Short-term Load Forecasting Based on Echo State Network and LightGBM

Abstract

At present, the continuous development of deep learning technology provides many new ideas for short-term load forecasting. In order to overcome the limitations of deep learning methods and further improve the accuracy of short-term load forecasting, a load forecasting model based on echo state network (ESN) and light gradient boosting machine (LightGBM) is proposed in this paper. Firstly, two load forecasting models based on ESN and LightGBM are developed respectively in this study. Characteristic data required for forecasting are input into each model and respective forecasts are obtained through training. A weighted combination of the two predictions is then performed using an optimal weighted combination method to determine the weight value of the combination, and the final combination forecast value is obtained. The proposed method is evaluated using open-source real load datasets and the results show that the method can effectively combine the advantages of the two models, incorporating both overall time series perception and effective processing of discrete data. The proposed method demonstrates can improve forecasting accuracy compared to using either model alone.