基于动态集成lssvr的超短期风电功率预测

《基于动态集成lssvr的超短期风电功率预测》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、基于动态集成LSSVR的超短期风电功率预测摘要：？对最小二乘支持向量回归(LeastSquareSupportVectorRegression,LSSVR)建模风电功率时变特性的局限性，提出了一种基于动态集成LSSVR的超短期风电功率预测模型.首先利用风电场监测控制与数据采集(SupervisoryControlAndDataAcquisition,SCADA)与数值天气预报(NumericalWeatherPrediction,NWP)系统的历史数据建立离线单体LSSVR模型库，然后根据预测时段与训练时段NWP序列的相似度从单体LSSVR模型库中动态选择候选集成成员，再后综合考虑正确性与多

2、样性确定集成成员.最后由预测时段与训练时段NWP序列间的相似度分配集成LSSVR成员的权重.通过对湖南省某风电场输出功率进行预测，验证了动态集成LSSVR预测模型的有效性，与持续法、自回归求和移动平均法、单体LSSVR模型、常权重LSSVR组合模型及BPNN动态集成模型相比，动态集成LSSVR模型具有更高的精度，在天气非平稳变化阶段更加明显.关键词：超短期风电功率预测；最小二乘支持向量回归；动态集成；动态时间弯曲距离；数值天气预中图分类号：TM614文献标志码：AUltra-short-termWindPowerPredictionBasedonDynamicalEnsembleLeastS

3、quareSupportVectorRegressionLIURongshengl，PENGMinfangl,ZHANGHaiyanl，WANXun2,SHENMeie3(1.CollegeofElectricalandInformationEngineering,HunanUniversity,Changsha410082，China；2.StateGridHunanElectricPowerCompanyElectricPowerResearchInstitute,Changsha410007，China；3.ComputerSchoolBeijingInformationScience

4、&TechnologyUniversity，Beijing100101，China)Abstract：Forthelimitationofleastsquaresupportvectorregression(LSSVR)inmodelingthetimevaryingfeatureofwindpower，anultra-short-termwindpowerprediction(USTWPP)modelbasedondynamicalensembleLSSVRwasproposed.Firstly，theoff-lineLSSVRmodellibrarywascreatedbymakingu

5、seofthehistoricaldatawhichwereobtainedfromNumericalWeatherPrediction(NWP)andsupervisorycontrolanddataacquisition(SCADA)systemofwindfarm.Then，thecandidatemembersofensembleLSSVRwereselectedfromoff-lineLSSVRmodellibrarydynamicallyaccordingtothesimilaritybetweentheNWPofforecastingperiodandtheNWPoftrain

6、ingperiod.Theensemblemembersweredecidedbyconsideringtheaccuracyanddiversity.Finally,theweightsofensembleLSSVRmemberswereassignedaccordingtothesimilaritybetweentheNWPoftrainingandNWPofpredictionperiod.ThevalidityofthedynamicalensembleLSSVRbasedpredictorwasverifiedbypredictingthewindpowerofawindfarmi

7、nHunanProvince.Comparedwithpersistencemethod(PM),autoregressiveintegratedmovingaverage(AGIMA)，LSSVR，constantweightensembleLSSVR,andensembleartificialneuralnetworks(ANN)，thedynamicalensembleLSSVRismoreaccura