资源描述：
《基于EWT-KELM方法短期风电功率组合预测》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、第44卷第504期电测与仪表Vol.44No.5042017年第12期ElectricalMeasurement&InstrumentationDec.2017基于EWT-KELM方法的短期风电功率组合预测卓泽赢1，曹茜2，李青3（1.国网新疆电力有限公司乌鲁木齐供电公司，乌鲁木齐830001；2.国网新疆电力有限公司经济技术研究院，乌鲁木齐830001；3.国网新疆电力有限公司电力科学研究院，乌鲁木齐830001）摘要：针对短期风电功率预测，提出一种基于经验小波变换(empiricalwavelettransform,EWT)预处理的核极限学习机(extremelearningmachi
2、newithkernels,KELM)组合预测方法。首先采用EWT对风电场实测风速数据进行自适应分解并提取具有傅立叶紧支撑的模态信号分量，针对每个分量分别构建KELM预测模型，最后对各个预测模型的输出进行叠加得到风速预测值并根据风电场风功特性曲线可得对应风电功率预测值，为验证本文方法的有效性，将其应用于国内某风电场的短期风电功率预测中，在同等条件下，与KELM方法、极限学习机(extremelearningmachine,ELM)方法、支持向量机(supportvectormachine,SVM)方法以及BP(backpropagationneuralnetwork)方法对比，实验结果表明
3、，本文所提方法具有较好的预测精度和应用潜力。关键词：经验小波变换；核极限学习机；组合预测；风电功率；风速-功率特性曲线中图分类号：TM641文献标识码：A文章编号：1001-1390（2019）00-0000-00Windpowershort-termcombinedforecastingbasedonEWT-KELMempiricalwavelettransformandextremelearningmachinewithkernelsmethodZhuoZeying1,CaoQian2,LiQing3(1.UrumqiPowerSupplyCompanyofStateGridXinji
4、angElectricPowerCo.,Ltd.UrumqiPowerSupplyCompany,Urumqi830001,China.2.EconomicsandTechnologyResearchInstituteofStateGridXinjiangElectricPowerCo.,Ltd.EconomicsandTechnologyResearchInstitute,Urumqi830001,China.3.ElectricPowerResearchInstituteofStateGridXinjiangElectricPowerCo.,Ltd.ElectricPowerResea
5、rchInstitute,Urumqi830001,China)Abstract:Aimingatshort-termwindpowerforecasting,akindofcombiningforecastingmethodforshort-termwindpowerbasedonempiricalwavelettransform(EWT)andextremelearningmachinewithkernels(KELM)isproposedinthispaper.firstlyFirstly,theEWTmethodisusedtodecomposethewindspeeddataan
6、dextractthedifferentmodeswhichhaveacompactsupportFourierspectrum.Secondly,differentKELMforecastingmodelsareconstructedforthesub-sequencesformedbytheeachmodecomponent.Simultaneously,theultimatewindspeedforecastingresultscanbeobtainedbythesuperpositionofthecorrespondingforecastingmodel,theforecastva
7、lueofwindpoweriscalculatedbythewindpowercharacteristiccurve.Inordertoverifytheeffectivenessoftheproposedmethods,itmethods,itisappliedtosomewindfarmsinChinaforshort-termwindpowerforcastingforecasting.Theexperiment
