Stochastic Dynamic Programming forLong TermHydrothermal Sche duling Considering Different Streamfiow Models

IntheSDP approach forLTHS,therandomvariable Abstract Thispaper isconcerned withtheperformance of (inflow) canberepresented through different models. The stochastic dynamic programming forlongtermhydrothermal simplest oneistorepresent theinflow byitsaverage value, scheduling. Different streamflow modelsprogressively more whichinpractice corresponds toconverting theproblem into a complex havebeenconsidered inorder toidentify thebenefits of deterministic one.Another modelmorecommonly usedisthe increasing sophistication ofstreamflow modeling on the independent model,whichconsiders theinflows as performance ofstochastic dynamic programming. Thefirst and independent randomvariables without timecorrelation. simplest modelconsiders theinflows given bytheir averageFinally, amoresophisticated modelisthedependent model, values; thesecond modelrepresents theinflows byindependent whichrepresents theinflow through aMarkovchain based on probability distribution functions; andthethird modeladopts a alag-one periodical autoregressive model. Markovchainbasedona lag-one periodical auto-regressive Besides thetimecorrelation oftheinflow series, thechoice model. Theeffects ofusing different probability distribution oftheprobability distribution oftheinflows isanother aspect functions havebeenalsoaddressed. Numerical results fora thatmay affect theSDP performance inLTHS.The hydrothermal test system composed byasingle hydroplant have periodically stationary Gaussian distribution isthemostoften beenobtained bysimulation with Brazilian inflow records. distribution considered (10). However, periodically stationary distributions donotproperly represent characteristics suchas IndexTerms-streamfiow models, longtermhydrothermalasmeranth no-edybavrofhe