Page - 221 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 Figure6.Thefitnesscurveof theparticleswarmoptimizationalgorithm. According to the above optimal parameter combination, the cooling load forecastingmodel basedonSVMcanbeobtained.Weuse theactualmeteorologicalparameters in Julyas test samples. Then, the forecasteddataareanti-normalized toobtain the load forecastvalue,which is compared with theactual loadsimulatedbyDesignBuilder toverify theaccuracyof theSVMmodel. Theresults of thecomparisonareshowninFigure7. 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1000.00 1100.00 Time(hour) Real Load P1(real weather data) Figure7.Comparisonbetweenthereal loadandforecast loadP1. P1 is the forecast loadadoptingreal weatherdatausingtheSVMmodel. Bycalculation, theMAPEof theSVMmodel is10.74%comparedto theactual situation.Dueto the24-h-aheadloadforecastmodelandthesource limitsofweather forecastdata,webelieve that the modelbasicallymeets the forecastingrequirements. Thenext researchcanbedoneusingthismodel. 221