Power load forecasting method based on mamba-2 model

With the acceleration of urbanization process, the load characteristics of urban power system increase, and the system structure changes accordingly. This presents a new challenge to accurately grasp the variation trend of urban power load. LSTM, Transformer and other models perform well in traditional time series forecasting tasks, but their high resource consumption and low computing efficiency limit their application in practical scenarios. Therefore, this paper proposed S-Mamba2 model for power load forecasting task, combined with selective state space model (SSM) and optimized it, adjusted the model structure to capture the correlation between variables and time dependence, so as to process multi-dimensional complex power load series data and achieve accurate and efficient prediction. The experimental results show that S-mamba2 has better forecasting performance than LSTM and Transformer models, with the highest load forecasting accuracy of 90.72%, which proves the feasibility and effectiveness of the proposed method. It is expected to provide technical support for power system maintenance and power dispatching optimization.