Accurate load forecasting for industrial users is crucial segment of promoting the economic operation and green transition of modern power systems, which is helpful to speeding up the process of achieving carbon peak and neutrality goals proposed by China. However, it is challenging to predict the individual industrial load precisely by causes of its significant volatility and randomness. This study proposes a novel short-term load forecasting hybrid model that is integrated of Transformer model and long short-term memory (LSTM) network for industrial users. Transformer model is capable to seize intricate dynamic relationship in long-term temporal data efficiently and accurately, which is used for feature extraction. LSTM is known to be good at time series forecasting, and it is used to forecast industrial users’ loads. The superiority of the proposed method is verified by experiments using three real data sets from different industries in China and Ireland. The analysis results of the three examples show that the proposed model has an obvious improvement in prediction performance compared with the comparison model.

Short-term load forecasting;industrial users;Transformer model;long short-term memory (LSTM)