美国电网介绍

ArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationModelingtheRestructuredIllinoisElectricityModelingtheRestructuredIllinoisElectricityMarketasaComplexAdaptiveSystemMarketasaComplexAdaptiveSystemCharlesM.MacalCharlesM.MacalGaleA.BoydGaleA.BoydRichardR.RichardR.CirilloCirilloGuenterGuenterConzelmannConzelmannMichaelJ.NorthMichaelJ.NorthPrakashPrakashR.R.ThimmapuramThimmapuramThomasD.ThomasD.VeselkaVeselkaCenterforComplexAdaptiveAgentSystemSimulation(CASCenterforComplexAdaptiveAgentSystemSimulation(CAS22))Decision&InformationSciencesDivisionDecision&InformationScience[e-mail]@anl.gov,,–Marketshavegreatercomplexity–Marketshavemoreplayers–Participantshavedifferent(oftenconflicting)objectivesanddecisioncharacteristicsorriskpreferenceszConventionalmodelingapproachesmustbeextendedtoaccountforthenewrealityofdecentralizedcompetitivedecision-makingzTheseextensionsmustbeconsistentwithexistingapproachesandoftenincludetheminthelargermodelingframeworkLoadAggregatorsLoadAggregatorsCustomersCustomersGenerationCompaniesBrokersBrokersBrokersBrokersLDCLDCLDCLDCLDCLDCLDCLDCPXISOPXISOPXISOPXISObidsbidsbidsbidscontractscontractscontractsArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationWhyAgents?zTherulesofbusinessandsocialinteractionareatleastasimportantastherulesofphysicswhenitcomestothegeneration,sale,anddeliveryofelectricalpowerzAgentsoperatingwithinanagentframeworkcanbeusedtomodeldecentralizedcompetitivedecision-makingzAgentframeworksallowgroupsofagentstointeractincomplex,dynamicwayszLearningandadaptationofagentbehaviorcanbemodeledzTransientconditionsofthesystemcanbestudiedinadditionothelong-run(equilibrium?)conditionszAlternativemarketrulescanbetestedArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationAgentsAreAutonomous,Self-directed,Decision-makingUnits•Agentsareheterogeneous•Goals•Behaviors•Attributes•Availableaccumulatedresources•Decisionrulesvarybyagent•Sophisticationofrules•Cognitive“load”•Internalmodelsoftheexternalworld•Memoryemployed•Whatistheeffectofindividualagentbehavior(rules)onthesystem?•Docertaintypesofagentsdominate?•Doesthesystemevolvetowardastablemixofagenttypes?Agent•Attributes•Rulesofbehavior•Memory•Sophistication•ResourcesArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationzPhysicalagents–Generators–Transmissionbuses–TransmissionlineszDecision-makingagents–Consumers–Demandagents–Distributioncompanies–Generationcompanies–Transmissioncompanies–ISO/RTOs–RegulatorsTheEMCASModelUsesAVarietyofAgentstoModelDecentralizedElectricityMarketsArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationGenerationCompanyAgentsConsiderManyFactorsinProposingBidsfortheDay-AheadMarketArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationEMCASOperatesatMultipleTimeScalesandDecisionCycleszPlanning:–Day-AheadPlanning:‹Agentsdeterminemarketallocationsforsellingproducts‹Bilateralcontractsareformedwithindividualdemandagents,andenergybidsaresentintotheISO‹AgentsmakeunitcommitmentschedulesforthenextdayzDispatch:Hourly/Real-TimeDispatch:–PowerplantsareoperatedasdirectedbytheISOinaccordancewithpriormarketarrangementsmadeunderbilateralcontractsandinenergyandancillaryservicemarketszWeeklyandMonthlyAdjustments–BilateralcontractsaremadewithindividualdemandagentsandaresenttotheISOforapproval–Marketingstrategiesareadjusted–AdjustmentscanbemadetounitmaintenanceschedulesArgonneNationalLaboratoryDecisionandInformationSciencesDivisionCenterforComplexAdaptiveAgentSystemsSimulationDECISIONOUTPUT•Bidaccepted/rejected•Unitutilization•Unitprofitability•Marketpricevs.bidprice•Weather&LoadEMCASAgentsMakeDecisionsBasedonBothPastExperiencesandFutureExpectations•Ownunitavailability•Prices•Weather•Load•Competingunitavailability•Owncoststructure•Marketrules•Bidstructure:capacityblocksfordifferentmarkets•BidpricesforeachblockandmarketLOOKBACK(ShortandLong-TermMemory)TIMEAgentAgentDecisionDecisionRulesRulesLOOKSIDEWAYSLOOKAHEADE