数据驱动的光伏功率超短期自适应优化预测

打开文本图片集

关键词：光伏发电；时序预测；变分模态分解；门控循环单元；雾淞优化DOI：10.15938/j. jhust.2025.05.007中图分类号：TM615 文献标志码：A 文章编号：1007-2683（2025）05-0070-08

Abstract：The stochastic natureof solar power has been complicating maters forshort-termandultra-short-term forecasting.To tacklethischalenge，thispaperproposesan IVMD-RIME-GRUmodelwithadaptiveoptimization.Firstly，inthedataprocesing procedure，theoriginalsolarpowersignalissubjectedtoImprovedVariationalModeDecomposition（IVMD），andthedecomposed modalsignalsareforecastedseparatelyindiferentmodels，whichmitigatesthenon-stationarityimpactofsolarpower.TheIVMD methoddetermines theoptimal numberofVMDmodes bychecking whethertheresidueapproximateswhitenoise，whichhelpsto aleviateinformationlossinmodaldecomposition.Inthetrainingandforecastingstage，theGatedRecurentUnit（GRU）andthe RIMEoptimizerarebothemployedfortheirdistinguishedperformanceintheirrespectivefields.TheGRUacquirestheoptimal parametersfromtheRIMEoptimizerandshowsbeterpeformance.Inthelaststep，theindependentforecastingresultsofeachmode compoundthectualforecastingoutputSimulationresultsindicatethatheproposedalgorithmnotonlyimprovesthepeformanceofthe original GRU forecasting model by more than 70% ，but also achieves good tracking of rapid peak-valley transitions in solar power generation.

Keywords：solarpower；time series forecasting；variational mode decomposition；gated recurrent unit；RIMEoptimizer

0 引言

在全球绿色低碳转型、我国“双碳”目标确立的背景下，光伏发电因其绿色环保及低成本优势，迅速发展为可再生能源的支撑性力量。（剩余12707字）