Xu Weight Is All You Need! Short-Term Power Load Forecasting Based on a Novel Adaptive Feature Selection Method and Xu Weight

As the new power system evolves, the electricity spot market demands increasingly higher accuracy in short-term power load forecasting. To address this, a short-term power load forecasting method based on a novel adaptive feature selection method and XU Weight is proposed. Firstly, after clustering historical data and selecting a model for each cluster, the novel adaptive feature selection method flexibly identifies key meteorological features based on the chosen model. Secondly, after training each model using the selected important meteorological features, the XU Weight method proposed for the first time, grounded on two hypotheses, leverages sample sizes and the similarity between the predicted time and historical samples, applying a combination prediction approach for short-term load forecasting. The proposed method is validated using actual load data from Hubei Province, China. Experimental results demonstrate that the novel adaptive feature selection method outperforms existing feature selection approaches in terms of accuracy. Additionally, the XU Weight method is shown to be both effective and generalizable in addressing the issue of low prediction accuracy caused by the predicted time belonging to a small-sample cluster.