Load Forecasting with Extreme Weather Based on ImTimeGAN-CGBM Architecture

In the case of frequent extreme weather, the increase in equipment such as air conditioning or electric heating leads to a continuous increase in electrical load. Load forecasting is helpful to accurately grasp the load trend under extreme weather and reduce the risk of overload. However, conventional prediction methods are difficult to adapt to extreme weather scenarios due to the influence of extremely high temperatures in summer. This paper proposes a Load Forecasting with Extreme Weather based on ImTimeGAN-CGBM, which includes ImTimeGAN and CGBM modules. First, ImTimeGAN is proposed to expand the small sample data for extremely hot weather, which can effectively reduce the risk of overfitting in model forecasting. Secondly, the CGBM module is used to predict the loads during high temperatures, which significantly improves the accuracy and generalization ability of the model load prediction. This paper conducts the experimental analysis using real electricity load data under extreme weather in 2023 and 2024, and the results confirmed that the proposed model has higher prediction accuracy on real datasets.