考虑可再生能源发电预测不确定性对优化混合电站的重要性：一种鲁棒的MILP方法

Abstract Hybrid Power Stations (HPS), integrating RES with storage systems, offer a promising solution to increase the penetration of renewable energy sources (RES) by converting intermittent production into dispatchable power. This study underlines the importance of considering the uncertainty in forecasts of power production to improve the management of these systems and proposes a Mixed-Integer Linear Programming (MILP) framework able to account for multiple possible outcomes, instead of a single one. Using a one-year dataset of historical forecasts and actual production, the study develops and tests three dispatch strategies: rule-based, standard MILP using raw forecasts, and robust MILP. The latter considers several production scenarios to enhance the reliability of the dispatch plan, valuable for grids with strict operational constraints. Unlike previous studies in the literature, this work dives into the aspect of tuning the span in which prediction scenarios are generated and the relaxation of constraints of the robust optimization. Sensitivity analyses showed that a forecast scenario span of ± 20 % for wind and ± 30 % for solar, paired with a moderate divergence penalty of €12/MWh, offered the best balance between system reliability and economic performance. The rule-based strategy exported 235.9 MWh/year but suffered from high undershooting, reducing net earnings by 16.3 %. The standard MILP approach improved performance, increasing annual energy exports and reducing undershooting, resulting in net earnings of €213.2 k/year compared to €197.5 k/year for the rule-based strategy. The robust MILP approach further optimized performance, achieving only 21.8 MWh/year of undershooting and net earnings of €223 k/year (only 3.1 % reduction from gross earnings), demonstrating that this strategy can improve the reliability of HPS operation while reducing penalties associated with forecast errors.