Understanding electricity prices beyond the merit order principle using explainable AI

Machine learning reveals electricity prices are more complex than traditional economic models suggest: The study found that a machine learning model substantially outperformed the standard "merit order principle" model used in market predictions, indicating that real-world electricity pricing involves factors and interactions that simplified economic models miss.

Renewable energy affects prices differently than expected: While both wind and solar generation influence prices as anticipated, wind power has a notably stronger effect on electricity prices than solar power does, suggesting energy policy and grid planning should account for these asymmetric impacts.

Fuel prices create hidden interactions that matter: Fuel costs rank among the most important price drivers, but they don't act in isolation—the analysis revealed "nontrivial dependencies" and "strong interactions" between fuel prices and other market factors, meaning policymakers cannot predict price impacts by analyzing fuel costs alone.

Grid inflexibility during rapid changes drives prices up: Large generation ramps (sudden changes in power supply) correlate with high prices, particularly because nuclear and lignite plants cannot quickly adjust their output, creating bottlenecks that increase costs—a finding relevant to planning grid transitions away from inflexible power sources.

Explainable AI makes energy markets more transparent: By using SHAP analysis to break down which factors drive prices and how they interact, this approach provides regulators and stakeholders with interpretable insights into market dynamics, supporting more informed policy decisions about electricity market design.