Optimization of multiple alkaline water electrolyzers coupled with solar photovoltaic power for green hydrogen production on a large scale

Abstract

The integration of intermittent renewable energy with alkaline water electrolyzer (AWE) for green hydrogen production faces challenges such as frequent start-stop cycle, uneven power distribution, and operational inefficiency. To address these issues, this study proposes a collaborative optimization framework combining a hybrid long short-term memory-convolutional neural network (LSTM-CNN) for solar photovoltaic (PV) power prediction and a discrete particle swarm optimization (DPSO) algorithm for dynamic scheduling of multiple alkaline water electrolyzers. The LSTM-CNN model achieves high-precision 15-min interval solar PV power forecasting, with root mean square error (RMSE) values as low as 0.106–2.657 and coefficients of determination ($R^2$) exceeding 0.95 under various weather conditions. The DPSO algorithm optimizes the states of electrolyzers by setting fitness functions and penalty functions or forcing solution to comply with the constraints. Experimental results demonstrate that the proposed method respectively achieves solar PV power utilization rates of 98.63 %, 96.11 %, and 94.13 % under weathers of Type I, Type II and Type III, outperforming simple start-stop (98.311 %, 96.021 %, and 86.594 %) and rotation algorithms (93.242 %, 96.021 %, and 86.62 %). The operational time discrepancies among electrolyzers are reduced to within 1–5 h, effectively mitigating overload risks and local hotspot effects. Furthermore, the algorithm minimizes daily start-stop cycle to once or twice under different weather types while maintaining uniformly power distribution. This work provides a scalable strategy for dispatching of multiple alkaline water electrolyzers, enhancing renewable energy utilization efficiency and supporting large-scale green hydrogen deployment toward carbon neutrality goals.