Multi-Scale Pyramid Networks and Dynamic Graph Learning for Short-Term Power Load Forecasting

Short-term power load forecasting is a crucial task in power system planning and operation, which is directly related to the stability and economy of power supply. Although several methods have been applied to this field, accurate forecasting still faces challenges, especially when dealing with complex spatiotemporal relationships. In this paper, we propose a short-term power load forecasting model based on the combination of multi-scale pyramid network and graph learning module. The multi-scale pyramid network can effectively capture multi-scale temporal features in the load data, while the multi-temporal graph generator further improves the prediction accuracy by constructing a map of spatiotemporal relationships among the load data. Experimental results on several public datasets show that the model outperforms both traditional methods and some of the latest deep learning models in terms of prediction accuracy and generalization ability. In particular, the contribution of each module to the overall performance of the model is verified through ablation experiments. The research in this paper not only provides new ideas for short-term power load forecasting, but also provides strong technical support for the optimal operation of smart grids.