Next day´s peak load forecasting using an artificial neural network

This paper presents a method of next day´s peak load forecasting using an artificial neural network (ANN). The authors combine the DSC search method (Davis, Swann, Campey search method) with the backpropagation learning algorithm (Bp) to reduce the training time and avoid converging at local minima. The forecasting results by the ANN is as good as human experts´ results and is better than the forecasting results by the regression model. The mean absolute percentage error (MAPE) of next day´s peak load forecasts using this method on actual utility data is shown to be 2.67% in the summer period and 1.52% in the winter period. The MAPE of forecasts using human experts´ experience is shown to be 2.86% and 1.59% in each period. The MAPE of forecasts using the regression model is shown to be 3.09% and 1.74% in each period.