Innovative Approaches to Long-term Power Load Forecasting with Autoformer

The significance of long-term power load forecasting stems from its profound impact on power system operation and planning, allocation of energy resources, and the overall stability of the social economy. Electric power load is influenced by various factors, including seasonal variations, meteorological conditions, technological advancements, among others. These intertwined factors render long-term load forecasting a complex task characterized by heightened uncertainty. While most existing methods primarily focus on short-term power load forecasting, there is presently a lack of effective approaches for addressing long-term power load forecasting challenges. To tackle this issue, this study uses the Autoformer model for long-term power load forecasting, and uses the Pearson model and random forest to explore the correlation between variables for the correlation degree between each characteristic variable and the power load. The research findings demonstrate that the Autoformer model outperforms other models, delivering the most accurate predictions with the lowest error values.