Wind-solar complementary hydrogen production strategy based on coordinated ratio and rotation optimization of heterogeneous electrolyzers

Power fluctuations in off-grid wind-solar-storage hydrogen production systems trigger unstable electrolyzer operation, frequent equipment start-stop cycles, and accelerated equipment performance degradation. To enhance the operational stability and durability of the system, this paper proposes a coordinated strategy combining capacity ratio optimization and rotational dispatching for arrays of alkaline electrolyzers (ALK) and proton exchange membrane electrolyzers (PEM). The cold-start sensitivity inherent to alkaline electrolyzers is fully integrated into system capacity allocation and global operational dispatching processes. First, a convolutional neural network-bidirectional long short-term memory (CNN-BiLSTM) model is constructed to conduct classified prediction of wind and solar power, delivering time-interval prior information to support electrolyzer dispatching. Afterwards, an optimization framework based on K-means-genetic algorithm (K-Means-GA) is developed. With comprehensive consideration of investment costs and start-stop losses, the optimal capacity ratio between ALK and PEM electrolyzers is solved, which realizes a balanced optimum among system economic performance, renewable energy accommodation capacity and equipment loss. In addition, a cooperative greedy optimal dispatching strategy (CO-GA) is designed by introducing the priority principle of hot-state operation range, anti-jitter judgment and local cost gradient decision-making mechanism. Simulation results verify that the proposed method effectively cuts down unnecessary equipment switching and cold-start behaviors, and raises the effective service time and hydrogen production power of facilities. The established optimization framework that accounts for cold-start characteristics can offer solid technical support for the safe, reliable, economical and efficient operation of off-grid renewable energy hydrogen production systems. © 2026 Elsevier Ltd.