Identifying drivers and mitigators for congestion and redispatch in the German electric power system with explainable AI

The transition to a sustainable energy supply challenges the operation of electric power systems in various ways. Transmission grid loads increase as wind and solar power is often installed far away from the consumers. System operators resolve grid congestion via countertrading or redispatch to ensure grid stability. While some drivers of congestion are known, the magnitude of their impact is unclear, and other factors might still be unidentified.
In this study, we conduct a data-driven investigation of congestion in the German transmission grid that reveals drivers and mitigators and quantifies their impact ex-post. Specifically, we used Gradient Boosted Trees and SHAP values to develop an explainable machine learning model for the hourly volume of redispatch and countertrade. As expected, wind power generation in northern Germany emerged as the main driver. Cross-border electricity trading, especially with Denmark, also plays an important role. German solar power has very little effect. Furthermore, our results suggest that run-of-river generation in the alpine region has a strong mitigating effect. Our results support the idea that market design changes, e.g., a bidding zone split, could contribute to congestion prevention.