Short-Term Load Forecasting of Integrated Energy Systems Based on Deep Learning

Abstract

Load forecasting is of great significance for the safety and economic operation of integrated energy systems. In this paper, a combined short-term load forecasting method of electric, thermal and gas systems based on deep learning is presented. Firstly, the deep learning architecture, which consists of a deep belief network(DBN) at the bottom and a back-propagation(BP) network at the top, is introduced. As an unsupervised learning method, the deep belief network extracts abstract high-level features, and the multitask regression layer is used for supervised prediction. Then, a two-stage load forecasting system with offline training and online prediction is established, and the indexes to verify the prediction accuracy of the model are presented. Finally, the effectiveness of the multi-load forecasting method is verified by the actual data of an integrated energy system. The results show that the proposed deep learning algorithm has excellent performances in both computational efficiency and prediction accuracy.