Multicomponent load forecasting of integrated energy system based on deep learning under low-carbon background

Abstract

Abstract In order to support the economic scheduling and optimal operation of integrated energy distribution system, a multiload forecasting method of integrated energy system based on deep learning is proposed. Firstly, Pearson coefficient is used to analyze the correlation between the three loads. Then, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) model is used to improve the hidden layer of recurrent neural network (RNN). GRU and LSTM adopt gate structure instead of hidden unit in traditional RNN structure, which can selectively remember important information, and then learn historical load parameter information efficiently, making the prediction result more accurate. Finally, the actual data of the integrated energy system is applied to verify the effectiveness of the algorithm. The experimental results show that the prediction accuracy of the LSTM-GRU model proposed in this article is more accurate, and the research results can provide a reference for the comprehensive load prediction of the integrated energy distribution system.