Predicting Energy Market Imbalance Prices with Random Forest: A Rolling Window Approach

Balancing supply and demand is essential for electricity grid stability, particularly with the increasing integration of renewable energy sources (RES). Accurate forecasting of imbalance prices can significantly enhance grid management and market efficiency. This study focuses on predicting imbalance prices in the Finnish market using a rolling window approach with a Random Forest (RF) model, addressing the variable nature of renewable energy sources (RES). The RF model demonstrated high forecasting accuracy across three cross-validation periods, achieving a highest coefficient of determination (R²) of 0.906 and a normalized root mean squared error (nRMSE) of 2.874% in the first period. When applied to the test set, the model yielded an R² of 0.90 and an nRMSE of 4.29%, effectively capturing non-linear and complex behaviors in imbalance prices. These results indicate that the RF model is a robust tool for imbalance price prediction, aiding market participants and system operators in enhancing decision-making and planning. Future research could explore the model's application in other markets and under various price conditions.