A Transformer-Based Load Forecasting Approach for Power Systems in Cold Regions

With the seasonal fluctuations in energy demand and the growing frequency of extreme weather events in cold regions, the significance of load forecasting for power grids in these areas has become increasingly prominent. This paper proposes a Transformer-based load forecasting model specifically designed for power systems in cold regions. The effectiveness of model is validated through comparisons with Random Forest (RF), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN) utilizing actual grid data. The Transformer model achieves a Mean Squared Error (MSE) of 0.0033 and a Root Mean Squared Error (RMSE) of 0.0575, significantly outperforming benchmark models. The proposed approach demonstrates superior accuracy in capturing the complex interactions within power systems in cold regions, enhancing system stability, energy efficiency, and resource allocation across system components. These results highlight the model’s potential as a robust tool for enhancing load forecasting in power systems within cold regions.