A Monthly Electricity Load Forecasting Model Integrating Transformer and Industry Mutual Information Graph

Le résumé fourni par la source

To enhance the capability of modeling inter-industry correlations and temporal dependencies in monthly electricity load forecasting, this paper proposes a multi-branch forecasting model integrating Transformer and Graph Convolutional Network (GCN). First, based on historical industry-specific electricity consumption data, an industry correlation graph is constructed using mutual information coefficients to capture potential collaborative patterns among industries. Then, a GCN branch is designed to extract structural dependency features, while a Transformer branch is employed to model temporal dynamics. These two types of features are fused to achieve accurate load forecasting. Experimental results show that the proposed method achieves a minimum mean squared error (MSE) of 0.0636 and a mean absolute error (MAE) of 0.1008, significantly outperforming traditional single-branch models. These results validate the effectiveness of multi-branch fusion and mutual information-driven industry graphs in improving model expressiveness and forecasting accuracy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.