Design and optimal scheduling of a forecasting-based wind-and-photovoltaic complementary electrolytic hydrogen production system

Hydrogen can effectively address energy shortages while reducing environmental pollution. Herein, for the first time, a complete complementary wind and photovoltaic (PV) hydrogen production system is designed, including an efficient power generation system model, accurate forecasting model, excellent optimal scheduling strategy, and effective catalyst. The off-grid complementary power generation system was implemented on a DC bus. A hybrid forecasting model, including a long short-term network (LSTM), quantile regression (QR), and regular vine copula, is developed. LSTM combined with QR can obtain the margin probability density function (PDF). The dependency of wind and PV sources is established using a regular vine copula, and the margin PDF is combined with the dependency to forecast based on wind and PV sources. A multi-objective optimal scheduling strategy, based on a stacked multilevel denoising autoencoder (SMLDAE), non-dominated sorting genetic algorithm-II (NSGA-II), and deep reinforcement learning (DRL), is proposed. SMLDAE, NSGA-II, and DRL are used to develop a surrogate model, establish the Pareto frontier set, and select the optimal solution, respectively. A three-dimensional hexagonal Co-Mn-S/Ni bifunctional catalyst was synthesized to reduce the power consumption of hydrogen production. This is the first study to consider the dependencies of different energy sources, use DRL to determine an optimal solution, and synthesize cobalt‑manganese sulfide composites with high catalytic performance in green hydrogen production. A 100 MW complementary wind and PV hydrogen production project demonstrates the superiority of our system. The total renewable hydrogen production cost is reduced to approximately $3.1/kg, enabling low-cost, large-scale, and highly efficient renewable hydrogen production.