您好,欢迎访问三七文档
EvaluationofstochasticreservoiroperationoptimizationmodelsAlcigeimesB.Celeste*,MaxBillibInstituteofWaterResourcesManagement,HydrologyandAgriculturalHydraulicEngineering,LeibnizUniversityofHanover,Appelstr.9A,30167Hanover,GermanyarticleinfoArticlehistory:Received6April2009Receivedinrevisedform17June2009Accepted19June2009Availableonline26June2009Keywords:ReservoiroperationImplicitstochasticoptimizationParameterization–simulation–optimizationabstractThispaperinvestigatestheperformanceofsevenstochasticmodelsusedtodefineoptimalreservoiroperatingpolicies.Themodelsarebasedonimplicit(ISO)andexplicitstochasticoptimization(ESO)aswellasontheparameterization–simulation–optimization(PSO)approach.TheISOmodelsincludemulti-pleregression,two-dimensionalsurfacemodelingandaneuro-fuzzystrategy.TheESOmodelisthewell-knownandwidelyusedstochasticdynamicprogramming(SDP)technique.ThePSOmodelscompriseavariantofthestandardoperatingpolicy(SOP),reservoirzoning,andatwo-dimensionalhedgingrule.ThemodelsareappliedtotheoperationofasinglereservoirdamminganintermittentriverinnortheasternBrazil.Thestandardoperatingpolicyisalsoincludedinthecomparisonandoperationalresultsprovidedbydeterministicoptimizationbasedonperfectforecastsareusedasabenchmark.Ingeneral,theISOandPSOmodelsperformedbetterthanSDPandtheSOP.Inaddition,theproposedISO-basedsurfacemodel-ingprocedureandthePSO-basedtwo-dimensionalhedgingruleshowedsuperioroverallperformanceascomparedwiththeneuro-fuzzyapproach.2009ElsevierLtd.Allrightsreserved.1.IntroductionThemajortaskofreservoiroperationistodecidehowmuchwatershouldbereleasednowandhowmuchshouldberetainedforfutureusegivensomeavailableand/orforecastedinformationatthebeginningofthecurrenttimeperiod.Inpractice,reservoiroperatorsusuallyfollowrulecurves,whichstipulatetheactionsthatshouldbetakenconditionedonthecurrentstateofthesystem.Rulecurvesaretypicallyconstructedfromsimulationmodelsthatprovidereservoirresponsestopredefinedoperatingpolicies.Sincethereareoftenalargenumberoffeasiblepolicies,mathe-maticaloptimizationtechniquesmayassistinidentifyingthebestoneastheyimplicitlylookatallpossiblealternatives[48].Theapplicationofoptimizationtosolvereservoiroperationproblemshasbeenatopicextensivelystudiedduringthelastfewdecades[19,47,48].Variousofthesestudiesdealwithdeterministicoptimi-zationmodels,whichdonotconsidertheuncertaintiesofsomevariablessuchasfuturereservoirinflows.Uncertaintybecomesimportantwhenexpectedvaluesofinflowscannotappropriatelyrepresenthighlyvariablehydrologicconditionsorwhenthein-flowscannotbereliablyforecastedforarelativelylongperiod.Insuchcases,theproblemistypicallyaddressedbystochasticdy-namicprogramming(SDP).SDPisthemostpopulartypeofexplicitstochasticoptimization(ESO),anapproachthatincorporatesprob-abilisticinflowmethodsdirectlyintotheoptimizationproblem.However,severalauthorshavepointedoutthat,despitethecontinuingresearch,thereisstillagapbetweentheoreticaldevel-opmentsandreal-worldimplementations[19,40,47,34,48].Reser-voiroperatorsareusuallyskepticaltoreplacesimulationwithsophisticatedoptimizationmodelssincethelatteraremathemati-callymorecomplex,especiallywhenstochasticityisexplicitlyincluded.Inordertocopewiththisissue,thepresentpaperinvestigatestwotechniques,namely,implicitstochasticoptimization(ISO)andparameterization–simulation–optimization(PSO),whicharebothabletoincorporateinflowuncertaintiesandtoproviderulecurvesinanarguablysimplerwaythanESO.Inbrief,ISOusesdeterministicoptimizationtooperatethereservoirunderseveralequallylikelyinflowscenariosandthenexaminestheresultingsetofoptimaloperatingdatatodeveloptherulecurves.Inadiffer-entway,PSOfirstpredefinesashapefortherulecurvebasedonsomeparametersandthenappliesheuristicstrategiestolookforthecombinationofparametersthatprovidesthebestreservoiroperatingperformanceunderpossibleinflowscenarios.Inthisway,moststochasticaspectsoftheproblem,includingspatialandtemporalcorrelationsofunregulatedinflows,areimplicitlyin-cluded[19].TheutilizationofISOforfindingreservoiroperatingpolicieswasfirstexploitedbyYoung[50]inastudythatutilizeddynamicprogrammingappliedtoannualoperations.Theoptimalreleasesfoundbythedynamicprogrammingmodelwereregressedon0309-1708/$-seefrontmatter2009ElsevierLtd.Allrightsreserved.doi:10.1016/j.advwatres.2009.06.008*Correspondingauthor.Presentaddress:DepartmentofCivilEngineering,FederalUniversityofCampinaGrande,Av.AprígioVeloso882,BairroUniversitário,CampinaGrande,Paraíba58.429-900,Brazil.E-mailaddresses:geimes@yahoo.com(A.B.Celeste),billib@iww.uni-hannover.de(M.Billib).AdvancesinWaterResources32(2009)1429–1443ContentslistsavailableatScienceDirectAdvancesinWaterResourcesjournalhomepage:flowfortheyear.Theregressionequationcouldbethususedtoobtainthereservoirreleaseatanytimegiventhepresentstorageandinflowcondi-tions.KaramouzandHouck[15]extendedYoung’sprocedurebyaddingoneextraconstrainttotheoptimizationmodelspecifyingthatthereleasemustbewithinagivenpercentageofthereleasedefinedbythepreviouslyfoundoperatingpolicy.There
本文标题:Celeste2009Evaluation of stochastic reservoir oper
链接地址:https://www.777doc.com/doc-3099581 .html