基于改进Res-BiGRU模型的风电功率预测*苏澈,钱驹李,陈益宁,张逸飞,肖汉(南京工程学院,江苏南京211100)摘要:在新能源研究领域中,风电功率预测是重点内容之一,准确的数值和变化趋势预测能使调度人员合理安排机组启停,改变基荷厂和主调频厂的运行方式,实现最优低碳发展。基于深度学习理论,设计了一种基于Res-BiG-RU的融合神经网络模型,同时采用改进自编码器的数据预处理方法对数据进行降维。仿真结果表明,经过数据降维的融合模型的预测精度优于单一神经网络。关键词:神经网络;风电功率预测;数据降维;自编码器中图分类号:TM93DOI:10.19768/j.cnki.dgjs.2023.04.016WindPowerPredictionBasedonImprovedRes-BiGRUFusionNeuralNetwork*SUChe,QIANJuli,CHENYining,ZHANGYifei,XIAOHan(NanjingInsitituteofTechnology,Nanjing211100,China)Abstract:Intheprocessofnewenergyresearch,windpowerforecastingisoneofthekeycontents.Accuratevaluesandchangetrendpredictioncanenabledispatcherstoreasonablyarrangethestartandstopoftheunit,changetheoperationmodeofthebaseloadplantandthemainfrequencyregulationplant,andachievethegoalofeconomicoptimalandlow-car-bondevelopment.Basedonthepreviousresearch,basedondeeplearningtheory,ares-BiGRU-basedfusionneuralnet-workmodelisdesigned,andthedatapreprocessingmethodofimprovedautoencoderisadoptedtoreducethedimensionali-tyofthedata.Simulationresults...