Seite - 408 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies 2018,11, 242 D H E G F 7KH YDOXHV RI WKH UHVLGXDO HUURUV NZK 7KH YDOXHV RI WKH UHVLGXDO HUURUV NZK 7KH YDOXHV RI WKH UHVLGXDO HUURUV NZK 7KH YDOXHV RI WKH UHVLGXDO HUURUV NZK 7KH YDOXHV RI WKH UHVLGXDO HUURUV NZK Figure 10. Prediction error histograms of the five predictors constructed by utilizing the energy-consumingpattern: (a)hybridDBNmodel; (b)BPNN;(c)GRBFNN;(d)ELM;and(e)SVR. Table3.Theperformancesof thefivemodels for theretail storeenergyconsumptionprediction. Methods DataType MAE(kwh) MRE(%) RMSE(kwh) r R2 MDBN Residualdata 47.71 5.03 76.83 0.94 0.89Originaldata 54.38 5.59 86.43 0.93 0.86 BPNN Residualdata 65.69 7.24 93.38 0.92 0.85Originaldata 75.45 8.20 100.40 0.94 0.87 GRBFNN Residualdata 54.60 5.75 83.87 0.93 0.87Originaldata 52.51 5.62 87.54 0.93 0.86 ELM Residualdata 58.54 6.29 88.62 0.93 0.86Originaldata 78.86 8.34 113.02 0.89 0.79 SVR Residualdata 48.28 5.19 81.31 0.93 0.87Originaldata 52.19 5.42 89.93 0.92 0.85 4.4. EnergyConsumptionPrediction for theOfficeBuilding In thissubsection,firstofall, theenergy-consumingpatternof theofficebuildingwillbeextracted fromtheofficebuildingdataset. Then, theconfigurationsof thefivepredictionmodels forpredicting theofficebuildingenergyconsumptionwillbeshownindetail. Finally, theexperimental resultswill begiven. 408