Short-Term Load Forecasting Based on SARIMAX-LSTM
Feng Sheng, Li Jia
Abstract
Load forecasting has been the focus of energy management system. In recent years, in addition to some traditional time series forecasting models, with the continuous development of machine learning, many models based on deep learning can also be applied to load forecasting. Different from the existing work, a hybrid model of SARIMAX-LSTM is presented in this paper, in which the SARIMAX model fits and predicts the data, obtains the fitting residual and prediction results, and then uses the LSTM network to predict the prediction error of the SARIMAX model, and modifies the prediction results of the SARIMAX model. In this paper, taking the actual load time series of a city as experimental data, this model is compared with SARIMAX model, LSTM model and SARIMAX-BP model. Experiments show that the model can be well adapted to short-term load forecasting and has the best forecasting effect.