Driven by the “dual-carbon” goals, this study proposes a two-layer capacity optimization method for compressed air energy storage (CAES) in abandoned mines, addressing both the challenges of renewable energy integration and the resource utilization of abandoned mine spaces. Unlike traditional static capacity designs that rely solely on roadway volume, this approach balances economic performance with renewable energy integration, providing a practical framework for planning energy storage systems in abandoned mines. The proposed model consists of a planning layer and an operation layer: the planning layer seeks overall economic optimality, while the operation layer aims to maximize renewable energy utilization. These layers interact iteratively, and the model is solved using an improved particle swarm optimization algorithm to determine the optimal configuration. Multi-scenario simulations based on the modified IEEE 33-node system show that, compared with traditional fixed-capacity configurations, the proposed model increases renewable energy absorption by an average of 6.53% and reduces total costs by 45.45% across four typical scenarios. The results verify the model’s effectiveness in improving both renewable energy utilization and economic performance.