Abstract: To effectively predict electricity consumption in steel enterprises and improve the sustainability of resource utilization, an electricity consumption prediction model based on the CNN-BiLSTM-Attention model is proposed. Firstly, the features of electricity consumption are extracted by the mutual information (MI) method to optimize the input variables of the model and improve its generalization and robustness. Then, a convolutional neural network (CNN) is introduced to capture the spatial correlation characteristics of multi-feature inputs, while a bidirectional long short-term memory (BiLSTM) network is used to extract temporal variation features. Finally, the extraction of key information is enhanced by the Attention mechanism. Comparison experiments are conducted with the actual electricity consumption data from a steel enterprise, and the results show that the model achieves a mean absolute error (MAE) and root mean square error (RMSE) of 3.984 0 and 7.690 7 on the dataset, respectively, outperforming other benchmark models in terms of prediction accuracy. The results of the study provide valuable insights for energy management, energy saving and emission reduction in steel enterprises.