第47卷第2期电网技术Vol.47No.22023年2月PowerSystemTechnologyFeb.2023文章编号：1000-3673（2023）02-0688-10中图分类号：TM721文献标志码：A学科代码：470·40基于时序残差概率的风电场超短期风速混合预测模型戴剑丰1,2，阎诚3，汤奕3（1．南京邮电大学自动化学院，江苏省南京市210023；2．南京邮电大学人工智能学院，江苏省南京市210023；3．东南大学电气工程学院，江苏省南京市210096）Ultra-short-termWindSpeedHybridForecastingModelforWindFarmsBasedonTimeSeriesResidualProbabilityModelingDAIJianfeng1,2,YANCheng3,TANGYi3(1.CollegeofAutomation,NanjingUniversityofPostsandTelecommunications,Nanjing210023,JiangsuProvince,China;2.CollegeofArtificialIntelligence,NanjingUniversityofPostsandTelecommunications,Nanjing210023,JiangsuProvince,China;3.SchoolofElectricalEngineering,SoutheastUniversity,Nanjing210096,JiangsuProvince,China)1ABSTRACT:Accuratepredictionofthewindspeedisofgreatsignificanceforimprovingtheaccuracyofwindpowerpredictionandthestableoperationofthepowergrid.Theprecisecharacterizationofthepredictionmodelresidualsisaprerequisitetoachieveaccuratepredictionofthewindspeedseriesinthewindfarm.Thispaperproposesanultra-short-termwindspeedhybridforecastingmodelbasedontheprobabilityoftimeseriesresiduals.First,thewindspeedisdecomposedintocomponentswithdifferentfrequencycharacteristicsbasedontheoptimizedvariationalmodaldecomposition.Then,adeterministicpredictionmodelisconstructedforthelinearcomponentswithregularchangesinthewindspeedcomponentsthroughthetimeseriesmodel.Forthefittingresidualcomponents,theconditionalkerneldensityestimationisusedtoestablishaprobabilityforecastingmodel.Thenbasedonthesuperpositionoftherecursiveresultsofthetwomodelsthewindspeedpredictionvalueisformed.Onthisbasis,inviewoftheproblemthattheresidualconditionalprobabilityofeachcomponentcannotdirectlyrepresenttheoriginalwindspeedprobabilityforecastingresult,thispaperproposesaprobabilitygenerationbasedonthesimulationtorealizethewindspeedprobabilityforecasting.Finally,takingtheoperatingdataofawindfarminNortheastChinaasanexample,i...
