‘West dian Journal of Engincring Vol. 25, No.2, Uanuary 2005) Techical Paper (Shama & Ramnansingh) 342 A Software-Based Tool for the Reliability Evaluation of a Utility’s Generation Capability C. Sharma* & V. Ramnansingh** ‘Analytical reliability evaluation analyses can be Broadly categorised nto two mal ‘areas, namely static or long-termand spline orshortterm evaluation The paper resents a PC-based Windows application software for evaluating the static ‘elibily indices of practical power systems using the well-established generation inadequacy probabilistic techniques. The name given to this software which was developed using MATLAB 5.3 is Reliability Bralaation Tool (REST). The effects of ‘the addition of energy limited systems to existing conventional power systems are ‘alsa incorporated into the package. The wo approaches for static evaluation; the loss of load and the frequeneidaraion have been applied to three power system configuration models The algorithms forthe building of the generation mode, ice, the eapacity outage probability table (COPT), equivalent assisting and te constrained models are also implemented as well ax the algorithms for the convolution of toad and capacity models. REST was tested using ty0 reliability fest systems, namety the IEEE Reliability Test System (RTS) and the Roy Bilinton Test Systems (RBTS). Selected resuls of the software using both test eases are resented inthis paper for Ulustraion purposes, = Introduction practical applications which fied to encapsulate the Probabilistic nature of a system since they were al 1.1 General Background ‘eterministically base. Power system reliability evaluation is an important consideration for planners, designers and operators. 1.2 Salient Features of REST ‘etiabiltyevaluaon sortware programmes have been developed and uilised in ll aspecs of power system planning, designing and operation. Tiss evident since ‘over the years, most generation planning ad expansion decisions made by utiles have been based on the Indices provide by the use of such programmes. Most of the software tools developed have been customised to the parcular power system and they have been implemented based on te probabilistic ‘approach to the evaluation of reliability indices a is evidenced in the published literature. This type of ccvaluation replaced the techniques First used in Psy o Egewg WL Se Ap, The generating capacity roqured to satisfy a particular Toad demand is a salient criterion which must be determined atthe planing and operating level Two requirements which must be satisfied for such a (2) Static Requirements (©) Spinning Requirements ‘The static requirement ean be considered as the installed capacity that must be planned and determined isa tun te Dune sted & Compu Ene 3‘West Indian Jura of Engineering Val. 25, No 2, anuary 2003) Te in advance ofthe system requirements and its reserve must include provision for overhaul (maintenance) purposes, unplanned outages, forced outages ad Hoa frowth requirements in excess of the estimates. ‘The spinning requirement consists of the reserve capacity that is spinning, synchronised and ready 0 {ake up load and analysis ofthis operating reserve encapsulates short-term evaluation ofthesyst [1] ‘Over the years, probabilistic concepts and techniques have been developed and applied to resolve these stati and spinning capacity problems. Forsmall powersystems, the implementation of these echnigues fan be easily achieved by hand caleulations, however, for larger practical systems, the computations involved ‘become complicated and time-demanding. This ‘problem can only be resolved by computerising the “ntire process or Computational methods. ‘With the increased trend towards the use of reliability evaluation packages by uilities, the need for programmes which are user-friendly, portable and cater for aiflerent power system models has become {ncreasingly important. These packages can also be ued as teaching tools for persons interested in the rellablity evaluation aspect of power systems. “This paper deseribes the development and functionality ofa genera-based, user-friendly software tool developed forstatc capacity reliability evaluation ‘The name given to this application is Reliability Evaluation Simulation Tool (REST). The specifications ‘ofthis package ae that it must include an interactive (Graphical User Interface (GUD that outputs te Loss “of Load Expectation (LOLE) and Expected Energy Not Supplied (HENS) indices. ’As part of acceptance ofthe final package, another salient objectives thatthe results of the entire Software must be verified with the Institute of Flectronie and Electrical Engineers Retiaility Test ‘System (IFEE RTS), The programme package must also be applied to produce te reliability indices of practical power system which has undergone varying ‘System models aver the lat eight years ~The Trinidad land Tobago Electricity Commission (TAXTEC), “Two well-established probabilistic techniques sein the sate reliablt evaluation of power systems fate implemented by REST, namely {#) Loss of Load, and () Frequency and Duration u nical Paper (Sharma & Ramnansingh) 33-42 “The three system models that are implemented bY [REST using the two aforementioned techniques at: Single area capacity model (ii) Two inerconnected systems model (ii) Thee interconnected systems mode! ‘The above models were chosen since they represent and encompass the various models of the case sty ‘exarople- T&CTEC over the last eight yeas. “The software chosen to implement the various system models is MATLAB 53 and this aplication ‘was prefered primarily because of the following (a) Its the current state-of the-art software adopted by industries worldwide. () teas powerful mathematical capabilities, (6) Iteliminates double-precision errors, and (@) Iefacttates the designing of GUIs from the individual m-iles developed foreach specific funetion. 2. Theoretical Considerations Reliability may be defined a “the probability of = ‘vce performing its purpose adequately forthe period Of time intended under the operating conditions encountered.” [2]. om the above statement, it s evident that probability concept and technique play an important folcin reliability evaluation analyses. A large number ‘of papers applying probability techniques to generating ‘opacity reliability evaluation have been published over the years, The two most widely adopted methods for bat reliability evaluation utilising these probability concep are (a) Loss of Load Approach, and (b) Frequency and Duration Approach, “The basic methodology used 1 evaluate the adequacy of particular generation configuration is depicted in the following representation shown on next page [1‘est Indian Joural of Engineering Vl. 25, No.2, Uanuary 2003) Technical Paper (Sharm & Rarnansiagh) 33-42 FIGURE sed fr apc ‘The system's risk model in the above Figure is determined by convolving the generation and load models. This technique can be considered as the fundamental principle used in the analyses of both aforementioned approaches. 21 Single Area Capacity Model - Loss of Load Approach Inthe single area capacity system, the generation mode! required for analysis using the loss of load approach is known as the Capacity Outage Probability Table (COPT), The COPT may he defined ¢s an array consisting ofthe posible capucity outage levels ofthe system and their respective probabilities (1). If all the unis in the system have identical sizes and Forced ‘Outage Rate! (FORS), this able can be easily obtained using the Binomial Expansion techaique. It is extremely unlikely however, that the COPT in a practical ease can be developed using the Binomial lechnique as most systems contain generator units of| varying sizes and FORS. In his cae, the units can be ‘combined using recursive algoritnn which the units are aided sequentially to produce the final model. The algorithm implemented by REST forthe case where a ‘multi-state wnit is added tothe capacity mode! i [1] Pay= YnPx-G) Ai) where: PC) ~ individval probabitity of the ‘capacity outage state of XMW after the unit is added 35 P*(X) ~ individual probability ofthe capacity outage state of XMW before unit is added = umber of states ofthe unit + capacity outage of states foe the unit being added probability ofexistence ofthe nie state ‘Three load models ate analysed by RESTin the single area capacity system namely a) Load Model 1 - Daily peak loads for a sven tie period and the munber af ‘occurrences where these daily peaks wore exceeded In this model, load data i nthe form ‘ofthe individual daily peak leeds and their corresponding number of ‘ceurrences over a given time period, b) Load Model 2 - Daity Peak Load Variation Curve (DPLVC) ‘The Daily Peak Load Variation Curve represents 2 cumulative load model here each day is presented by its individual daly peak Toad, Le, readings are taken every day forthe ‘peak load on the system, From this, the DPLYC is formed by aranging all the values in descending order. ‘Load Model 3- Load Duration Curve (LDC ‘The Lond Duration Curve is similar to the DPLVC except that each hours represented by is individual hourly peak load, The readings are then ranged in descending order Io form the final load mode o‘West nian Journal of Engineering Vl. 25, No.2, anunry 2008) Technical Paper Sharmin & Ramnansingh) 33-42 2.2 Loss of Load Expectation (LOLE) ‘This generation model (capacity outage probability {able} can then be convolved with the appropriate load mode to produce an expected risk of loss of loud that istermed the Loss of Load Expectation Index (LOLE) index. For each load level L, the contribution tothe LOLE index is found by summing all the individual probabilities where the eapacity available, Cis less than the particular load level. The entre proces i repeated fr all lad levels inthe time period andthe final LOLE index is deduced by summing all these individual contributions [3] “The equation vsing load model | is shown below’ LoL $ rect 1) dap a where C= Cepacity available [= Load level “The LOLE analysis without maintenance using load models 2 and 3 (DPLVC and LDC respectively) is ‘sven by Equation (3) LoLe= SAC) @) 23° Sensitivity Study Analysis {A sensitivity study analysis demonstrates the variation {reliably index (LOLE) with the system parameter ‘The analyses performed by REST allows for the variation of two systom parameters, namely, the peak Toad and the generators’ FORS, These can be summarised a follows 8) LOLE asa function of the system peak load “The LOLE index is evaluated by REST {fora range of system peak loads using load model 2 deseribed in section 21 ‘The results from this analysis ean then be displayed in a graphical format were log, (LOLE) vs. System peak Toad is plated b) LOLE asa function of system peak load and FOR ‘The system's risk index fora given pacity and forecast peak load is ‘dependent onthe forced tage ate values for the individ wits (1 Tn his analysis, the LOLE values for «range of peak load levels are evaluated ‘function of unit forced outage rates. ‘The results from this analysis are also displayed ina graphical format where Jog, (LOLE) vs. System peak loa is plotted foreach FOR value use. 24 Loss of Energy Analysis Im addition othe LOLE analysis, load mode 3 (load oration curve) when convolved with the COPT can also be utilised to calculate the expected energy not supplied by the system due to insufficient installed Capacity. This principle is explained by considering ‘te following load duration curve in Figure 2 below. att Cty {Rees [eomome. [FIGURE 2 ery Curent ane wa nen Capac Oge Condi “The total area under the curve represents the energy demanded over he piven time period and consequently, ‘portion of ths energy required F,, would not Be supplied forall oad levels which are greater than a particular capacity outage 0, (3). “This analysis can be done with the addition of exch generator to the system and the results of such evaluation can be used to determine the following ‘energy related indices vi‘West nian Journal of Enginsring Vol 25, No, 2, aman 2008) Technical Paper (Sharma Ramiansingh) 33 42 (1) Expected Energy Not Supplied (ZENS) (@) Loss of Eneray Expectation (LOEE) (3) Berg Index of Reliability (EIR) ‘An advantage given by this energy-based evaluation method is that produetion costs simulations ean be performed since the output energy generated by each tnt can be determined from the aforementioned indices. 24.1 Expected Energy Not Supplied (GENS) ‘The basi principle described in section 2.4.can be used to determine the expected energy not supplied by the system, with the addition ofeach generator unit to the capacity move In this analysis, foreach state of the capacity model, C, the enerey not supplied, E,. is ‘obtained by summing all load levels which ee greater than this particular capacity state ofthe system, The ‘expecation is then obtained from the product of the ‘energy not supplied F, as aresultof each capacity state and the associated probability ofthe system residing in that particular state (3) ‘The expected energy not supplied, PENS is ‘then evaluated by numerically summing the individual mathematical expectations for esch sate of te capacity mode In this analysis, iis important thatthe merit ‘order ofthe units be maintained since the EENS is ‘dependent on the particular loading order chosen for {he units. Tis relates tothe fact that the states of the generation model (COPT) vary with the adaition of| ach unit of a paniculr capacity. The expected energy not supplied can be subsequently evaluated by REST using the following equation: ens-Sa%, a 2.42 Loss of Energy Expectation (LOE) ‘When all the units have been added to the system's capacity model the inal value ofthe FENS represents the system loss ofenergy expectation (LOBE) [3]. This index depicts the probable energy not supplied by the system even afterall generator units have been loaded to their respective capacities. 7 24.3. Energy Index of Reliability (EIR) ‘The EIR represents the probable ratio between the total energy that will be supplied by the system afterall units have been added and the total energy demanded by the system to supply all dhe load levels in the specified time period, The following formula isis (wore) t=) - LOL) _ Toral Energy Demanded © 25 Production Cost Evaluation The expected energy not supplied concept discussed in section 2.41 provides an approach t0 production cost modelling. The total expected production cost of the system can be determined by summing the expected ‘energy output ofeach unit inthe sytem muti by their respective energy costs per MWh, [4]. The algorithm implemented by REST for this analysi shouin in equation 6: “Taal Potton Cost & ceneray oupuceyof wit 4) x (CosuMWh of unit &) oy 26 Energy Limited Systems ‘The BENS index discussed in section 241 can also beevaluated withthe inclusion of an energy limited or ‘4 non-conventional system to the existing capacity ‘model. An energy limited unit may be considered as & tunic whose outpot at any point i time i dictated by the energy available or derated due to a technical problem. The analyses carried out by REST using energy limited systems are sensitivity studies where the system’ FENS is evaluated for varying penetration levels and for varying FORS ofthe vnit, The results of such analyses are displayed in a graphical format bythe software tol Single Areas Capacity Model - Frequency and Duration Approach ‘The mestiod requires additional system information that used in the los of Toad approach technique. ‘The latter metho utilises the steady state availability, AA, and the unavailability, U, parameters for its 27“West Indian Jour of Enginering Vl, 25, No.2, nuary 2008) Technical Paper (Shama Ramnansingh) 33-42 generation model, The F&D technique utilises in fadition to the A&U parameters, the transmission rate parameters 2 and (11 "The generation model is similar in form to the capacity outage probability table required forthe loss ‘of load approach, This model, however, provides ‘additional information on the frequency of fecountering every capacity outage state and departure fates from each state to higher and lower capacity ‘outage states, In the ease where a multi-state unit is talded wo the capacity model, the algorithms are shown jn equations (7) 0 2). rarEea-on sas) 2an-Zo W- Cm O-C)+9.409) 8) poo 2a) = Baap oC). or~C) +40 o) Oe ERA rete o Peo [REST implements one Toad model which is refered to asthe individual load state model In this model, the sequence of daily peaks is assumed to be a Sationary random process and is represented by a ‘Markov chain model 2.8 Two lnterconnected Systems ‘The adequacy of the generating capacity in a single area power system - System A, is normally improved ‘when this system is interconnected to another single area capacity system - System B, via @ te line For implementation purposes i this projec, this te Tine is assumed 10 be 100% reliable, Each interconnected system can then operat ata piven risk level with a lower reserve than would be required ‘without the interconnection. The technique used by REST for developing the model of the assisting system taking into account its peak load, isthe equivalent assisting unit which ean then be moved through the tietnes and added tothe existing capacity model of the asisted system. The computation of the LOLE index inthe assisted system proceeds asin the case of ‘normal single area system study. 2.9 Three interconnected Systems “The adequacy of the generating capacity ina single area power system ~ System A, is futher improved ‘when this system is interconnected to two single area capacity systems - Systems B and C, vi two te ics In this analysis, che equivalent assisting unit method ‘is applied wo develop the eguivalent assisting units for systems B and C. The maximum asitances provided by systems B and C are however limited to the tie capacities AB and AC respectively. As aresul, wote constrained equivalent assisting unit moels were ‘veloped taking these tie capacities into account ‘These tle constrained unit models are then ‘added tothe existing capacity model of System A and ‘new generation mode is produced for the modified system A. 3, Software Design and implementation ‘The software model adopted in the designing of REST ‘was the Waterfall Model with Prototyping (5). The ‘MATLAB design for this tool consisted of designing ‘sequence of GUIs which would allow the user (0 Select a particular system model for analysis and perform all the evaluations pertaining to that mode "The three options provided forthe user are: (a) Single area capacity model {(b) Two interconnected systems model (©) Three interconnected systems mode! In the case where the single area capacity model is selected, the user has the option to select between the two approaches: Loss of Load and Frequency and ‘Duration. For the two and three interconnected systems model, the loss of lond approach is suiomttically assumed ‘Figure 3 shows the programme's main sereen “This window allows the user to select the particular typeof system model to be evaluated and inthe ease ‘of the single area system, the approach to be utilised. Once a selection is made from the meno, the wer is taken through the analysis via software proms, 4. Software Testing ‘One of the major objectives of this tol was thatthe emt software must be verified against the results of the Institute of Electronic and Electrical Engineering‘West Indian Journal of Engineering Vo.25, No. anuary 2008) Tete Paper (Sharma & Ramnansingh) 33-42 pais Sy of nes WOME AAD ass anoay SWAJSAS JOMOd JEIICIG 4oj auiwesBol4 uonenjenz Auiqeyay & Jo juaudojenaq Guyoous6ug seyndwoy ¥ je219913 Jo juounsedag| SeIPUT 35994 OUR JO ASIEN SLL & Taas URW (57 w Pen vonerms woneree3 Ace 9‘West nan Journal of Eginerng Vol. 25, No, 2, January 2003) Tenicl Paper (Sharma & Ramnansingh) 33-42 Reliability Test System (IEEE RTS). This est system ccnsists of 32 generators and an appropriate bulk transmission network. The results of this model Drimarily served asa reference case, which allows the results from programmes developed by users to be benchmarked or yertied. For certain functions where there were no published IEEE RTS results, the Roy Billinton Test System (RBTS) was utilised for verification parposes. "The ra data roquired for simulation ofthe various functionality by REST must be obtained from the particular utility under study and consists of the Following: (a) The value in MW and the probability associated with each distinct tate of the _geneators in the system. This is used Tor the generation of the COPT. (b) Fifty-two (52) weekly peak load values ‘expressed asa percentage of the annual oak load, seven daily peak loads texpressesd as a percentage ofthe weekly peak load and 24-hourly peak loads tepressed asa percentage ofthe dally peak load. This data is used in the LOLEE find EENS analyses, (©) The weekly maintenance schedule for the particalar uty andthe corresponding unis on maintenance. This i used inthe LLOLE analysis which incorporates the elfet of a scheduled maintenance. (4) The range of peak loads or FOR values required forthe sensitivity study analyses () The merit loading order ofthe units on the system used forthe EENS analysis (This is based on an economic despatch programme), (9) The equations ofthe two or three straight-line approximations wo the lead duration curve als utilised forthe EENS analysis (g) The model ofthe energy limited uni its comesponding rating in MW and the range of FORS and penesration levels "This data is used forthe analysis of an mite unit when added vo an existing system, (0) The cost per MWh of each unit in merit Toading order. This data is used for ‘production cost analyses () The Mean Time To Failure (MTTF) and ‘the Mean Time To Repair (MTTR) values which are used inthe generation ofthe COPT utilising the Frequeney and oration method The peek loads and the ti capacity (es) ofthe systems used in the two or three {nerconnected systems models respectively 4.1 Results of the IEEE ATS and ABTS Test Cases Single Area Capacity Model - Loss of Load ‘Approach (See Table 1) ‘TABLE: Rens for Capa Outage Prob Table (COPD) ‘apacy Out naval Cumuaive india Cumulative ci Prooaity Probabity Probabity —_Probaity o 2aswst19E-001 1.00000000E+0000.29630406 10 100 2ope1SSSIE-002 SA76OIT44E-0O1 —O.020ENTSH 054760141 200 1 2868ss686-000 © 8192R1006-001 001866 | 0.g8ta2840 00 —a576919026-004 GzrT2—KOSE-ONe — .00057E9O.OHeNT297 250 400590026005 7 40105264E-002 .00006K81_0.00740107 1200 —-2arsea6ssE-one — 7912822596-008 | .00000413 90079128 1900 ‘.90000030 2.981800156-007 «4048507706 008 00004043, 0‘West ndian Journal of Engineering Vl. 25,'No, 2, January 2005) Techical Pape (Sharma & Raransinghy 33-42 TAMLE2: Selected Sas or Capacty Oatase Probably Tate for Poweraen Capacity Out nda ‘Cumulative nw), Probably Prot 0.0 .405616266-001 1.000000002.000 200 7. s0971845E-002 6594989746-001 7080 4.7o2sH008E-012 o.neseA7BIEONT 7090 3.479059506-012 9.905€08705-011 ttsn0 — sezase2s0c-028 seriessi2e-o2s M1510 47887158E-208 4 7e807 1586-028 42° Results of Practical Case Study: T&TEC ‘This case presets the saliont results of a testcase performed on the T&TEC system - results for Powergen Trinidad Generation System (21 Generator tis), (See Table 2). 5. Limitations of REST ‘The entire software package was developed and {implemented using MATLAB Version 5.3. This softvare was chosen primarily because ofits powectal ‘mathematical and GUI building capabilities. However, ‘ne limitation of using MATLAB isthe method by hich programmes are executed. Everytime MATLAB ‘uns a programme, tas to check the code for syntax eors, compile and then execute it. In programmes such as C and C+, stnd alone exceutable files are ‘created which does not require recompilation every Lime the programme i un. The effet of MATLAB's imitation is noticed when programmes containing a large numberof inputs, eg. IEEE RTS are run since now the execution time is significantly incressed, 6. Recommendations ‘The proposals offered to extend the functionality of the present programme so that it can eater to more options for system planners areas follows 8) Implement the Frequency and durstion technique to the analysis ofthe two and three interconnected systems models 1b) Implement the concept of tie Hine reliability inthe development of the tie constrained models forthe two and three interconnected systems modes, 4 ©). Implement the concepts of spinning ‘capacity reliability evaluation and apply techniques such asthe Pennsylvania New Jersey Maryland (PIM) approach to develop an analogous counterpart forthe short-term reliability evaluation of a power system. 7. Conclusion ‘This paper deseribed the development of a MATLAB- based software tool, REST, implementing the probability concepts of static capacity reliability evaluation. Three system models were developed, namely single area generation system, two imerconnected and thee interoonnected systems, The ‘developed software enabled the evaluation of two fundamental indices used in generating capacity planning, namely the Loss of Load Expectation (LOLE) and Expected Energy Not Supplied (ENS). ‘The software was implemented in different modules before being integrated into a single application, The developed sofiware was highly user Iiendly had many capabiliis to create and view the generation and load models, performed various sensitivity studies, production cost simulations, slisplayed graphical plots and examined the effects of ‘energy Timited systems. The results from the software also had a high degre of precision, The individual m-files coded for the programme had been successfully linked with the graphical user interface (GUD to provide an appealing Software package for power utilities. The entire programme is portale ince al files were easly stored ‘oma single compact dise. ‘The entire software was tested using two reliability test systems, namely IEEE Reliability Test System (RTS) and the Roy Billiton Test System (RSTS). The simulation results obtained from RES ‘were in good agreement with the published reslls for the two fest systems, ‘The package was also successfully applied to determine the reliability indices of a practical power system, namely the Trinidad and Tobago Electricity Commission (T&TEC),et Indian Jounal of Engineering Vol. 25, No 2, Canary 2003) Technical Paper (Sharma & Rarmnansngh) 33-42 References {1} Billiton, R, and Allan, RN. (1984). Reliability valuation of Power ystems. Pian Advanced Publishing Programme, Great Briain, pp. 6-128. {2} Billiton, R, (1980). Power System Reliability Evaluation. Gordon and Breach Science Publishers, New York, pp. 40-42. {3} Allan, RAN, Billiton, R. and Abdel-Gawad, NM. (1980). The IEEE Reliability Test System Exjensions to and Evaluation ofthe Generating ‘Sytem”. IEEE Transactions on Power Systems, Vol. PWRS-1, No.4 pp 1-7 [4} Billiton, R. and Cheung, LLCH. (1987) Load ‘Modification: A Unified Approach fr Generating Capacityrellabilty Evaluation and Production cost Modelling, IEE Proceedings, Vol. 134, Pt. C.No. 4 pp. 273-275. [5] Goel, L. and Gupsa,R. (1997). Windows-hased Simulation Tool for Electricity Generating ‘Capacity In, J, Engng. Ed, Vol. 13, No.5, pp. 347.357.