisting of ten inputs for ten unit system to the neural network produce fine
tuned sixty outputs. The outputs consist of one to ten power generations depending upon
forecasted load.
The best fine tuned results generated from both the A.N.N models amongst the sixty
samples and corresponding near optimal output generated by DP have been tabulated. The plot
of Absolute percentage error against the numbers of samples is shown in Figure 5.8.
The output results of test system – III are shown in the following tables:
Table (5.9) shows the target values for 10 unit system.
Table (5.10) presents the best results obtained by the proposed Hybrid Model-I (DP-PSO-ANN)
amongst the sixty samples.
Table (5.11) presents the unit commitment schedule and power sharing (MW) of the best
solution obtained from proposed Hybrid Model-I (DP-PSO-ANN) considering MUT and MDT
constraints with 10 % spinning reserve.
Table (5.12) shows the load demand and total production cost (TPC) obtained by proposed
Hybrid Model-I (DP-PSO-ANN).
Table (5.13) presents the comparison of the best results of the Hybrid Model-I (DP-PSO-ANN)
with other approaches available in the literature.
Table (5.14) Comparison between proposed Hybrid model-I and other approaches for daily
saving and Percentage saving in fuel cost.
Table (5.15) Comparison of the results of the proposed Hybrid Model-I with Genetic
Algorithm, dynamic programming, simulated annealing and Lagrange relaxation method:
98
Table (5.16) presents the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained by using Hybrid Model-I (DP-BP-ANN), considering MUT and MDT
constraints with 10 % spinning reserve:
Table (5.17) gives the hourly load demand, operating fuel cost, start-up cost and Total
Production Cost obtained by using (DP-BP-ANN), considering MUT and MDT constraints with
10 % spinning reserve:
The Salient features of the proposed approach in the light of the observations from the results
are as follows:
1. The total production cost for 10 unit system is $563942.3.
2. When compared with other thirty four approaches available in the literature, the
proposed hybrid models-I gives low total production cost compared as given in Table
5.13.
3. The daily costs saving of the proposed Hybrid model-III compared with other
approaches have a range of $0.02 to $24812.02 per day. Which is equivalent to a saving
of 0.00000355% to 4.399% as given in Table 5.14?
4. The daily cost saving compared with GA, DP, SA and LR is $1877 0.33 per day. Which
is equivalent to saving of 0.333% as given in Table 5.15?
99
Table 5.9 Targets for the 10 unit system: Test system-III
Targets
Hour Load (MW)
P-1 P-2 P-3 P-4 P-5 P-6 P-7 P-8 P-9 P-10
1 700 455 245 0 0 0 0 0 0 0 0 2 750 455 295 0 0 0 0 0 0 0 0 3 850 455 370 0 0 25 0 0 0 0 0 4 950 455 455 0 0 40 0 0 0 0 0 5 1000 455 390 0 130 25 0 0 0 0 0 6 1100 455 360 130 130 25 0 0 0 0 0 7 1150 455 410 130 130 25 0 0 0 0 0 8 1200 455 455 130 130 30 0 0 0 0 0 9 1300 455 455 130 130 85 20 25 0 0 0 10 1400 455 455 130 130 162 33 25 10 0 0 11 1450 455 455 130 130 162 73 25 10 10 0 12 1500 455 455 130 130 162 80 25 43 10 10 13 1400 455 455 130 130 162 33 25 10 0 0 14 1300 455 455 130 130 85 20 25 0 0 0 15 1200 455 455 130 130 30 0 0 0 0 0 16 1050 455 310 130 130 25 0 0 0 0 0 17 1000 455 260 130 130 25 0 0 0 0 0 18 1100 455 360 130 130 25 0 0 0 0 0 19 1200 455 455 130 130 30 0 0 0 0 0 20 1400 455 455 130 130 162 33 25 10 0 0 21 1300 455 455 130 130 85 20 25 0 0 0 22 1100 455 455 0 0 145 20 25 0 0 0 23 900 455 425. 0 0 0 20 0 0 0 0 24 800 455 345 0 0 0 0 0 0 0 0
100
Table 5.10 Best results obtained by the proposed Hybrid Model-I (DP-PSO-ANN) amongst the sixty samples: Test system-III
Outputs obtained by the proposed
DP-PSO-ANN model P-1 P-2 P-3 P-4 P-5 P-6 P-7 P-8 P-9 P-10455 245 0 0 0 0 0 0 0 0 455 295 0 0 0 0 0 0 0 0 455 370 0 0 25 0 0 0 0 0 455 455 0 0 40 0 0 0 0 0 455 390 0 130 25 0 0 0 0 0 455 360 130 130 25 0 0 0 0 0 455 410 130 130 25 0 0 0 0 0 455 455 130 130 30 0 0 0 0 0 455 455 130 130 85 20 25 0 0 0 455 455 130 130 162 33 25 10 0 0 455 455 130 130 162 73 25 10 10 0 455 455 129.96 129.92 162.14 80 25.01 42.96 10 10 455 455 130 130 162 33 25 10 0 0 455 455 130 130 85 20 25 0 0 0 455 455 130 130 30 0 0 0 0 0 455 310 130 130 25 0 0 0 0 0 455 260 130 130 25 0 0 0 0 0 455 360 130 130 25 0 0 0 0 0 455 455 130 130 30 0 0 0 0 0 455 455 130 130 162 33 25 10 0 0 455 455 130 130 85 20 25 0 0 0 455 455 0 0 145 20 25 0 0 0 455 425.01 0 0 0 20 0 0 0 0 455 345 0 0 0 0 0 0 0 0
101
Table 5.11 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from proposed Hybrid Model-I (DP-PSO-ANN): Test system-III
Power output of each unit (MW)
Hour UC schedule
Load (MW)
1 2 3 4 5 6 7 8 9 10
1 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 295 0 0 0 0 0 0 0 0 3 1100100000 850 455 370 0 0 25 0 0 0 0 0 4 1100100000 950 455 455 0 0 40 0 0 0 0 0 5 1101100000 1000 455 390 0 130 25 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 20 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 129.96 129.92 162.14 80 25.01 42.96 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 20 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 20 25 0 0 0 22 1111111000 1100 455 455 0 0 145 20 25 0 0 0 23 1100010000 900 455 425.01 0 0 0 20 0 0 0 0 24 1100000000 800 455 345 0 0 0 0 0 0 0 0
102
Table 5.12 Load demand and Total Production Cost (TPC) obtained by proposed Hybrid Model-I (DP-PSO-ANN): Test system-III:
Hour Load
(MW)
Fuel cost ($.h)
Startup Cost ($/h)
Total Production Cost ($/h)
1 700 13683.13 0 13683.13 2 750 14554.50 0 14554.50 3 850 16809.45 900 17709.45 4 950 18597.67 0 18597.67 5 1000 20020.02 560 20580.02 6 1100 22387.04 1100 23487.04 7 1150 23261.98 0 23261.98 8 1200 24150.34 0 24150.34 9 1300 27251.06 860 28111.06 10 1400 30057.55 60 30117.55 11 1450 31916.06 60 31976.06 12 1500 33890.28 60 33950.28 13 1400 30057.55 0 30057.55 14 1300 27251.06 0 27251.06 15 1200 24150.34 0 24150.34 16 1050 21513.66 0 21513.66 17 1000 20641.82 0 20641.82 18 1100 22387.04 0 22387.04 19 1200 24150.34 0 24150.34 20 1400 30057.55 490 30547.55 21 1300 27251.06 0 27251.06 22 1100 22735.52 0 22735.52 23 900 17645.54 0 17645.54 24 800 15427.42 0 15427.42
Sum 27100 559847.98 4090.00 563937.98
103
Table 5.13 Comparison of the best results of the Hybrid Model-I (DP- PSO-ANN) with other approaches available in the literature: Test system-III
Approach DP
[56] LR
[56] ELR [36] EPL [15] SPL [16] RPACO
Total production cost($)
565825 565825 563977 563977 564950 565302
Approach HPSO [80]
IPSO LR-PPSO GA-UCC [57]
EMOALHN
TPC ($) 563942 563954 563977 563977 563977 Approach FPGA
[60] PSO [65]
MRCGA [61]
GA-LR LR-PSO ALR
TPC ($) 564094 564212 564244 564800 565275 565508 Approach GA
[56] GRASP MA ICGA [59] DPHNN BPSO
TPC ($) 565825 565825 565827 566404 588750 565804 Approach AG EP [52] ESA [44] ASSA [45] AG SA TPC ($) 564005 564551 565828 563938 564005 565825
Approach PLEA [82]
DP-LR EP-LR ACSA PLEA [82]
TPC ($) 563977 564049 564049 564059 563977 Proposed Hybrid Model-I (DP-PSO-ANN) 563937.98
• Abbreviations are given in Appendix-B.
104
Table 5.14 Comparison between proposed Hybrid model-I and other approaches for daily saving and Percentage saving in fuel cost: ten unit system
S.
No.
Approach
Total Production
Cost ($)
Amount of daily saving ($) compared
with proposed hybrid model –I
(DP-PSO-ANN).
% saving in fuel cost compared with proposed
model. (DP-PSO-ANN).
1 PSO-SA 563938 0.02 3.55E-06 2 ASSA 563938 0.02 3.55E-06 3 HPSO 563942 4.02 0.000713 4 IPSO 563954 16.02 0.002841 5 ELR 563977 39.02 0.006919 6 EPL 563977 39.02 0.006919 7 LR-PPSO 563977 39.02 0.006919 8 GA-UCC 563977 39.02 0.006919 9 EMOALHN 563977 39.02 0.006919 10 PLEA 563977 39.02 0.006919 11 AG 564005 67.02 0.011884 12 DP-LR 564049 111.02 0.019687 13 EP-LR 564049 111.02 0.019687 14 ACSA 564059 121.02 0.021460 15 FPGA 564094 156.02 0.027666 16 PSO 564212 274.02 0.04859 17 MRCGA 564244 306.02 0.054265 18 EP 564551 613.02 0.108703 19 GA-LR 564800 862.02 0.152857 20 SPL 564950 1012.02 0.179456 21 LR-PSO-1 565275 1337.02 0.237086 22 RPACO 565302 1364.02 0.241874 23 ALR 565508 1570.02 0.278403 24 BPSO 565804 1866.02 0.330891 25 DP 565825 1887.02 0.334615 26 LR 565825 1887.02 0.334615 27 SA 565825 1887.02 0.334615 28 GA 565825 1887.02 0.334615 29 GRASP 565825 1887.02 0.334615 30 MA 565827 1889.02 0.334969 31 LR-PSO-2 565869 1931.02 0.342417 32 ICGA 566404 2466.02 0.437286 33 BCGA 567367 3429.02 0.608049 34 DPHNN 588750 24812.02 4.399778 35 Proposed DP-PSO-ANN 563937.98
105
Table 5.15 Comparison of the results of the proposed Hybrid Model-I with Genetic Algorithm, dynamic programming, simulated annealing and Lagrange relaxation method:
Approach/Model Total
ProductionCost ($)
Amount of Daily Saving ($)
% Saving
GA,DP,SA and LR 565825.00 - - Hybrid Model-I (SI learning) 563937.98 1887.02 0.333 Hybrid Model-I (BP learning) 563935.58 1889.42 0.333
Table 5.16 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained by using (DP-BP-ANN), considering MUT and MDT constraints with 10 % spinning
reserve: Test system-III
Power output of each unit (MW) Hour UC schedule
Load (MW)
1 2 3 4 5 6 7 8 9 10 1 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 294.99 0 0 0 0 0 0 0 0 3 1100100000 850 455 370 0 0 24.99 0 0 0 0 0 4 1100100000 950 455 455 0 0 40 0 0 0 0 0 5 1101100000 1000 455 390 0 130 25 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 19.99 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 130 130 162 80 25 42.99 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 19.99 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 19.99 25 0 0 0 22 1111111000 1100 455 455 0 0 145 19.99 25 0 0 0 23 1100010000 900 454.99 424.99 0 0 0 19.99 0 0 0 0 24 1100000000 800 455 345.000 0 0 0 0 0 0 0 0
106
Table 5.17 Load demand, operating fuel cost, startup cost and Total Production Cost obtained by Using (DP-BP-ANN), considering M.U.T and M.D.T Constraints with 10 %
spinning reserve: Test system-III:
Hour Load (MW)
Fuel cost ($)
Startup Cost ($)
Total Production
Cost ($) 1 700 13683.13 0 13683.13 2 750 14554.33 0 14554.33 3 850 16809.25 900 17709.25 4 950 18597.67 0 18597.67 5 1000 20020.02 560 20580.02 6 1100 22387.04 1100 23487.04 7 1150 23261.98 0 23261.98 8 1200 24150.34 0 24150.34 9 1300 27250.83 860 28110.83 10 1400 30057.55 60 30117.55 11 1450 31916.06 60 31976.06 12 1500 33889.90 60 33949.90 13 1400 30057.55 0 30057.55 14 1300 27250.83 0 27250.83 15 1200 24150.34 0 24150.34 16 1050 21513.66 0 21513.66 17 1000 20641.82 0 20641.82 18 1100 22387.04 0 22387.04 19 1200 24150.34 0 24150.34 20 1400 30057.55 490 30547.55 21 1300 27250.83 0 27250.83 22 1100 22735.30 0 22735.30 23 900 17644.7968 0 17644.80 24 800 15427.42 0 15427.42
Sum 27100 559845.58 4090.00 563935.58
0.0000.0020.0040.0060.0080.0100.012
1 7 13 19 25 31 37 43 49 55
No. of Samples
AP
E
Series1
Figure 5.8 Absolute Percentage Error (APE) for a load of 900MW (DP-BP-ANN Model):
Ten unit system: Test system-III
107
5.4 Hybrid Model-II: Neuro-Genetic Hybrid Approach
The proposed Hybrid Model-II (Neuro-Genetic approach) combines the Genetic
Algorithm (GA) with Artificial Neural Networks (ANN) using Swarm Intelligence (SI) learning
rule. Unit commitment problem has been addressed independently by Genetic Algorithm [56],
Particle Swarm Optimization and Artificial Neural Network. GA has the ability to search better
for non-convex fuel cost function than convex fuel cost function. Three machine standard test
system has been tested for validation of the proposed model. The discussion to follow is with
reference to context of the discussion in chapter 3.
In this model Genetic Algorithm produces near optimal solution based training data for
neural network model. The neural network fine tunes the data subject to the target values. The
best fine tuned solution is considered the required solution. The target values may be taken by
randomly generated values by GA around the near optimal solutions satisfying both equality and
inequality constraints. The SI-learning rule based feed forward neural network has been used for
fine tuning the near optimal GA results. The standard back propagation learning rule has also
been used in this hybrid approach for the comparison of results. Flow chart in Figure 5.9
highlights the steps in the hybrid methodology
In this model the test data and training data are generated by using Genetic Algorithm. For
proper training of ANN, a pair of load as input and their corresponding generation schedules as
output are prepared off-line by using Genetic Algorithm (GA) and are stored in a data base. Each
pair is referred as input/output database pair.
108
No Yes
Figure 5.9 Flowchart for GA-PSO-ANN, Neuro-Genetic Hybrid Model-II
Start
Read System Data
GA is used to generate training and test data for 24 hour based on forecasted load.
Initialization
ANN training Using SI learning to obtain UC schedules
Testing of test cases and Calculation of fuel cost
End
Check for comparison ANN using BP
Learning
ANN using BP learning to obtain UC Schedules
109
5.5 Case Studies --- non convex cost function Following two standard test systems have been selected for the illustration of the
effectiveness of the proposed approaches.
a. Test System IV --- 3 units system with 24 hours load.
b. Test System V --- 3 units system with 24 hours load.
Input data: The description and input data of test systems used for investigation in the case
studies is given in Table A. 4 and Table A.5 placed in Appendix A.
Computer Implementation: The algorithms have been implemented in C++ on P-IV Personal
Computer.
Output Results: The summary and comparison of results is given in this chapter.
5.5.1 Numerical Results of Test Systems –IV and V: 3 units systems: Hybrid Model-II
Sixty samples consisting of three inputs for three unit system to the neural network produce
fine tuned sixty outputs. The outputs consist of one to three power generations depending upon
forecasted load.
The best fine tuned results generated from both the ANN models amongst the sixty samples
and corresponding near optimal output generated by GA have been tabulated.
The output results of non convex test system-IV and V are shown in the following tables.
Table (5.18) presents the comparison of the best operating fuel cost ($) obtained amongst the
sixty samples for 3 unit system IV and V.
Table (5.19) presents the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained from the Hybrid Model-II: (GA-PSO-ANN) for Test system-IV.
Table (5.20) presents the Unit Commitment Schedule and Power Sharing (MW) of the solution
obtained from the Hybrid Model-II (GA-BP-ANN) for Test system-IV.
Table (5.21) presents the Unit Commitment Schedule and Power Sharing (MW) of the solution
obtained from the Hybrid Model-II: (GA-PSO-ANN): for Test system-V.
110
Table (5.22) presents the Unit Commitment Schedule and Power Sharing (MW) of the solution
obtained from the Hybrid Model-II: (GA-BP-ANN) for Test system-V.
The Salient features of the proposed approach in the light of the observations from the
results are as follows:
1. For non-convex cost function of test systems IV and V the hybrid model-II gives better results than hybrid model-III.
Table 5.18 Comparison of the best Operating fuel cost ($) obtained amongst the sixty samples for 3 unit system by using Proposed Hybrid Model-II: (Non-convex)
(GA-PSO-ANN) (GA-BP-ANN) Operating fuel cost Test system-IV $ 207773.614 $ 207772.477 Operating fuel cost Test system-V $ 200665.137 $ 200662.723
111
Table 5.19 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the Hybrid Model-II: (GA-PSO-ANN): Test system-IV
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 2 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 3 111 600.00 400.00 150.00 1150 6175.105 3767.125 1660.862 11603.092 4 111 600.00 400.00 100.00 1100 6175.105 3767.125 1109.101 11051.331 5 110 600.00 400.00 0 1000 6175.105 3767.125 0 9942.230 6 110 551.64 348.36 0 900 5430.872 3449.030 0 8879.902 7 110 542.05 257.95 0 800 5377.510 2532.705 0 7910.216 8 100 600.00 0 0 600 6175.105 0 0 6175.105 9 100 550.00 0 0 550 5399.592 0 0 5399.592 10 100 500.00 0 0 500 5211.370 0 0 5211.370 11 100 500.00 0 0 500 5211.370 0 0 5211.370 12 100 500.00 0 0 500 5211.370 0 0 5211.370 13 100 500.00 0 0 500 5211.370 0 0 5211.370 14 100 500.00 0 0 500 5211.370 0 0 5211.370 15 100 600.00 0 0 600 6175.105 0 0 6175.105 16 110 542.05 257.95 0 800 5377.510 2532.705 0 7910.216 17 110 459.37 390.63 0 850 4623.302 3742.912 0 8366.214 18 110 551.64 348.36 0 900 5430.872 3449.030 0 8879.902 19 110 550.00 400.00 0 950 5399.592 3767.125 0 9166.717 20 110 600.00 400.00 0 1000 6175.105 3767.125 0 9942.230 21 111 600.00 400.00 50.00 1050 6175.105 3767.125 586.260 10528.490 22 111 600.00 400.00 100.00 1100 6175.105 3767.125 1109.101 11051.331 23 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 24 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773
SUM 207773.614
112
Table 5.20 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from the Hybrid Model-II (GA-BP-ANN): Test system-IV
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 599.999 399.999 199.999 1200 6175.096 3767.107 2241.522 12183.724 2 111 599.999 399.999 199.999 1200 6175.096 3767.107 2241.522 12183.724 3 111 599.999 399.999 149.999 1150 6175.096 3767.107 1660.841 11603.044 4 111 599.999 399.999 99.999 1100 6175.096 3767.107 1109.081 11051.284 5 110 599.984 400 0 1000 6174.954 3767.125 0 9942.079 6 110 551.64 348.36 0 900 5430.872 3449.030 0 8879.902 7 110 542.05 257.95 0 800 5377.510 2532.705 0 7910.216 8 100 599.999 0 0 600 6175.096 0 0 6175.096 9 100 549.999 0 0 550 5399.573 0 0 5399.573 10 100 499.999 0 0 500 5211.361 0 0 5211.361 11 100 499.999 0 0 500 5211.361 0 0 5211.361 12 100 499.999 0 0 500 5211.361 0 0 5211.361 13 100 499.999 0 0 500 5211.361 0 0 5211.361 14 100 499.999 0 0 500 5211.361 0 0 5211.361 15 100 599.999 0 0 600 6175.096 0 0 6175.096 16 110 542.05 257.95 0 800 5377.510 2532.705 0 7910.216 17 110 459.37 390.63 0 850 4623.302 3742.912 0 8366.214 18 110 551.64 348.36 0 900 5430.872 3449.030 0 8879.902 19 110 550.009 399.969 0 950 5399.764 3766.573 0 9166.337 20 110 599.984 400 0 1000 6174.954 3767.125 0 9942.079 21 111 599.999 399.999 49.99 1050 6175.096 3767.107 586.253 10528.456 22 111 599.999 399.999 99.999 1100 6175.096 3767.107 1109.081 11051.284 23 111 599.999 399.999 199.999 1200 6175.096 3767.107 2241.522 12183.724 24 111 599.999 399.999 199.999 1200 6175.096 3767.107 2241.522 12183.724
SUM 207772.477
113
Table 5.21 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from the Hybrid Model-II: (GA-PSO-ANN): Test system-V
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5887.927 3767.125 1868.583 11523.635 2 111 600.00 400.00 200.00 1200 5887.927 3767.125 1868.583 11523.635 3 111 600.00 400.00 150.00 1150 5887.927 3767.125 1384.472 11039.524 4 111 500.36 400.00 199.63 1100 4928.400 3767.125 1861.425 10556.950 5 110 600.00 400.00 0 1000 5887.927 3767.125 0 9655.052 6 110 500.00 400.00 0 900 4921.587 3767.125 0 8688.712 7 110 408.20 391.80 0 800 4138.137 3744.593 0 7882.730 8 100 600.00 0 0 600 5887.927 0 0 5887.927 9 100 550.00 0 0 550 5689.290 0 0 5689.290 10 100 500.00 0 0 500 4921.587 0 0 4921.587 11 100 500.00 0 0 500 4921.587 0 0 4921.587 12 100 500.00 0 0 500 4921.587 0 0 4921.587 13 100 500.00 0 0 500 4921.587 0 0 4921.587 14 100 500.00 0 0 500 4921.587 0 0 4921.587 15 100 600.00 0 0 600 5887.927 0 0 5887.927 16 110 408.20 391.80 0 800 4138.137 3744.593 0 7882.730 17 110 593.47 256.53 0 850 5860.308 2508.824 0 8369.131 18 110 500.00 400.00 0 900 4921.587 3767.125 0 8688.712 19 110 599.48 350.52 0 950 5877.847 3477.965 0 9355.812 20 110 600.00 400.00 0 1000 5887.927 3767.125 0 9655.052 21 111 600.00 400.00 50.00 1050 5887.927 3767.125 488.550 10143.602 22 111 600.00 400.00 100.00 1100 5887.927 3767.125 924.461 10579.513 23 111 600.00 400.00 200.00 1200 5887.927 3767.125 1868.583 11523.635 24 111 600.00 400.00 200.00 1200 5887.927 3767.125 1868.583 11523.635
SUM 200665.137
.
114
Table 5.22 Unit Commitment Schedule and Power Sharing (MW) of the solution obtained from the Hybrid Model-II: (GA-BP-ANN) Test system-V
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600 400 199.999 1200 5887.927 3767.125 1868.564 11523.615 2 111 600 400 199.999 1200 5887.927 3767.125 1868.564 11523.615 3 111 600 400 149.999 1150 5887.927 3767.125 1384.453 11039.505 4 111 500.36 399.99 199.63 1100 4928.400 3766.947 1861.425 10556.772 5 110 599.99 399.99 0 1000 5887.735 3766.947 0 9654.682 6 110 499.99 399.99 0 900 4921.398 3766.947 0 8688.344 7 110 408.20 391.80 0 800 4138.137 3744.593 0 7882.730 8 100 599.999 0 0 600 5887.908 0 0 5887.908 9 100 549.999 0 0 550 5689.281 0 0 5689.281 10 100 499.999 0 0 500 4921.568 0 0 4921.568 11 100 499.999 0 0 500 4921.568 0 0 4921.568 12 100 499.999 0 0 500 4921.568 0 0 4921.568 13 100 499.999 0 0 500 4921.568 0 0 4921.568 14 100 499.999 0 0 500 4921.568 0 0 4921.568 15 100 599.999 0 0 600 5887.908 0 0 5887.908 16 110 408.20 391.80 0 800 4138.137 3744.593 0 7882.730 17 110 593.47 256.53 0 850 5860.308 2508.824 0 8369.131 18 110 499.99 399.99 0 900 4921.398 3766.947 0 8688.344 19 110 599.48 350.52 0 950 5877.847 3477.965 0 9355.812 20 110 599.99 399.99 0 1000 5887.735 3766.947 0 9654.682 21 111 600.00 400.00 49.99 1050 5887.927 3767.125 488.560 10143.612 22 111 600.00 400.00 99.99 1100 5887.927 3767.125 924.277 10579.329 23 111 600.00 400.00 199.99 1200 5887.927 3767.125 1868.389 11523.441 24 111 600.00 400.00 199.99 1200 5887.927 3767.125 1868.389 11523.441
SUM 200662.723
115
5.6 Hybrid Model –III: Scaleable deterministic hybrid approach
The proposed hybrid model-III combines the Maximum Power Output (PMAX) and Full
Load Average Production Cost (FLAPC) of each unit. This model is applicable for both convex
and non convex fuel cost functions.
The priority list method has been used independently [15, 16] and by incorporating the
priority list in evolutionary programming (PLEA) [82]. The extended priority list (EPL) in [15]
is one of the most significant works. The main drawback in these methods is that the final
solution is non deterministic. The choice of the merit-order method for the hybridization process
was based on the following merits:
• Simple to implement
• Computationally fast
• Most widely used technique by the electricity utilities
The proposed unit commitment algorithm hybrid model-III incorporates the unit
commitment solution by the deterministic priority list scheme for generation of initial schedules.
This hybrid model solves the UC problem for convex and non-convex fuel cost functions. The
UC schedule is prepared according to PMAX and FLAPC. Unit with higher PMAX will be at
higher priority. If the two units have the same PMAX then unit with lower FLAPC will be of
higher priority. Economic Dispatch is based on lambda iteration method and average load
assigned methods. Priority order is given in Table 5.23. Flow chart is represented in Figure 5.10.
Table 5.23 Proposed Priority order based on Hybrid Model-III (PMAX-FLAPC)
Unit No.
PMAX FLAPC (S/MWH)
Proposed Priority order
(PMAX-FLAPC)
Unit No.
PMAX FLAPC (S/MWH)
Proposed Priority order (PMAX-
FLAPC) 1 455 18.6062 1 6 80 27.4546 7
2 455 19.5329 2 7 85 33.4542 6
3 130 22.2446 5 8 55 38.1472 8
4 130 22.0051 4 9 55 39.4830 9
5 162 23.1225 3 10 55 40.0670 10
116
Figure 5.10 Flow chart for Hybrid Model–III (PMAX-FLAPC)
117
5.7 Case Studies ---Convex fuel cost function
The following standard test systems have been selected for the validation of the proposed
approach.
a. Test System I --- 3 units with 24 hours load.
b. Test System-II --- 3 units with 24 hours load
c. Test System III --- 10 units with 24 hours load
Input data: The description and data of test system used for investigation in the case studies is
given in Table A.1, Table A.2 and Table A.3 placed in Appendix A.
Computer Implementation: The algorithms have been implemented in C++ on P-IV Personal
Computer.
Output Results: The summary and comparison of results is given in this section.
5.7.1 Numerical results of test system –I: three unit system: Hybrid Model-III
The output results of the test system-I are shown in the following tables:
Table (5.24) presents the Comparison of proposed hybrid model-III with Genetic Algorithm,
Conventional Priority List and Hopfield Neural Network methods.
Table (5.25) presents the comparison of the Summary of unit commitment schedules of the
proposed Hybrid Models I and III for 3 units System.
Table (5.26) gives the comparison of Number of Units in Operation for 3 unit systems for hybrid
models I and III with three single approaches.
Table (5.27) shows the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained from the Proposed Hybrid Model-III: (PMAX-FLAPC2, ED based on average
load assigned method).
Table (5.28) shows the Unit Commitment Schedule and Power Sharing (MW) of the best
solution obtained from the proposed Hybrid Model-III (PMAX-FLAPC3) (ED based on lambda
Iteration method).
The Salient features of the conventional approaches and the proposed hybrid model-I in the light
of the observations from the results are as follows:
118
1. Table 5.24 presents the comparison of the proposed hybrid model-III with other methods
available in the literature i.e. Genetic Algorithm (GA), Conventional priority list (PL)
method and Hopfield Neural Network (HFNN).
2. Hybrid model-III gives better results than GA, conventional priority list and HFNN.
3. The daily cost saving of Hybrid Model-III amount to $389.28 compared with Genetic
Algorithm. This is equivalent to a percentage saving of 0.1926 as shown in Table 5.24.
4. The daily cost saving of Hybrid Model-III amount to $474.28 compared with Hopfield
Neural Network. This is equivalent to a percentage saving of 0.2346 as shown in Table
5.24.
5. The daily cost saving of Hybrid Model-III amount to $761.26 compared with
Conventional priority list. This is equivalent to a percentage saving of 0.37612 as shown
in Table 5.24.
6. Hybrid Model-III (PMAX-FLAPC2, average load assigned method) is applicable to both
convex and non-convex fuel cost curves. The operating fuel cost is $201640.215. This
method is very fast in speed in calculating the optimal schedule of the generators. This
method gives fuel cost $8.13 higher than hybrid model-III (ED method) as given in Table
5.24.
7. In the proposed hybrid models-I and III when the load is 500 MW only unit 1 is in
operation thereby giving saving in startup cost.
119
Table 5.24 Comparison of proposed Hybrid Model-III with Genetic Algorithm, Conventional Priority List and Hopfield Neural Network methods for 3 unit systems:
Test system-I:
Algorithm Daily Operating Cost ($)
Amount of Daily Saving (comparedwith GA)
% saving in fuel cost compared with GA
Amount of Daily Saving (compared with conventional PL)
% saving in fuel cost compared with PL
Amount of Daily Saving (compared with Hopfield Neural Network
% saving in fuel cost compared with Hopfield Neural Network
Genetic Algorithm
202021.360 - - - - - -
Conventional Priority List (FLAPC)
202393.346 - - - - - -
Hopfield Neural Network
202106.36 - - - - - -
Hybrid Model–III (PMAX-FLAPC2)
201640.215 381.15 0.1886 753.13 0.37211 466.15 0.2306
Hybrid Model–III (PMAX-FLAPC3)
201632.089 389.28 0.1926 761.26 0.37612 474.28 0.2346
120
Table 5.25 Comparison of the Summary of unit commitment schedules of the proposed Hybrid Models I and III for 3 Unit Systems: Test System –I
Single
approach- I
Single approach –
II
Proposed Single
approach-III
Hybrid Model-
III
Hybrid Model-
III
Hybrid Model-
I
Hybrid Model-I
Load (MW)
Complete Enumeration
Conventional priority list (FLAPC) with ED
UC(2,1,3)
Merit Order
(PMAX) with ED
UC(1,2,3)
Merit Order
(PMAX-FLAPC-
2) UC(1,2,3)ED(2,1,3)
Merit Order
(PMAX-FLAPC-3) with
ED UC(1,2,3
)
DP-PSO-ANN (SI-
learning)
DP-BP-ANN (BP-
learning)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500 500 500 600 800 850 900 950 1000 1050 1100 1200 1200
111 111 111 111 110 110 110 100 100 011 011 011 011 011 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 110 110 110 110 110 110 110 110 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
201415. 789
202393. 346
201632. 089
201640. 215
201632. 089
201632. 089
201630. 923
No. of Transitions
No. of Transitions
No. of Transitions
No. of Transitions
No. of Transitions
No. of Transitions
No. of Transitions
10 2 4 4 4 4 4
1 on state 0 off state
121
Table 5.26 Comparison of Number of Units in Operation for 3 unit systems for hybrid models I and III with three single approaches: Test System I
Single approach-
I
Single approach-
II
Proposed Single
approach-III
Hybrid Model-
I
Hybrid Model-
I
Hybrid Model-
III
Hybrid Model-
III
Complete Enumerat-
ion
(Conventi-onal priority
list )
(PMAX) with ED
DP- PSO-ANN
DP- BP-ANN
(PMAX-FLAPC2)
(PMAX-FLAPC3)
Hr.
Load
units units units Units units units units
1 1200 3 3 3 3 3 3 3 2 1200 3 3 3 3 3 3 3 3 1150 3 3 3 3 3 3 3 4 1100 3 3 3 3 3 3 3 5 1000 2 2 2 2 2 2 2 6 900 2 2 2 2 2 2 2 7 800 2 2 2 2 2 2 2 8 600 1 2 1 1 1 1 1 9 550 1 2 1 1 1 1 1
10 500 2 2 1 1 1 1 1 11 500 2 2 1 1 1 1 1 12 500 2 2 1 1 1 1 1 13 500 2 2 1 1 1 1 1 14 500 2 2 1 1 1 1 1 15 600 1 2 1 1 1 1 1 16 800 2 2 2 2 2 2 2 17 850 2 2 2 2 2 2 2 18 900 2 2 2 2 2 2 2 19 950 2 2 2 2 2 2 2 20 1000 2 2 2 2 2 2 2 21 1050 3 3 3 3 3 3 3 22 1100 3 3 3 3 3 3 3 23 1200 3 3 3 3 3 3 3 24 1200 3 3 3 3 3 3 3
122
Table 5.27 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the Proposed Hybrid Model-III: (PMAX-FLAPC2, ED based on average load
assigned method): Test system-I
Power Output of each unit(MW)
Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 3 111 600.00 400.00 150.00 1150 5875.320 3760.400 1658.340 11294.060 4 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 400.00 400.00 0 800 3978.920 3760.400 0 7739.320 8 100 600.00 0 0 600 5875.320 0 0 5875.320 9 100 550.00 0 0 550 5389.505 0 0 5389.505 10 100 500.00 0 0 500 4911.500 0 0 4911.500 11 100 500.00 0 0 500 4911.500 0 0 4911.500 12 100 500.00 0 0 500 4911.500 0 0 4911.500 13 100 500.00 0 0 500 4911.500 0 0 4911.500 14 100 500.00 0 0 500 4911.500 0 0 4911.500 15 100 600.00 0 0 600 5875.320 0 0 5875.320 16 110 400.00 400.00 0 800 3978.920 3760.400 0 7739.320 17 110 450.00 400.00 0 850 4441.305 3760.400 0 8201.705 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 600.00 400.00 50.00 1050 5875.320 3760.400 586.260 10221.980 22 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480
SUM 201640.215
123
Table 5.28 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the proposed Hybrid Model-III (PMAX-FLAPC3) (ED based on lambda Iteration
method): Test system-I
Power output of each unit(MW)
Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 3 111 600.00 400.00 150.00 1150 5875.320 3760.400 1658.340 11294.060 4 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 100 600.00 0.00 0 600 5875.320 0.000 0 5875.320 9 100 550.00 0.00 0 550 5389.505 0.000 0 5389.505 10 100 500.00 0.00 0 500 4911.500 0.000 0 4911.500 11 100 500.00 0.00 0 500 4911.500 0.000 0 4911.500 12 100 500.00 0.00 0 500 4911.500 0.000 0 4911.500 13 100 500.00 0.00 0 500 4911.500 0.000 0 4911.500 14 100 500.00 0.00 0 500 4911.500 0.000 0 4911.500 15 100 600.00 0.00 0 600 5875.320 0.000 0 5875.320 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 600.00 400.00 50.00 1050 5875.320 3760.400 586.260 10221.980 22 111 600.00 400.00 100.00 1100 5875.320 3760.400 1107.840 10743.560 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 2237.760 11873.480
SUM 201632.089
5.7.2 Numerical Results of Test System-II: Three units system: Hybrid Model-III:
(Convex Fuel Cost Curve)
In this study the fuel cost of the unit-3 which is coal fired unit is decreased from 1.2 to
1.0, whereas the fuel cost of unit 1 and 2 is same as in Test System-I. By decrease in fuel cost of
unit 3, the unit 3 which is uneconomical unit in study-1 and has the lowest priority now becomes
124
the most economical unit in this study. The priority of units based on FLAPC in this study is
Unit 3, 2, 1. Whereas the priority in test system-I is unit 2, 1, 3.
The output results of the test system-II are shown in the following tables:
Table (5.29) shows the comparison of the operating fuel cost ($) for proposed Hybrid Models -I
and III for 3 unit system.
Table (5.30) gives the summary of Unit Commitment Schedules for 3 unit systems
Table (5.31) presents the summary of number of Units in Operation for 3 unit system.
Table (5.32) presents the Unit commitment schedule and Power Sharing (MW) of the best
solution obtained by Hybrid Model-III (PMAX-FLAPC with ED based on average load) for 3
unit systems.
Table (5.33) presents the Unit commitment schedule and Power Sharing (MW) of the best
solution obtained by Hybrid Model-III (PMAX-FLAPC with ED) for 3 unit systems.
The salient features of the proposed approaches in the light of the observations from the
results are as follows:
1. Hybrid model-I using SI-learning hits the target values while the BP-learning give results
very close to the target values.
2. Hybrid model-III gives high quality solution very quickly.
Table 5.29 Comparison of the operating fuel cost ($) for proposed Hybrid Models -I and III
for 3 unit system: Test System –II (Convex Fuel Cost Curve)
Proposed Hybrid Approaches
Hybrid Model-I (SI-Learning)
Hybrid Model-I (BP-Learning)
Hybrid Model-III (ED based on load to Economic unit)
Hybrid Model-III (Lambda Iteration
Method) DP-PSO-ANN DP-BP-ANN (PMAX-FLAPC2)
UC(1,2,3) ED(3,2,1)
(PMAX-FLAPC3) with ED
Operating fuel cost ($)
199262.671 199262.310 199299.8125 199262.671
125
Table 5.30 Summary of Unit Commitment Schedules for 3 unit systems: Test System II:
Hybrid Model-I
Hybrid Model-I
Hybrid Model-III
Hybrid Model-III
Load (MW)
DP-PSO-ANN DP-BP-ANN (PMAX-LAPC2)
(PMAX-FLAPC3)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500 500 500 600 800 850 900 950 1000 1050 1100 1200 1200
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
111 111 111 111 110 110 110 100 100 100 100 100 100 100 100 110 110 110 110 110 111 111 111 111
199262.671 199262.310 199299.812 199262.671 No. of
Transitions No. of
Transitions No. of
Transitions No. of
Transitions 4 4 4 4
126
Table 5.31 Summary of number of Units in Operation for 3 unit system: Test system II:
Hybrid Model-I
Hybrid Model-I
Hybrid Model-III
Hybrid Model-III
DP-PSO-ANN DP-BP-ANN (PMAX-FLAPC2)
(PMAX-FLAPC3)
Hr Load units units units units 1 1200 3 3 3 3 2 1200 3 3 3 3 3 1150 3 3 3 3 4 1100 3 3 3 3 5 1000 2 2 2 2 6 900 2 2 2 2 7 800 2 2 2 2 8 600 1 1 1 1 9 550 1 1 1 1 10 500 1 1 1 1 11 500 1 1 1 1 12 500 1 1 1 1 13 500 1 1 1 1 14 500 1 1 1 1 15 600 1 1 1 1 16 800 2 2 2 2 17 850 2 2 2 2 18 900 2 2 2 2 19 950 2 2 2 2 20 1000 2 2 2 2 21 1050 3 3 3 3 22 1100 3 3 3 3 23 1200 3 3 3 3 24 1200 3 3 3 3
127
Table 5.32 Unit commitment schedule and Power Sharing (MW) of the best solution obtained by Hybrid Model-III (PMAX-FLAPC with ED based on average load) for 3 unit systems:
Test System-II
Power output of each unit(MW)
Fuel Cost of each unit($/h)
Hr.
UC Sch.
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 3 111 550.00 400.00 200.00 1150 5389.505 3760.400 1864.800 11014.705 4 111 500.00 400.00 200.00 1100 4911.500 3760.400 1864.800 10536.700 5 111 600.00 400.00 0.00 1000 5875.320 3760.400 0.000 9635.720 6 111 500.00 400.00 0.00 900 4911.500 3760.400 0.000 8671.900 7 110 400.00 400.00 0.00 800 3978.920 3760.400 0.000 7739.320 8 100 600.00 0.00 0.00 600 5875.320 0.000 0.000 5875.320 9 100 550.00 0.00 0.00 550 5389.505 0.000 0.000 5389.505 10 100 500.00 0.00 0.00 500 4911.500 0.000 0.000 4911.500 11 100 500.00 0.00 0.00 500 4911.500 0.000 0.000 4911.500 12 100 500.00 0.00 0.00 500 4911.500 0.000 0.000 4911.500 13 100 500.00 0.00 0.00 500 4911.500 0.000 0.000 4911.500 14 100 500.00 0.00 0.00 500 4911.500 0.000 0.000 4911.500 15 100 600.00 0.00 0.00 600 5875.320 0.000 0.000 5875.320 16 110 400.00 400.00 0.00 800 3978.920 3760.400 0.000 7739.320 17 110 450.00 400.00 0.00 850 4441.305 3760.400 0.000 8201.705 18 110 500.00 400.00 0.00 900 4911.500 3760.400 0.000 8671.900 19 110 550.00 400.00 0.00 950 5389.505 3760.400 0.000 9149.905 20 110 600.00 400.00 0.00 1000 5875.320 3760.400 0.000 9635.720 21 111 450.00 400.00 200.00 1050 4441.305 3760.400 1864.800 10066.505 22 111 500.00 400.00 200.00 1100 4911.500 3760.400 1864.800 10536.700 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520
SUM 199299.825
128
Table 5.33 Unit commitment schedule and Power Sharing (MW) of the best solution obtained by Hybrid Model-III (PMAX-FLAPC with ED) for 3 unit systems:
Test System-II
Power output of each unit(MW)
Fuel Cost of each unit($/h)
Hr.
UC Sch.
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 2 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 3 111 570.35 400.00 179.65 1150 5586.289 3760.400 1665.372 11012.061 4 111 532.59 400.00 167.41 1100 5222.177 3760.400 1547.344 10529.921 5 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 6 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 7 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 8 100 600.00 0 0 600 5875.320 0 0 5875.320 9 100 550.00 0 0 550 5389.505 0 0 5389.505 10 100 500.00 0 0 500 4911.500 0 0 4911.500 11 100 500.00 0 0 500 4911.500 0 0 4911.500 12 100 500.00 0 0 500 4911.500 0 0 4911.500 13 100 500.00 0 0 500 4911.500 0 0 4911.500 14 100 500.00 0 0 500 4911.500 0 0 4911.500 15 100 600.00 0 0 600 5875.320 0 0 5875.320 16 110 433.18 366.82 0 800 4284.887 3450.577 0 7735.464 17 110 460.88 389.12 0 850 4542.955 3658.336 0 8201.290 18 110 500.00 400.00 0 900 4911.500 3760.400 0 8671.900 19 110 550.00 400.00 0 950 5389.505 3760.400 0 9149.905 20 110 600.00 400.00 0 1000 5875.320 3760.400 0 9635.720 21 111 494.83 400.00 155.17 1050 4862.520 3760.400 1430.760 10053.679 22 111 532.59 400.00 167.41 1100 5222.177 3760.400 1547.344 10529.921 23 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520 24 111 600.00 400.00 200.00 1200 5875.320 3760.400 1864.800 11500.520
SUM 199262.671 5.7.3 Numerical Results of Test System-III: Ten units system: Hybrid Model-III: (Convex
fuel cost function)
Using the priority list in Table 5.23, the units are committed based on their priority with the
highest priority (higher capacity) being on first followed by other units. Units are committed
until the load demand plus the spinning reserve requirements are fulfilled.
129
The output results of test system – III are shown in the following tables:
Table (5.34) shows the comparison of the best results of the Hybrid Model-III with other
approaches available in the literature.
Table (5.35) presents the daily saving and percentage saving in fuel cost compared with other
approaches.
Table (5.36) and Table (5.37) presents the Unit Commitment Schedule, Power Sharing (MW) ,
Load demand and Total Production Cost (TPC) of the best solution obtained from proposed
Hybrid Model-III, with MUT and MDT constraints with 10 % spinning reserve: (ED by load to
economic unit).
Table (5.38) and Table (5.39) shows the detailed results of Unit Commitment Schedule, Power
Sharing (MW), Load demand and Total Production Cost (T.P.C) of the best solution obtained
from proposed Hybrid Model-III, with MUT and MDT constraints with 10 % spinning reserve.
(ED by lambda iteration method)
Table (5.40) presents the summary of Unit Commitment schedules for 10 unit systems with 10%
spinning reserve and considering minimum up time and down time constraints.
Table (5.41) shows the Comparison of Transition Cost for 10 unit system without considering s.r
and MUT and MDT.
Table (5.42) shows the Comparison of operating fuel cost ($) for 10 unit systems considering
10% spinning reserve and considering minimum up/down time constraints without transition
cost.
Table (5.43) presents the Comparison of the operating fuel cost ($) for 10 unit systems
considering 10% spinning reserve, considering minimum up/down time constraints and transition
cost.
Table (5.44) gives the Comparison of the results of the proposed Hybrid Models-I and III with
Genetic Algorithm.
130
The Salient features of the proposed approach in the light of the observations from the results
are as follows:
1. When compared with other 30 approaches available in the literature. The proposed hybrid
models III gives low total production cost compared with other approaches as given in
Table 5.33.
2. Hybrid Model-I explores more search space than Hybrid model-III. At hour 23, Table
5.34 when the load is 900 MW units 6 is committed and unit 5 is off in model-I, thus
gives a optimum unit commitment schedule at hour 23 and reduction in operating fuel
cost. The total production cost in this case is $563937.98.
3. The daily costs saving of the proposed Hybrid model-III compared with other approaches
have a range of $28.00 to $24773.00 per day. Which is equivalent to a percentage saving
of 0.00496 to 04.3925 as given in Table 5.34?
4. Table 5.40 gives the comparison of transition cost. The operating fuel cost in
Enumeration approach is $543479.1, which is less than the proposed Hybrid model-III,
but the start up cost is $8180, which is very high compared with proposed model. The
startup cost of the proposed model is $4900. The total production cost of the proposed
Hybrid model-III is $1604 less than single approach-I. Therefore, the proposed hybrid
model-III fulfills the requirement of the objectives of the UC problem, to minimize the
fuel cost, and transition cost.
5. The total production cost (operating fuel cost + transition cost) of the hybrid models-I and
III is low compared with single approaches.
6. The daily costs saving of the proposed Hybrid model-III compared with Genetic
Algorithm has $ 1847.98 per day. Which is equivalent to a percentage saving of 0.326 as
given in Table 5.43?
7. The unit scheduling problem is basically depends upon the spinning reserve
requirements. With less spinning reserve required less number of units is committed. For
5% or 10% spinning reserve the unit schedule is different. The UC schedule and hence
the total production cost depends upon spinning reserve. In this study 10 % spinning
reserve is assumed.
8. Comparing the results of the proposed hybrid model-III for Test System-III, it is clear
that the hybrid model III, provides a better quality of solution at a faster speed. Hybrid
131
Model -III is simple and more efficient than conventional priority list method. The fast
preparation of UC schedules and fast economic dispatch calculation leads to a
deterministic and efficient method.
Table 5.33 Comparison of the best results of the Hybrid Model-III with other approaches
available in the literature: Test system-III
Approach DP [56]
LR [56]
ELR [36]
EPL [15]
SPL [16]
Total production cost($)
565825 565825 563977 563977 564950
Approach HPSO [80]
RPACO LR-PPSO GA-UCC [57]
EMOALHN
TPC ($) 563942 565302 563977 563977 563977 Approach FPGA
[60] PSO [65]
MRCGA [61]
GA-LR LR-PSO ALR
TPC ($) 564094 564212 564244 564800 565275 565508 Approach GA
[56] GRASP MA ICGA
[59] DPHNN BPSO
TPC ($) 565825 565825 565827 566404 588750 565804 Approach AG EP
[52] ESA [44]
ASSA [45]
AG SA
TPC ($) 564005 564551 565828 563938 564005 565825 Approach PLEA
[82] DP-LR EP-LR ACSA RPACO
TPC ($) 563977 564049 564049 564059 565302 Proposed Hybrid Model-III 56397564875.617.02 Proposed Hybrid Model-III 56397563977.027.02
• Abbreviations are given in Appendix-B.
132
Table 5.35 Daily Saving and Percentage saving in fuel cost compared with other approaches
S. No.
Approach
Total Fuel Cost ($/h)
Amount of daily saving ($/h) compared with proposed hybrid
model –III
% Cost saving in fuel cost compared
with proposed model-III
1 ELR 563977 0 0 2 EPL 563977 0 0 3 LR-PPSO 563977 0 0 4 GA-UCC 563977 0 0 5 EMOALHN 563977 0 0 6 PLEA 563977 0 0 7 AG 564005 28.00 0.00496 8 DP-LR 564049 72.00 0.01276 9 EP-LR 564049 72.00 0.01276 10 ACSA 564059 82.00 0.01453 11 FPGA 564094 117.00 0.0207 12 PSO 564212 235.00 0.0416 13 MRCGA 564244 245.00 0.0434 14 EP 564551 574.00 0.0107 15 GA-LR 564800 823.00 0.1459 16 SPL 564950 973.00 0.1725 17 LR-PSO-1 565275 1298.00 0.2301 18 RPACO 565302 1325.00 0.2349 19 ALR 565508 1531.00 0.2714 20 BPSO 565804 1827.00 0.3239 21 DP 565825 1848.00 0.3276 22 LR 565825 1848.00 0.3276 23 SA 565825 1848.00 0.3276 24 GA 565825 1848.00 0.3276 25 GRASP 565825 1848.00 0.3276 26 MA 565827 1850.00 0.3280 27 LR-PSO-2 565869 1892.00 0.3354 28 ICGA 566404 2427.00 0.4303 29 BCGA 567367 3390.00 0.6010 30 DPHNN 588750 24773.00 4.3925 31 Proposed approach 563977
133
Table 5.36 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from proposed Hybrid Model-III, with MUT and MDT constraints with 10 % spinning
reserve: (ED by average load to economic unit): Test system-III:
Power output of each unit (MW) Hour Unit Commitment
schedule 12543768910
Load (MW)
1 2 3 4 5 6 7 8 9 10 1 1100000000 700 455 245 0 0 0 0 0 0 0 02 1100000000 750 455 295 0 0 0 0 0 0 0 03 1100100000 850 455 370 0 0 25 0 0 0 0 04 1100100000 950 455 455 0 0 40 0 0 0 0 05 1101100000 1000 455 455 0 65 25 0 0 0 0 06 1111100000 1100 455 455 35 130 25 0 0 0 0 07 1111100000 1150 455 455 85 130 25 0 0 0 0 08 1111100000 1200 455 455 130 130 30 0 0 0 0 09 1111111000 1300 455 455 130 130 85 20 25 0 0 010 1111111100 1400 455 455 130 130 162 33 25 10 0 011 1111111110 1450 455 455 130 130 162 73 25 10 10 012 1111111111 1500 455 455 130 130 162 80 58 10 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 014 1111111000 1300 455 455 130 130 85 20 25 0 0 015 1111100000 1200 455 455 130 130 30 0 0 0 0 016 1111100000 1050 455 455 20 95 25 0 0 0 0 017 1111100000 1000 455 455 20 45 25 0 0 0 0 018 1111100000 1100 455 455 35 130 25 0 0 0 0 019 1111100000 1200 455 455 130 130 30 0 0 0 0 020 1111111100 1400 455 455 130 130 162 33 25 10 0 021 1111111000 1300 455 455 130 130 85 20 25 0 0 022 1111111000 1100 455 455 20 100 25 20 25 0 0 023 1100100000 900 455 420 0 0 25 0 0 0 0 024 1100000000 800 455 345 0 0 0 0 0 0 0 0
134
Table 5.37 Load demand , fuel cost , start up cost and Total Production Cost (TPC) obtained from proposed Hybrid Model-III, with MUT and MDT constraints with 10 % spinning reserve.
(ED by average load to economic unit): Test system-III
Hour Load (MW)
Fuel cost ($/h)
Startup Cost ($/h)
Total Production
Cost ($)
Spinning reserve (MW)
1 700 13683.13 0 13683.13 210 2 750 14554.50 0 14554.50 160 3 850 16809.45 900 17709.45 222 4 950 18597.67 0 18597.67 122 5 1000 20059.70 560 20619.70 202 6 1100 22442.40 1100 23542.40 232 7 1150 23284.40 0 23284.40 182 8 1200 24150.34 0 24150.34 132 9 1300 27251.06 860 28111.06 197 10 1400 30057.55 60 30117.55 152 11 1450 31916.06 60 31976.06 157 12 1500 33945.16 60 34005.16 162 13 1400 30057.55 0 30057.55 152 14 1300 27251.06 0 27251.06 197 15 1200 24150.34 0 24150.34 132 16 1050 21597.63 0 21597.63 282 17 1000 20757.86 0 20757.86 332 18 1100 22442.40 0 22442.40 232 19 1200 24150.34 0 24150.34 132 20 1400 30057.55 490 30547.55 152 21 1300 27251.06 0 27251.06 197 22 1100 23674.23 0 23674.23 137 23 900 17684.69 0 17684.69 172 24 800 15427.42 0 15427.42 110
Sum 21700 561253.53 4090.00 565343.53 4357
135
Table 5.38 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from proposed Hybrid Model-III, with MUT and MDT constraints with 10 % spinning
reserve. (ED by lambda iteration method): Test system-III
Power output of each unit (MW) Hr. Unit Commitment
Schedule
Load (MW)
1 2 3 4 5 6 7 8 9 101 1100000000 700 455 245 0 0 0 0 0 0 0 0 2 1100000000 750 455 295 0 0 0 0 0 0 0 0 3 1100100000 850 455 370 0 0 25 0 0 0 0 0 4 1100100000 950 455 455 0 0 40 0 0 0 0 0 5 1101100000 1000 455 390 0 130 25 0 0 0 0 0 6 1111100000 1100 455 360 130 130 25 0 0 0 0 0 7 1111100000 1150 455 410 130 130 25 0 0 0 0 0 8 1111100000 1200 455 455 130 130 30 0 0 0 0 0 9 1111111000 1300 455 455 130 130 85 20 25 0 0 0 10 1111111100 1400 455 455 130 130 162 33 25 10 0 0 11 1111111110 1450 455 455 130 130 162 73 25 10 10 0 12 1111111111 1500 455 455 130 130 162 80 25 43 10 1013 1111111100 1400 455 455 130 130 162 33 25 10 0 0 14 1111111000 1300 455 455 130 130 85 20 25 0 0 0 15 1111100000 1200 455 455 130 130 30 0 0 0 0 0 16 1111100000 1050 455 310 130 130 25 0 0 0 0 0 17 1111100000 1000 455 260 130 130 25 0 0 0 0 0 18 1111100000 1100 455 360 130 130 25 0 0 0 0 0 19 1111100000 1200 455 455 130 130 30 0 0 0 0 0 20 1111111100 1400 455 455 130 130 162 33 25 10 0 0 21 1111111000 1300 455 455 130 130 85 20 25 0 0 0 22 1111111000 1100 455 455 0 0 145 20 25 0 0 0 23 1100100000 900 455 420 0 0 25 0 0 0 0 0 24 1100000000 800 455 345 0 0 0 0 0 0 0 0
136
Table 5.39 Load demand , fuel cost , start up cost and Total Production Cost (TPC) of the best solution obtained from proposed Hybrid Model-III, with MUT and MDT constraints with 10
% spinning reserve: Test system-III (ED by Lambda iteration method)
Hour Load (MW)
Fuel cost ($/h)
Startup Cost ($/h)
Total Production
Cost ($)
Spinning reserve (MW)
1 700 13683.13 0 13683.13 210 2 750 14554.50 0 14554.50 160 3 850 16809.45 900 17709.45 222 4 950 18597.67 0 18597.67 122 5 1000 20020.02 560 20580.02 202 6 1100 22387.04 1100 23487.04 232 7 1150 23261.98 0 23261.98 182 8 1200 24150.34 0 24150.34 132 9 1300 27251.06 860 28111.06 197 10 1400 30057.55 60 30117.55 152 11 1450 31916.06 60 31976.06 157 12 1500 33890.16 60 33950.16 162 13 1400 30057.55 0 30057.55 152 14 1300 27251.06 0 27251.06 197 15 1200 24150.34 0 24150.34 132 16 1050 21513.66 0 21513.66 282 17 1000 20641.82 0 20641.82 332 18 1100 22387.04 0 22387.04 232 19 1200 24150.34 0 24150.34 132 20 1400 30057.55 490 30547.55 152 21 1300 27251.06 0 27251.06 197 22 1100 22735.52 0 22735.52 137 23 900 17684.69 0 17684.69 172 24 800 15427.42 0 15427.42 110
Sum 27100 559887.02 4090.00 563977.02 4357
137
Table 5.40 Summary of Unit Commitment schedules for 10 unit systems with 10% spinning reserve and considering minimum up time and down time constraints:
Test System III
Hybrid Model-I
Hybrid Model-I
Hybrid Model-III
Hybrid Model-III
DP-PSO-ANN SI-Learning
DP-BP-ANN BP-Learning
(PMAX-FLAPC2)
(PMAX-FLAPC3)
1100000000 1100000000 1100100000 1100100000 1101100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100010000 1100000000
1100000000 1100000000 1100100000 1100100000 1101100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100010000 1100000000
1100000000 1100000000 1100100000 1100100000 1101100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100100000 1100000000
1100000000 1100000000 1100100000 1100100000 1101100000 1111100000 1111100000 1111100000 1111111000 1111111100 1111111110 1111111111 1111111100 1111111000 1111100000 1111100000 1111100000 1111100000 1111100000 1111111100 1111111000 1111111000 1100100000 1100000000
559847.98 559845.58 560775.61 559887.01
138
Table 5.41 Comparison of Transition Cost for 10 unit system without considering s. r and MUT and MDT: Test System III (10 unit system)
Operating
fuel cost ($/h)
Start-upcost ($/h)
Total Production
Cost ($)
Difference in T.P.C
($)
% Saving in total production cost(T.P.C)
Single Approach-I (Enumeration)
543479.10
8180.00
551659.10
0
0
Proposed Hybrid Method-III 545154.82
4900.00
550054.82
1604.28
0.29081
Table5.42 Start up cost comparison
0
2000
4000
6000
8000
10000
Enumeration Proposed approach
Series1
Table 5.43 Comparison of operating fuel cost ($) for 10 unit systems considering 10% spinning reserve and considering minimum up/down time constraints without transition cost:
Test System III
Single Approaches
Single approach–I
Single approach-II
Single approach–III
UC Schedule
UC12534768910 UC12435678910 UC12534768910
Operating fuel cost ($)
559887.02 561682.98 560775.6
Hybrid Models
Hybrid Model–I Hybrid Model–I Hybrid Model–III
DP-PSO-ANN
DP-BP-ANN
(PMAX-FLAPC)
Operating fuel cost ($)
559847.98 559845.58 559887.02
139
Table 5.44 Comparison of the operating fuel cost ($) for 10 unit systems considering 10% spinning reserve, considering minimum up/down time constraints and transition cost: Test
System III
Single Approaches
Single approach- I
(Enumeration)
Single approach-II
(Conventional Priority List)
Single approach– III
(Proposed Priority List)
UC Schedule
12543768910 12435678910 12534768910
Operating fuel cost ($)
560744.47 561682.98 560775.6
Transition cost ($) 4090 4440 4100 Total Production Cost ($)
564834.47 566122.98 564875.61
% Cost Saving compared with conventional priority list
-
-
0.22
Hybrid Approaches
Proposed Hybrid Model –I
Proposed Hybrid Model –II
Proposed Hybrid Model–III
DP-PSO-ANN DP-BP-ANN (PMAX-FLAPC3) UC Schedule
12543768910 12543768910 12543768910
Operating fuel cost ($)
559847.98 559845.58 559887.02
Transition Cost ($) 4090 4090 4090 Total Production Cost ($)
563937.98 563935.58 563977.02
% Cost Saving compared with conventional priority list
0.385
0.386
0.379
5.8 Case Studies: Hybrid model-III ---Non-Convex fuel cost function
The following two standard test systems have been selected for the illustration of the
effectiveness of the proposed approach.
a. Test System IV --- 3 units with 24 hours load.
b. Test System-V --- 3 units with 24 hours load
Input data: The description and data of test system used for investigation in the case studies is
given in Table A.4 and Table A.5 placed in Appendix A.
140
Computer Implementation: The algorithms have been implemented in C++ on P-IV Personal
Computer.
Output Results: The summary and comparison of results is given in this section.
5.8.1 Numerical results of test systems – IV and V: 3 units systems: Hybrid Model- III: (non-convex fuel cost function)
The output results of test systems IV and V are shown in the following tables:
Table (5.45) gives the Comparison of the Operating fuel cost ($) for 3 unit system obtained by
using Proposed Hybrid Model-III compared with Hybrid model –II.
Tables (5.46) and (5.47) showing the load demand, Unit Commitment Schedule, Power Sharing
(MW) of the committed units and Total Production Cost (TPC) in the 24 hours.
The salient features of the proposed approaches in the light of the observations from the results are as follows:
1. The G.A. approach is suitable for non convex curves. Lambda iteration method fails for
non convex curves. Hybrid Model-III (load to economic unit) is also applicable for non
convex curves, but the results obtained by fine tuning by using Hybrid Model-II (GA-
PSO-ANN and GA-BP-ANN) are encouraging.
2. Hybrid model-II gives better results in terms of solution quality and gives a reduction of
$ 318.49 and $ 478.42 per day in operating fuel cost compared with Hybrid Model-III for
test systems IV and V respectively as given in Table 5.45.
Table 5.45 Comparison of the Operating fuel cost ($) for 3 unit system obtained by using
Proposed Hybrid Model-III compared with Hybrid model -II: (Non-convex)
Hybrid Model-II Hybrid Model-III
Reduction in fuel cost ($/day)
(GA-PSO-ANN)
(GA-BP- ANN)
Test system-IV
Operating fuel cost
$ 207773.614 $ 207772.477 $ 208092.107 $318.49
Test system-V
Operating fuel cost
$ 200665.137 $ 200662.723 $ 201143.551 $478.72
141
Table 5.46 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the Hybrid Model-III (PMAX-FLAPC3): (non-convex):
Test system-IV:
Power Output of each unit(MW)
Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($/h)
1 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 2 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 3 111 600.00 400.00 150.00 1150 6175.105 3767.125 1660.862 11603.092 4 111 600.00 400.00 100.00 1100 6175.105 3767.125 1109.101 11051.331 5 110 600.00 400.00 0 1000 6175.105 3767.125 0 9942.230 6 110 500.00 400.00 0 900 5211.370 3767.125 0 8978.495 7 110 400.00 400.00 0 800 4278.854 3767.125 0 8045.978 8 100 600.00 0 0 600 6175.105 0 0 6175.105 9 100 550.00 0 0 550 5399.592 0 0 5399.592 10 100 500.00 0 0 500 5211.370 0 0 5211.370 11 100 500.00 0 0 500 5211.370 0 0 5211.370 12 100 500.00 0 0 500 5211.370 0 0 5211.370 13 100 500.00 0 0 500 5211.370 0 0 5211.370 14 100 500.00 0 0 500 5211.370 0 0 5211.370 15 100 600.00 0 0 600 6175.105 0 0 6175.105 16 110 400.00 400.00 0 800 4278.854 3767.125 0 8045.978 17 110 450.00 400.00 0 850 4448.871 3767.125 0 8215.995 18 110 500.00 400.00 0 900 5211.370 3767.125 0 8978.495 19 110 550.00 400.00 0 950 5399.592 3767.125 0 9166.717 20 110 600.00 400.00 0 1000 6175.105 3767.125 0 9942.230 21 111 600.00 400.00 50.00 1050 6175.105 3767.125 586.260 10528.490 22 111 600.00 400.00 100.00 1100 6175.105 3767.125 1109.101 11051.331 23 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773 24 111 600.00 400.00 200.00 1200 6175.105 3767.125 2241.543 12183.773
SUM 208092.107
142
Table 5.47 Unit Commitment Schedule and Power Sharing (MW) of the best solution obtained from the Hybrid Model-III (non-convex UC123, ED 321): Test system-V
Power Output of each
unit(MW) Fuel Cost of each unit($/h)
Hour
UC Schedule
P-1 P-2 P-3
Load (MW)
F-1 F-2 F-3
Operating Fuel Cost
($)
1 111 600.00 400.00 200.00 1200 5887.927 3767.125 1869 11523.635 2 111 600.00 400.00 200.00 1200 5887.927 3767.125 1869 11523.635 3 111 550.00 400.00 200.00 1150 5689.290 3767.125 1869 11324.998 4 111 500.00 400.00 200.00 1100 4921.587 3767.125 1869 10557.294 5 110 600.00 400.00 0 1000 5887.927 3767.125 0 9655.052 6 110 500.00 400.00 0 900 4921.587 3767.125 0 8688.712 7 110 400.00 400.00 0 800 3986.486 3767.125 0 7753.610 8 100 600.00 0 0 600 5887.927 0 0 5887.927 9 100 550.00 0 0 550 5689.290 0 0 5689.290 10 100 500.00 0 0 500 4921.587 0 0 4921.587 11 100 500.00 0 0 500 4921.587 0 0 4921.587 12 100 500.00 0 0 500 4921.587 0 0 4921.587 13 100 500.00 0 0 500 4921.587 0 0 4921.587 14 100 500.00 0 0 500 4921.587 0 0 4921.587 15 100 600.00 0 0 600 5887.927 0 0 5887.927 16 110 400.00 400.00 0 800 3986.486 3767.125 0 7753.610 17 110 450.00 400.00 0 850 4741.175 3767.125 0 8508.300 18 110 500.00 400.00 0 900 4921.587 3767.125 0 8688.712 19 110 550.00 400.00 0 950 5689.290 3767.125 0 9456.415 20 110 600.00 400.00 0 1000 5887.927 3767.125 0 9655.052 21 111 450.00 400.00 200.00 1050 4741.175 3767.125 1869 10376.883 22 111 500.00 400.00 200.00 1100 4921.587 3767.125 1869 10557.294 23 111 600.00 400.00 200.00 1200 5887.927 3767.125 1869 11523.635 24 111 600.00 400.00 200.00 1200 5887.927 3767.125 1869 11523.635
SUM 201143.551
143
CHAPTER 6
Unit Commitment of National Transmission & Despatch
Company Limited (NTDC)
6.1 Introduction
This chapter gives discussion of the National Transmission and Despatch Company
(NTDC) network, its operational problem, and the results of proposed hybrid models-II and III
for convex and non convex cost function.
6.2 WAPDA --- Brief Overview
The Pakistan Water and Power Development Authority, (WAPDA) was created in 1958
as a Semi-Autonomous Body with water and power wing. WAPDA worked in vertically
integrated environment up to 2001.
Since October 2007, WAPDA has been bifurcated into two distinct entities i.e. WAPDA
and Pakistan Electric Power Company (PEPCO). WAPDA is responsible for water and
hydropower development whereas PEPCO is vested with the responsibility of thermal power
generation, transmission, distribution and billing.
WAPDA is now fully responsible for the development of Hydel Power and Water Sector
Projects. PEPCO has been fully responsible for the management of all the affairs of nine
Distribution Companies (DISCOs), four Generation Companies (GENCOs) and a National
Transmission and Despatch Company (NTDC. The public sector hydel and thermal generation
are in the control of WAPDA and GENCOs respectively. Independent Power Producers (IPPs)
are in private sector.
144
6.3 National Transmission and Despatch Company
National transmission and despatch company links the power generation units and load
centers dotting the entire country, thus creating one of the largest contiguous grid systems of the
world. At present NTDC is operating and maintaining nine 500 kV and twenty four 220 kV grid
stations along with 10,167 km length of associated transmission lines [111]. NTDC power
system has following mainly five types of power stations connected to the National Grid system:
a. Hydro power stations
b. Steam power stations
c. Gas turbine power stations
d. Combined cycle power plant stations.
e. Nuclear power stations.
The generating units are loaded according to the merit order determined by their cost of
operation and synchronized with system with the rising trend of load curve. However sometimes
it is necessary to take generation at high operational cost subjected to constraints such as less
transmission or transformer capacity. All the functions of 500/220 kV power system and power
houses are monitored by SCADA system through R.T.Us installed at the grid/power stations.
NTDC is responsible to purchase the power from hydel stations in the north, thermal
units in public and private sectors installed mostly in the central southern regions of the country
and to sell power to distribution companies through its large network of transmission lines and
grid stations of 500 kV and 220 kV voltage capacities.
6.4 Operational Constraints in NTDC System
The various constraints in the NTDC system are:
6.4.1 Hydro-Electric Generation Constraints
Pakistan is the one of the most fortunate countries of the world having the lot of water
potentials. Their total estimated capacity is about 30,000 MW, and all are almost in the northern
areas of Pakistan. WAPDA's hydel power stations consists of five major station located at
Tarbela, Mangla, Warsak, Chashma, Ghazi Boratha and nine small hydel power stations. Two
145
hydel power stations are presently commissioned in private sector namely Jagran (AJK) and
Malakand ( NWFP).
Tarbela and Mangla dams are multi-purpose projects with main emphasis on irrigation
under the indus water treaty. Tarbela and Mangla reservoirs are the major hydel power
generation sources of the WAPDA system, representing the only significant capacity for the
seasonal storage of water. There are two principal effects of wide seasonal variations in hydel
generation capability. These are the variations in water releases from the reservoirs and in
hydraulic heads available for power generation. Operation of the NTDC system is in practice
dominated by these variations.
Water management i.e., the use of reservoir storage and planning of water release is
dominated by irrigation needs rather than power requirements. Seasonal water management
schedules are derived on the basis of data of current and historic water levels in the reservoirs.
The seasonal pattern of reservoir and hydro electric plant operation has considerable
implications for WAPDA system operation in general. The system experiences shortages of:
• Generation capacity and energy during late winter to early summer when reservoir levels
are low.
• Capacity during the period later in summer, before the reservoirs are filled.
• Energy during the canal closure period in January to early February when reservoir
releases are severely restricted.
Considerable load shedding takes place on NTDC system during the above periods. The load
distribution pattern of NTDC system is such that 75% of the total load is located to the north of
the Multan. So primary transmission system has to transfer the blocks of power from north to
south when full hydel generation capacity is available and from south to north in winter when
water is in short supply and reliance has to be placed on thermal generation.
146
6.4.2 Thermal Generation Constraints
Seasonal variations in thermal plant capability also have considerable influence on the
system operation. Gas turbines are inherently sensitive to ambient temperature. Hence the power
capabilities of gas turbine components of combined cycle plant are significantly lower in
summer.
6.4.3 Transmission line Constraints
Various transmission constraints have effected the system operation in recent years.
Power flows on NTDC network are from north to south in summer and from south to north in
winter and both the times power transfer is limited by transmission capability subjected to the
constraints such as line or transformer overloading, stability problems and difficulty with control
of voltage or reactive power.
6.4.4 Seasonal Variations in Power Demand
The seasonal variation in electricity demand also effects the system operation. This
variation arises principally from increased air-conditioning loads in summer and heating load in
winter.
6.4.5 Spinning Reserve Constraint
The load demand is more than the power generated. There is a short fall of about
5000MW.
6.4.6 Minimum up and down time constraints
For a 210 MW Thermal Power Station the minimum up and down times is approximately
3.5 and 1.5 hours respectively.
6.4.7 Start up cost considerations
Start up costs depends on the number of hours the unit has been off.
• Up to 8 hours : hot state
• After 8 hours and up to 150 hours : warm state
• After 150 hours : cold state
147
For a thermal power station the cold startup cost may be Rs.1, 53,869.52 and warm
Rs. 1, 17,890.08 approximately.
6.4.8 Maintenance cost
The maintenance cost consists of fixed maintenance cost and variable maintenance cost.
6.4.9 Fuel constraint
Due to limited fuel this constraint presents a most challenging unit commitment problem.
6.4.10 Ramping rates
Ramping rate is the maximum rate of change of the output in MW/min. Steam units have
ramping rate of 1MW/min.
6.4.11 Unit deration
Older units give output less than their installed capacity.
6.5 Test systems for NTDC system
The following four test systems with cost curves close to the original machines in the
system have also been prepared for unit commitment.
1. 12-Unit NTDC Test Circuit
2. 15-Unit NTDC Test Circuit
3. 25-Unit NTDC Test Circuit
4. 34-Unit NTDC Test Circuit
However in the discussion to follow the Hybrid models IV and V for unit commitment will be
investigated.
6.6 Case Studies
The cost curve characteristics and load data for 12 machine system are given in Table A.7
placed in Appendix A. The cost curves of the 15, 25, 34 units are given in Tables A. 8, A. 9 and
A.10 respectively placed in appendix A.
148
6.7 Numerical results
Table 6.1 Comparison of the results for NTDC 12, 15, 25 and 34 unit systems
Test System Hybrid Model-
III(ED by load to
economic unit
Hybrid Model-III (ED by
lambda iteration method)
Difference in cost
12-unit 120700024 120692850.78 7173.22 15-unit 150549504 150548559.52 944.48 25-unit 176237968 176234285.37 3682.63
Operating Fuel Cost
($) 34-unit 208266768 208266739.93 28.07
Table 6.2 Summary of UC schedule and operating fuel cost for 12 unit NTDC systems
S. No.
Load demand (MW)
Unit Commitment
schedule
Operating fuel cost by Hybrid Model-II ($)
(load to economic unit)
Operating fuel cost by Hybrid Model-III ($)
(Lambda iteration method) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
1600 1610 1620 1630 1625 1650 1705 1710 1705 1680 1675 1700 1750 1800 1850 1900 1950 1975 2000 1980 1970 1800 1750 1700
101011001110 101011001110 101011001110 101011101110 101011101110 101011101110 101011101110 101011101110 101011101110 101011101110 101011101110 101011101110 101011111110 101011111110 101111111110 101111111110 111111111110 111111111110 111111111111 111111111110 111111111110 101011111110 101011111110 101011101110
4649995.0 4701900.5 4754483.0 4524162.0 4503106.5 4609900.0 4864742.5 4890442.0 4864742.5 4743059.5 4720487.0 4839212.5 4845135.0 5074447.5 5108391.5 5337704.0 5625712.0 5780787.5 5939454.0 5817434.5 5745929.0 5074447.5 4845135.0 4839212.5
4649994.13 4701899.88 4754482.21 4524161.51 4503106.29 4609899.74 4864742.03 4890441.19 4864742.03 4743059.13 4720486.56 4839212.02 4845134.73 5074447.01 5108390.86 5337703.14 5625710.97 5780786.05 5932296.44 5817433.40 5745927.70 5074447.01 4845134.73 4839212.02
SUM 120700024 120692850.78
149
Table 6.3 Summary of UC schedule and operating fuel cost for 15 unit NTDC systems
S. No.
Load demand (MW)
Unit Commitment
schedule
Operating fuel cost by Hybrid Model-III ($) (load to economic unit)
Operating fuel cost by Hybrid Model-III ($)
(Lambda iteration method) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2400 2425 2450 2475 2500 2525 2550 2575 2600 2625 2650 2675 2700 2725 2750 2775 2800 2825 2850 2875 2900 2925 2950 2958
000011111111111 000011111111111 000011111111111 000011111111111 000011111111111 000011111111111 001011111111111 001011111111111 001011111111111 001011111111111 001011111111111 001111111111111 001111111111111 001111111111111 001111111111111 001111111111111 001111111111111 011111111111111 011111111111111 011111111111111 111111111111111 111111111111111 111111111111111 111111111111111
5605847.0 5666131.5 5737777.0 5836529.0 5939074.0 6045412.5 5909131.5 5970236.5 6056239.5 6156508.5 6260571.0 6153930.5 6213804.5 6278376.0 6376369.0 6478155.5 6583735.5 6538807.5 6640594.0 6746174.0 6701246.0 6803032.5 6908612.5 6943199.0
5605524.78 5666130.72 5737776.32 5836528.06 5939073.22 6045411.80 5909082.52 5970236.13 6056239.05 6156508.16 6260570.69 6153381.67 6213803.84 6278375.29 6376368.35 6478154.83 6583734.73 6538806.72 6640593.20 6746173.10 6701245.09 6803031.57 6908611.47 6943198.20
SUM 150549504 150548559.52
150
Table 6.4 Summary of UC schedule and operating fuel cost for 25 unit NTDC systems
S. No.
Load demand (MW)
Unit Commitment schedule Operating fuel cost by Hybrid Model-III ($)
(Load to economic unit)
Operating fuel cost by Hybrid Model-
III ($) (Lambda iteration
method) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2800 2825 2850 2875 2900 2925 2950 2975 3000 3025 3050 3075 3100 3125 3150 3175 3200 3225 3250 3275 3400 3425 3450 3555
000011111111111110000000000001111111111111000000000000111111111111100000000000011111111111110000000000101111111111111000000000010111111111111100000000001011111111111110000000000101111111111111000000000010111111111111100000000001011111111111110000000000111111111111111000000000011111111111111100000000001111111111111110000000000111111111111111000000000011111111111111100000000011111111111111110000000001111111111111111000000000111111111111111100000000111111111111111110000000011111111111111111000000001111111111111111111000000111111111111111111100000111111111111111111110000011111111111111111111111111
6620140.5 6713427.0 6814454.5 6919276.0 6802118.0 6862402.5 6934048.0 7032800.0 7135345.0 7241683.5 7106462.5 7167567.5 7253570.5 7353839.5 7457902.0 7416009.0 7516278.0 7620340.5 7578447.5 7678716.5 8092939.0 8116494.0 8221770.5 8581921.0
6620130.05 6713426.00 6814453.80 6919275.01 6800094.40 6862047.31 6934047.41 7032799.16 7135344.32 7241682.90 7105270.37 7167496.82 7253569.77 7353838.88 7457901.41 7416008.14 7516277.25 7620339.78 7578446.51 7678715.62 8092938.01 8116493.14 8221769.56 8581919.76
SUM 176237968 176234285.37
151
Table 6.5 Summary of UC schedule and operating fuel cost for 34 unit NTDC systems
S. No.
Load demand (MW)
Unit Commitment schedule Operating fuel cost by Hybrid Model-III ($)
(load to economic unit)
Operating fuel cost by Hybrid Model-III ($)
(Lambda iteration method)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
3500 3490 3480 3475 3470 3480 3750 3650 3700 3680 3690 3695 3695 3696 3698 3700 3715 3700 3690 3575 3775 3760 3785 3700
111111111111111111100000100100000011111111111111111110000010010000001111111111111111111000001001000000111111111111111111100000100100000011111111111111111110000010010000001111111111111111111000001001000000111111111111111111111101100111111111111111111111111110000010011111111111111111111111111010011001111111111111111111111111100001100111111111111111111111111110000110011111111111111111111111111010011001111111111111111111111111101001100111111111111111111111111110100110011111111111111111111111111010011001111111111111111111111111101001100111111111111111111111111111100110011111111111111111111111111010011001111111111111111111111111100001100111111111111111111111111110000010010001111111111111111111111111111011111111111111111111111111111111100111111111111111111111111111111111111111111111111111111111111010011001111111
8354777.5 8311605 8269039 8247984 8227080 8269039 8965442 8609265 8786917 8719249 8762240 8765619 8765619 8769866 8778379 8786917 8833262 8786917 8762240 8582748 9049066 8989955 9086636 8786917
8354776.62 8311603.89 8269038.11 8247982.83 8227079.28 8269038.11 8965440.36 8609263.51 8786914.93 8719247.69 8762238.33 8765616.87 8765616.87 8769864.34 8778377.5 8786914.93 8833259.84 8786914.93 8762238.33 8582746.22 9049064.11 8989953.09 9086634.29 8786914.93
SUM 208266768 208266739.93 Execution time 3370.49 s
152
CHAPTER 7
Conclusions and Suggestions
The present work deals with scheduling of thermal units, which could be the major part
of the hydrothermal power system. The scheduling of thermal generating units is considered as
two linked optimization problems as it consists of the unit scheduling problem and economic
dispatch sub problem.
Unit commitment is the essential and vital step in the daily operational planning of the
power system. The UCP is a combinatorial optimization problem with equality and inequality
constraints and ED is a nonlinear optimization problem.
The applications of artificial intelligence (AI) techniques have the potential in solving the
UCP. These techniques have the ability to handle nonlinearities and discontinuities commonly
found in power systems.
Research has been focused on UCP techniques with respect to dimensionality problem,
difficult constraint handling, spinning reserve considerations, computational time, and quality of
solution. The exact solution of the UCP may be obtained from exhaustive enumeration because it
calculates all the feasible and infeasible combinations of units. Due to the high dimensionality of
the search space it is impractical.
In the present study two new AI-based hybrid algorithms have been developed to solve
the UCP for convex and non convex fuel cost functions. The algorithms are presented in chapter
5. The other hybrid algorithm based on classical approaches is applied first time to solve the
UCP.
153
These algorithms are:
• Hybrid Model – I: A hybrid of Particle Swarm Optimization (PSO), Artificial Neural
Network (ANN) and Dynamic Programming (DP).
• Hybrid Model-II: Neuro-Genetic Hybrid Approach
• Hybrid Model –III: Scaleable Deterministic Hybrid Approach
The proposed Hybrid Model-I combines the Dynamic Programming (DP) with Artificial
Neural Networks (ANN) using Swarm Intelligence (SI) learning rule. In this model dynamic
programming produces near optimal solution based on training data for neural network model.
The neural network fine tunes the data subject to the target values of power output of units. The
best tuned solution is considered the required solution. The swarm intelligence learning rule
based feed forward network has been used for fine tuning the near optimal dynamic
programming results. The standard back propagation learning rule neural network has also been
used in this hybrid approach for the comparison of results. The particle swarm optimization
algorithm is applied to the neural network to obtain a set of weights that will minimize the error
function. Weights are progressively updated until the convergence criterion is satisfied. The UC
problem has been decomposed in to discrete loads on hourly basis. These small models get
trained faster due to simple network structure and perform efficiently due to swarm intelligence
learning rule. Three and ten unit standard test systems have been tested for validation of the
proposed approach. The results of the standard test systems obtained using this approach show
the comprehensive reduction in total production cost thus indicating the promise of the approach.
The proposed algorithm explores wider solution space, and gives better quality solution for
convex cost function.
The proposed Hybrid Model-II (Neuro-Genetic approach) combines the Genetic Algorithm
(GA) with Artificial Neural Networks (ANN) using Swarm Intelligence (SI) learning rule. GA
has the ability to search better for non-convex fuel cost function than convex fuel cost function.
Three machine standard test system has been tested for validation of the proposed model. In this
model Genetic Algorithm works as global optimizer and produces near optimal solution based
training data for neural network model. The neural network fine tunes the data subject to the
target values. The best fine tuned solution is considered the required solution. The target values
may be taken by randomly generated values by GA around the near optimal solutions satisfying
154
both equality and inequality constraints. The SI-learning rule based feed forward neural network
has been used for fine tuning the near optimal GA results. The standard back propagation
learning rule has also been used in this hybrid approach for the comparison of results. In this
model the test data and training data are generated by using Genetic Algorithm (GA). For proper
training of ANN, a pair of load as input and their corresponding generation schedules as output
are prepared off-line by using Genetic Algorithm (GA) and are stored in a data base. Each pair is
referred as input/output database pair. The case studies for three unit test system IV and V shows
that SI learning based ANN produces better results than back propagation learning network. The
total production cost results are also better than conventional approaches.
The proposed hybrid model-III combines the Maximum Power Output (PMAX) and Full
Load Average Production Cost (FLAPC) of each unit. This model is applicable for both convex
and non convex fuel cost functions. The proposed unit commitment algorithm hybrid model-III
incorporates the unit commitment solution by the deterministic priority list scheme for
generation of initial population. This hybrid model solves the UC problem with valve point
effect and without valve point effect. The UC schedule is prepared according to PMAX and
FLAPC. Unit with higher PMAX will be at higher priority. If the two units have the same
PMAX then unit with lower FLAPC will be of higher priority. Economic Dispatch is based on
lambda iteration method and average load assigned methods. Five test systems consisting of
three and ten units have been tested. Three test systems are convex, and two test system are non
convex. Final results are better than other many approaches available in the literature. The issue
in solving the combinatorial optimization problem is to have good feasible neighbour/trial
solutions from an existing feasible solution. The proposed algorithm is fast and efficient in
generating feasible initial and trial solutions and is applicable of solving a large scale power
system in reasonable computational time.
A major step in solving the UCP is the solution of the economic dispatch sub problem. In
this regard, a new GA based real power search algorithm has been implemented and tested.
PAKISTANI utility system and its operational problems have been reviewed with a view to
carry out its unit commitment studies. The data of 12, 15, 25 and 34 unit systems have also been
prepared to carry out unit commitment studies.
155
The demonstration of validity of applying the proposed models to the solution of UCP in
this thesis gives rise to the number of topics for further research in this area. Some of the
recommendations for future work may be summarized as follows:
• Exploration of new AI based hybrid models which give good quality of solution,
explore a wider search space, able in solving a large scale system in acceptable
computational time and is applicable to PAKISTANI utility system.
• Some other costs could also be taken into consideration in the objective function
such as: maintenance cost etc.
• The objective function may be taken as Profit Based Unit Commitment (PBUC)
instead of Cost Based Unit Commitment (CBUC).
• Some other constraints could also be taken into consideration such as: security,
transmission line capacity, fuel and emission.
• The hydro generation system can be included in the proposed algorithms to solve the
hydrothermal scheduling problem.
156
References
[1] A. J. Wood, and B. F. Wollenberg, Power generation, operation and control, New York:
John Wiley & Sons, Inc, 1996.
[2] B.Xiaomin, and S. M. Shaidehpour, “Extended neighbourhood search algorithm for
constrained unit commitment,” Electric Power and Energy Systems, Volume 19, No.5 pp.
349–356, 1997.
[3] H. Y. Yamin, “Review on methods of generation scheduling in electric power systems,”
Electric Power System Research., Volume 69, pp. 227–248, 2004.
[4] L. Chao-An, R. B. Johnson, and A. J. Svoboda, “A new unit commitment method,” IEEE
Transactions on Power Systems, Volume 12, No. 1, pp.113 – 119, Feb. 1997.
[5] N. P. Padhy “Unit commitment – a bibliographical survey” IEEE Transactions on Power
Systems, Volume 19, No. 2, pp. 1196-1205, 2004.
[6] B. F. Hobbs, M. H. Rothkopf, and R. P. O’Neill, The Next Generation of Electric Power
Unit Commitment Models, Boston: Kluwer Academic Publishers, 2001.
[7] J. J. Shaw, “A direct method for security-constrained unit commitment,” IEEE
Transactions on Power Systems, Volume 10, No. 3, pp. 1329–1242, 1995.
[8] X. Guan, S. Guo, and Q. Zhai, “The conditions for obtaining feasible solutions to
security-constrained unit commitment problems,” IEEE Transactions on Power Systems,
Volume 20, No. 4, pp. 1746–1756, 2005.
[9] C. Wang and S. M. Shahidehpour, “Ramp-rate limits in unit commitment and economic
dispatch incorporating rotor fatigue effect,” IEEE Transactions on Power Systems,
Volume 9, pp. 1539–1545, Aug. 1994
[10] C. Wang and S. M. Shahidehpour “Effects of ramp-rate limits on unit commitment and
economic dispatch” IEEE Transactions on Power Systems, Volume 8, pp. 1341–1350,
Aug. 1993.
[11] G. B. Sheble and G. N. Fahd, “Unit commitment literature synopsis,” IEEE Transactions
on Power Systems, Volume 9, pp. 128–135, Feb. 1994.
[12] F. N. Lee, “A fuel-constrained unit commitment method” IEEE Transactions on Power
Systems, Volume 4, pp. 1208–1218, Aug. 1989.
157
[13] N.P. Padhy, “Unit commitment using hybrid models: A comparative study for dynamic
programming, expert system, fuzzy system and genetic algorithms,” International Journal
of Electrical Power & Energy Systems, Volume 23, No. 8, pp. 827–836, 2001.
[14] A. H. Mantawy, Y. L. Abdel-Magid, and S. Z. Selim, “Integrating genetic algorithms,
tabu search, and simulated annealing for the unit commitment problem,” IEEE
Transactions on Power Systems, Volume 14, No.3, pp. 829–836, 1999
[15] T. Senjyu, K. Shimaukuro, K. Uezato, and T. Funabashi, “A Fast Technique for Unit
Commitment Problem by Extended Priority List," IEEE Transactions on Power Systems,
Volume 18, No. 2, pp0882-888, May 2003.
[16] T. Senjyu, T. Miyagi, A. Y. Saber, N. Urasaki, and T. Funabashi, “Emerging solution of
large-scale unit commitment problem by stochastic priority list,” Elect. Power Syst. Res.,
Volume 76, pp. 283–292, Mar. 2006.
[17] P. G. Lowery, “Generating Unit Commitment by Dynamic Programming,” IEEE
Transactions on Power Apparatus and Systems, Volume PAS-85, Issue 5, pp. 422-426,
May 1966.
[18] C. K. Pang, G. B. Sheble, F. Albuyeh, “Evaluation of Dynamic Programming Based
Methods and Multiple area Representation for Thermal Unit Commitments,” IEEE
Transactions on Power Apparatus and Systems, Volume PAS-100, Issue 3, pp. 1212-
1218, March 1981.
[19] S. D. Bond, and B. Fox, “Optimal thermal unit scheduling using improved dynamic
programming algorithm,” IEE Proceedings Generation, Transmission and Distribution,
Vol. 133, No.1,pp. 1-5, January 1986.
[20] W. L. Snyder, H.D. Powel, J.C. Rayburn, “Dynamic programming approach to unit
commitment,” IEEE Transactions on Power Systems, No. 2, pp. 339–350, May 1987.
[21] W. J. Hobbs, G. Hermon, S. Warner, and G.B. Sheble “An enhanced dynamic
programming approach for unit commitment,” IEEE Transactions on Power Systems,
No.3 pp. 1201–1205, August 1988.
[22] C. C. Su; and Y. Y. Hsu, “Fuzzy dynamic programming: an application to unit
commitment,” IEEE Transactions on Power Systems, Vol. 6, No 3, pp. 1231-1237, Aug.
1991.
158
[23] Z. Ouyang, S. M. Shahidehpour, “An intelligent dynamic programming for unit
commitment application” IEEE Transactions on Power Apparatus and Systems, Volume
6, Issue 3, pp. 1203-1209, Aug. 1991.
[24] A. I. Cohen, and M. Yoshimura, “A Branch-and-Bound Algorithm for Unit
Commitment,” IEEE Transactions on Power Apparatus and Systems, Vol. 102, No.2, pp.
444-451, Feb. 1983.
[25] T. S. Dillon, K. W. Edwin, H. D. Kochs, and R. J. Taud, “Integer Programming
Approach to the Problem of Optimal Unit Commitment with Probabilistic Reserve
Determination,” IEEE Transactions on Power Apparatus and Systems, Vol.97, No. 6, pp.
2154-2166, Nov. 1978. [26] S. Takriti, and J. R. Birge, “Using integer programming to refine Lagrangian-based unit
commitment solutions,” IEEE Transactions on Power Systems, Vol. 15, No. 1, pp. 151-
156, Feb. 2000.
[27] Li Tao, and S. M. Shahidehpour, “Price-based unit commitment: a case of Lagrangian
relaxation versus mixed integer programming,” IEEE Transactions on Power Systems,
Vol. 20, No. 4, pp. 2015-2025, Nov. 2005.
[28] B. Venkatesh, T. Jamtsho, and H. B. Gooi, “Unit commitment a fuzzy mixed integer
Linear Programming solution,” IET Generation, Transmission & Distribution, Vol. 1, No.
5, pp. 836-846, September 2007.
[29] A. Merlin, and P. Sandrin, “A New Method for Unit Commitment at Electricite De
France’ IEEE Transactions on Power Apparatus and Systems, Volume 102, No 5, pp.
1218 - 1225, May 1983.
[30] R. Nieva, and A. Inda, I. Guillen, I., “Lagrangian Reduction of Search-Range for Large-
Scale Unit Commitment,” IEEE Transactions on Power Systems, Volume 2, No. 2, pp.
465-473, May 1987.
[31] F. Zhuang, and F. D. Galiana, “Towards a more rigorous and practical unit commitment
by Lagrangian relaxation”, IEEE Transactions on Power Systems, Volume 3, No. 2, pp.
763-773, May 1988.
159
[32] S. Virmani, E. C. Adrian, K. Imhof, and S. Mukherjee, “Implementation of a Lagrangian
relaxation based unit commitment problem,” IEEE Transactions on Power Systems,
Volume 4, No.4, pp. 1373-1380, Nov. 1989.
[33] R. Baldick, “The Generalized Unit Commitment Problem,” IEEE Transactions on Power
Systems, Volume 10, No. 1, pp. 465-475, Feb. 1995.
[34] W.L Peterson and S. R.Brammer, “A Capacity Based Lagrangian Relaxation Unit
Commitment with Ramp Rate Constraints,” IEEE Transactions on Power Systems,
Volume 10, No. 2, pp. 1077-1084, May 1995.
[35] A. G. Bakirtzis, and C. E. Zoumas, “Lambda of Lagrangian relaxation solution to unit
commitment problem,” IEE Proceedings on Generation Transmission and Distribution,
Volume 147, pp. 131–136, 2000.
[36] W. P. Ongsakul, and P. Nit, “Unit commitment by enhanced adaptive Lagrangian
relaxation,” IEEE Transactions on Power Systems, Volume 19, No. 1, pp. 620–628,
2004.
[37] D. Murtaza and S. Yamashiro, “Unit Commitment Scheduling by Lagrange Relaxation
Method Taking into Account Transmission Losses,” Electrical Engineering in Japan, Vol.
152, No. 4, pp. 27-33, 2005
[38] S. H. Hosseini, A. Khodaei, and F. Aminifar, “A Novel Straightforward Unit
Commitment Method for Large-Scale Power Systems,” IEEE Transactions on Power
Systems, Volume 22, No. 4, November 2007
[39] K. Chandram, N. Subrahmanyam, and M. Sydulu, “New approach with Secant method
for solving Unit Commitment problem,” IEEE/PES Transmission and Distribution
Conference and Exposition, pp. 1-7, April 2008.
[40] A. H. Mantawy, Y. L. Abdel-Magid, and S. Z. Selim, “Unit commitment by tabu search,”
IEE Proceedings Generation Transmission and Distribution, Volume 145, No. 1, pp. 56–
64, 1998.
[41] F. Zhuang, and F .D. Galiana, “Unit commitment by simulated annealing,” IEEE
Transactions on Power Systems, Volume 5, No.1, 311–318, 1990.
[42] A. H. Mantawy, Y. L. Abdel-Magid, and S. Z. Selim, “A simulated annealing algorithm
for unit commitment,” IEEE Transactions on Power Systems, Volume 13 No.1, pp. 197–
204, 1998.
160
[43] S. Y. W Wong, “An enhanced simulated annealing approach to unit commitment,”
International Journal of Electrical Power & Energy Systems, Volume 20, No. 5 pp. 359–
368, 1998.
[44] D. N. Simopoulos, S. D. Kavatza, and C. D. Vournas, “Unit Commitment by an
Enhanced Simulated Annealing Algorithm,” IEEE Transactions on Power Systems,
Volume 21, No. 1, February 2006.
[45] A. Y. Saber, T. Senjyu, T. Miyagi, N. Urasaki, and T. Funabashi, “Fuzzy Unit
Commitment Scheduling Using Absolutely Stochastic Simulated Annealing,” IEEE
Transactions on Power Systems, Volume 21, No. 2, pp. 955-964, May 2006.
[46] S. Mokhtari, J. Sing, and B. Wollenberg “A unit commitment expert system (power
system control),” IEEE Transactions on Power Systems, Volume 3, No. 1, pp. 272-277,
1988.
[47] S. K. Tong, S. M. Shahidehpour, and Z. Ouyang, "A Heuristic Short-Term Unit Commitment," IEEE Transactions on Power systems, Vol.-6, No. 3, pp. 1210-1216, August 1991.
[48] S. Li, S.M. Shahidehpour, C. Wang “Promoting the application of expert systems in
short-term unit commitment,” IEEE Transactions on Power Systems, Volume 8, No. 1,
pp. 286–292, 1993.
[49] H. Sasaki, Y. Fujii, M. Watanbe, J. Kubokawa, and N. Yorino, “A Solution Method of
Unit Commitment by Artificial Neural Networks,” IEEE Transactions on Power Systems,
Volume 7, No. 3, pp. 974-981, August 1992.
[50] T. Yalcinoz, M. J. Short, and B. J. Cory, “Application of Neural Networks to Unit
Commitment,” IEEE Transactions on Power Systems, pp. 649-654, 1999.
[51] M. H. Wong, T. S. Chung, and Y. K. Wong, “An Evolving Neural Network Approach in
Unit Commitment Solution,” Journal of Microprocessors and Microsystems, Volume 24,
pp. 251–262, 2000.
[52] K. A. Juste, H. Kita, E. Tanaka, and J. Hasegawa, “An evolutionary programming
solution to the unit commitment problem,” IEEE Transactions on Power Systems,
Volume 14, No. 4, pp. 1452–1459, 1999.
[53] D. Dasgupta, and D. R. McGregor, “Short term unit-commitment using genetic
algorithms” Fifth International Conference Proceedings, on Tools with Artificial
Intelligence, pp. 240-247, Nov. 1993
161
[54] X. Ma, A. A. El-Keib, R. E. Smith,, and H. Ma, “A genetic algorithm based approach to
thermal unit commitment of electric power systems,” Electric Power System Research,
Volume.34, pp. 29–36, 1995.
[55] S. O. Orero, and M. R. Irving, “A genetic algorithm for generator scheduling in power
systems,” Electric Power Energy Systems, Volume 18, No. 1, pp. 19–26, 1996.
[56] S. A. Kazarlis, A. G. Bakirtzis, and V. Petridis, “A genetic algorithm solution to the unit
commitment problem,” IEEE Transactions on Power Systems, Volume 11, No.1, pp. 83–
92, February 1996.
[57] T. Senjyu, K. Shimabukuro, K. Uezato, and T. Funabashi, “A unit commitment problem
by using genetic algorithm based on unit characteristic classification,” IEEE Power
Engineering Society Winter Meeting, Volume l, pp. 58-63, 2002.
[58] E. Gil, J. Bustos and H. Rudnick, “Short-Term Hydrothermal Generation Scheduling
Model Using a Genetic Algorithm,” IEEE Transactions on Power Systems, Volume 18,
No. 4,pp.1256-1264, November 2003.
[59] I. G. Damousis, A. G. Bakirtzis, and P. S. Dokopoulos, “A Solution to the Unit
Commitment Problem Using Integer-Coded GA,” IEEE Transactions on Power Systems,
Volume 19, No. 2, May 2004.
[60] C. Dang, and M. Li, “A Floating-Point Genetic Algorithm for Solving the Unit
Commitment Problem,” European Journal of Operational Research, No. 181 pp.1370–
1395, May 2006.
[61] L. Sun, Y. Zhang, and C. Jiang, “A Matrix Real-Coded Genetic Algorithm to the unit
commitment problem,” Electric Power Systems Research, No.76 pp. 716–728, 2006.
[62] S. Saneifard, N. R. Prasad, and H. A. Smolleck, “A Fuzzy Logic Approach to Unit
Commitment,” IEEE Transactions on Power Systems, Volume 12, No. 2, May 1997.
[63] S. C. Pandian, and K. Duraiswamy, “Fuzzy Logic Implementation for Solving the Unit
Commitment Problem,” International Conference on Power System Technology, pp.413-
418, November 2004.
[64] Z. L. Gaing; "Discrete particle swarm optimization algorithm for unit commitment,"
IEEE Power Engineering Society General Meeting, 2003, Volume 1, pp.418-424, 13-17
July 2003.
162
[65] B. Zhao, C. X. Guo, B. R. Bai, and Y. J. Cao, “An improved particle swarm optimization
algorithm for unit commitment,” Int. J. Elect. Power Energy Syst., Volume 28, pp. 482–
490, Sep. 2006.
[66] T. Y. Lee, and C. L. Chen, “Unit commitment with probabilistic reserve: An IPSO
approach,” Energy Conversion and Management, Vol. 48, No. 2, pp. 486-493, February
2007.
[67] X. Yuan, H. Nie, A. Su, L. Wang, and Y. Yuan, “An improved binary particle swarm
optimization for unit commitment problem,” Expert Systems with Applications, Vol. 36,
No. 4, pp. 8049-8055, May 2009.
[68] T. Sum-im, W. Ongsakul, "Ant Colony Search Algorithm for Unit Commitment," IEEE
Conference, ICIT, 2003.
[69] A. Y. Saber, and A. M. Alshareef, “Scalable unit commitment by memory-bounded ant
colony optimization with A* local search” International Journal of Electrical Power and
Energy Systems, Vol. 30, No. 6-7 , pp. 403-411, July-September 2008.
[70] A. Viana, G. Pinho de Sousa and M. Matos, “Using GRASP to Solve the Unit
Commitment Proble,” Annals of Operations Research, No. 120, pp. 117–132, Kluwer
Academic Publishers, 2003.
[71] Z. Ouyang, and S. M. Shahidehpour, “A Hybrid Artificial Neural Network-Dynamic
Programming Approach to Unit Commitment,” IEEE Transactions on Power Systems,
Vol. 7, No.2, pp. 236-242, 1992.
[72] Z. Ouyang, and S. M. Shahidehpour, “A Multi-Stage Intelligent System for Unit
Commitment,” IEEE Transactions on Power Systems, Vol. 7, No. 2, pp.639-646, May
1992.
[73] D. P. Kothari, A. Ahmad, “An expert system approach to the unit commitment problem,”
Energy Conversion and Management, Volume 36 No. 4, pp. 257–261, 1995.
[74] H. Shyh-Jier, and H. Ching-Lien, “Application of Genetic- Based Neural Networks to
Thermal Unit Commitment,” IEEE Transactions on Power Systems, Volume 12, No. 2,
pp. 654-660, May 1997.
[75] M. H. Wong, T. S. Chung, Y. K. Wong, “An evolving neural network approach in unit
commitment solution,” Microprocessors and Microsystems, Volume 24, No. 5, pp. 251–
262, 2000.
163
[76] R. Nayak, and J. D. Sharma, “A hybrid neural network and simulated annealing approach
to the unit commitment problem,” Computers & Electrical Engineering, Volume 26, No.
6, pp. 461–477, 2000.
[77] C. P. Cheng, C.W. Liu, and C. C. Liu, "Unit commitment by Lagrangian relaxation and
genetic algorithms," IEEE Transactions on Power Systems, Volume 15, No. 2, pp. 707-
714, May 2000.
[78] C. P. Cheng, C. W. Liu, and C. C. Liu, “Unit commitment by annealing-genetic
algorithms,” Electric Power Energy Systems, No. 24, pp. 149-158, 2002.
[70] J. Valenzuela and A. E. Smith, “A seeded memetic algorithm for large unit commitment
problems,” Journal of Heuristics, Volume 8, No. 2, pp. 173–195, 2002.
[80] T. O. Ting, M.V.C. Rao, C.K. Loo and S.S. Ngu “Solving Unit Commitment Problem
Using Hybrid Particle Swarm Optimization,” Journal of Heuristics, Volume 9, pp. 507–
520, 2003.
[81] C .C. A. Rajan, M. R. Mohan, and K. Manivannan, “Neural based tabu search method for
solving unit commitment problem,” IEE Proc., Gener. Transm. Distrib., Volume 150,
No.4, pp. 469–474, 2003.
[82] D. Srinivasan and J. Chazelas, “A priority list-based evolutionary algorithm to solve large
scale unit commitment problem,” Proceedings Int. Conf. on Power System Technology,
pp. 1746–1751, Nov. 2004.
[83] L. Shi, J. Haoa, J. Zhou, and G. Xu , “Ant colony optimization algorithm with random
perturbation behavior to the problem of optimal unit commitment with probabilistic
spinning reserve determination,” Electric Power Systems Research, No. 69, pp 295-303,
2004.
[84] H. H. Balci, and J. F. Valenzuela, “Scheduling Electric Power Generators Using Particle
Swarm Optimization combined with the Lagrangian Relaxation method,” International
Journal of Applied Math Computer Science, Volume 14, No. 3, pp.411–421, 2004
[85] T.A.A. Victoire, and A.E. Jeyakumar, “Unit commitment by a tabu-search-based hybrid-
optimisation technique,” IEE Proceedings-Generation, Transmission and Distribution,
Volume 152, Issue 4, pp. 563-574, July 2005.
164
[86] S. Chusanapiputt, D. Nualhong, S. Jantarang, and S. Phoomvuthisarn, “Relative Velocity
Updating in Parallel Particle Swarm Optimization Based Lagrangian Relaxation for
Large-scale Unit Commitment Problem,” IEEE Region TENCON 2005
[87] P. Sriyanyong, and Y. H. Song, “Unit commitment using particle swarm optimization
combined with Lagrange relaxation,” IEEE Power Engineering Society General Meeting,
Volume 3, pp. 2752 - 2759 , June 2005.
[88] T. Aruldoss, A. Victoire, A. Ebenezer, and Jeyakumar, “A Simulated Annealing Based
Hybrid Optimization Approach for a Fuzzy Modelled Unit Commitment Problem,”
International Journal of Emerging Electric Power Systems Volume 3, No.1, Article 1054,
2005.
[89] T.O. Ting, M.V.C. Rao, and C. K. Loo, “A novel approach for unit commitment problem
via an effective hybrid particle swarm optimization,” IEEE Transactions on Power
Systems, Volume 21, Issue 1, pp. 411-418, Feb. 2006.
[90] V.N. Dieu and W. Ongsakul, “Enhanced augmented Lagrangian Hopfield network for
unit commitment,” IEE proceedings, Generation, Transmission and Distribution, Volume
153, No.6, pp. 624-632, November 2006.
[91] A. Y. Saber, T. Senjyu, A. Yona, N. Urasaki, and T. Funabashi “Fuzzy unit commitment
solution—A novel twofold simulated annealing approach,” Electric Power Systems
Research, No. 77, pp. 1699-1712, January 2007.
[92] S. Nasser, H. Mahdi, and Z. Majid, “Unit Commitment Using Particle Swarm-Based-
Simulated Annealing Optimization Approach,” IEEE Swarm Intelligence Symposium,
pp. 297-302, 1-5 April 2007.
[93] A. Y. Saber, T. Senjyu, A. Yona, and T. Funabashi, “Unit commitment computation by
fuzzy adaptive particle swarm optimization,” IET Generation, Transmission and
Distribution, Volume 1, No. 3, pp. 456-465, May 2007.
[94] S. S. Kumara and V. Palanisamy “A dynamic programming based fast computation
Hopfield neural network for unit commitment and economic dispatch,” Electric Power
Systems Research, No. 77, pp. 917–925, 2007.
[95] R. C. Eberhart, and J. Kennedy, “A New Optimizer Using Particle Swarm Theory,”
Proceedings of the Sixth International Symposium on Micro Machine and Human
Science, pp. 39-43, Nagoya, Japan, 1995.
165
[96] J. Kennedy, and R. C. Eberhart, “Particle Swarm Optimization,” Proceedings of the IEEE
International Conference on Neural Networks, Volume 4, pp. 1942-1948. Perth,
Australia, 1995.
[97] Y. H. Shi, and R. C. Eberhart, “A Modified Particle Swarm Optimizer,” IEEE
International Conference on Evolutionary Computation, Anchorage, Alaska, 1998.
[98] Y. H. Shi, and R. C. Eberhart, “Parameter Selection in Particle Swarm Optimization,”
The 7th Annual Conference on Evolutionary Programming, San Diego, USA, 1998.
[99] M. Clerc, “The Swarm and the Queen, Towards a Deterministic and Adaptive Particle
Swarm Optimization,” Proceedings of Congress on Evolutionary Computation, pp 1951-
1957, Washington DC, 1999.
[100] E. A. Grimaldi, F. Grimaccia, M. Mussetta, and R. E. Zich, “PSO as an Effective
Learning Algorithm for Neural Network Applications,” Proceedings of the International
Conference on Computational Electromagnetic and its Applications, 2004.
[101] J. J. Grainger and W. D. Stevenson Jr, Power System Analysis, McGraw-Hill.
[102] D. E. Gold berg, Genetic Algorithms in Search, Optimization and Machine Learning,
Addison-Wesley Publishing Company, Inc. Reading Massachusetts, 1989.
[103] H. Sasaki, Y. Fujii, M. Watanbe, J. Kubokawa and N. Yorino, “A Solution Method using
Neural Network for the Generator Commitment Problem”, Journal of Electrical
Engineering in Japan, Volume 112, pp. 55-61, 1992.
[104] R. C. Eberhart, P. K. Simpson, and R. W. Dobbins, Computational Intelligence PC Tools,
Academic Press Professional, 1996.
[105] Y.D. Valle, G. K. Venayagamoorthy, S. Mohagheghi, J.C. Hernandez, and R.G. Harley,
“Particle Swarm Optimization: Basic Concepts, Variants and Applications in Power
Systems,” IEEE Transactions on Evolutionary Computation, Volume 12, No. 2, pp.171-
195, April 2008.
[106] S. Mohagheghi, Y. D. Valle, G. Venayagamoorthy, and R. Harley, “A Comparison of
PSO and Backpropagation for Training RBF Neural Networks for Identification of a
Power System with STATCOM,” IEEE Proceedings in Swarm Intelligence Symposium,
pp. 381–384, June 2005.
166
[107] V. G. Gudise, and G. K. Venayagamoorthy, “Comparison of Particle Swarm
Optimization and Backpropagation as Training Algorithms for Neural Networks,” IEEE
Proceedings in Swarm Intelligence Symposium, pp. 110–117, April 2003.
[108] J. Chia-Feng, “A Hybrid of Genetic Algorithm and Particle Swarm Optimization for
Recurrent Network Design,” IEEE Transactions on Systems, Man, Cybern., Part B:
Cybern, Volume 34, No. 2, pp. 997–1006, April 2004.
[109] K. S. Swarup and S. Yamashiro, “A Genetic Algorithm Approach to the Solution of the
Basic Unit Commitment Problem,” IEE Japan national Convention Conference, PP. 29-
34, October 1997.
[110] K. S. Swarup, and P.V. Simia, “Neural computation using discrete and continuous
Hopfield networks for power system economic dispatch and unit commitment,”
Neurocomputing, Vol. 70, No. 1-3, pp. 119-129, December 2006.
[111] WAPDA Annual Report 2004. Public Relations Division, WAPDA, Lahore.
167
APPENDIX A
Test Systems --- Description and Data
A.1 Three Unit Test Systems --- Convex Fuel Cost Curve A.1.1 Test System-I --- 3 -Unit System
The data for the 3 unit base system for comparison was taken from [1].
Unit 1: Coal-fired steam unit: Input = output curve:
600P150 00142.02.7510 12
111 ≤≤++=⎟⎠⎞
⎜⎝⎛ PP
hMBtuH
Unit 2: Coal-fired steam unit: Input = output curve:
400P100 00194.085.7310 22
222 ≤≤++=⎟⎠⎞
⎜⎝⎛ PP
hMBtuH
Unit 3: Coal-fired steam unit: Input = output curve:
MWPPh
MBtuH 200P50 00482.097.778 32
333 ≤≤++=⎟⎠⎞
⎜⎝⎛
Unit 1: Fuel cost = 1.1 $/MBtu Unit 2: Fuel cost = 1.0 $/MBtu Unit 3: Fuel cost = 1.2 $/MBtu
TABLE A.1 Unit Data:
Unit No. P max (MW)
P min (MW)
a ($/hr)
b ($/MWhr)
c ($/MW2 hr)
1 600 150 561 7.92 0.001562 2 400 100 310 7.85 0.001940 3 200 50 93.6 9.564 0.005784
Load Data Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load (MW)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load (MW)
500 500 600 800 850 900 950 1000 1050 1100 1200 1200
168
A.1.2 Test System-II
The data for the 10 unit base system for comparison was taken from [1].
Unit 1: Coal-fired steam unit: Input = output curve:
600P150 00142.02.7510 12
111 ≤≤++=⎟⎠⎞
⎜⎝⎛ PP
hMBtuH
Unit 2: Coal-fired steam unit: Input = output curve:
400P100 00194.085.7310 22
222 ≤≤++=⎟⎠⎞
⎜⎝⎛ PP
hMBtuH
Unit 3: Coal-fired steam unit: Input = output curve:
MWPPh
MBtuH 200P50 00482.097.778 32
333 ≤≤++=⎟⎠⎞
⎜⎝⎛
Unit 1: Fuel cost = 1.1 $/MBtu Unit 2: Fuel cost = 1.0 $/MBtu Unit 3: Fuel cost = 1.0 $/MBtu
TABLE A.2
Unit Data: Unit No. P max
(MW) P min
(MW) a
($/hr) b
($/MWhr) c
($/MW2hr) 1 600 100 561 7.92 0.001562 2 400 100 310 7.85 0.001940 3 200 50 78 7.97 0.004820
Load Data: Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load (MW)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load (MW)
500 500 600 800 850 900 950 1000 1050 1100 1200 1200
169
A.2 Ten Unit Standard Test System
A.2.1 Test System-III
The data for the 10 unit base system for comparison was taken from [56].
TABLE A.3 Unit Data:
Unit 1 Unit 2 Unit 3 Unit 4 Unit5 P max (MW) 455 455 130 130 162 P min (MW) 150 150 20 20 25 a ($/h) 1000 970 700 680 450 b ($/MWh) 16.19 17.26 16.60 16.50 19.70 c (S/MW2-h) 0.00048 0.00031 0.002 0.002211 0.00398 Min up time (MUT) (h) 8 8 5 5 6 Min down time (MDT) (h) 8 8 5 5 6 Hot start cost ($) 4500 5000 550 560 900 Cold start cost ($) 9000 1000 1100 1120 1800 Cold start hrs (h) 5 5 4 4 4 Initial status (h) 8 8 -5 -5 -6
Unit 6 Unit 7 Unit 8 Unit 9 Unit10 P max (MW) 80 85 55 55 55 P min (MW) 20 25 10 10 10 a ($/h) 370 480 660 665 670 b ($/MWh) 22.26 27.74 25.92 27.27 27.79 c (S/MW2-h) 0.00712 0.0079 0.00413 0.00222 0.00173 Min up time (MUT) (h) 3 3 1 1 1 Min down time (MDT) (h) 3 3 1 1 1 Hot start cost ($) 170 260 30 30 30 Cold start cost($) 340 520 60 60 60 Cold start hrs (h) 2 2 0 0 0 Initial status (h) -3 -3 -1 -1 -1
Load Data: Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load (MW)
700 750 850 950 1000 1100 1150 1200 1300 1400 1450 1500
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load (MW)
1400 1300 1200 1050 1000 1100 1200 1400 1300 1100 900 800
170
A.3 Three Unit Test Systems --- Non-Convex Fuel Cost Curve A.3.1 Test System-IV (Non-Convex Fuel Cost Curve)
TABLE A. 4
Unit Data:
Unit No. P max (MW)
P min (MW)
a $/h
b $/ MWhr
c $/ MW2hr
e f
1 600 150 561 7.92 0.001562 300 0.0315 2 400 100 310 7.85 0.001940 200 0.0420 3 200 50 93.6 9.564 0.005784 150 0.0630
Load Data: Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load (MW)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load (MW)
500 500 600 800 850 900 950 1000 1050 1100 1200 1200
A.3.2 Test System-V (Non-Convex Fuel Cost Curve)
TABLE A.5
Unit Data: Unit No. P max
(MW) P min
(MW) a
$/h b
$/ MWhr c
$/ MW2hr e f
1 600 100 561 7.92 0.001562 300 0.0315 2 400 100 310 7.85 0.001940 200 0.0420 3 200 50 78 7.97 0.004820 150 0.0630
Load Data: Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load (MW)
1200 1200 1150 1100 1000 900 800 600 550 500 500 500
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load (MW)
500 500 600 800 850 900 950 1000 1050 1100 1200 1200
171
A.4 Pakistani Utility NTDC Systems
A.4.1 12 Unit NTDC System
A.4.2 15 Unit NTDC System
A.4.3 25 Unit NTDC System
A.4.4 34 Unit NTDC System
TABLE A.6
12 Unit NTDC system data: Unit No.
P max (MW)
P min (MW)
a ($/hr)
b ($/MWhr)
c ($/MW2hr)
1 210 170 117965.5955 3871.2446 03.3829 2 65 25 38044.3947 3930.3768 35.7800 3 300 230 21835.2917 1300.7907 02.0633 4 100 40 121270.2094 1364.9392 00.5492 5 120 40 27983.7444 1143.7230 10.6067 6 250 125 117845.1040 3073.0320 03.0347 7 110 70 21927.9678 1631.1822 03.2305 8 100 25 88181.4304 964.2456 05.0629 9 210 100 59533.4194 1920.6626 01.6412 10 210 110 65184.4252 1701.8310 01.8056 11 320 240 42799.8444 2212.5550 00.0078 12 65 25 38044.3947 3930.3768 35.7800
Load Data:
Hour 1 2 3 4 5 6 7 8 9 10 11 12 Load
(MW) 1600 1610 1620 1630 1625 1650 1705 1710 1705 1680 1675 1700
Hour 13 14 15 16 17 18 19 20 21 22 23 24 Load
(MW) 1750 1800 1850 1900 1950 1975 2000 1980 1970 1800 1750 1700
172
APPENDIX B
List of Abbreviations
EP Evolutionary Programming
DP Dynamic Programming
LR Lagrange Relaxation
ESA Enhanced Simulated Annealing
MA Memetic Algorithm
GA Genetic Algorithm
SGA Standard Genetic Algorithm
FPGA Floating Point Genetic Algorithm
ICGA Integer Coded Genetic Algorithm
BCGA Binary Coded Genetic Algorithm
PSO Particle Swarm Optimization
BPSO Binary Particle Swarm Optimization
IPSO Improved Particle Swarm Optimization
PSO-B-SA1 Particle Swarm Optimization Based Simulated Annealing
ELR Enhanced Lagrange Relaxation
ALR Augmented Lagrange Relaxation
EPL Extended Priority List
FAPSO Fuzzy Adaptive Particle Swarm Optimization
HPSO Hybrid Particle Swarm Optimization
LR-PPSO Lagrange Relaxation Parallel Particle Swarm Optimization
LR-PRPSO Lagrange Relaxation Parallel Relative Particle Swarm
Optimization
GA-UCC Genetic Algorithm -Unit Characteristics Classification
TS-HPSO-SQP-TS-IRP Tabu Search based Hybrid Particle Swarm Optimization -
Sequential Quadratic Programming- Tabu Search- Improved
Random Perturbation
173
TS-HPSO-SQP-TS-RP Tabu Search based Hybrid Particle Swarm Optimization -
Sequential Quadratic Programming- Tabu Search - Random
Perturbation
AG Annealing Genetic Algorithm
ACSA Ant Colony Simulated Annealing
DP-LR Dynamic Programming- Lagrange Relaxation
LR-GA Lagrange Relaxation- Genetic Algorithm
LR-PSO Lagrange Relaxation-Particle Swarm Optimization
ASSA Absolutely Stochastic Simulated Annealing
EMO-ALHN Enhanced Merit Order- Augmented Lagrange Hopfield Network
SPL Stochastic Priority List
PLEA Priority List based Evolutionary Algorithm
MRCGA Matrix Real Coded Genetic Algorithm
SF Straight Forward
MA-LR Memetic Algorithm seeded with Lagrange Relaxation
DPHNN Dynamic Programming based Hopfield Neural Network
FPGA Floating Point Genetic algorithm
Twofold- SA Twofold Simulated Annealing
RPACO Ant Colony Optimization with Random Perturbation
MA seeded with LR Memetic Algorithm seeded with Lagrange Relaxation
SA-PSO-SQP Simulated Annealing based Hybrid Particle Swarm Optimization
- Sequential Quadratic Programming
LR-EP Hybrid Lagrange Relaxation Evolutionary Programming
MACO memory-bounded ant colony optimization
GRASP Greedy Randomized Adaptive Search Procedure
EALHN Enhanced Augmented Lagragian Hopfield Network
174
APPENDIX C
Notation
The following notation is throughout the thesis:
Pih real power output of unit i at hour h,(MW)
Uih the on/off status of unit i at hour h. Uih = 0 when OFF, Uih = 1 when ON,
Fi (Pih ) fuel cost function or fuel cost rate of unit i, ($/h)
ai, bi, ci positive fuel cost coefficients of unit i measured in $/h, $/MW h and $/MW2 h,
respectively,
ei , fi cost coefficients from the valve point effect of the unit i,
Xi on h duration during which the unit i is continuously ON, (h)
Xi off h duration during which the unit i is continuously OFF, (h)
h-costi hot start cost of unit i,($)
c-costi cold start cost of unit i, ($)
c-s-houri cold start time of unit i, (h)
N the number of units,
H the number of hours, (24 h)
Dh load demand at hour h, (MW)
SRh spinning reserve at hour h, (MW)
Tiup minimum up time (MUT) of unit i, (h)
Tidown minimum down time (MDT) of unit i, (h)
TPC total production cost, ($/h)
STih start up cost of unit i in hour h, ($/h)
SDih shut down cost of unit i in hour h, ($/h)
Pimin minimum generations limit of unit i, (MW)
Pimax maximum generations limit of unit i, (MW)
HR heat rate (BTU/KWH)
FLAPC Full Load Average Production Cost ($/MWH)
ELD Economic load dispatch
175
APPENDIX D Derived Publications [1] Engr. Aftab Ahmad , Engr. Tahir Nadeem Malik, and Dr. Aftab Ahmad,” Unit
Commitment Problem Of Thermal Generation Units For Short Term Operational Planning using simple genetic algorithm” IEEEP New Horizon 2006 vol.52,April to June 2006 pp.22-26.
[2] Prof. Aftab Ahmad , Prof. Tahir Nadeem Malik, and Engr. Shahid-ur-Rehman Farooque
“Reconfiguration of Distribution Feeder to Reduce Active Losses” IEEEP New Horizons Vol. No.56, April to June 2007 pp.3-8.
[3] Prof. Aftab Ahmad , Prof. Tahir Nadeem Malik, and Engr. Shahid-ur-Rehman Farooque
“Hybrid Approach to Reduce Voltage Drop and Loading of a Problematic Feeder” IEEEP New Horizons Vol. No.56, April to June 2007 pp.14-17.
[4] Engr. Aftab Ahmad, Dr. Aftab Ahmad, and Engr. Tahir Nadeem Malik, “Unit
Commitment Problem of Thermal Generation Units for Short Term Operational Planning” IEEEP New Horizons Vol. No.56, April to June 2007 pp.20-23.
[5] Aftab Ahmad and Azzam ul Asar, “A PSO based Artificial Neural Network approach
for short term unit commitment problem” accepted for publication in Mehran University Research Journal of Engineering and Technology on June, 2009.