Research on cooperative scheduling strategy of wind-solar-compressed air energy storage system

Efficient energy storage scheduling technology has become crucial for ensuring grid stability and enhancing system economy as the increasing proportion of renewable energy in the energy structure. This study proposes a data-driven dispatch strategy for compressed air energy storage (CAES), aimed at achieving the dual objectives of combined cooling, heating, and power (CCHP) and reducing operational costs. First, a dynamic modular approach is employed to model the CAES system, accurately representing the physical characteristics and coupling relationships of its components, thereby minimizing mutual interference among them. Secondly, to address the inadequacies of traditional control strategies in adapting to dynamic environments and their low efficiency in energy management, this research introduces an improved model-free deep reinforcement learning (DRL) algorithm—TD3-AC. This algorithm innovatively combines self-attention mechanisms and behavior cloning models on the basis of the traditional twin delayed deep deterministic policy gradient (TD3) algorithm, significantly enhancing the stability and robustness of energy dispatch. Experimental results demonstrate that, compared to traditional dispatch methods, the proposed algorithm successfully reduces operational costs by 8.3 % and improves dispatch accuracy by 29.13 %. This achievement greatly enhances the system's stability and economy, providing an innovative technical pathway for the intelligent management of large-scale energy storage systems.