Optimal control of hybrid wind-storage-hydrogen system based on wind power output prediction

In off-grid wind-storage‑hydrogen systems, energy storage reduces the fluctuation of wind power. However, due to limited energy storage capacity, significant power fluctuations still exist, which can lead to frequent changes in the operating status of the electrolyzer, reducing the efficiency of hydrogen production and the lifespan of the electrolyzer. To cope with fluctuating wind power, it is necessary to exercise reasonable control and adjustment of the electrolyzer operating status and power distribution, and to effectively set the charging and discharging time and power of energy storage. To utilize the energy storage flexibility and improve hydrogen production efficiency, this paper proposes a multi-objective rolling optimization for the electrolyzer control. Firstly, use a long short-term memory network to predict the wind power output in the future time cycle, and based on this, calculate the number of electrolyzer that need to be kept in the startup state within the predicted time cycle. Then, based on real-time wind power output, determine the operating status and power distribution of the electrolyzer, as well as the charging and discharging of energy storage. Ultimately, during the real-time control phase, we implement a multi-objective rolling optimization strategy to dynamically adjust the power distribution sequence. This approach optimizes electrolyzer performance, ensuring seamless adaptation to both the current realized wind power and its accurate future output predictions. The effectiveness of the proposed control strategy is validated by comparison with the classic start-stop control and array rotation control strategy. This article uses SimuNPS for simulation verification, the simulation results indicate that the proposed method achieves fewer start-stop times and higher hydrogen production regarding the operation of electrolyzers, thereby effectively enhancing the wind-storage‑hydrogen system's economic viability. •Wind power predictions is employed to compute the average necessary operating units within the electrolyzer array.•Eelectrolyzers are operated at varying power levels.•A multi-objective rolling optimization is adopted to control the electrolyzers.