搜索
：912121.1000842.300072：，。BP（），，4，，。，，。：；；；：TM615：A：1004-9649（2010）09-0075-04：2010-04-01：（1970—），，，，、。E-mail:ydzhanglan@139.com。0，，。，，。，，，，，。1。，，、、、［1-2］。，，，、。，，，。1.1。06：00—19：00。1，2d。1，，。2、3。1~3，。，，：、、、［3］。1.2：Ps=ηSI［1-0.005（t0＋25）］（1）：η；S，m2；I，kW/m2；t0，℃［4］。1Fig.1PoweroutputofsunnydaysELECTRICPOWER43920109Vol．43，No．9Sep.201075435Fig.5Dailyaveragepoweroutput2Fig.2Poweroutputofcloudydays3Fig.3Poweroutputofrainydays4Fig.4Dailypoweroutputandinsolation4。，，。1.3。，。，。5、62008923—927（）。，，。，，。22.1BP（）BP（）［5］，。。BP7。，、；［6］。2。2.22.2.16Fig.6Dailyaveragetemperature7BPFig.7TopologicalstructureofBPnetwork76：916-23-1416-19-1416-20-1416-21-1416-22-14R0.22231220.995840.22391280.995900.2229880.995830.2201750.995930.2216950.995881Tab.1Thetestingresultofdifferentnetworkstructure，。，，。，；。4，。2.2.2，16。06：00—19:001414，，16。06：00—19：0014，14。2.2.31614，。，，；，。，，R，1，。1。2116-21-14BP，。2.3，，。，［7］，：P*=p-pminpmax-pmin（2）：p；pmax，pminp；P*。S（logsig），：f（x）=11+exp（-x），01，。8BP。w1、b1w2、b2。P，a1，a2。3，。BP［8］，9。10，A，T，R。910，，R。8BPFig.8ForecastingmodelofBPnetwork9BPFig.9TrainingcurveofresilientBPmethod10Fig.10Linearregressionanalysisbetweenoutputvaluesandtargetvalues77432Tab.2ErrorsofPVsystempowerforecastingresults2，。4，。。：［1］CHAKRABORTYS,WEISSMD,SIMOESMG.Distributedintelligentenergymanagementsystemforasingle-phasehigh-frequencyACmicrogrid［J］.IEEETransactionsonIndustrialElectronics,2007,54（1）:97-109.［2］YONAA,SENJYUT,FUNABASHIT.Applicationofrecurrentneuralnetworktoshort-term-aheadgeneratingpowerforecastingforphotovoltaicsystem［C］//IEEEPowerEngineeringSocietyGeneralMeeting,Tampa,FL,USA,2007:1-6.［3］,,.［J］.,2009,24（9）:153-158.CHENChang-song,DUANShan-xu,YINJin-jun.Designofphotovoltaicarraypowerforecastingmodelbasedonneutralnetwork［J］.TransactionsofChinaElectrotechnicalSociety,2009,24（9）:153-158.［4］TSIKALAKISAG,HATZIARGYRIOUND.Centralizedcontrolforoptimizingmicrogridsoperation［J］.IEEETransactionsonEnergyConversion,2008,23（1）:241-248.［5］MELLITA,ARABAH,KHORISSIN,etal.AnANFIS-basedforecastingforsolarradiationdatafromsunshinedurationandambienttemperature［C］//IEEEPowerEngineeringSocietyGeneralMeeting,Tampa,FL,USA,2007:7-12.［6］YONAA,SENJYUT,FUNABASHIT,etal.Applicationofneuralnetworktoone-day-ahead24hoursgeneratingpowerforecastingforphotovoltaicsystem［C］//IntelligentSystemsApplicationstoPowerSystems,TokiMesse,Niigata,Japan,2007:1-6.［7］,.［J］.,2008,41（2）:74-78.LIRan,LIGuang-min.Photovoltaicpowergenerationoutputforecastingbasedonsupportvectormachineregressiontechnique［J］.ElectricPower,2008,41（2）:74-78.［8］,.MATLAB［M］.:,2004.ZHOUKai-li,KANGYao-hong.NeuralnetworkmodelandMATLABemulationprogramming［M］.Beijing:TsinghuaUniversityPress,2004./%19：0006：0007：0008：0009：0010：0011：0012：000.00180.0014-25.10220.00000.03670.00000.0389-6.18900.10410.18070.09990.1802-3.9711-0.30560.70711.00000.74630.93165.5423-6.84060.91000.97056.655813：0014：000.76830.39440.81920.38036.6283-3.590815：0016：0017：0018：000.58500.32990.21010.04660.58540.34860.21310.05430.06085.67011.418316.5725PhotovoltaicsystempowerforecastingbasedonneutralnetworksZHANGLan1,ZHANGYan-xia2,GUOChang-min1,ZHAOJie21.BeijingNewTechnologiesforEnergyResearchandDevelopmentCenter,YunnanPowerGridCorporation,Beijing100084,China;2.KeyLaboratoryofPowerSystemSimulationandControlofMinistryofEducation,TianjinUniversity,Tianjin300072,ChinaAbstract：Thecharacteristicsofaphotovoltaic（PV）systemandvariousfactorsthataffectthepowergenerationofaPVsystemwereanalyzed.AndthemodelofthePVsystempowerforecastingwasestablishedbasedonneuralnetworks.Inthemodel,thethree-layerbackpropagation（BP）networkmodelwasappliedforitspowerfulabilityofnonlinearmappingandgeneralization.Fourseason-forecastingsubmodelswereestablished.ThehistoricalpowerdataofaPVsystemwasclassifiedindifferentweatherconditionsandputintotheneuralnetworksto-getherwiththeweatherconditions.Therefore,theneuralnetworkscouldbetrainedforthepowerforecasting.Theresultsshowthatthismodelhasahighaccuracy,andcanprovideaneffectiveandfeasiblewaytoforecastthePVsystempoweroutput.Keywords：photovoltaicsystem；powerforecasting；neutralnetwork；weatherconditionimpact78