Seite - 16 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 2226 Figure3.ForecastingvaluesofLS-SVR-CQFOAandothermodels forGEFCom2014(Jan.). 7LPH Actualvalue BPNN LSSVRCQPSO LSSVRCQTS LSSVRCQGA LSSVRCQBA LSSVRFOA LSSVRQFOA LSSVRCQFOA 7LPH Figure4.ForecastingvaluesofLS-SVR-CQFOAandothermodels forGEFCom2014(July). Tables 7–9 indicate the evaluation results from different forecasting accuracy indexes for IDAS2014, GEFCom2014 (Jan.) andGEFCom2014 (July), respectively. For Table 7, the proposed LS-SVR-CQFOAmodel achieves smaller values for all employedaccuracy indexes than the seven othermodels: RMSE (14.10),MAPE (2.21%), andMAE(13.88), respectively. ForTable 8, similarly, theproposedLS-SVR-CQFOAmodelalsoachievessmallervalues forall employedaccuracy indexes compared to thesevenothermodels: RMSE(40.62),MAPE(1.02%), andMAE(39.76), respectively. Similarly inTable9, theproposedLS-SVR-CQFOAmodelalsoachievessmallervaluesforallemployed 16