Causal-Oriented Machine Learning for Industrial Load Forecasting in Integrated Energy Systems

The integration of renewable energy has intensified research on source-grid-load-storage systems, where industrial load forecasting remains challenging due to multi-scale variability and complex factor interdependencies. Causal learning addresses these challenges by analyzing causality within complex systems, providing more interpretable insights into the relationships between industrial loads and influencing factors. This study leverages causal learning through a practical, easily integrated causal-oriented machine learning framework for industrial load modeling and forecasting. A probabilistic algorithm is employed to identify causal relationships and quantify the impact of each factor. This causal knowledge is then embedded into a causal-oriented method, enhancing model accuracy, robustness, and interpretability. Simulation results demonstrate that the proposed approach effectively captures causal relationships and significantly improves forecasting performance.