The lack of multi-time-scale dynamic coupling and multi-energy flow coordination has limited the economic operation and flexibility of the wind power hydrogen storage hybrid system. This paper proposes an improved neural network collaborative scheduling algorithm. A nonlinear dynamic model including electrolyzer thermodynamics, hydrogen storage tank pressure-temperature coupling and fuel cell dynamic response is constructed; an LSTM-CNN hybrid prediction architecture is designed to achieve accurate prediction of wind power; a two-layer scheduling framework is established, and a chaotic map-initialized CAPSO algorithm is proposed, which realizes rapid convergence of mixed integer nonlinear programming through dynamic inertia weights and boundary mutation strategies. Experiments show that after optimization, the maximum scheduling cost is reduced to 24,300 yuan, the wind abandonment rate is reduced, and the hydrogen energy utilization rate is improved, providing a scheduling paradigm for wind-solar-hydrogen storage systems.
Zhang et al. (Thu,) studied this question.