第39卷第2期2023年2月电网与清洁能源PowerSystemandCleanEnergyVol.39No.2Feb.2023清洁能源CleanEnergy——————————基金项目：国家自然科学基金项目（61876097）。ProjectSupportedbytheNationalNaturalScienceFoundationofChina（61876097）.ABSTRACT：Therandomnessanduncertaintyofphotovoltaicpowergenerationarethemainrestraintstothedevelopmentofphotovoltaic.Inordertoimprovetheaccuracyofshort-termpredictionofphotovoltaicpower，ashort-termphotovoltaicoutputpowerpredictionmethodconsideringweathertypesandhistoricalsimilardaysisproposedinthispaper.Firstly，thehistoricaldataaredividedaccordingtodifferentseasonsandweathertypes，andthesimilardaysaredeterminedaccordingtothecalculationresultsofgraycorrelationdegree.Secondly，chaosinitialization，nonlinearadjustmentofcontrolfactors，Levyflightandotherstrategiesareusedtoimprovethespottedhyenaoptimizationalgorithm.Theimprovedspottedhyenaalgorithmisusedtooptimizethekernelextremelearningmachineandashort-termphotovoltaicoutputpowerpredictionmodelbasedontheimprovedspottedhyenaalgorithmisestablishedtooptimizethekernelextremelearningmachine.Finally，themonitoringdataofanactualphotovoltaicpowerstationisusedforsimulationanalysis.Theresultsshowthattheshort-termphotovoltaicoutputpowerpredictionmodelbasedonISHO-KELMcanreducethevolatilityinthephotovoltaicoutputpowerpredictionprocess，improvethepredictionaccuracy，andthushadverifiedthecorrectnessandpracticalityofthephotovoltaicpredictionmethodproposedinthispaper.KEYWORDS：photovoltaicoutputpower；improvedspottedhyenaoptimizeralgorithm；kernelextremelearningmachine；weathertype；historicalsimilarday摘要：光伏发电的随机性和不确定性是制约光伏发展的主要原因。为了提高短期光伏发电功率预测精度，提出了一种考虑天气类型和历史相似日的短期光伏输出功率预测方法。针对不同季节和天气类型划分历史数据，根据灰色关联度计算结果确定相似日。采用混沌初始化、控制因子非线性调整和莱维飞行等策略对斑点鬣狗优化（spottedhyenaoptimizer）算法进行改进，采用改进斑点鬣狗算法（improvedspottedhyenaoptimizer）对核极限学习机进行优化，建立...
