Short-Term Load Forecast for Energy Management System Using Neural Networks with Mutual Information Method of Input Selection

Energy management system (EMS) provides efficient planning and management of power balance. The load forecast is one of the important tools for EMS. The short-term load forecasting (STLF) has a significant impact on efficient operation of EMS. Particularly, the accuracy of STLF can directly impact to EMS operating performance. This work applies the nonlinear feature selection called Mutual information (MI) method to select significant features of input to improve model performance of Artificial neural network (ANN) and Recurrent neural network (RNN). Then, we compare the STLF with Mae Hong Son data set using the ANN and RNN models. Root mean squared error (RMSE) and Mean absolute error (MAE) are used to determine STLF performance. To illustrate the effectiveness of the improved STFL models, we compare performance between models with MI method and models using stepwise regression method.