Chalise, Sushant2022-03-022022-03-022021-09MASTERS OF SCIENCE IN CLIMATE CHANGE AND SUSTAINABLE DEVELOPMENT PROGRAMMEhttps://hdl.handle.net/20.500.14540/8666Solar energy has immense promise as a source of renewable energy. Itisabundantthroughout theyear,althoughitissubjecttouncertaintyduetovariousparameters.Sunenergysources’Solarenergyhasimmensepromiseasasourceofrenewableenergy.Itisabundantthroughout theyear,althoughitissubjecttouncertaintyduetovariousparameters.Sunenergysources’ affectivityandproductivitycanbeimprovedbyaccurateforecastingofsolarradiation.Fore- castingGlobalSolarRadiation(GSR)inthefieldofresearchhasattractedwidespreadattention fromtheresearchcommunityinmanypracticalfieldsincludingenergy.Differentmodelsfor predictingGSRpotentialhavebeenusedintheliterature.Oneofthemostprominentlinear modelsfortimeseriesforecastingistheAutoregressiveIntegratedMovingAverage(ARIMA). Therearealsodifferentmachinelearningmodelswhichshowpromisingforecastingresults.To takeadvantageoftheuniquebenefitsofARIMAandmachinelearningmodelsinlinearand nonlinearmodelingthedataofsolarradiationpotential,weproposeahybridmethodcombining ARIMAandmachinelearningmodelsANN(ArtificialNeuralNetwork)andLSTM(LongShort TermMemory)modelsinthisstudy.ThedatasetwasobtainedforthelocationofKushma, Parbatfordurationbetween1990to2014.Forthesupplieddatasets,theARIMAplusANN hybridmodelwasseentobethebestmethodforpredictingsolarradiationpotential.Thecor- relationcoefficient(Rsquare)iscalculated0.847.Theerrorvaluesforthismodelareaccessed asRMSE,MAPEandMAEof1.719,6.456and1.330respectively.Theexperimentalresults ofrealdatasetsshowthatthecombinedmodelcaneffectivelyimprovethepredictionaccuracy achievedbyanymodelusedalone.TTheacquiredresultsalsodemonstratedthatthecreated modelcouldbeutilizedtocalculatethesolarradiationpotentialofanygeographicregionwith knownclimaticparameters.enThesis