An Improved Load Forecasting Method Based on Optimal Weighted Combination

Abstract—Load forecasting is the basis for the planning and safe operation of electric power. To satisfy all forecasting conditions as much as possible and improve the results of load forecasting, the combination forecasting method is proposed. The new method integrates the advantages of residual gray theoretical model and back propagation (BP for short) neural network model with additional momentum-adaptive learning rate. And the weighted coefficient optimization process is introduced in the study. Basing on the electricity consumption of 2001-2011of a power supply company, the optimal weighted combination method are applied to forecast and analyze. The simulation results show that the mean square error of optimal weighted combination forecasting method is smaller than that of either a single forecasting method, which can improve the prediction precision significantly. Keywords-load forecasting; grey model; BP neural network; additional momentum-adaptive learning rate; optimal weighted combination I