Robust dispatch of multi-electrolyzer systems for renewable energy hydrogen production under wind forecast uncertainty

To address real-time control failures in renewable energy hydrogen production systems (REHPS) caused by forecasting errors and optimization delays, this paper proposes a rolling horizon framework embedded with a rule-based compensation mechanism to enhance the real-time performance and controllability of optimal scheduling in practical applications. A mixed-integer linear programming model is formulated using multi-state nonlinear electrolyzer modeling and time-of-use electricity prices to optimize the start-stop sequences and power allocation of electrolyzer clusters. By integrating the rule-based compensation within the rolling window, dispatch decisions are effectively corrected to achieve real-time response. Clustering analysis based on a real-world wind power dataset shows that the proposed strategy increases the wind utilization rate by an average of 7.224% and the system hydrogen production efficiency by 5.97%, compared to the best rule-based strategy (S1). Furthermore, the sensitivity analysis of the strategy demonstrates its robustness across varying prediction accuracy levels. The proposed rule-based compensator enables the deployment of a centralized optimizer onto an industrial-grade real-time controller without requiring additional hardware. This solution has been applied in the preliminary design study of a hundred-megawatt-scale REHPS in Northeast China, providing a deployable optimization solution for large-scale renewable hydrogen production.