Seite - 387 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1138 Table10.AverageandmaximalMAPEsoftheproposedandcurrentmethodsforshort-termloadforecasting. Method BPNNs SAEs RNNs LSTM ProposedMethod AverageMAPE 26.12% 23.23% 16.85% 13.26% 10.98% MaximalMAPE 32.42% 28.51% 20.97% 17.45% 15.12% Table 11. MAPEs of compared methods for nine customers’ short-term load forecasting on 30November2013. Customer Category Feeder BPNNs SAEs RNNs LSTM ProposedMethod 37148000 1 2 26.96% 23.28% 16.56% 13.09% 10.80% 51690000 1 7 27.56% 25.61% 15.56% 10.77% 9.25% 37165000 1 7 28.55% 24.81% 13.35% 14.55% 11.12% 53990001 2 2 24.23% 22.23% 17.87% 11.27% 10.07% 54265001 2 3 26.33% 27.56% 17.23% 12.22% 11.91% 54265002 2 3 29.23% 24.89% 15.63% 14.63% 10.76% 31624001 3 35 30.45% 22.56% 19.93% 12.29% 13.56% 41661001 3 34 32.23% 25.65% 16.72% 14.65% 12.23% 76242001 3 33 25.36% 23.22% 14.66% 13.27% 9.98% Indetail, the forecasting loadcurvesofCustomers37148000and53990001on30November2013 basedonthesemethodsareshowninFigures16and17.Wecanobserve theclosest curve to theactual curve is theproposedmethodintheresultsof theseexperiments. Timeinformation is important in short-termload forecastingwhich theBPNNsandSAEscannotextract. Therefore, theygetpoorer performance in theexperiments. Thevanishinggradientproblemlimits theperformanceofRNNs becauseof thedecreasingperceptionofnodes. Thearchitecture is simpler and theparameters are fewer inGRUneuralnetworkcomparedtoLSTMnetworks (Section2.1). Therefore, theperformances ofGRUneuralnetworksarebetter than theothercurrentmethods. Ingeneral, theavailabilityand improvementof theproposedmethodareprovenbythereal-worldexperiments. 6DPSOLQJ 3RLQWV 6DPSOLQJ 3HULRG PLQXWHV $FWXDO 3URSRVHG PHWKRG /670 511V 6$(V %311V Figure16.The loadcurvesofCustomer37148000basedontheproposedmethodandtheothercurrentmethods. 387