Page - 223 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 0.00% 20.00% 40.00% 60.00% 80.00% 100.00% 120.00% 0 20 40 60 80 100 120 140 Temperature(Ԩ) Frequency Cumulative probability Figure9.Histogramandthecumulativeprobabilitydistributionof theT(h–1)*. Table3 lists theexampleof thecorrectedweather forecastvaluesusingMCMandactualvalues to predict thecooling loadat9o’clockon3July. It canbeseenthat theweather forecastdatacorrectedby MCMarecloser to theactualdata. Theothersampleshavethesameeffectsandwillnotberepeated hereduetospace limitations. Table3.The inputparametersof the loadforecastingmodel. DataType L(d–1,h)kw T(h)◦C T(h–1)◦C T(h–2)◦C T(h–3)◦C RH(h)% RH(h–1)% RH(h–2)% Actualdata −584.99 27.4 26.0 25.2 24.7 78.9 84.0 87.8 Forecastdata −584.99 28.9 27.2 25.6 24.4 58.0 66.0 78.0 Reviseddata −584.99 28.2 26.6 25.1 24.1 68.0 75.1 86.0 Table4 shows theprobabilitydistribution functionsobeyedby theerrorsof inputparameters obtainedfromthestatisticalanalysisbySPSS.Theprocedures for thecorrectionofotherparameters aresimilar to thatof theparameterT(h–1).Wenolongerdescribemoredetailshere. Table4.Theprobabilitydistributionofeach inputparameter. Factor Parameter ProbabilityDistribution X2 T(h) N[0.677,1.8892] X3 T(h–1) N[0.588,1.7992] X4 T(h–2) N[0.494,1.6762] X5 T(h–3) N[0.358,1.5392] X6 RH(h) N[−9.424,10.3792] X7 RH(h–1) N[−8.882,9.8222] X8 RH(h–2) N[−8.038,9.1782] Weimportedtheforecastdata for Julybeforeandafter thecorrection into theSVMmodel for load forecasting. Bothof theresultsarecomparedwiththerealcoolingloadsimulatedbytheDesignBuilder, whichareshowninFigure10. 223