Abstract: Under the strategies of "Carbon peaking" and "Carbon neutrality," the high energy-consuming industries in western China must optimize their energy intensity and industrial structure,vigorously develop new energy systems,and achieve clean transformation.In this context,the uncertainty of the evolution process of electricity consumption,which is influenced by multiple dynamic factors,has increased.Therefore,there is an urgent need to predict the evolution process of electricity consumption in the western region in the context of the "3060" target.Firstly,an index system for predicting electricity consumption in the western region is established based on the carbon peak and carbon neutrality targets.This system helps neural network models adapt to the uncertainty in long-term load forecasting due to political and economic events.Secondly,a corresponding combined forecasting model for electricity demand in western China is proposed.On the one hand,this model integrates the advantages of traditional electricity consumption forecasting improvement models,enhancing the applicability of the combined model to factors affecting carbon peak and carbon neutrality targets.On the other hand,the bial gated recurrent unit and long-short term memory（BiGRULSTM）neural network model proposed in this paper takes full advantage of the bidirectional time series feature extraction capability of the bi-directional gated recurrent unit（BiGRU）model and the applicability of the long-short term memory（LSTM）network to time series prediction,greatly improving the prediction accuracy.Finally,the electricity consumption evolution under different scenarios in the western Ningxia region under the carbon peak and carbon neutrality targets is predicted,and the model is comprehensively evaluated.Based on the prediction results,specific suggestions are provided for achieving the "3060" targets in the local area.