Seite - 224 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 (a) (b) 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1000.00 1100.00 Time(hour) Real Load P2˄Forecast weather data) 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1000.00 1100.00 Time(hour) Real Load P3(Revised weather data) Figure 10. In (a), P2 is the forecasting load adoptingweather forecast data, and in (b), P3 is the forecasting loadadoptingtheweather forecastdatadealtwithMCM. Ascanbeseenfromthefiguresabove, thepredictionresultsofP2andP3arestilldifferent from theactual loadatsomepoints,butP3 iscloser to thereal loadfromtheoverall level thanP2,especially from10thto24thJuly. Thisproves that theuncertaintyanalysisofweather forecastdata forcooling loadforecastingbasedonMCMisbeneficial to improvetheaccuracyof loadprediction. According to the results of load forecasting under the two scenarios, the evaluation of the prediction performance is shown in Table 5. As can be seen from the table, the 24-h-ahead load forecastingmodelbasedonSVMhasgoodpredictionaccuracy.WeestablishedtheSVMmodeland usedactualmeteorologicaldata for loadforecasting. TheMAPEofP1comparedwith theactual load 224