基于CEEMDAN-LSTM模型的大伙房水库入库流量中长期预报

打开文本图片集

Medium- and Long-term Forecast of Inflow of Dahuofang Reservoir Based on CEEMDANLSTMModel

WANGChunyu'，ZHANGJing1*，YANGXu²，YANBin （1.CollgeofWaterResoures，ShenyangAgriculturalUniversityhenyangO866，China; 2.LiaoningDahuofangReir AdministrationLimitedLiabilityCompany，Fushun113oo7，China）

Abstract：Astocsisbcttinfallmpatufasanctivityteedultilectosf complexityitslwofhngesossigificantadomssnducertaintyreatlyiceasingtdifultyfedi-ndlote forecastandrestrictingitsaplicationectinproductionpractice.Therefore，howtobreakthroughthetechnicalbotleneckof medium-andlong-termflowforecasthasbecomeakeyproblemtobesolvedinthecurrnthydrologicalscienceresearchInviewof this，withDahuofang Reservoirastheresearchobject，the monthlyflow forecastof thereservoirwascariedoutbyusing gray correlationanalysisandprincipalcomponentanalysiscombinedwiththeCEEMDAN-LSTMmodel.Monthlyflow，rainfalland meteorologicaldatafrom1961to2008wereselectedtocalibratethemodelparameters.The monthlyflowdatafrom2009to2O20was usedtovalidatetemodel.Theidicatorsofeterinationcoefient，ootmansquareeandaverageelativeerorreapplied to evaluate theforecastresults.Aording totheresults，whenfourprincipalcomponentsareselectedbyusing principalcomponent analysis toreducethedimensionalityof theforecastfactorsetsaftertheadditionoftheprevious average temperatureandmaximum temperaturedata，theCEEMDAN-LSTMmodelcanefectivelyimprovetheforecastaccuracyandbecometheoptimalmodelforthe monthlyflowforecastofDahuofangReservoir.Technicalsupportisprovidedfortheformulationofthefuturemedium-andlong-te dispatching plan of Dahuofang Reservoir.

Keywords：iflow; medium-andlong-termforecast;principalcomponentanalysis;CEEMDAN-LSTMmodel; DahuofangReservoir

中长期流量预报在水资源规划与管理中具有重要的理论价值与实践意义，其预测结果可为流域水文情势分析、水库调度方案优化以及水资源合理配置提供科学依据。（剩余9574字）