神经网络英文文献

ARTIFICIALNEURALNETWORKFORLOADFORECASTINGINSMARTGRIDHAO-TIANZHANG,FANG-YUANXU,LONGZHOUEnergySystemGroup,CityUniversityLondon,NorthamptonSquare,London,UKE-MAIL:[e-mail],[e-mail],[e-mail]:Itisanirresistibletrendoftheelectricpowerimprovementfordevelopingthesmartgrid,whichappliesalargeamountofnewtechnologiesinpowergeneration,transmission,distributionandutilizationtoachieveoptimizationofthepowerconfigurationandenergysaving.Asoneofthekeylinkstomakeagridsmarter,loadforecastplaysasignificantroleinplanningandoperationinpowersystem.ManywayssuchasExpertSystems,GreySystemTheory,andArtificialNeuralNetwork(ANN)andsoonareemployedintoloadforecasttodothesimulation.ThispaperintendstoillustratetherepresentationoftheANNappliedinloadforecastbasedonpracticalsituationinOntarioProvince,Canada.Keywords:Loadforecast;ArtificialNeuronNetwork;backpropagationtraining;Matlab1.IntroductionLoadforecastingisvitallybeneficialtothepowersystemindustriesinmanyaspects.Asanessentialpartinthesmartgrid,highaccuracyoftheloadforecastingisrequiredtogivetheexactinformationaboutthepowerpurchasingandgenerationinelectricitymarket,preventmoreenergyfromwastingandabusingandmakingtheelectricitypriceinareasonablerangeandsoon.Factorssuchasseasondifferences,climatechanges,weekendsandholidays,disastersandpoliticalreasons,operationscenariosofthepowerplantsandfaultsoccurringonthenetworkleadtochangesoftheloaddemandandgenerations.Since1990,theartificialneuralnetwork(ANN)hasbeenresearchedtoapplyintoforecastingtheload.“ANNsaremassivelyparallelnetworksofsimpleprocessingelementsdesignedtoemulatethefunctionsandstructureofthebraintosolveverycomplexproblems”.Owingtothetranscendentcharacteristics,ANNsisoneofthemostcompetentmethodstodothepracticalworkslikeloadforecasting.Thispaperconcernsaboutthebehaviorsofartificialneuralnetworkinloadforecasting.AnalysisofthefactorsaffectingtheloaddemandinOntario,CanadaismadetogiveaneffectivewayforloadforecastinOntario.2.BackPropagationNetwork2.1.BackgroundBecausetheoutstandingcharacteristicofthestatisticalandmodelingcapabilities,ANNcoulddealwithnon-linearandcomplexproblemsintermsofclassificationorforecasting.Astheproblemdefined,therelationshipbetweentheinputandtargetisnon-linearandverycomplicated.ANNisanappropriatemethodtoapplyintotheproblemtoforecasttheloadsituation.Forapplyingintotheloadforecast,anANNneedstoselectanetworktypesuchasFeed-forwardBackPropagation,LayerRecurrentandFeed-forwardtime-delayandsoon.Todate,Backpropagationiswidelyusedinneuralnetworks,whichisafeed-forwardnetworkwithcontinuouslyvaluedfunctionsandsupervisedlearning.Itcanmatchtheinputdataandcorrespondingoutputinanappropriatewaytoapproachacertainfunctionwhichisusedforachievinganexpectedgoalwithsomepreviousdatainthesamemanneroftheinput.2.2.ArchitectureofbackpropagationalgorithmFigure1showsasingleNeuronmodelofbackpropagationalgorithm.Generally,theoutputisafunctionofthesumofbiasandweightmultipliedbytheinput.Theactivationfunctioncouldbeanykindsoffunctions.However,thegeneratedoutputisdifferent.Owingtothefeed-forwardnetwork,ingeneral,atleastonehiddenlayerbeforetheoutputlayerisneeded.Three-layernetworkisselectedasthearchitecture,becausethiskindofarchitecturecanapproximateanyfunctionwithafewdiscontinuities.ThearchitecturewiththreelayersisshowninFigure2below:Figure1.NeuronmodelofbackpropagationalgorithmFigure2.Architectureofthree-layerfeed-forwardnetworkBasically,therearethreeactivationfunctionsappliedintobackpropagationalgorithm,namely,Log-Sigmoid,Tan-Sigmoid,andLinearTransferFunction.TheoutputrangeineachfunctionisillustratedinFigure3below.Figure.3.Activationfunctionsappliedinbackpropagation(a)Log-sigmoid(b)Tan-sigmoid(c)linearfunction2.3.TrainingfunctionselectionAlgorithmsoftrainingfunctionemployedbasedonbackpropagationapproachareusedandthefunctionwasintegratedintheMatlabNeuronnetworktoolbox.TABLE.I.TRAININGFUNCTIONSINMATLAB’SNNTOOLBOX3.TrainingProcedures3.1.BackgroundanalysisTheneuralnetworktrainingisbasedontheloaddemandandweatherconditionsinOntarioProvince,CanadawhichislocatedinthesouthofCanada.TheregioninOntariocanbedividedintothreepartswhicharesouthwest,centralandeast,andnorth,accordingtotheweatherconditions.Thepopulationisgatheredaroundsoutheasternpartoftheentireprovince,whichincludestwoofthelargestcitiesofCanada,TorontoandOttawa.3.2.DataAcquisitionTherequiredtrainingdatacanbedividedintotwoparts:inputvectorsandoutputtargets.Forloadforecasting,inputvectorsfortrainingincludealltheinformationoffactorsaffectingtheloaddemandchange,suchasweatherinformation,holidaysorworkingdays,faultoccurringinthenetworkandsoon.Outputtargetsaretherealtimeloadscenarios,whichmeanthedemandpresentedatthesametimeasinputvectorschanging.Owingtotheconditionalrestriction,thisstudyonlyconsiderstheweatherinformationandlogicaladjustmentofweekdaysandweekendsasthefactorsaffectingtheloadstatus.Inthispaper,factorsaffectingtheloadchangingarelistedbelow:(1).Temperature(℃)(2).DewPointTemperature(℃)(3).RelativeHumidity(%)(4).Windspeed(km/h)(5).WindDirection(10)(6).Visibility(km)