Development of Electric Load Forecasting System using Neural Networks

Electric load forecasting in power system is very important for insuring reliability and for economical operation. Especially, daily peak load forecasting and daily load curve for following day are the basic operation of generation scheduling. Therefore, many statistical methods and empirical methods have been conventionally used for such forecasting. However, it is difficult to obtain the accurate forecasting. Moreover, it has been difficult to construct a proper functional model and to investigate factors related to electric load and weather conditions. This paper presents an electric load forecasting system using artificial neural networks. The system forecasts electric load curves for following day and current day using two forecasted loads previously, weather conditions and the information of holiday. The system has been practically used in the central load dispatching center of Tohoku Electric Power Co., since July 1998. The forecasting results show the effectiveness of the proposed methods.