Anda di halaman 1dari 245
COMPOSITE SYSTEM RELIABILITY EVALUATION USING SEQUENTIAL MONTE CARLO SIMULATION A Thesis Submitted to the College of Graduate Studies and Research in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in the Department of Electrical Engineering University of Saskatchewan by Annapoorani Jonnavithula Saskatoon, Saskatchewan Fas. 1997 © Copyright Annapoorani Jonnavithula, 1997. All rights reserved. i+l National Library of Canada Acquisitions and Bibliographic Services {395 Welington Steet ‘Grawa ON KIA ONG Canaca Bibliotheque nationale ‘du Canada Acquisitions et services bibliographiques 305, rue Wellington ‘Qeawa ON KHAO Canada The author has granted a non- L’auteur a accordé une licence non exclusive licence allowing the exclusive permettant la National Library of Canada to Bibliothéque nationale du Canada de reproduce, loan, distribute or sell _reproduire, préter, distribuer ou copies of this thesis in microform, vendre des copies de cette thése sous paper or electronic formats. la forme de microfiche/film, de reproduction sur papier ou sur format électronique. The author retains ownership of the _L’auteur conserve la propriété du copyright in this thesis. Neither the droit d’auteur qui protége cette thése. thesis nor substantial extracts from it Nila thése ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent étre imprimés reproduced without the author’s ou autrement reproduits sans son permission. autorisation. 0-612-23941-1 Canada UNIVERSITY OF SASKATCHEWAN College of Graduate Studies and Research SUMMARY OF DISSERTATION ‘Submitted in partial fulfillment of the requirements for the DEGREE OF DOCTOR OF PHILOSOPHY by Annapoorani Jonnavithula Department of Electrical Engineering University of Saskatchewan 1997 Examining Committee: Or. P.R. Ukrainetz Dean’s Designate, Chair College of Graduate Studies and Research Dr.T.S. Sidhu Chair of Advisory Committee Department of Electrical Engineering Dr. R. Billinton Supervisor, Associate Dean of Graduate Studies Research and Extension, College of Engineering Dr. J. Kells Professor, Department of Civil Engineering Dr. N. A. Chowdhury Professor, Department of Electrical Engineering Dr. K. Takaya Professor, Department of Electrical Engineering External Examiner: Dr.R.N. Allan Professor of Electrical Energy Systems Head, Department of Electrical Engineering & Electronics UMIST, Manchester. United Kingdom COMPOSITE SYSTEM RELIABILITY EVALUATION USING SEQUENTIAL MONTE, CARLO SIMULATION Monte Carlo simulation methods can be effectively used to assess the adequacy of a composite power system network. The sequential simulation approach is the most fundamental tble and can be used to provide a wide range of indices. It can also be used to technique av. provide estimates which can serve as benchmarks against which other approximate methods can be compared. The focus of this research work is on the reliability evaluation of composite generation and transmission systems with special reference to frequency and duration related indices and estimated power interruption costs at each load bus. One of the main objectives is to use the sequential simulation method to create a comprehensive technique for composite system adequacy evaluation. This thesis recognizes the need for an accurate representation of the load model at the load buses which depends on the mix of customer sectors at each bus. Chronological hourly load curves are developed in this thesis, recognizing the individual load profiles of the customers at each load bus. Reliability worth considerations are playing an ever increasing role in power system planning and operation. Different methods for bus outage cost evaluation are proposed in this thesis. It may not be computationally feasible to use the sequential simulation method with time varying loads at each bus in large electric power system networks. Time varying load data may also not be available at each bus. This research work uses the sequential methodology as a fundamental technique to calibrate other non sequential methods such as the state sampling and state transition sampling methods. Variance reduction techniques that improve the efficiency of the sequential simulation procedure are investigated as a part of this research work. Pertinent features that influence reliability worth assessment are also incorporated. All the proposed methods in this thesis are illustrated by application to two reliability test systems. In addition to the basic studies, a number of sensitivity analyses are conducted to show the impact of selected modeling assumptions. PERMISSION TO USE ‘The author has agreed that the Library, University of Saskatchewan, may make this thesis freely available for inspection. Moreover, the author has agreed that permission for extensive copying of this thesis for scholarly purposes may be granted by the professor or professors who supervised the thesis work recorded herein or, in their absence, by the Head of the Department or the Dean of the College in which the work was done. It is understood that due recognition will be given to the author of this thesis and to the University of Saskatchewan in any use of the material in this thesis. Copying or publication or any other use of the thesis for financial gain without approval of the University of Saskatchewan and the author's written permission is prohibited. Requests for permission to copy or to make any other use of the material in this thesis in whole or in part should be addressed to: Head of the Department of Electrical Engineering University of Saskatchewan Saskatoon, Saskatchewan STN 5A9, Canada ABSTRACT Monte Carlo simulation methods can be effectively used to assess the adequacy of composite power system networks. The sequential simulation approach is the most fundamental technique available and can be used to provide a wide range of indices. It can also be used to provide estimates which can serve as benchmarks against which other approximate techniques can be compared. The focus of this research work is on the reliability evaluation of composite generation and transmission systems with special reference to frequency and duration related indices and estimated power interruption costs at each load bus. One of the main objectives is to use the sequential simulation method to create a comprehensive technique for composite system adequacy evaluation. This thesis recognizes the need for an accurate representation of the load model at the load buses which depends on the mix of customer sectors at each bus. Chronological hourly load curves are developed in this thesis, recognizing the individual load profiles of the customers at each load bus. Reliability worth considerations are playing an ever increasing role in power system planning and operation. Different methods for bus outage cost evaluation are proposed in this thesis. It may not be computationally feasible to use the sequential simulation method with time varying loads at each bus in large electric power system networks. Time varying load data may also not be available at each bus. This research work uses the sequential methodology as a fundamental technique to calibrate other non sequential methods such as the state sampling and state transition sampling techniques. Variance reduction techniques that improve the efficiency of the sequential simulation procedure are investigated as a part of this research work. Pertinent features that influence reliability worth assessment are also incorporated All the proposed methods in this thesis are illustrated by application to two reliability test systems. In addition to the basic studies, a number of sensitivity analyses are conducted to show the impact of selected modeling assumptions. ii ACKNOWLEDGMENTS ‘The author would like to express her sincere gratitude and appreciation to her supervisor, Dr. Roy Billinton, for his guidance and encouragement during the course of this research worl. ‘The author also expresses her indebtedness to her parents, her husband Satish, her brother Sriram and relatives and friends for their constant encouragement and moral support. Financial assistance in the form of a University of Saskatchewan Graduate Scholarship is gratefully acknowledged. Dedicated To My Parents S. Vimala & V. Sankarakrishnan iv TABLE OF CONTENTS PERMISSION TO USE ... DEDICATION .......... TABLE OF CONTENTS . LIST OF TABLES . LIST OF FIGURES LIST OF ABBREVIATIONS 1. INTRODUCTION 1.1. Introduction 1.2. Power System Reliability And Related Concepts 1.3. Concepts Of Composite System Analysis 1.4, Scope And Objectives Of The Thesis aawe 2. CONCEPTS OF SEQUENTIAL MONTE CARLO SIMULATION AND ‘TIME VARYING LOADS AT LOAD BUSES peed 2.1, Introduction .. 2.2. Monte Carlo Simulation Techniques ...... 2.2.1, State Sampling Technique .. 22.2, State Transition Sampling Technique ... 2.2.3. Sequential Simulation Method ... 2.3. Basic Methodology Of Sequential Simulation 2.3.1, Convergence Criterion 2.3.2. Advantages Of The Sequential Simulation Approach .. 2.3.3, Disadvantages Of The Sequential Simulation Method .. Sey 12 12 13 13 15 18 19 19 2.4. Sequential Monte Carlo Simulation For Generation Adequacy Assessment 7 2.5. Accurate Representation Of The Load Model At Each Load Bus ee 2.6. Development Of A Chronological Load Model At Each Load Bus 2.6.1. Customer Characteristics 2.7. Developing The Load Curves For The Buses Of The Two Test Systems on 2.7.1, The Roy Billinton Test System 2.7.1.1. Calculation Of Hourly Load At The Load Buses In The RBTS 2.7.2. The IEEE Reliability Test System ... 2.7.2.1. The Hourly Load Model For The Buses In The IEEE-RTS .. 23 28, Summary... COMPOSITE GENERATION AND TRANSMISSION SYSTEM ADEQUACY EVALUATION USING SEQUENTIAL MONTE CARLO. SIMULATION 40 3.1. Introduction 3.2, Adequacy Indices 3.2.1. Load Point Indices ... 3.2.2. System Indices .. 3.3. Composite System Adequacy Assessment Using Sequential Monte Carlo Simulation 3.3.1. Minimization Model For Load Curtailment . 8.3.2. Dual Upper Bound Linear Programming Algorithm 3.3.3. Solution Algorithm For Composite System Adequacy Assessment Using Sequential Monte Carlo Simulation ... 53 3.4, System Studies Utilizing an Annualized Load ..... 3.4.1. Application To The RBTS 3.4.2. Application To The IEEE-RTS ....... 3.5. System Studies Using Exponential State Residence Times . 40 41 43 44 46 47 49 3.6. 3.7. 3.8. 3.9. With Time Varying Loads 3.5.1. Application To The RBTS 3.5.1.1. Adequacy Indices 3.5.1.2. Distributions Of Adequacy Indices 3.5.2. Application To The IEEE-RTS ... Adequacy Indices . Distributions Of Adequacy Indices .. Effect Of The Repair Process Distribution On The Adequacy Indices .. 3.6.1, Application To The IEEE-RTS ... 3.6.1.1. Weibull Distribution .. 3. Normal Distribution .. 3.6.1.3. Log normal Distribution 3.6.2. System Sensitivity Studies . Load Forecast Uncertainty .. 3.7.1. Case Studies on the IEEE-RTS . Derated State Model For Generating Units 3.8.1. Application To The IEEE-RTS Summary And Conclusions .. 60 61 61 63 65 65 67 69 70 70 1 73 14 78 . 81 83 85 88 RELIABILITY WORTH ASSESSMENT IN COMPOSITE ELECTRIC POWER SYSTEMS 41 42 43. 44. 45. 46. Introduction Customer Damage Function Three Methods For Evaluating The Cost Of Interruption At ABus . 43.1, Method A 4.3.2, Method Al 4.3.3. Method B Concept of Interrupted Energy Assessment Rate . 44.1, IBAR Evaluation Using Analytical Methods 4.4.2, TEAR Evaluation Using Sequential Monte Carlo Simulation Annualized Expected Outage Cost Using Method A Application Of The Methods A, Al and B With Time Dependent Loads 46.1. Application To The RBTS 4.6.2. Application To The IEEE-RTS . 463. Distribution Of ECOST Index For The RBTS 104 104 105 And IEEE-RTS . 107 4.7. Bffect OF A Specific Curtailment Strategy On The ECOST For The Two Test Systems - 109 4.8. Sensitivity Studies On The IEEE-RTS .. es 112 4.8.1. System Studies With Non Exponential State Residence Times .. 112 4.8.2, Effect Of Load Forecast Uncertainty . 4.8.3. Effect Of Derated Units On The Reliabil 4.9. Summary And Conclusions .. 115 116 118s APPLICATION OF NON SEQUENTIAL MONTE CARLO METHODS TO COMPOSITE SYSTEM ADEQUACY EVALUATION +. 120 5.1. Introduction 120 5.2. Composite System Adequacy Assessment Using State Sampling Technique . a 121 5.2.1. Estimation of Adequacy Indices Using The State Sampling Method 123 5.3. Composite System Adequacy Assessment Using State ‘Transition Sampling Technique = . 124 5.3.1. Estimation of Adequacy Indices Using The State ‘Transition Sampling Method .. . 127 5.4. Comparison of Monte Carlo Simulation Methods With Reference To An Annualized Load . 128 5.5. Application Of The Non Sequential Methods To The | Load Duration Curve . 130 5.5.1. Application To The RBTS . 130 5.5.2. Application To The IEEE-RTS . 135 5.5.3. Discussion on the Frequency Index .. 141 5.6. Summary And Conclusions 143 6 VARIANCE REDUCTION TECHNIQUES FOR USE WITH SEQUENTIAL MONTE CARLO SIMULATION 144 6.1. Introduction 144 6.2. Control Variates - Concepts 146 6.3. Case Studies Using Control Variate Method .. 151 6.3.1. Impact Of Control Coefficients On The Composite System LOLP And System EENS Of The RBTS ... . 151 6.3.2. Effect Of A Control Coefficient On The System ECOST For The RBTS .. 155 6.3.3. Analysis With The IEEE -RTS . 156 6.4. Antithetic Variates - Concept: . 158 6.4.1. Application To The RBTS 158 6.4.2. Application To The IEEE-RTS 161 6.5. Conclusions .... 163 7. FEATURES THAT INFLUENCE RELIABILITY WORTH ASSESSMENT .. 165 7.1. Introduction ........ erry 165 7.1.4. ‘Temporal Variation In Customer Interruption Costs 166 7.2. Respondent Estimates Of The Variation In Cost Of Interruption With Time of Year, Week and Day 170 7.3. Case Studies. . 173 7.3.1. Application To The RBTS 174 7.3.2. Application To The IEEE-RTS 176 7.8.3. Discussion Of The Results With Reference To The ‘Two Test Systems 179 7.4. Two Types Of Cost Models 180 7.4.1. Customer Damage Function Method - A Overview 181 7.4.2. Probability Distribution Approach . 183 7.4.2.1. Distributions At Intermediate Durations 184 7.4.2.2. Cost Transformation Results soe 186 742.3. Cost Sampling Method .. . 187 7.5. Case Studies .. . 188 7.5.1. Application To The RBTS 7.5.2. Application To The IEEE-RTS .. 7.5.3. Comparison Of The IEAR Using Both Types of Cost Models 7.6. Summary And Conclusions 188 189 191 192 8. SUMMARY AND CONCLUSIONS 195 8.1. Introduction 8.2. Conclusions 195 . 199 REFERENCES .. - 201 207 207 A. RANDOM NUMBER GENERATION A.1. Linear Congruential Method . A.1.1. Prime Number Multiplicative Congruential Generator . 208 B. DATA OF THE ROY BILLINTON TEST SYSTEM . 209 C. DATA OF THE IEEE - RELIABILITY TEST SYSTEM -. 210 213 217 D. TIME VARYING LOAD MODEL DATA .. D.1. Calculation of Sector Load Factors E, TEMPORAL VARIATIONS IN COST OF INTERRUPTION DATA. 219 F. SECTOR PEAK LOAD ALLOCATION FOR USE WITH THE PROBABILITY DISTRIBUTION COST MODEL .... 221 ‘Table 2.1: Table 2.2: Table 2.3: Table 2.4: Table 2.5: Table 3.1: ‘Table 3.2: Table 3.3: ‘Table 3.4: Table 3.5: Table 3.6: Table 3.7: Table 3.8: ‘Table 3.9: ‘Table 3.10: ‘Table 3.11: Table 8.12: Table 3.13: ‘Table 8.14: LIST OF TABLES Generation System Adequacy Indices for the RBTS Generation System Adequacy Indices for the IEEE-RTS ......_ 23 Sector Peak Load Allocation in MW at each Load Bus of the RBTS ...... - Allocation Factors in Per unit for Hour #1 .. Sector Peak Load Allocation in MW for the IEEE-RTS ........._ 37 Annualized Indices for The RBTS Annualized indices for the IEEE-RTS ....... Annual Load Bus Indices for the RBTS Annual System Indices for the RBTS . Annual System Indices for the IEEE-RTS Annual Load Bus Indices for the IEEE-RTS Comparison of System Indices Using Different Repair Distribution Processes ... Load Bus Indices Including Load Forecast Uncertainty in the IEEE-RTS . : System Indices of the IEEE-RTS with Load Forecast Uncertainty .. Unit Derating Data for the IEEE-RTS .. Load Bus Indices with Two Derated Units ... System Indices Including Two Derated Units Load Bus In 87 With Three Derated Units ‘System Indices Including Three Derated Units -..csccscss. 88 Table 4.1: Table 4.2: Table 4.3: Table 4.4: ‘Table 4.5: Table 4.6: Table 4.7: Table 4.8: Table 4.9: Table 5.1: Table 5.2: Table 5.3: Table 5.4: ‘Table 5.5: Table 5.6: Table 5.7: Table 5.8: Table 5.9: ‘Table 5.10: 15 Step Load model Data Sector Interruption Cost Estimates Expressed In ($/kW) Annualized Reliability Worth Indices ..... Comparison of the ECOST (k$/yr) for the RBTS. - 104 Comparison of the ECOST (k$/yr) for the IEEE-RTS 106 ECOST (k$/yr) for the RBTS .. 110 ECOST (k$/yr) for the IEEE-RTS a 110 ‘System ECOST for Different Repair Process Distributions 113 Effect of Load Forecast Uncertainty on the ECOST (k$/yr) for the IEEE-RTS - 116 Annual Load Bus ECOST (k§/yr) After Introducing Derated Generating Units for the IEEE-RTS soy System Reliability Indices for an Annualized Peak Load of 185MW - RBTS . 128 System Reliability Indices for an Annualized Peak Load of 2850MW - IEEE-RTS . 129 10 Step Load model Data .. sevens 181 Comparison of The System Indices Using The Two ‘Methods - 10 Uniform Load Steps for The RBTS .. . 181 Comparison of The System Indices Using The Two Methods - 15 Non-Uniform Load Steps for The RBTS.......... 132 Load Bus Indices Using The State Sampling (SS) Method - 70 Uniform Load Steps for The RBTS ... . 138 Load Bus Indices Using The State Transition Sampling (STS) Method - 70 Uniform Load Steps for The RBTS.......... 133 System Indices Using The State Sampling(SS) and State Transition Sampling (STS) Methods - 70 Uniform Load Steps for The RBTS . . 138 Comparison of The System Indices Using The Two Methods - 10 Uniform Load Steps for The IEEE-RTS......... 136 136 Table 5.11: Table 5.12: Table 5.13: Table 5.14: ‘Table 5.15: Table 6.1: Table 6.2: Table 6.3: Table 6.4: Table 6.5: Table 6.6: Table 7.1: Table 7.2: Table 7.3: Table 7.4: Table 7.5: Table 7.6: Table 7.7: Comparison of The System Indices Using The Two Methods - 15 Non-Uniform Load Steps for The IEEE-RTS Hag? Comparison of The System Indices Using The Two Methods - 10 Uniform Load Steps for The IEEE-RTS «....0uen 137 State Sampling Results with 49 Load Levels for The IEEE-RTS .. .. 188 State Transition Sampling Results with 49 Load Levels for The IEEE-RTS . .. 139 Comparison of The System Indices Using The Two Methods - 49 Non-uniform Load Steps for The IEEE-RTS, .. 140 Comparison of The System Indices for the RBTS 152 HLI System Indices for the IEEE-RTS using an Analytical Method .. 157 Load Bus Indices Using Antithetic Variates for the RBTS .. 159 ‘System Indices Using Antithetic Variates for the RBTS ..... 159 Load Bus Indices Using Antithetic Variates for the IEEE-RTS . . 161 Comparison of System Indices for the IEEE-RTS .... . 162 Variation of Interruption Cost with Time of Year, Week and Day .. : - 471 Variation of the ECOST (k$/yr) for the Different Cases of the RBTS .. 174 Variation of the ECOST (k/yr) for the Different Cases of the IEEE-RTS aT Customer Damage Function: 1991 $/kW Cost of Interruption Data . 182 ‘Transformation Factor, Distribution Parameters and Probability of Zero Costs corresponding to each Interruption Duration of the Industrial Sector ..... Comparison of the ECOST (k$/yr) Between Two Different Cost Models for the RBTS .. Comparison of the ECOST (k$/yr) Between Two Different Cost Models for the IEEE-RTS ... Table 7.8: Table B.1: ‘Table B.2: Table B.3: Table C.1: Table C.2: Table C.3: Table D.1: Table D.2: Table D.3: Table D.4: Table D.5: Table D.6: Table E.1: ‘Table E.2: Table E.3: Table E.4: Table E.5: Table E.6: Table F.1: Table F.2: Comparison of the IEAR ($/kWh) for the Test Systems Generator Data Bus Data Line Data sess Generator Data Bus Data Line Data Weekly Residential Sector Allocation Hourly Percentage of the Sector Peak Load for All Sectors . ee a Hourly Percentage of the Sector Peak Load for All Sectors .. 215 Daily Percentage of the Sector Peak Load 216 Daily Percentage of the Sector Peak Load . 216 Sector Average Load Value in MW at Each Load Bus of the RBTS . se 218 Interruption Cost Variation with Time of the Year zig Interruption Cost Variation with Day of the Week .. 219 Interruption Cost Variation with Time of the Day . 219 Interruption Cost Variation with Time of the Year for Residential and Agricultural sectors . 220 Interruption Cost Variation with Day of the Week for Residential and Agricultural sectors ............. 220 Interruption Cost Variation with Time of Day for Residential and Agricultural sectors ... 220 Sector Peak Load Allocation (MW) for the RBTS «0.2... 221 Sector Peak Load Allocation (MW) for the IEEE-RTS 221 Figure 2.8: Figure 2.9: Figure 2.10: Figure 3.1: Figure 3.2: Figure 3.3: Figure 3.4: Figure 3.5: Figure 3.6: LIST OF FIGURES Functional Zones in a Power System ‘Superimposition of the System Available Capacity ‘Model on the Load Modei Variation of the LOEE with Simulation Years for the RBTS ... 22 27 Load Profile for Agricultural and Industrial Sectors .. Load Profile for Commercial and Large User sectors ........_ 28 Load Profile for Residential and Office & Building Sectors 28 Load Profile for Government & Institution Sector Single Line Diagram of the RBTS with Customer Compositions 29 .. BL Load Curves for the RBTS . Single Line Diagram of the IEEE-RTS . A Daily Load Curve for Bus 2 and Bus 18 of the IEEE-RTS Flowchart for The Dual Upper Bounding Linear Programming Algorithm ... Variation of the Annualized System EENS with Simulation Years for the RBTS .. . 84 36 . 38 Variation of the Annualized System EFLC with Simulation Years for the IEEE-RTS .. Variation of the e System EENS with Simulation Years for the RBTS ... a Variation of the System EFLC with Simulation Years for the RBTS . 63 Distribution of the EENS index for the RBTS .......... xv Figure 3.’ Figure 3.8: Figure 3.9: Figure 3.10: Figure 3.11: Figure 3.12: Figure 3.13: Figure 3.14: Figure 3.15: Figure 3.16: Figure 3.17: Figure 3.18: Figure 3.19: Figure 3.20: Figure 4.1: Figure 4.2: Figure 4.3: Figure 4.4: Figure 4.5: Figure 4.6: Distribution of the LOLE index for the RBTS .. Distribution of the EFLC index for the RBTS Variation of the System EFLC with Simulation Years for the IEEE-RTS ... Variation of the System EENS with Simulation Years for the IREE-RTS .. Distribution of the LOLE Index for the IEEE-RTS Distribution of the EENS Index for the IEEE-RTS ... Distribution of the EFLC Index for the IEEE-RTS Variation of the System EENS with Simulation Years for the IEEE-RTS while using a Weibull Repair Process Distribution .. Distributions of the EENS Index ..... Distributions of the Frequency Index Distributions of the LOLE Index .. Variation of the System EENS with Simulation Years including Load Forecast Uncertainty for Nie TERE RTS sass esessereaeas Derated State Model for Generating Unit .... Operating Cycle for a Generating Unit with Derated States cae d85) Consumer, Utility and Total Costs As a Function of System Reliability . Sector Customer Interruption Costs Convergence of the System ECOST for the RBTS 105 Convergence of the System ECOST for the IEEE-RTS .... 107 Distribution of the ECOST index for the RBTS Distribution of the ECOST index for the IEEE-RTS sevseee 108 108 xvi Figure 4.8: Figure 4.9: Figure 4.10: Figure 5.1: Figure 5.2: Figure 5.3: Figure 5.4: Figure 5.5: Figure 5.6: Figure 6.1: Figure 6.2: Figure 6.3: Figure 6.4: Figure 6.5: Figure 6.6: Figure 6.7: Figure 6.8: Variation of the ECOST with Simulation Years for the RBTS = a Variation of the ECOST with Simulation Years for the IEEE-RTS 112 Variation of the System ECOST with Simulation Years while using a Weibull Distribution for the Repair Process of the IEEE-RTS ... — sevens M4 Distribution of the ECOST index for Non Exponential Distributions 114 Explanation of System State Transition Sampling 126 Load Duration Curve for The RBTS .. w 130 Comparison of the System EENS for the RBTS 134 Comparison of the System ECOST for the RBTS ..... 135 Comparison of the System EENS for the IEEE-RTS ........ 140 Comparison of the System ECOST for the IEEE-RTS ....... 141 Variation of the System LOLP in the RBTS with Simulation Years for an Optimal Control Coefficient ....... 152 Variation of the System EENS in the RBTS with Simulation Years for an Optimal Control Coefficient ......_ 153 Variation of the System LOLP in the RBTS with Simulation Years for a Constant Control Coefficient 153 Variation of the System EENS in the RBTS with Simulation Years for a Constant Control Coefficient 154 Variation of the System ECOST in the RBTS with Simulation Years for a Constant Control Coefficient 156 Variation of the System EENS in the RBTS with Simulation Years using Antithetic Variates .. 160 Variation of the System ECOST in the RBTS with Simulation Years using Antithetic Variates .. 160 Variation of the System EENS in the IEEE-RTS with Simulation Years using Antithetic Variates . 162 xvii Figure 6.9: Figure 7.1: Figure 7.2: Figure 7.4: Figure 7.5: Figure 7.6: Figure 7.8: Figure 7: Figure 7.10: Variation of the System ECOST in the IEEE-RTS with Simulation Years using Antithetic Variates . 163 Mean Attribute Importance 167 Breakup of Time of Interruption Comparison of the ECOST for the RBTS Obtained using ‘Temporal Variations for Cost of Interruption Data 175 . 168 Convergence of the System ECOST for the RBTS - Case G o..cccenve Comparison of the ECOST for the IEEE-RTS Oktained using Temporal Variations for Cost of Interruption Data 178 . 176 Convergence of the System ECOST for the IEEE-RTS - Case G 179 Customer Damage Function for 1991 Cost of Interruption Data .. see 182 Normal curves for the 1991 Industrial Sector's ‘Transformed Cost ... wae 186 Variation of the System ECOST for the RBTS with Simulation Years using Two Different Cost Models 189 Variation of the System ECOST for the IEEE-RTS with Simulation Years using Two Different Cost Models 191 xviii ac APM av Bldg. CCDF CEA Com. COMREL Diff. dist. ECOST EDNS EENS EFLC ENLC ENS F&D HLIL HLULI IEEE Indus. LOLE LOEE LIST OF ABBREVIATIONS Alternating Current Agricultural IEEE sub-committee on the Application of Probability Methods Antithetic Variate Building Composite Customer Damage Function Canadian Electrical Association Commercial COMposite system RELiability evaluation Control Variate Difference disturbance Expected outage COST Expected Demand Not Served Expected Energy Not Served Expected Frequency of Load Curtailments Expected Number of Load Curtailments Energy Not Served Frequency and Duration. Government Hierarchical Level I Hierarchical Level IT Hierarchical Level III Interrupted Energy Assessment Rate Institute of Electrical and Electronics Engineers Industrial Institution Load Forecast Uncertainty Loss Of Load Expectation Loss Of Energy Expectation xix MECORE NSERC rdf PLC RBTS RTS SCDF sIc Ss TS an a8 vRT Monte carlo Enumeration and simulation approach for COmposite REliability evaluation Mean Time To Failure Mean Time To Repair Natural Sciences and Engineering Research Council probability density function Probability of Load Curtailed Roy Billinton Test System Residential Reliability Test System Sector Customer Damage Function Standard Industrial Classification State Sampling State Transition Sampling ‘Time To Failure ‘Time To Repair Variance Reduction Technique Xx 1. INTRODUCTION 1.1. Introduction The increasing dependence of modern society on electrical energy puts a heavy pressure on electric power utilities to provide an energy supply of acceptable quality with reasonable assurance of continuity [1]. It is not economical and technically feasible to attempt to design a power system with one hundred percent reliability. Power engineers, however, attempt to achieve an acceptable level of reliability within existing economic constraints. Attempts have been made over the last three decades to resolve the dilemma between the economic, operating and reliability constraints, by developing a wide range of techniques. The criteria and techniques first used in practical application, and in use even today, are deterministically based. The basic drawback of deterministic criteria is their inability to consider the probabilistic or stochastic nature of system behavior, customer demands or of component failures. The need for probabilistic evaluation was recognized many years ago but these techniques were not widely used in the past due to lack of data and computational resources, ete. A wide range of probabilistic techniques have now been developed which attempt to recognize the severity of an outage event, its impact on system behavior and operation, together with the likelihood (probability) of its occurrence. The data required to support these techniques is generally available and computational resources are not a problem. 1.2. Power System Reliability And Related Concepts ‘The word reliability when used in the context of power networks is generally defined as the concern regarding the ability of the power system to provide an adequate supply of electrical energy [2]. The term reliability has a wide range of meaning and cannot be associated with a single specific definition. In general terms, it is related to the existence of sufficient facilities within a system so that the system is capable of supplying electric power to its customers under both static and dynamic conditions, with a mutually acceptable assurance of continuity and quality [1]. It is necessary to recognize the extreme generality of the term and therefore to use it to indicate, in a general rather than a specific sense, the overall ability of the power system to perform its function. A simple but reasonable subdivision of the concern designated as system reliability can be made by considering the two basic and functional aspects of the system, adequacy and security [1]. System adequacy relates to the existence of sufficient facilities within the system to satisfy the consumer load demand or system operational constraints. These include the facilities necessary to generate sufficient energy and the associated transmission and distribution facilities required to transport the energy to the consumer load points. Adequacy is therefore associated with static conditions which do not include system disturbances. System security relates to the ability of the system to respond to disturbances arising within it. These include the conditions associated with both local and widespread disturbances and the loss of major generation and transmission facilities ete. This thesis is restricted to the adequacy assessment of electric power ion of the electrical facilities within a power system can be performed to provide the three segments of generation, transmission and distribution facilities. The adequacy assessment of a power network in its entirety is quite complex and computationally expensive. All the basic techniques for adequacy assessment can be classified in terms of their applications to the three segments. These segments are also called functional zones [2] as shown in Figure 1.1. This division of the power system is somewhat simplistic. Nevertheless it is appropriate, as most power utilities are either divided into these three zones for purposes of organization, planning, operation and/or analysis or are solely responsible for one of these functions. GENERATION | be Hierarchical Level 1 FACILITIES I [TRANSMISSION jug—|— Hierarchical Level II FACILITIES DISTRIBUTION <— Hierarchical Level I FACILITIES Figure 1.1: Functional Zones in a Power System Figure 1.1 also shows the combination of these functional zones into hierarchical levels which represent important functional boundaries. Generation facilities alone form hierarchical level-I (HLI) and together with the transmission facilities constitute hierarchical level-II (HLII). Hierarchical level-IIT (HLIID is composed of all three functional zones. The focus of thio thesis is on adequacy assessment at HLII. The total problem of adequacy assessment of the generation and the transmission facilities in regard to supplying an adequate, dependable and suitable energy to the major customer load centers is designated as composite system or HLII adequacy evaluation. The reliability performance of a particular configuration can best be expressed by calculating appropriate indices. These indices can be calculated for each hierarchical level. Some indices provide attributes of reliability as seen by a customer, but many indices are calculated for internal use in a power utility. These indices assist utility system planners and operators to compare alternate plans or operating procedures in specific segments of a power system. Reliability criteria can be broadly divided into: * Deterministic criteria. © Probabilistic criteria. In general, deterministic criteria indicate a concern regarding certain properties of the power system and use discrete measures of the network components such as the number of generating units, transmission lines and transformers ete. that must fail before it is considered impossible for the system to supply the load demand at a particular load center. Deterministic criteria for generation assessment such as the percentage reserve approach [1] have been used by many power utilities for many years and are still being used by a small group of utilities. The drawback with a deterministic approach is that it is not responsive to many of the parameters which actually influence system reliability. Some of these parameters are the failure and repair rates of various system components, dependent outage events, maintenance schedules and the load profile in the system. The criteria are therefore not directly related ta the electric system performance experienced by system customers. The deterministic criteria are now giving way to more realistic probabilistic criteria. Probabilistic criteria are concerned with the probability (and other relevant indices) that a power network can supply the required load within an acceptable range of quality (the voltage at the bus) to a customer. Probabilistic indices directly consider the uncertainty which is inherent in the power system reliability domain. These indices reflect the response of various parameters which actually affect the system reliability. This presents a more realistic picture of the performance of a system than that provided by a deterministic approach and is the basic reason why these indices are becoming popular in modern power utilities [1]. 1.3. Concepts Of Composite System Analysis ‘The economic, social and political climate in which the electric power industry now operates has changed considerably during the last few decades. It is now widely recognized that statistical assessment of past performance is an important aspect in the planning and operation of power systems. ‘The quantitative assessment of each of the functional zones of an electric power system is widely conducted using existing techniques. Application of these techniques in the planning of generation systems and distribution networks is fairly well advanced and is widely used. The development of a suitable transmission network to convey the energy to the major load points is an important part of the planning process and is termed as composite or bulk power system expansion planning. The application of quantitative reliability assessment techniques to bulk power systems (composite generation and transmission systems) is in its infancy and relatively little use is made of these techniques in practical decision-making. The need, however is widely recognized and interest is expanding rapidly although deterministic criteria are still the norm [8]. ‘The procedures used for adequacy assessment at HLII can be broadly classified into two general areas: * Analytical methods © Monte Carlo simulation techniques. Analytical techniques represent the system by a mathematical model and evaluate the reliability indices from this model using mathematical solutions. Monte Carlo simulation methods, however, estimate the reliability indices by simulating the actual process and random behavior of the system. The method therefore treats the problem as a ceries of experiments. There are merits and demerits in both methods. Generally Monte Carlo simulation requires a large amount of computing time compared to analytical methods. However, it can include any system effect or system process which may have to be approximated in an analytical approach. ‘There are two basic techniques utilized when Monte Carlo methods are applied to power system reliability evaluation. These methods are known as the sequential and non sequential techniques. In the basic state sampling technique, the states of all components are sampled and a non-chronological system state is obtained. Research work in the area of composite system adequacy evaluation using the non sequential Monte Carlo simulation approach has been done at the University of Saskatchewan [4] and a computer program MECORE (Monte carlo-Enumeration-COmposite system REliability evaluation) [5] has been developed. This program utilizes a hybrid approach consisting of a non sequential Monte Carlo simulation and an enumeration technique for the reliability evaluation of large scale composite generation-transmission systems. There is frequently a need to know the likely range of the reliability indices, the likelihood of a certain value being exceeded, and similar parameters. These can only be assessed from a knowledge of the probability distribution associated with the expected value, and this is rarely achievable using an analytical or non sequential approach. In such cases, sequential simulation can be utilized. In sequential simulation, each subsequent system state sample is related to the previous set of system states. A sequential time evolution of system behavior is created which enables a wide range of reliability indices to be evaluated. 14. Scope And Objectives Of The Thesis ‘The focus of this research work is on the reliability evaluation of composite generation and transmission systems with special reference to frequency and duration related indices and estimated power interruption costs at each load bus. The sequential simulation technique permits sampling random variates from a wide range of probability distributions, to evaluate expected system adequacy indices and their distributions. ‘The main objective of this research work is to use the sequential simulation method with time varying bus load models to create a comprehensive technique for HLII adequacy evaluation. This includ + the development of a comprehensive time varying load model at each bus in the Roy Billinton Test System (RBTS) and the IEEE Reliability Test System (IEEE-RTS) * examining the effect of the time varying load model on composite system adequacy indices * extending the analysis to reliability worth studies in systems with time varying bus loads * comparing the detailed sequential simulation results with more conventional approximate approaches * examining the use of variance reduction techniques in sequential simulation analysis of composite power systems. The basic concepts and different types of Monte Carlo simulation [4, 5] are presented in Chapter 2. The sequential Monte Carlo simulation is discussed in detail and applied to generation adequacy assessment of two test systems, the RBTS [6] and the IEEE-RTS [7]. ‘The need for a more accurate representation of the individual bus loads is illustrated and procedures used to create hourly loads for all the load buses of the two test systems are presented in this chapter. Composite generation and transmission (HLII) adequacy evaluation is concerned with the total problem of assessing the ability of the generation and transmission system to supply adequate and suitable electrical energy to the major system load points. Quantitative assessment of the adequacy of a composite system can be performed using a Monte Carlo simulation approach. Chapter 3 presents a general review of the types of indices, namely the load point and system indices. The software package developed as a part of this research work to utilize a sequential Monte Carlo simulation method together with time varying loads is explained in this chapter. This program was applied to the composite system adequacy analysis of the RBTS and IEEE-RTS. Both the adequacy indices and distributions of these indices are presented with reference to the test systems. Reliability worth assessment plays an important role in electric power system planning and operation. Chapter 4 proposes a technique for evaluating the costs of interruption and hence, the reliability worth in a composite power system with time varying loads at load buses using sequential Monte Carlo simulation. A generalized methodology of determining the interruption costs, without considering the time varying aspect of the loads is also discussed in this chapter. The procedures developed were used to determine the reliability worth of the test systems. The effect of utilizing a specific curtailment strategy on the reliability worth is also illustrated. Sensitivity analyses was performed by considering various distributions to represent the repair process of components, introducing bus load forecast uncertainty aspects and considering derated generating units. The two non sequential Monte Carlo simulation methods, namely the state sampling and state transition sampling techniques used for composite system analysis are briefly described in Chapter 5. The effect of load model approximations for non sequential methods is illustrated in this chapter. An approximate method utilizing a load duration curve of the system load and an enumeration process was applied to the developed load model, and the results are compared with those obtained from using a sequential Monte Carlo simulation approach for the test systems. Estimation of composite system adequacy indices using the sequential Monte Carlo simulation approach with time varying loads at each load bus is computationally expensive. Variance reduction techniques can be used together with the sequential simulation process to enhance the efficiency of 10 the simulation. Chapter 6 discusses two commonly used variance reduction techniques’ - control variates and antithetic variates with reference to the sequential simulation technique. The efficiencies of these variance reduction techniques are illustrated by application to the test systems. ‘Two aspects that influence reliability worth assessment are dealt with in Chapter 7. The first consideration is the variation in interruption costs with time of interruption occurrence. This chapter illustrates the effect on the reliability worth, of temporal variations in the interruption costs for the residential, agricultural, industrial, commercial and large user sectors of the test systems. ‘The second aspect considered is the use of a probability distribution approach to model the interruption costs, in the evaluation of reliability worth. The probability distribution approach was incorporated for reliability worth analysis of the test systems and the results are compared with those from a conventional customer damage function cost model. Chapter 8 summarizes the thesis and presents the conclusions. 2. CONCEPTS OF SEQUENTIAL MONTE CARLO SIMULATION AND TIME VARYING LOADS AT LOAD BUSES 2.1. Introduction ‘There are two main categories of power system reliability evaluation techniques: analytical and simulation. Analytical techniques represent the system by analytical models and evaluate the indices from these models using mathematical solutions. Monte Carlo simulation methods, however, estimate the indices by simulating the actual process and random behavior of the system. The method, therefore treats the problem as a series of experiments. There are merits and demerits in both methods [4]. When complex operating conditions are involved and /or the number of severe events is relatively large, Monte Carlo methods are often preferable. The analytical approach is usually restricted to the evaluation of expected values and sometimes to a limited range of system parameters. There is frequently a need to know the likely range of the reliability indices, the likelihood of a certain value being exceeded, and similar parameters. These can only be assessed from a knowledge of the probability distribution associated with the expected value, and this is rarely achievable using an analytical approach. In such cases, simulation can be utilized. Simulation can be used to experiment with new situations about which little or no information is available. Simulations can sometimes be valuable in breaking down a complicated u 12 system into subsystems, each of which may then be modeled and analyzed separately. 2.2, Monte Carlo Simulation Techniques There are two basic techniques utilized when Monte Carlo methods are applied to power system reliability evaluation. These methods are known as the sequential and non-sequential techniques. 2.2.1. State Sampling Technique In the basic state sampling technique (5, 10], the states of all components are sampled and a non-chronological system state is obtained. In the random simulation approach, each consecutive sample of system states is randomly selected quite independently from previous and subsequent samples. The basic sampling procedure is conducted by assuming that the behavior of each component can be categorized by a uniform distribution (0,1). A major disadvantage of the state sampling technique is the difficulty associated with attempting to calculate frequency and duration related indices. This is due to the fact that the frequency calculation related to a failure state requires the recognition of all no load curtailment states which can be reached from the failure state in one transition. This cannot be achieved using the state sampling technique by itself and it is necessary to add an enumeration procedure which may be computationally very expensive [11]. One approach to this problem is the state transition sampling technique, in which an actual frequency index can be obtained without requiring an additional enumeration procedure. 13 2.2.2, State Transition Sampling Technique This technique [12, 13] focuses on state transitions of the system as a whole rather than on component states or durations. The major advantage of this method is that it can be used to calculate an actual frequency index by creating a system state transition chain. Generating unit derated states can easily be included without additional calculations. This method requires the assumption of exponential distribution residence times, which may not always be true. State sampling techniques in general do not involve sampling component state duration distribution functions and the storage of chronological information as required in the sequential method. The state sampling and state transition sampling techniques were used in the adequacy assessment of composite power systems as a part of the author's M.Se. research work [11, 14]. The next section discusses the opportunities for reliability evaluation using the sequential simulation method. 2.2.3. Sequential Simulation Method ‘The term sequential simulation is often used in engineering literature [15, 16] to designate the technique in which the history of a system is simulated in fixed discrete time steps. Specialized literature on simulation techniques calls this "discrete-event simulation with fixed-increment time advance" [17] and reserves the term sequential for the procedure that takes samples sequentially until a certain measure of accuracy is achieved. This thesis uses the term sequential simulation in the engineering sense (18], in which any event occurring within a particular time step is considered to occur at the end of the time step, and the system states are updated accordingly. A time step 14 of 1 hour is considered adequate for power system reliability analysis since the number of changes within that period is generally small. In sequential simulation, each subsequent system state sample is related to the previous set of system states. A sequential time evolution of system behavior is created which enables a wide range of reliability indices to be evaluated. The random factors affecting the capacity and energy states of history dependent systems and various operating scenarios are more appropriately taken into account by using sequential Monte Carlo simulation. Sequential simulation is very useful when the system to be analyzed is past- dependent, i.e., the state of the system at any given time is partially determined by the historical evolution of the system. Sequential simulation becomes a necessity when the operating system is history-dependent or time correlated. This particularly applies to hydro generation systems when the reservoir storage capacity is relatively small and the use of water has to be carefully controlled [19], since the available power at any moment is dependent on, among other factors, the past water inflows, past operation policies and historical evolution of the load. At the present time, sequential simulation is the only realistic option available to develop the distributional aspects associated with system index mean values. This approach can be used to represent most of the contingencies and operating characteristics inherent in the system and also provide the most comprehensive range of reliability indices. Sequential Monte Carlo simulation is the only method that can deal with the time varying aspect of the loads at each bus. ‘The probability distributions associated with times-to-failure and times- to-repair of each system component are often not known and only the mean time to failure (MTTF) and the mean time to repair (MTTR) are available. In these cases, it is generally assumed that the underlying distributions are 15 exponential and the expected values are calculated on that basis. The sequential simulation method can, however, be used to sample the failure and repair times from any statistical distribution, such as Weibull, normal ete. 2.3. Basic Methodology Of Sequential Simulation ‘The sequential simulation approach is based on sampling the probakility distribution of the component state duration. In this approach, chronological component state transition processes for all components are first simulated by sampling. The chronological system state transition process is then created by combination of the chronological component state transition processes. This approach uses the component state duration distribution function. In a two-state component representation, these are the operating and repair state duration distribution functions and are usually assumed to be exponential. Other distributions can also be used and these are discussed later. The sequential method can be summed up in the following ste Step 1: The initial state of each component is specified. Generally, it is assumed that all components are initially in the success or up state. Step 2: The duration of each component residing in its present state is sampled from its probability distribution. For example, an exponentially distributed random variate has the probability density function [15], frit) =he It (2.1) where 4 is the mean value of the distribution. Its cumulative probability distribution function is F(t)=1-e -At (2.2) 16 Using the inverse transform method the random variate T' [15] is given by: In. (1-U) (2.3) where U is a uniformly distributed random number obtained from a multiplicative congruential pseudo-random number generator. The procedure used to generate the pseudo-random number is described in Appendix A. Since 1-U distributes uniformly in the same way as U in the interval [0,1], T zn@. (24) If the present state is the up state, 2 is the failure rate of the component. If the present state is the down state, Ais the repair rate of the component. Step 3: Step 2 is repeated in the given time span, i.e., usually a year and sampling values of each state duration for all components are recorded. The chronological system state transition process can be obtained by combining the chronological component state transition processes of all components. Step 4: System analysis is conducted for each different system state to obtain the reliability index function 9(S). Assuming that each system state has the probability P(S), the mathematical expectation of the index or test function of all system states is given by: E(¢)=_¥ 9(S) P(S) where G is the set of system states. (2.5) SeG 7 Substituting the sampling frequency of the state S for its probability P(S) gives n(S) E(o)= X (5) (2.6) SeG where N is the total number of samples and n(S) is the number of occurrences of state S. 9($) can be obtained by appropriate system analysis. For example, to determine the system probability of load curtailment, the index function 9(5) is given as [20]: 1 if there is a load curtailment associated with state S 9(S) = { an 0 if there is no load curtailment Equations 2.5 and 2.6 are associated with the random state sampling approach. This same concept when extended for use in sequential technique gives the estimate of the expected value NS [ni(S) zr] > sv) (9) (2.8) where: n(S) = Number of occurrences of state $ in year i, $(S:) = Index function corresponding to j th occurrence in year i, NS = _—_ Number of simulation years. 18 2.3.1. Convergence Criterion If @ denotes the unavailability (failure probability) of a system and X; is a one-zero indicator variable or index function which states that fr then, the estimate or the expectation of system unavailability is given by [4] N 0 ifthe system is in the up state 1 ifthe system is in the down state oe (2.10) where N is the number of system sample states. ‘The sample variance is N. =e S(xi - 3)? Wel vOO = (2.41) and the uncertainty around the estimate can be measured by the variance of the expectation estimate: = V(x) v(a)= ——. (2.12) (Q)= =F (2.12) ‘The accuracy level of Monte Carlo simulation can be expressed by the coefficient of variation, f, which is defined as. B £2) (2.18) The coefficient of variation shown in Equation 2.13 is often used as the convergence criterion in Monte Carlo simulation studies. In power system reliability evaluation, different reliability indices have different convergence 19 speeds. It has been found that the coefficient of variation of the EENS index has the lowest rate of convergence. This coefficient of variation is therefore used as the convergence criterion in order to guarantee reasonable accuracy in a multi-index study (4). It can be observed from Equation 2.13, that for a desired level of accuracy B, the required number of samples N depends on the system unavailability and is independent of the size of the system. For this reason, Monte Carlo methods are quite suitable for handling large systems with complex features. It should also be noticed that the computational effort is affected by the value being estimated, i-e., the more reliable the system is, the harder it is to estimate the value. Finally, computational effort is heavily affected by the accuracy (B value) sought. 2.3.2. Advantages Of The Sequential Approach © The sequential method can be easily used to calculate the actual frequency index. * Any state duration distribution can be easily considered. ‘* The statistical probability distributions of the reliability indices can be calculated in addition to the expected values. 2.3.3. Disadvantages Of The Sequential Method * Compared to the basic state sampling approach, this method requires more computation time and storage because it is necessary to generate a random variate following a given distribution for each component and 20 store information on chronological component state transition processes of all components in a long time span. * This approach also requires parameters associated with all component state duration distributions. In some cases, especially for a multistate component representation, it may be difficult to provide all transition rates between states of each component. 4. Sequential Monte Carlo Simulation For Generation Adequacy Assessment In this section, the sequential Monte Carlo simulation approach is applied to a single area generating system adequacy assessment in order to illustrate the basic approach. The first step is to generate operating histories for each generating unit by drawing sample values of TTF (Time-to-Failure) and TTR, (Time-to-Repair) of the unit. The operating history of each unit is in the form of chronological up-down-up operating cycles. The system available capacity is then obtained by combining the operating cycles of all units. ‘The sceond step is to superimpose the system available capacity curve on the chronological system hourly load curve to obtain the system available margin model. A positive margin denotes that the system generation is sufficient to meet the system load, while a negative margin implies that the system load has to be curtailed. This superposition is indicated in Figure 2.1 Appropriate adequacy indices can then be estimated [4]. 2 system available capacity wll u hourly system load ENS L_ ENS: Energy Not Served ‘TIME (hours) Figure 2.1: Superimposition of the System Available Capacity Model on the Load Model The sequential simulation approach is illustrated by applying this technique to a HLI adequacy evaluation. The generating system adequacy assessment has been performed for two test systems. The test systems used for adequacy studies in this thesis are: 1. Roy Billinton Test System - (RBTS) 2. IEEE Reliability Test System - (IEEE-RTS) In these studies, the generating unit state durations are assumed to be exponentially distributed and all generating units are two state units. The 22 annual chronological hourly load model of the IEEE-RTS [7] has been used to represent the load model for both the test systems. ‘The RBTS [6] has eleven generating units, ranging from 5 MW to 40 MW. The generating unit data are given in Appendix B. Table 2.1 describes the generating system adequacy indices for the RBTS with an annual peak load of 185 MW. Table 2.1: Generation System Adequacy Indices for the RBTS Loss of Load Expectation -LOLE (h/yr) |__1.071_| | Loss of Energy Expectation - LOEE. 8.818 (MWhiyr) Loss of load frequency (oce/yr) 0.2325 Simulation years 800 CPU time (sec) 53 Figure 2.2 shows the convergence of the loss of energy expectation (LOEE) index over 800 simulation years. 12 g a 10 & Ee § 22 Be . g 2 ° 0 200 400 600 800 Simulation years Figure 2.2: Variation of the LOEE with Simulation Years for the RBTS 23 ‘The IEEE-RTS [7] has 32 generating units with sizes ranging from 12 MW to 400 MW. The generating unit reliability data are described in Appendix C. ‘The generation system adequacy or HLI indices for the IBEE-RTS with an annual peak load of 2850 MW obtained using the sequential Monte Carlo simulation approach is described in Table 2.2. Table 2.2: Generation System Adequacy Indices for the IEEE-RTS Loss of Load Expectation -LOLE (h/yr) | __ 9.213 Loss of Energy Expectation - LOEE| 1168.2 (MWh/yr) Loss of load frequency (oce/yr) 1.86 Simulation years 400 CPU time (see) 157 The sequential Monte Carlo simulation method can be used +o perform the system analysis in a chronological manner. This method allows time dependent loads to be considered in the analysis. The next section discusses the need for the development of such time varying load models. 2.5. Accurate Representation Of The Load Model At Each Load Bus ‘The most readily available comprehensive load model in terms of overall system load is that proposed in the IEEE Reliability Test System (RTS) [7] This system was published in 1979 by the IEEE Subcommittee on the Application of Probability Methods (APM). The creation of the IEEE-RTS also provided a starting point in collecting relevant data required in reliability studies. The IEEE-RTS has been used extensively, since it was proposed, in 24 various reliability studies conducted by reliability engineers in power industries and universities. The report [7] by the IEEE Reliability Test System Task Force describes a load model, generation system and transmission network. The load model information provided can be used to. calculate hourly loads for one year on a per unit basis, expressed in a chronological fashion so that daily, weekly and seasonal patterns can be developed. ‘The annual peak load for the test system is 2850 MW. The system load is described by specifying the weekly peak loads in percent of the annual peak load, the daily peak load in percent of the weekly peak load and the hourly peak load in percent of the daily peak load. This load model is sufficient for system reliability studies at HLI such as loss of load expectation (LOLE) and loss of energy expectation (LOEE) as illustrated in Section 2.4. The published information, however, is not adequate for estimating costs of interruption which require additional information for each customer class. The IEEE-RTS load data is specified as total system demand and does not indicate how individual customer class loads should vary during the period concerned. All the earlier studies [4, 18] have used the IEEE-RTS hourly load model for the system as a whole. Individual bus loads, at any hour were assumed to be proportional to the ratio of peak Icad at that bus to the peak load of the system. This procedure is not absolutely correct as individual buses follow different load curves depending on the mix of customers at that bus. Different hourly load curves at each bus should be developed so that detailed estimates of the cost of interruption can be obtained using sequential simulation. 25 2.6. Development Of A Chronological Load Model At Each Load Bus One of the difficulties in applying probabilistic methods in the area of cost of interruption studies is that these methods require extensive load information for each customer sector [21]. These data are not usually available. The increased popularity of applying stochastic methods in system reliability evaluation has created a demand for the collection of outage data and other relevant information. Detailed load consumption and demand information is, however, still not available. ‘The application of probabilistic methods in the area of determining cost of interruption at a bus requires detailed knowledge of each customer load. In the absence of this information, it was therefore necessary to create a database which contains relevant information about each customer sector load. This was accomplished using some available data and a series of realistic assumptions. ‘The Standard Industrial Classification (SIC) has been used to identify seven types of customer sectors [22]. These sectors are + Large user + Industrial * Commercial * Agricultural + Residential * Government & institutions * Office & buildings. ‘The load at each bus has been allocated to these different sectors (22] for the test systems used in this work. The general load shapes of these sectors are quite unique. Their characteristics are described in the following section. 26 2.6.1. Customer Characteristics Industrial loads are considered to be base loads that contain little weather dependent variation [23]. However, depending on the type of industry, these loads may have unique characteristics because of shift operations, etc. The electricity use characteristics of large users and industrial customers are similar. Large industrial customers normally have a relatively large demand for electric power that remains quite stable from day to day or season to season. In general, larger industrial customers, with more continual production activities, have the most uniform demand for electrical energy. Smaller industrial customers who may run only two shifts per day with minimal or no weekend production have lower demands during evenings and weekends. However, these smaller industrial customers exhibit a fairly constant demand during production hours. Commercial and government & institutional demand curves are relatively high but constant during the daylight hours of the normal business day and fall off during the nighttime hours. In the case of commercial establishments, evening demand may fall off gradually due to the accommodation of evening shopping hours in many retail outlets. This class of customers also show seasonal variations as a result of space conditioning and seasonal differences in lighting, which constitute their major energy requirements. Residential [24, 25] and agricultural customers show greater temporal variability in their demand for electrical power than do commercial and industrial customers. Demand, particularly by residential customers, is very strongly dependent upon seasonal weather variations and also exhibits very pronounced daily peak demands during the early morning and early evening. a Daily load variation in the residential sector is primarily as a result of domestic uses of cooking equipment, hot water and lighting. Residential loads have the most seasonal fluctuations. The seasonal variations of the residential components in many cases are responsible for the seasonal variations in system peak, the extent of the residential influence depending on the percentage of the total system load that is residential (251. This characteristic is due to the widespread use of weather sensitive devices such as space heaters and air conditioners. Other high-energy devices used by residential customers are water heaters, refrigerators and dryers. Refrigeration loads tend to have constant characteristics compared to the cyclical load characteristics of dryers and water heaters. ‘The load profiles of these seven customer sectors for a typical day are shown in Figures 2.3 to 2.6. 2 08 206 Boa 2 02 ° 0 6 12 18 24 Agricultural. —€9— Industrial Figure 2.3: Load Profile for Agricultural and Industrial Sectors g 08 3 0. B04 B02 0. 0 6 12 18 24 Hour —4— Commercial ——— Large users Figure 2.4: Load Profile for Commercial and Large User sectors 2 08 3 06 5 04 = 02 o o 6 2 18 24 Hour —o— Residential —— Office & Buildings Figure 2.5: Load Profile for Residential and Office & Building Sectors 29 Per unit value eecee of FOS - 0 6 12 18 24 Hour Figure 2.6: Load Profile for Government & Institution Sector 2.7. Developing The Load Curves For The Buses Of The Two Test Systems The load curves for the customer sectors were developed taking into account daily, weekly and seasonal patterns. Seasonal influences have also been considered in the load model. The three seasons used are summer, winter and fall/spring. The 8736 hours are distributed into the three seasons as follows: . Winter hours: _1-1344 and 7225-8736. . Fall/Spring hours: 1345-2856 and 5041-7224. . Summer hours: 2857-5040. The weekly, daily and hourly percent of the sector peak load attributed to the various sectors is described in Appendix D. These hourly load curves were 30 developed with reference to the two published test systems, the RBTS [6] and the IEEE-RTS (7]. 2.7.1. The Roy Billinton Test System - RBTS The RBTS is an educational test system developed by the Reliability Section of the Power Systems Research Group at the University of Saskatchewan [6]. It is sufficiently small to permit the conduct of a large number of reliability studies with reasonable solution time yet sufficiently detailed to reflect the actual complexities involved in a practical reliability analysis. This system is abbreviated as the RBTS throughout this thesis. The single line diagram of the test system together with customer composition at load buses is shown in Figure 2.7. It can be seen from this figure that there are some residential and commercial sector customers at every load bus. As an example, Bus 2 has industrial, commercial, residential and government and institutional users allotted to it. The bus data and generator data of this system are given in Appendix B. The system has 2 generator buses, 5 load buses, 9 lines and 11 generating units. The voltage limits for this test system are assumed to be 1.05 to 0.97 p.u. The total system load is 185 MW and the total generation is 240 MW. The maximum and minimum ratings of the generating units are 5 MW and 40 MW respectively. The transmission network shown in Figure 2.7 has been drawn to give a geographical representation. 3I BUS2 Industrial Commercial Residential Govt. & Inst. Bus 2 Bus 3 Bus 4 85 MW 40MW BUS 3 BUS 4 Large users Tadusteial Industrial ‘Commercial ‘Commercial Residential Residential Office & Bldg! Bus6 20MW BUSS ‘ommerciall Industrial Residential Commercial | Bus 6 Govt. & Inst. ‘Agricultural Office & Bldg. Residential 20MW Figure 2.7: Single Line Diagram of the RBTS with Customer Compositions 32 2.7.1.1, Caleulation Of Hourly Load At The Load Buses In The RBTS Let Yim be the proportion of the sector peak load contributed by sector m during hour j, to the load at bus k. Yjm is also referred to as the allocation factor. The load at bus &, for hour jis given by Equation 2.14. Load at bus k for hourj = all sectors in bus k (Yim x sector peak load allocation at bus k). (2.14) m=i The system peak load for the RBTS is 185 MW. The sector peak load allocation for each bus of the RBTS is given in Table 2.3. Table 2.3: Sector Peak Load Allocation in MW at each Load Bus of the RBTS User Sector _|Bus2 |Bus3 |Bus4 |Bus5 | Bus6 Large user 55.50 Industrial 3.50 | 3.05 | 16.30 3.05 Commercial 3% | 4.70 | 4.70 | 3.70 | 1.70 Agricultural 7.40 Residential 7.25 | 19.90 | 19.00 | 8.90 | 7.85 Govt. & Institution | 5.55 5.55 | Office & Bldg. 1.85 1.85 In order to illustrate the procedure, a sample calculation for determining the bus loads for hour #1 in a year is described. Depending upon the season, day of week and time of day, the allocation factor in per unit for 33 the various sectors is obtained. Table 2.4 describes the allocation factor for the seven customer sectors during the first hour of the year. Table 2.4: Allocation Factors in Per unit for Hour #1 User Sector Allocation factor Large users 0.337 Industrial 0.1037 Commercial 0.01 Agricultural 0.001 Residential 0.531072 Govt. & Institution o4 Office & Bldg. 0.59 Weighting these allocation factors by the respective bus sector peak load results in the sector hourly load at a bus. The bus loads are then calculated as the summation of these sector hourly loads at the bus. Equations 2.15 to 2.19 illustrate this sample calculation for determining the bus and sector loads at hour #1 of the year. Bus 2 = (0.531072 x7.25) + (0.1037 x 3.50) + (0.01 x 3.75) + (0.4 x 5.55) 46 MW (2.15) Bus 3 = (0.837 x 55.50) + (0.1087 x 3.05) + (0.01 x 4.70) + (0.631072 x 19.90) + (0.59 x 1.85) = 30.727 MW (2.16) Bus 4 = (0.531072 x 19) + (0.1087 x 16.30) + (0.01 x 4.70) = 11.828 MW 24n 34 (0.01 x 3.70) + (0.531072 x 8.90) + (0.4 x5.55) + (0.59 x1.85) = 8.075 MW (2.18) (0.1087 x 3.05) + (0.01 x 1.70) + (0.001 x 7.40) + (0.531072 x 7.85) = 4.509 MW (2.19) The load curves obtained for the first week of the year for the RBTS is presented in Figure 2.8. It can be observed that bus 3, has a flat segment during daytime, since the major contribution to the load at this time is from the large user sector. 90 80 70 60 50 40 Load (MW) 30 20 10 o 24 48 n 96 120 144 168 Hour -----Bus2 ‘Bus 3. —O-—— Bus 4 Bus 5 —#—Bus 6 Figure 2.8: Load Curves for the RBTS 35 .2, The IEEE Reliability Test System. ‘The single line diagram of the 24 bus IEEE-RTS is given in Figure 2.9. This system was established in 1979 by an IEEE Task Force (7] as a reference network which can be used to test or compare methods for system reliability studies. The bus, line, transformer and generator data of this system are included in Appendix C. The system has 10 generator buses, 17 load buses, 33 transmission lines and 5 transformers. The total number of generating units is 32. The minimum and maximum rating of the generating units are 12 MW and 400 MW respectively. The voltage limits for the buses are assumed to be 1.05 and 0.95 p.u. The total system peak load is 2850 MW and the total generation is 3405 MW. ‘The system can be divided into two subsystems on the basis of voltage levels. The north subsystem is at 230 kV while the south subsystem is 138 KV. The north region has a surplus in power while the south subsystem is a power deficient area. The north subsystem load is 1518 MW while that of the south subsystem is 1332 MW. The installed capacity in the north region is 2721 MW and that of the south is 684 MW. The number of generating units in the north and south subsystems are 21 and 11 respectively. ae eos Pas Bais 3 35 Bor Figure 2.9: Single Line Diagram of the IEEE-RTS 36 37 2.7.2.1. The Hourly Load Model For The Buses In The IEEE-RTS ‘The sector peak load allocation for the buses of this test system is described in Table 2.5. Table 2.5: Sector Peak Load Allocation in MW for the IEEE-RTS Bus | Large | Indus. |Com. | Agri. |Res. | Govt. | Office | Total 0.0} 39.90] 14.25} 0.0 36.85] 17.0] 0.0] 108.0 0.0} 0.0) 14.25] 0.0) 48.45) 34.30] 0.0} 97.0) 0.0} 59.80} 14.25] 11.45) 94.50} 0.0| 0.0| 180.0 0.0} 0.0} 14.25] 0.0] 25.55} 3420| 0.0 74. 0.0) 19.90] 14.25] 6.0] 3685) 00; 0.0} 710 0.0} 39.95| 14.25] 11.45] 67.50] 0.0| 2.85] 136.0 0.0} 39.95) 14.25] 22.70) 48.10) 0.0/ 0.0) 125.0] 0.0} 19.90) 28.55! 6.0] 94.05) 25.65] 285] 171.0 85.50| 0.0| 8.50| 33.80) 4150) 0.0/ 5.70| 175.0 10 | 42.75} 39.95) 14.25] 17.90) 80.15) 0.0| 0.0| 195.0 13 | 42.75] 59.80/ 28.55| 16.70] 80.15] 25.65) 11.40] 265.0| 14 | 85.47/ 3995) 5.60] 0.0 62.98) 0.0) 0.0) 194.0] 15 | 213.75] 0.0| 34.50} 0.0] 5450] 0.0] 14.25 317.0 | 16 | 42.75] 0.0) 14.25] 0.0] 2590] 17.10] 0.0] 100.0 18 | 188.20) 39.90] 22.55) 0.0] 62.40) 0.0} 19.95] 333.0 19 |110.97/ 0.0} 14.25] 0.0) 55.78) 0.0] 0.0] 181.0 20 | 42.86! 0.0| 14.25) _0.0| 53.79 17.10| 0.0) 128.0 Total! 855.0} 399.0| 285.0| _114.0| 969.0 171.0] _ 57.0) 2850.0 CoIaneRowe Equation 2.14 is, once again used to calculate the hourly load at the sectors and at all buses for the IEEE-RTS. For instance, Load at bus 1 during hour 1 of the year = (0.531072 x 36.85) + (0.337 x 39.90) + (0.01 x 14.25) + (0.4 x 17.00) = 39.958 MW. (2.20) 38 ‘The hourly load curves for all the buses can be obtained in a manner similar to that described above and in the previous section. Figure 2.10 shows the load curves for buses 2 and 18 of the IEEE-RTS for a typical day of the year. Bus 18 has a pronounced flat region compared to bus 2 as it has more large user sector customers. Bus 2, on the other hand, is predominantly a residential service area with some government & institution and commercial establishments. The hourly load curves developed for all the load buses of the IEEE-RTS are used as the load model for subsequent adequacy analysis in the rest of this thesis. a ee _____—-_— o ° 6 12 18 m4 Hour Bus 2 Bus 18 Figure 2.10: A Daily Load Curve for Bus 2 and Bus 18 of the IEEE-RTS 39 2.8. Summary The state sampling and state transition sampling Monte Carlo simulation methods are briefly introduced in this chapter. The basic concepts of sequential Monte Carlo simulation method is discussed and the application of this technique to generation system adequacy assessment has been illustrated with reference to two test systems, the Roy Billinton Test System (RBTS) and the IEEE-Reliability Test System (IEEE-RTS). The necessity for developing time varying load curves at load buses is also discussed in this chapter. Seven types of customer sectors namely, agricultural, industrial, commercial, large users, residential, government and institutions, office and buildings have been identified and the load characteristics of these customer sectors are presented. The procedure used for developing the hourly load curves with reference to the two test systems is illustrated. The chronological hourly load curves developed for the RBTS and IBEE-RTS are used as the load models for all the adequacy studies performed in this thesis. 3. COMPOSITE GENERATION AND TRANSMISSION SYSTEM ADEQUACY EVALUATION USING SEQUENTIAL MONTE CARLO SIMULATION 3.1, Introduction Composite generating and transmission system adequacy evaluation is concerned with the total problem of assessing the ability of the generation and transmission system to supply adequate and suitable electrical energy to the major system load points. This form of analysis can also be designated as. hierarchical level II (HLII) evaluation. Evaluation at HLI provides an assessment of the ability of the generating facilities to generate sufficient energy to satisfy the total system requirement. HLII evaluation extends the analysis by including the major bulk transmission facilities and provides an assessment of the composite system's ability to serve the major loads. HLIII evaluation further extends the analysis by including the distribution facilities and provides an assessment of actual customer load point adequacy. This thesis is restricted to adequacy evaluation in composite generation and transmission systems. ‘The transmission configuration which links the generating units to the major load buses is usually relatively complicated and it is seldom possible to model the configuration using simple series/parallel reduction techniques. Quantitative assessment of the adequacy of a composite system can be 41 performed using the Monte Carlo simulation approach. There is a wide range of indices which can be caleulated at each major load point and for the overall system. There is, however, no consensus in the industry as to which adequacy indices are the best. This may simply reflect the actual complexity of the problem which has limited the progress of adequacy assessment, bi it may also indicate the variety of purposes for which adequacy indices may be used. It may therefore be appropriate depending upon the failure criteria, to study a variety of adequacy indices which convey meaningful information regarding the performance of the system and are also well suited to making system design/alteration decisions. Individual load point adequacy indices are necessary to identify weak points in the system and to help establish optimum response to system investment. These indices when obtained using a valid load model can provide bulk system supply values for use in overall customer adequacy evaluation and provide a quantitative indication of the relative contribution of the three basic segments of generation, transmission and distribution. Overall system adequacy indices provide an appreciation of global adequacy which may prove more appealing to management and may be useful in the comparison of one system with another. They may not, however, be as sensitive to the addition of a line or generating element as the individual bus indices in the immediate proximity of the added element. 3.2. Adequacy Indices There are three fundamental parameters in the calculation of system adequacy indices. These parameters are: 42 © Frequency of events. * Duration of events. © Severity of events. ‘The severity of an event depends on the components under outage, their relative importance and their location in the network configuration. An outage event may affect a wide area of the system or it may affect only a small group of buses or perhaps a single bus. The adequacy indices should focus attention on those parts of the system that are directly affected by the outage event. The total contribution of all possible outage contingencies should indicate those areas in the system which have low reliability and are prone to disturbances. Overall system indices cannot provide this information. It is therefore not desirable to draw conclusions regarding the adequacy of a particular system load point from overall system indices or bus average indices. ‘The need for determining individual load point indices is also necessitated by the fact that the effect of high level outages is not uniformly distributed over the entire system. Moreover, varying the load at each bus of a power system in equal proportion may not result in a proportionate variation of the indices at each bus. This is due to the fact that load flow studies involve the solution of non-linear simultaneous equations. The effect of load variation may therefore be not uniform at each bus, depending upon the network configuration and the system component parameters. The two fundamental approaches to composite system reliability evaluation are analytical enumeration and Monte Carlo simulation. Monte Carlo methods are more flexible when complex operating conditions and system considerations such as chronology, reservoir operating rules, bus load 43 uncertainty need to be incorporated. Several models (9, 10, 20, 26] have been implemented to perform composite system analysis using Monte Carlo methods. A comprehensive list of the indices [1, 27] considered is as follows. The evaluation of these adequacy indices using sequential Monte Carlo simulation is described in the following sections. 3.2.1. Load Point Indices Loss of load expectation at Bus k : NST Nw DY] Laie LOLEk = 1 (hour/yr) an NS where: Nic = Number of interruptions occurring at Bus & , in year i, djx = Duration of an interruption, in hours for year iat Bus k, NS = Number of simulation years. Probability of load curtailed at Bus k : LOLE ¢ (hour / yr) (3.2) PLC = : 8736 Expected frequency of interruption/load curtailment at Bus k : ‘NS 2. EFLC = a (occurrences/yr) (3.3) Expected energy not served at Bus k : NS [ Nix ‘E(BUSENS;x) 1 iS = EENS & 7 (MWhiyr) (3.4) where: BUSENSjx = Energy not served in MWh for interruption j, in year i at Bus k 3.2.2, System Indices System loss of load expectation : NST ny | fo LOLE = a (hour/yr) 5) where: ni = Number of system interruptions in year i, Di = Duration of system interruption j, in hours, for year i System expected frequency of load curtailments : NS ni EFLC pee (oceurrences/yr) (3.6) 45 System expected energy not served : NS| ni =| S(sysensi) EENS = 1 (Mwhiyr) aD NS where: SYSENSji = System energy not served in MWh for interruption j, in year i System probability of load curtailed : LOLE (hour / yr) Pl a eo 7 LC 7 (3.8) ‘These indices when calculated for a single load level over a period of one year are referred to as "annualized indices". In practical systems, the load does not remain constant throughout the period and the effect of a variable load level can be included to produce more representative "annual indices”. ‘The basic values for the annual indices will be different from the annualized indices obtained using the peak load levels. It is important to appreciate that the two sets of load point and system indices do not replace each other but actually complement each other. The load point indices are very useful in system design and in comparing alternative system configurations and system alternatives. The overall system indices indicate the adequacy of the composite system to meet its total load demand and energy requirements and therefore quite useful to the system planner and to the system manager. It must be recognized, however, that it may be difficult and sometimes misleading to draw conclusions regarding the adequacy of a particular system load point from the overall 46 system indices. The indices described above are obtained in this thesis using the sequential Monte Carlo simulation approach. 4.3. Composite System Adequacy Assessment Using Sequential Monte Carlo Simulation Analytical contingency enumeration [3] or Monte Carlo simulation techniques can be used to assess the adequacy of a composite system. The non sequential Monte Carlo method is based on combining the state sampling or state transition sampling techniques with a direct analytical approach for system analysis and the utilization of a minimization model for load cartailment. This method is suited to the analysis of large systems and can be used to include multi-state representation of generating units. Extensive research has been done at the University of Saskatchewan in the area of Monte Carlo simulation for composite system adequacy assessment. This has resulted in the development of a computer program designated as MECORE [4, 5] which uses a non sequential Monte Carlo simulation method to calculate the annualized indices and a hybrid method utilizing an enumeration approach for aggregated states to calculate annual indices. Both the state sampling and state transition sampling procedures have been incorporated in the MECORE program for composite system adequacy assessment. One basic aim of this research work was to develop a sequential Monte Carlo simulation technique for composite system adequacy assessment. A computer program has been developed to achieve this purpose. The various modules of this software are described in the following sections. a7 3.3.1. Minimization Model For Load Curtailment The objective of the minimization model is to minimize the total load curtailed at each bus while simultaneously satisfying the power balance, DC load flow relationships and the limits of line power flows and generation outputs [5]. The two major load curtailment philosophies used are: * Loads are curtailed at the buses which are as close to the elements on outage as possible. * Loads are classified according to their importance and divided into categories of firm load and curtailable load. ‘The minimization model with the load curtailment philosophies incorporated is described in these equations. min x lw £ (rptc») 9) tend p= such that NB a Ti(Xj) = LZ Au(X;)| PGe+ LY LCw- PDe ite (3.10) REL p= EPG + | Excp)|= = PD @ap ieNG eND | p=1 ieND PGrin $ PG < PGmax (3.12) 48 [r(x)| < Trax (8.13) OS LCp< ap PD: (i € NDip = ba) (aay where: ND =Set of load buses; NG = Set of generator buses; NB = Number of buses; L = Number of lines; LC = Load curtailment vector; PG = Generation vector; PD = Load vector; 1(x;) = Line flow vector for state X,; A(X,) = Relation matrix between line flows and power injections for state X;; PGrin ! PGrax = Minimum / Maximum limits for the generation vector; Tmac = Limit vector for the line flows 7(x,) The load curtailment variable at each load bus is divided into q sub variables and ap is the load percentage associated with each sub variable. In these studies, the values used are: g = 3 and a= 20%, a2=20%, a3= 60%. The least important load corresponds to the smallest Y and the most important load to the largest 7. W; are the weighting factors associated with each load bus. The buses closest to the elements on outage have relatively small W; and those farther from the outage elements have relatively larger ‘Wi. A dual upper bounding linear programming algorithm is used to solve the minimization model [23]. 49 3.3.2. Dual Upper Bound Linear Programming Algorithm Linear programming seeks to find the optimum value of a linear objective function while meeting a set of linear constraints, ie., we wish to find the optimum set of X's that minimize the following objective function. Z = CX] + COX > cesses CNEN subject to a set of linear constraints: + AINKN $ By AQ 1X] + AQQX9 + voce AQNKN S BQ Ay1X] +ApoXo +. In addition, the variables themselves may have specified upper and lower limits, i.e., xymin 0,%; IC; < 0, Xj=Xjmax 2. IfCj =0,X; is usually set to its minimum value. xjmin 8. A slack variable is added to each constraint. Using the above Xj values, the slack variables are set to make each constraint equal to its limit. Variable Exchange: Step 1. The basis variable with the greatest violation determines the row, R, to be pivoted on. If there are no limit violations among the basis variables, the process is terminated. Otherwise, the most violated variable leaves the basis and is set equal to the limit that was violated. Step 2. The variable that enters the basis is selected using one of the following column selection procedures. Column Selection Procedure P1: The Procedure P1 is used when the most violated variable is below its minimum. The aim is to pick column S, given constraint row R, whose basis variable is below its minimum and is the worst violation, so that Cs/CAR,9) is minimum for all S that meet the following rules: a) S is not in the current basis. b) AR, is not equal to zero. ©)IfXg is at its minimum, then Ap,g must be negative and Cg must be positive or zero. 4) IfXg is at its maximum, then Ap,g must be positive and Cg must be negative or zero. SL Column Selection Procedure P2: This Procedure P2 is used when the most violated variable is above its maximum. The aim is to pick column S, given constraint row R, whose basis variable is above its maximum and is the worst violation, so that Cg/(-Ap,g) is minimum for all S that meet the following rules: a) S is not in the current basis. b) ARG is not equal to zero. ©) IfXg is at its minimum, then Ap,s must be positive andCg must be positive or zero. 4) IfXg is at its maximum, then Ap,g must be negative and Cg must be negative or zero. Step 3. When a column has been selected, pivoting at the selected row R (from Step 1) and column S (from Step 2) is performed. The pivot column's variable, $ goes into the basis. If no column fits the column selection criteria, an infeasible solution results. Setting the Variables After Pivoting: 1. All nonbasis variables, except Xg remain as they were before pivoting. 2. The most violated variable is set to the limit that was violated. 8. Since all nonbasis variables are determined, each basis variable is set to whatever value is required to make the constraints balance. Return to Step 1 of the variable exchange procedure. These steps are shown in the flowchart in Figure 3.1. 92 START +] ‘Search among the basis variables for the variable with the worst violation. This determines the row selection, R. No violation among a basis variables Any violation Yes Most violated variable is Most violated varieble is low its minimum ation PS above its maximum Pick column using columa selection procedure P2 Pick column S using column selection procedure PI Infeasible solution Infeasible solution Pivot on selected row and column Flowchart for The Dual Upper Bounding Linear Programming Algorithm 53 3.3.3. Solution Algorithm For Composite System Adequacy Assessment Using Sequential Monte Carlo Simulation Composite system adequacy assessment using sequential Monte Carlo simulation can be summarized in the following steps. Step 1: The first step is to generate operating histories for all generating units and transmission lines by drawing sample values of TIF (Time-to- Failure) and TTR (Time-to-Repair) from their appropriate probability distributions as described in Section 2.3 for a year. The operating history of all components are in the form of chronological up-down-up or up-derated- down-up operating cycles. Step 2: The generation is allocated for every simulated hour according to demand and available generating units. The load at each bus for the simulated hour is known, from the time varying load model. Step 8: A linearized load flow similar to that used by all other known , SICRET and CONFTRA [9, programs based on Monte Carlo simulation, e. 10] is used to represent the transmission system. This DC load flow has to be performed for every simulated hour of operation. The DC load flow representation only takes into account active power flows, and voltage problems are not considered. The following relationship is obtained using the linearized load flow formulation: T(S) = A(S))x(PG- PD) (3.15) where: PG = generation output vector, PD = load power vector, ‘Sj = system state vector for j th hour, T(S)) = line flow vectors under state Sj, and (Sj) = relation matrix between line flows and power injections under state Sj. The m th row of A(S;) can be calculated as follows: ee acid) (3.16) where: p,q = bus numbers of the m th line, Xm = reactance of the m th line, Z(S)) = bus impedance matrix of the system under state Sj in which the resistance of all lines are neglected, and Zp(Si), Za(Sj) = p th and q th rows of Z(S)). ‘The bus impedance matrix Z need not be completely calculated every time the network configuration changes due to line outages. The bus impedance matrices after the outage of specified lines, 2(Si) can be directly obtained from the bus impedance matrix of the normal state without any line outages, ie, 2(S0), Z(Si) = Z(So)+Z(So)MOMTZ(So) (j # 0) @.n whe: @ = [x- mTZ(soym] : (3.18) 55 X is a diagonal matrix whose dimension is the same as the number of outaged lines and whose diagonal elements are the reactance's of outage circuits. M is a submatrix composed of the columns corresponding to the outage lines of the bus-line incidence matrix. In the case of a single line outage @ becomes a scalar quantity and is found equal to Q = [Xa —Zpr(So) - ZiqlS0) + 2Zpa(So)]t (3.19) where: Xpq_= reactance of the outaged line which is connected between Buses p and q, and Zep(S0), Zpq(S0), Zpa(S0) = elements of matrix Z(S0). ‘The above simplifications enables a faster DC load flow calculation. Step 4: If the system state is one without any generating unit and line outages, and if there is no line overloading, the process proceeds to the adequacy analysis of the next hour. If a generating unit is in the outage state, then it is assumed that other spare generation units at that bus are immediately available if they are in operation. However, if it is a contingency state, a minimization model of load curtailment described in Section 3.3.1 is used to reschedule generation outputs in order to maintain generation- demand balance and alleviate line overloads and, at the same time, to avoid load curtailment if possible or to minimize total load curtailment if unavoidable. It is assumed in the minimization model analysis that, if a 56 generating unit is in a contingency state (partially derated state or full forced out state), the spare generating units available at that bus are immediately put on line. Step 5: The adequacy indices are accumulated depending on whether the system state for the simulated hour leads to any load curtailment. Step 6: Steps 2 - 5 are performed for the yearly sequence of system states. ‘The yearly adequacy indices are accumulated as F(Xj) where X; is the sequence of system states in year j and F(X;) is the reliability index function over the year j. In order to evaluate the system probability of load curtailed for instance, F(X;) is the sum of the durations of all failure states, i.e., load curtailment states, divided by 8736 in the year j. Step 7: If the coefficient of variation of the chosen index is greater than the tolerance level, steps 1 - 6 are repeated until convergence is achieved. The coefficient of variation B is calculated as given in Section 2.3.1. AWE TNS EP) (8.20) where V (F) = Variance of the test function, E(F) = Expected estimate of the test function and NS =Number of simulated years. The simulation procedure is stopped when the estimated coefficient of variation of the expected energy not served is less than 5%. 37 Step 8: After convergence is achieved, the adequacy indices are estimated as described in Section 3.2 and summarized in Equation 3.21. NS & F(X) Expected value = ae (3.20 where F(X}) is the test or indicator function for the index to be estimated The methodology described above has been applied to the adequacy analysis of two test systems, the RBTS and the IEEE-RTS. 3.4. System Studies Utilizing An Annualized Load ‘The sequential Monte Carlo simulation method was used with an annualized peak load to obtain the adequacy indices for the two test systems. 3.4.1. Application To The RBTS ‘The annualized system peak load for the RBTS is 185 MW. Table 3.1 presents some of the basic load point and system indices for the test system using the sequential approach. The system required 150 years of simulation to obtain a 5% coefficient of variation of the expected energy not served. The CPU time utilized on a SUN station was 733 sec. 38 ‘Table 3.1: Annualized Indices for The RBTS Bus | Probability | Frequency EENS No: Vyr) Mwhiyr) 2 0.0011 0.45 39.84 3 0.0039 1.72 482.26 4 0.0066 2.30 357.57 5 0.0002 0.23 3.85 6 0.0011 0.88 197.49 System | 0.0099 | _4.13 1081.01 Figure 3.2 shows the convergence of the system EENS for 150 simulation years. 1400 1200 8 800 600 400 200 EENS (MWhi/yr) 0 50 100 150 Simulation years Figure 3.2: Variation of the Annualized System EENS with Simulation ‘Years for the RBTS 3.4.2. Application To The IEEE-RTS. 59 ‘The annualized peak system load for the IEEE-RTS is 2850 MW. The load point and system indices for this system are given in Table 3.2. ‘Table 3.2: Annualized Indices for the IEEE-RTS Bus Probability ‘Frequency EENS No: ye) (MWhi/yr) i 0.0022, 1.06 787.06 2 0.0083 3.38 2972.30 3 0.0042 1.87 2323.15 4 0.0063 2.50 2009.76 5 0.0058 2.56 1610.04 6 0.0079 3.00 4561.50 7 0.0069 2.25 2751.86 8 0.0091 2.88 5417.76 9 0.0001 0.18 131.36 10 0.0002 0.13 163.51 13 0.0258 6.50 21938.29 14 0.000007, 0.06 1.35 15 0.0102 3.25 7992.46 16 0.0163 5.50 5173.23 18 0.0507 9.75, 62611.67 19 0.0152 5.25 7188.68 20 0.0083 3.31 3044.91 ‘The sequential simulation method used 15 simulation years to obtain a convergence of less than 5% of the expected energy not served. Figure 3.3 60 shows that the expected frequency of load curtailment (EFLC) settles to 18 occurrences/year within 15 simulation years. 30 25 22 Be o Is z 10 3 ° 0 5 10 15 Simulation years Figure 3.3:Variation of the Annualized System EFLC with Simulation ‘Years for the IEEE-RTS 3.5. System Studies Using Exponential State Residence Times With ‘Time Varying Loads The chronological hourly load models developed for the two test systems are used together with a sequential simulation approach to obtain the adequacy indices of the RBTS and IEEE-RTS respectively. The adequacy indices estimated are the probability of load curtailed (PLC), expected frequency of load curtailments (EFLC) and expected energy not supplied (EENS) at the load points and system . Exponential state residence times [2] have been used to represent the failure and repair process of all the components in the two test systems. 61 3.5.1. Application To The RBTS The adequacy indices obtained for the RBTS using the sequential simulation method are presented in the following sections. 3.5.1.1. Adequacy Indices Tables 3.3 and 3.4 present the estimates of the load bus and system indices after simulating the operation of the system for 1000 simulation years at which point the coefficient of variation of the expected energy not served (EDNS) reaches 5%. The simulation procedure utilized 5157 sec of CPU sime in order to obtain the indices when utilizing a time varying load, while the annualized indices were obtained for a 150 year simulation period in 733 sec. A detailed representation of the load model needs extensive computational requirements in terms of CPU time and memory. Table 3.3: Annual Load Bus Indices for the RBTS Bus No: | PLC EFLC (/Yr) EENS (MWh / Yr) 2 0.00005, 0.09 1.09 3 0.00028 0.64 19.05 4 0.00029 0.49 14.96 5 0.00001 0.02 0.27 6 0.00118 0.93 111.73 62 ‘Table 3.4: Annual System Indices for the RBTS Probability of load curtailed 0.001599 Expected energy not supplied (MWh/yr) 147.113 Expected frequency of load curtailments (/yr) 1.70 Average duration of load curtailment (hrs /dist.) LITT Simulation years 1000 CPU time (sec) 5157 Figures 3.4 and 3.5 show the system EENS and EFLC versus the number of simulation years respectively. 250 8 150 EENS (MWhiyr.) 8 8 ° 0 250 500 750 1000 Simulation years Figure 3.4: Variation of the System EENS with Simulation Years for the RBTS 63 EFLC (/yr.) ° = cea a 0 250 500 750 1000 ‘Simulation years Figure 3.5: Variation of the System EFLC with Simulation Years for the RBTS 3.5.1.2. Distributions Of Adequacy Indices ‘The sequential simulation approach provides the ability to obtain probability distributions of the estimated indices. The distribution of the system energy not served is illustrated in Figure 3.6. Figures 3.7 and 3.8 present the distributions of the system loss of load index and system frequency of interruption index. 450 400 350 300 250 200 150 100 50 0 50 300 550 = 800-1100 Energy not served (MWh) period) Frequency (yrs. /simulation Figure 3.6: Distribution of the EENS index for the RBTS 64 In Figure 3.6, the first interval indicates an energy not served equal to or greater than 0 but less than 50 MWh. It can be observed from this figure that 418 years out of 1000 years, experience a load curtailment of less than 50 MWhiyr. period) 8 Frequency (yrs, /simulation 8 10 40 70 100 130 160 190 Loss of Load (hrs) Figure 3.7: Distribution of the LOLE index for the RBTS Frequency (yrs. /simulation Petiod) of 88888 2 4 6 8 0 2 Frequency of interruption (Occ.) Figure 3.8: Distribution of the EFLC index for the RBTS 65 It can be seen from Figure 3.8 that 553 years experience a system interruption of less than 2 occurrences. Figures 3.6, 3.7 and 3.8 indicate that the distributions of all the indices follow an exponential shape. 8.5.2. Application To The IEEE-RTS ‘The adequacy indices obtained using the time varying load model and the sequential simulation approach for the IEEE-RTS are presented and discussed in the following sections. 8.5.2.1. Adequacy Indices ‘The system and load bus indices obtained using the sequential simulation technique are described in Tables 3.5 and 3.6 respectively. ‘Table 3.5: Annual System Indices for the IEEE-RTS Probability of load curtailed 0.003643 Expected energy not supplied (MWh/yr) 4092.66 Expected frequency of load curtailments (yr) 8.1725 Average duration of load curtailment (hrs /dist) 3.8954 Simulation years 400 CPU time (sec) 14757. Table 3.6: Annual Load Bus Indices for the IEEE-RTS Bus PLC EFLC EENS No: vy) (MWhiyr) a 0.00011 0.25 29.69 2 0.00022 0.54 55.22 3 0.00024 0.63 97.28 4 0.00017 0.43 33.42 5 0.00031 0.81 61.05, 6 0.00028 0.71 94.69 7 0.00017 0.49 55.84 8 0.00032 0.78 138.35 9 0.000004 0.01 1.48 10 0.000003, 0.02. 0.91 13 0.00085 1.95 564.80 14 0.0 0.00 0.00 15 0.00029 0.71 192.32 16 0.00052 1.32 147.05 18 0.00273 5.68 2349.85 19 0.00006 1.07 228.55 20 0.00014 0.57 42.17 66 Figures 3.9 and 3.10 show the convergence of the expected frequency of load curtailments (EFLC) and expected energy not served (EENS) within 400 simulation years. A 5% coefficient of variation of the EENS was used as the convergence criterion. 67 12 10 EFLC (/yr) on kOe 0 100 200 300 400 Simulation years Figure 3.9: Variation of the System EFLC with Simulation Years for the IEEE-RTS EENS (MWh/yr) 7828883 ° 100 200 300 400 Simulation years Figure 3.10: Variation of the System EENS with Simulation Years for the IEEE-RTS 3.5.2.2. Distributions Of Adequacy Indices ‘The distributions of the estimated indices provide interesting insight into the behavior of the system. Figures 3.11, 3.12 and 3.13 indicate the distributions of the LOLE, EENS and EFLC indices. 8 8 8 Frequency (yrs. simulation period) 68 8 ° 10 50 90 130 170 Loss of load (h) Figure 3.11:Distribution of the LOLE Index for the IEEE-RTS 200 = 160 £ Be 120 a? 2k 80 ¢ 40 & o 2000 10000 18000-26000» 34000 Energy not served (MWh) Figure 3.12: Distribution of the EENS Index for the IEEE-RTS Figure 3.12 shows that 188 years out of 400 simulation years experience a load curtailment of less than 2000 MWh. Only 3 years experience a load curtailment of more than 26000 MWh. The average or expected energy not served for this system is 4092.7 MWhisr. 0 period) 3 Frequency (yrs. /simulation 10 20 30 40 50 60 Frequency of interruption (oce.) Figure 3.13: Distribution of the EFLC Index for the IEEE-RTS It can be observed from Figure 3.13 that 261 years or more than 50% of the entire simulation period of 400 years result in system load curtailment of 10 or less occurrences per year. Distribution On The Adequacy An integral part of the sequential simulation approach is to sample random variates from probability distributions with the purpose of evaluating expected values and probability distributions of the composite system adequacy indices. This section shows the effect that repair probability distributions have on these indices. The failure processes are assumed to follow an exponential distribution. 10 3.6.1. Application To The IEEE-RTS The main objective of these studies was to perform sensitivity analyses with different distributions for the repair process in order to identify the effect of the distributions on the system adequacy indices of the IEEE-RTS. ‘The mean values of the MTTF and MTTR of each component are kept the same in order to maintain consistency. It is also assumed that all components have the same underlying distribution for each case study. It should be noted, however, that no added complexity is imposed on the simulation if different components have different distributions. A brief outline of the sampling process for each of the three types of distribution is described in the following sections. 3.6.1.1. Weibull Distribution ‘The pdf of the Weibull distribution [15] is: 8 frit) = s Ode +4) (3.22) where ais the scale factor and f is the shape parameter. Using the inverse transform method, the random variate Tis given by: T=a( nus (8.23) Ifo and B are known, then times to failure or repair can easily be found using Equation 3.23. If the shape and scale parameters are not known, 7 Equation 3.23 cannot be used directly. These parameters are related to the mean and variance of the distribution which, as in the case of log normal or normal distributions, are either known or can be assumed. The mean value and variance for the Weibull distribution are given by: neo (i) (3.24) o@er{i+2) ra (8.25) where I’ is the gamma function. Since solution of Equations 3.24 and 3.25 for a and B in terms of p and o are not straight forward, a reasonable value for the shape parameter B is assumed and the corresponding value of a. is calculated from Equation 3.24. Equation 3.23 is then used to calculate the random variates representing times to failure or repair. 3.6.1.2. Normal Distribution ‘The pdf of a normal variate T with mean 1 and variance 62, denoted as N(y,02), is given by: =u) frit) = ee we

Anda mungkin juga menyukai