Optimal regression tree based rule discovery for short-term load forecasting

This paper proposes a new method for discovering rules in short-term load forecasting. The method is based on a hybrid technique of the optimal regression tree (ORT) and an artificial neural network (ANN). ORT contributes to clustering input data while ANN is used to predict one-step ahead loads. Short-term load forecasting plays an important role to smooth power system operation and control. As a result, more exact models are required to handle it appropriately. This paper puts an emphasis on clarifying the nonlinear relationship between input and output variables in a prediction model. As a prefiltering technique, ORT is used to discover some rules from actual data. To enhance the accuracy of the regression tree, tabu search is used to solve a combinational problem of the ORT structure efficiently. This paper applies ANN to data classified by ORT. The proposed method is demonstrated with actual data