70|·2023年第42卷第4期方案设计|EngineeringDesign基于深度学习集成方法的日用电最大负荷预测杨敏,马燕如,朱刘柱,王宝(国网安徽省电力有限公司经济技术研究院,安徽合肥230022)摘要:电力负荷受气温等多重因素影响,具有短期波动性和非线性特征,给电力系统调度带来了极大不确定性和挑战。为提前做好电力生产计划和调度预案,开展短期电力日最大负荷预测具有重要的应用价值。基于深度学习集成(SDAE-B)方法的计算准确度和计算效率优势,对较大变化的外界因素具有高鲁棒性。选取某省级电网2018—2020年三年的日度电力数据和气温数据,利用SDAE-B方法对该地区2020年任意15天日最大负荷进行预测,并与运用SDAE方法和支持向量回归(SupportVectorRegression,SVR)方法得到的测试结果进行比较。结果显示SDAE-B方法的预测误差最小,且该深度学习方法具有强大的特征提取能力,能在最大程度上减少数据典型特征的损失,且很好地跟踪电力日最大负荷的非线性特征。关键词:电力负荷;峰值预测;深度学习;SDAE-B方法DailyelectricpeakloadforecastingbasedonSDAE-BYANGMin,MAYanru,ZHULiuzhu,WANGBao(InstituteofEconomicandTechnology,StateGridAnhuiElectricPowerCompany,Hefei230022,China)Abstract:Theelectricloadinpowermarketisofsignificantrandomnessandnon-linearity,whichbringsgreatuncertaintyandchallengetopowersystemscheduling.Inordertomakepowerdispatchingplan,theelectricloadpeakforecasthaveawidepotentialapplicationinpowerareas.TheStackedDenoisingAutoEncoder--Bagging(SDAE-B)methodhastheadvantagesofcalculationaccuracyandefficiency,andhighrobustnesstolargechangesinexternalfactors.Inthispaper,thedailypowergridunifiedadjustmentdataandtemperaturedataofaProvinceinChinafrom2018to2020areselectedtopredictthemaximumloadofany15-daysdispatchingin2020intheProvinceusingSDAE-Bmethod,andthetestresultsarecomparedwiththoseobtainedbySDAEmethodandSVRmethod.Theresultsshowthatthedeeplearningintegrationmethod(SDAE-B)hastheminimumerrorintheprediction,sotheSDAE-Bmethodisusedtoforecasttheshort-termpowerloadpeakwithlesserror,andarelativelysatisfactoryresultisachievedinthepredictionaccuracy.Theempiricalresultsshowthattheproposeddeeplearningmethodhasstrongfeatureextractionabilitya...