Different types of energy storage have differentiated costs and operating characteristics, so they are suitable for stabilizing the volatility of wind, solar and other renewable energy in different time scales. The existing research does not fully consider the differences of stakeholders and optimization objectives, and lacks the research on multiple energy storage planning from the perspective of source and network coordination. To address this issue, this paper proposes a two-level optimization framework for multi-type hybrid energy storage systems. The method first employs Empirical Mode Decomposition (EMD) to decompose the wind–solar output into distinct frequency components. High-frequency components are absorbed on the source side by electrochemical energy storage, while low-frequency components are regulated on the grid side by compressed-air energy storage (CAES). On this basis, a multi-objective optimization model is formulated that simultaneously minimizes source-side costs and, on the grid side, accounts for the storage’s comprehensive performance assessment, economic efficiency, and system stability. The model is efficiently solved via a hierarchical iterative scheme. Case studies show that, compared with conventional single-layer storage configuration, the proposed source–grid collaborative scheme reduces voltage fluctuations and net load fluctuations by 24.86% and 18.62%, thereby significantly enhancing system stability and renewable accommodation capability.