电力配电网络的负荷模型与智能电表数据

0HAL Id: tel-015577620https://theses.hal.science/tel-015577620提交日期：2017年7月6日0HAL是一个多学科开放存取档案，用于存储和传播科学研究文献，无论其是否已发表。这些文献可以来自法国或国外的教育和研究机构，或来自公共或私人研究中心。0HAL多学科开放存取档案，旨在存储和传播研究级科学文献，无论是否已发表，来自法国或国外的教育和研究机构，公共或私人实验室。0用于电力配电网络运行和规划的负荷模型与智能电表数据0Ni Ding0引用此版本：0Ni Ding. 用于电力配电网络运行和规划的负荷模型与智能电表数据. 电力. Grenoble大学, 2012. 英文. �NNT :2012GRENT092�. �tel-01557762�TH�SEpourobtenirlegradedeDOCTEURDEL'UNIVERSIT�DEGRENOBLESp�ialit�:G�nie�letriqueArr�t�minist�riel:7Aut2006Pr�sent�eparNiDINGTh�sedirig�eparYvonB�SANGERetodirig�eparFr�d�riWURTZpr�par�eauseinduLaboratoireG2ELABdansl'�oleDotorale:EEATSLoadmodelsforoperationandplanningofeletriitydistributionnetworkswithsmartmeteringdataTh�sesoutenuepubliquementle30Novembre2012,devantlejuryompos�de:Pr.NouredineHadjsaidGrenobleINP,Pr�sidentPr.CarloAlbertoNuiUniversit�deBologne,RapporteurPr.CorinneAlonsoUniversit�deToulouse,RapporteurPr.DidierMayerMinedeParis,MembrePr.YvonB�sangerGrenobleINP,MembreDr.Fr�d�riWurtzCNRSGrenoble,MembreInvit�s:M.OlivierDevauxEDFR&DM.AlainGlatignyShneiderEletriContentsAknowledgmentsxiiiNotationsxix1Generalintrodution:thenewproblematiofloadmodelsinthesmartgridontext11.1Bakground:smartgridandsmartmetersforloadmodeling..21.2Motivationandobjetives...........................31.2.aFornetworkoperationneed...........................41.2.bFornetworkplanningneed...........................51.3Contributionsofthethesis..........................61.4Sopeandorganizationofthedissertation...............7AShort-termloadforeastingmodelsformonitoringandstatees-timator112Loadforeastingtehniquesandshort-termmodelframework132.1Literaturereview.................................142.1.aForeastingleadtimesandin�uenefators..................142.1.bForeastingmethods..............................162.1.b-iClassialapproah..........................182.1.b-iiArti�ialintelligentapproah....................252.1.b-iiiHybridmodels............................352.1.Literaturereviewonlusionsandperspetives................372.2Datadesription..................................402.2.aMV/LVsubstation...............................402.2.a-iTemperaturein�uene........................412.2.a-iiDaytypein�uene..........................422.2.a-iiiTimein�uene............................422.2.bMVfeeder....................................452.3ChoiesofTimeseriesandNNmethods..................452.4Performaneriteriaandreferenease................462.4.aPerformaneriteria:MAPEandMAE....................462.4.bReferenease:thenaivemodel........................472.5Conlusion......................................47ivCONTENTS3Timeseriesmodel493.1Additivetimeseriesmodelandproedureoverview.........503.2Statistialtools..................................513.2.aDummyVariableRegression..........................513.2.bTrendComponentEstimation.........................523.2.CyliComponentEstimation.........................523.2.dTestsofStationarity..............................533.2.eSmoothedPeriodogram.............................533.2.fRegressionModelwithFourierComponents.................543.2.gANOVANullityTest..............................543.2.hCompleteForeastingModel..........................553.3Appliationexampleresults..........................553.3.aTrainingset...................................553.3.bTestset.....................................573.3.ResidualAnalysis................................603.3.-iNormality...............................603.3.-iiIndependene.............................613.4Weatherunertainty...............................623.5Conlusion......................................644Neuralnetworkmodel674.1Mahinelearningtehnique..........................684.2MultiLayerPereptronsandtrainingproess............694.3Modeldesign.....................................714.3.aVariableseletion................................714.3.bModelseletion.................................734.3.b-iModelseletionmethodology....................734.3.b-iiAssessmentofthegeneralizationabilityofthemodels......744.4Numerialillustration.............................764.4.aFramework....................................764.4.bModeldesign:anillustrativeexample.....................774.4.b-iVariableseletionexample......................774.4.b-iiSeletingthebestmodelforagivenomplexity..........814.4.b-iiiComplexityseletionexample....................814.4.Results......................................834.5Overallomparisonwiththetimeseriesmodel............854.6Conlusionandperspetive..........................86BLoadestimationmodelsfordistributionnetworkplanning895Loadresearhprojetsindistributionnetworks:stateoftheart915.1Deisionmakingindistributionnetworkplanning..........925.1.aCoinidentload.................................93CONTENTSv5.1.bTypialLoadPro�le(TLP)..........................955.2Loadresearhprojetsindifferentountries............965.2.aFinlandDSOmodel...............................975.2.bDenmarkDongEnergy.............................985.2.NorwaySINTEFEnergyResearh.......................995.2.dTaipowersystem.................................995.3Frenhloadresearhprojet........................1005.3.aDatadesription.................................1025.3.bEDFBAGHEERAmodel............................1035.3.b-iTMBtemperatureandbasimodel.................1045.3.b-iiCommonoe�ientestimation...................1055.3.b-iiiSpei�parameterestimation....................1055.3.b-ivIllustrativeexampleandmodel'soutput..............1075.4Conlusion......................................1116Nonparametrimodel1136.1Nonparametrimodel...............................1156.1.aStatistialtests.................................1166.1.bKerneldensityestimation............................1176.1.CUSUMalgorithm...............................1176.1.dKernelregression................................1186.1.eSmoothingparameterseletion:ross-validationtehnique..........1196.2Computationalexample.............................1206.2.aIllustrativeexampleresults...........................1216.2.bComparisonwiththeBAGHEERAmodel...................1246.3Validationstudy..................................1276.4Disussion.......................................1296.4.aCitationsoftheupper-boundde�nitionsinEDFreports...........1306.4.bUpperboundinthenonparametrimodels..................1316.4.Validationtrialontheupper-boundestimation................1326.5Conlusionandperspetive..........................1377Generalonlusionandperspetive1397.1Conlusion......................................1397.2Perspetive......................................140Bibliographie154Appendies155ATimeseriesmodel'sresultsummary155BBinaryhypothesistest157CExampleofANOVAnullitytest159viCONTENTSDComparisonresultsofnaivemodel,timeseriesmodelandneu-ralnetworkmodel161ER�sum�fran�ais169E.1Introdutiong�n�rale:lanouvelleprobl�matiquedumod�ledehargedansleontextedur�seauintelligent..........169E.1.aR�seauintelligentetompteursintelligentspourlesmod�lesdeharge...169E.1.bObjetifsetplandur�sum�fran�ais......................170E.1.Contributiondeth�se..............................172E.2Mod�ledehargepr�ditifourttermepourlaonduiteetl'estimateurd'�tat.................................173E.2.aM�thodesdelapr�visiondehargedanslalitt�rature............174E.2.bDesriptiondedonn�es.............................178E.2.Choixdesm�thodes:s�riehronologiqueetr�seaudeneurones.......180E.2.dCrit�resdeperformaneetmod�leder�f�rene................181E.2.eMod�les�riehronologique...........................182E.2.fMod�ler�seaudeneurones...........................187E.2.f-iConeptiondumod�le........................188E.2.f-iiComparaisonglobaleavelemod�ledes�riehronologique....191E.3Mod�led'estimationdehargepourlaplanifiationdur�seaudedistribution...................................193E.3.aMod�leBAGHEERA..............................195E.3.bMod�lenonparam�trique............................198E.4Conlusionsetperspetivesg�n�rales..................199ListofFigures1.1AvailablemeasurementsintheFrenhdistributionnetworks..........31.2Relationshipamongforeastingmodels,SE,andADAfuntions........52.1Summaryofloadforeastingmethodsintwodimensions............172.2Singlepereptronstruture..............................272.3One-hidden-layernetworkstruture.........................272.4Supervisedlearningproedure............................282.5Reurrentneuralnetworkstruture.........................292.6FuzzyLogiproess..................................342.7Fuzzylogi:inputvariablesmembershipfuntion................352.8Fuzzylogi:outputvariablesmembershipfuntion................352.9Dailyaverageloadandtemperaturedatathrough 414days(fromSept.9,2009toOt.27,2010)ofsubstationCE_MOU(mainlyresidential).....402.10Dailyaverageloadthrough 414days(fromSept.9,2009toOt.27,2010)ofsubstationVI_LOG(mixedserviesetorandindustrial)..........412.11Dailyaverageloadthrough 414days(fromSept.9,2009toOt.27,2010)ofsubstationCE_FRO(anindustriallient)...................412.12NormalweekomparedtotheweekwithanationalholidayofSubstationCE_FRO(anindustriallient)...........................432.13SimilarityindexalulatedbasedonalldaysofsubstationCE_MOU....442.14SimilarityindexwithoutweekendsandpubliholidaysofsubstationCE_MOU442.15MVfeedersandpositionofonnetedMV/LVsubstations...........453.1Stepsofthedesignedtimeseriesforeastingmethod...............513.2Trainingsetandtestsetperiodsoftheavailabledata..............553.3Aweeklyonsumptionpattern(Otober5,2009toOtober11,2009)ofamixedindustrialandserviesetorsubstationVI_LOG.............563.4SubstationVI_LOG,MAEriteriaalulatedonthetrainingset(117days)fordi�erentslidingwindowsizes(weeks)......................563.5PeriodogramofthedetrendedtrainingdatasetsmoothedbytheDaniellkernel573.6SubstationVI_LOG,omparisonoftheforeastingresultswiththerealmeasurementsonthetestsetperiod(297days)..................583.7SubstationVI_LOG,two-day-aheadloadforeastingresultsonweekdays..583.8SubstationVI_LOG,two-day-aheadloadforeastingresultsonweekends..593.9SubstationVI_LOG,densityfuntionplotandumulativedensityfuntionplotoftheresidual...................................613.10SubstationVI_LOG,evolutionofautoorrelationfuntionsofeahstep...62viiiLISTOFFIGURES3.11HistogramoftheGaussiandistributedtemperatureunertaintyaddingtotheatualtemperature................................633.12Three-dayforeastingtemperaturesomparedtotheatualtemperatures..644.1Orthogonalforwardrankingproess.........................734.2Neuralnetworkseletionproedure.........................744.3Separationoftheloadurveintothedailyaveragepowerandtheintradaypowervariation.....................................774.4Generationofseondaryvariablesandprobevariables..............784.5Cumulativeprobabilityforaprobevariabletohaveabetterrankthanaandidatevariable....................................804.6Modelseletionfortheintradaypowervariationmodel.............824.7NeuralnetworkomplexityseletionstrategieswithVLOOsoreandlever-agedistribution.....................................825.1Networkdeisionmakingproedure.........................935.2Exampleofoinidenefatoralulation.....................945.3DistributionLoadEstimation(DLE)proess...................975.4Voltage-dropandtaphangeradjustment......................1015.5Two-year(July01,2004 ∼June30,2006)dailyaverageloadsofo�-peak/on-peakoptionlientno.5..............................1035.6Two-year(July01,2004 ∼June30,2006)dailyaverageloadsofbasioptionlientno.18.......................................1035.7O�-peak/on-peakoptionlientno.5:urve�ttingono�-peakdailyenergyuse............................................1085.8O�-peak/on-peakoptionlientno.5:urve�ttingonon-peakdailyenergyuse............................................1085.9Basioptionlientno.18:urve�ttingondailyenergyuse...........1095.10O�-peak/on-peakoptionlientno.5:outputsoftheBAGHEERAmodel,TMBloadestimationsonweekdays.........................1105.11O�-peak/on-peakoptionlientno.5:omparisonofTMBweekend'sandweekday'sloadestimation...............................1116.1Overviewofthenonparametrimodel.......................1166.2Statistialtestsproedure...............................1176.3Datadiagram:historialdata,1st-yeardata,and2nd-yeardata........1206.4O�-peak/on-peakoptionlientno.5:statistialtestsresultofthermosensi-tivehek........................................1216.5O�-peak/on-peakoptionlientno.5:CUSUMhartofdailyaveragepower.1216.6O�-peak/on-peakoptionlientno.5:separationresultofoneyear'spowerdatabyCUSUMalgorithm..............................1226.7O�-peak/on-peakoptionlientno.5:weekdayminimumpowerestimations.1226.8O�-peak/on-peakoptionlientno.5:statistialtestsresultforthedataoherenehek.....................................1236.9O�-peak/on-peakoptionlientno.5:ross-validationresultonthesmooth-ingparameterseletionofthekernelestimation.................124LISTOFFIGURESix6.10NW,LL,andLL2regressors,indiatingtherelationshipbetweenthevaria-tionoftemperatureandthelient'sdailypoweronsumption.........1246.11O�-peak/on-peakoptionlientno.5:presentationofunertaintyofasample1256.12O�-peak/on-peakoptionlientno.5:maximumpowerestimationofweek-dayloads.........................................1256.13SumSquareErrors(SSE)softheBAGHEERAestimator,NW,LL,andLL2estimatorsonthetestdata..............................1266.14Studyasesandsenariosinthevalidationstudy.................1276.15Studyaseno.1,senario1:o�-peak/on-peakoptionlients,omparisonofSSEsofBAGHEERA,NW,LLandLL2estimatorsonthedaysbelow0degreeduringtheseondyear............................1286.16Studyaseno.1,senario2:o�-peak/on-peakoptionlients,omparisonofSSEsofBAGHEERA,NW,LLandLL2estimatorsontheo�-peakhoursofthedaysbelow0degreeduringtheseondyear..................1286.17Studyaseno.2,senario1:o�-peak/on-peakoptionlients,omparisonofSSEsofBAGHEERA,NW,LLandLL2estimatorsonthe30oldestdaysoftheseond-yeardata................................1296.18Studyaseno.2,senario2:o�-peak/on-peakoptionlients,omparisonofSSEsofBAGHEERA,NW,LLandLL2estimatorsontheo�-peakhoursofthe30oldestdaysoftheseond-yeardata....................1296.1910%hourlypowerexessprobabilitythresholdandmedianvalueforeverytimestep.........................................1326.20Summaryoftheupper-boundomparisonoftherealmeasurements,theBAGHEERAmodel,andnonparametrimodels.................1336.21Poweronsumptionoflientno.22duringtwoyears(July01,2004 ∼June30,2006).........................................1346.22Poweronsumptionoflientno.17duringtwoyears(July01,2004 ∼June30,2006).........................................1356.2330-minutetimestepstandarddeviation(sd)oflientNo.17..........135E.1Relationentrelesmod�lesdehargepr�ditifs,l'estimateurd'�tat,etlesfontionsavan�esdur�seau.............................171E.2R�sum�desm�thodesdehargepr�ditivesendeuxdimensions........176E.3Courbedehargeettemp�raturejournali�rependant 414jours(du9/9/2009au27/10/2010)duposteHTA/BTCE_MOU(onnet�prinipalement�deslientsr�sidentiels)..........................