翌日需要曲線予測手法と精度向上の検討

抄録

This paper presents a result of study for a next day electric load curve forecasting method and its accuracy improvement. Electric load curve forecasting is one of the most important tasks for insuring reliability and economic electric power supply. Because start-stop time of generators is decided using the load curve forecasting. This paper proposes three methods. First one is a new structure of forecasting models. It consists of single-output neural networks and a multi-output neural network. Second one corrects load curve using the operation information of commercial-scale utility customers. Last one selects the accuracy forecasting models adaptively using recent forecasting error. Input data of each forecasting model is different type of weather information. The simulation results reveal the effectiveness of the proposed methods for a next day electric load curve forecasting.