Abstract: As the low-cost nature of renewable energy has reduced the operating costs of power systems, its inherent uncertainty has significantly increased peak shaving costs. Therefore, establishing an effective cost allocation mechanism for peak shaving has become increasingly important. In this context, energy storage, with its excellent peak shaving capabilities, has become an indispensable key element in system investment. Analyzing the distribution of system investments and capacity evolution involves complex interactions among various stakeholders on the generation, grid, and load sides, making system dynamics an effective tool for addressing such issues. A peak shaving cost allocation mechanism based on the marginal peak shaving contribution of units is proposed, effectively evaluating the peak shaving contributions of various units. Additionally, since typical system dynamics models cannot achieve the multiple iterations required for marginal contribution calculations at a single time node, a two-level modeling approach based on system dynamics is designed. The upper-level model analyzes the interaction between energy storage and other system components and its impact on overall investment decisions, while the lower-level model designs a cost allocation mechanism based on the marginal peak shaving contributions of units and energy storage facilities, thereby deriving long-term investment strategies for grid participants. This approach addresses the limitation of general system dynamics models in performing iterative calculations at a single time node. Case studies validate the scientific validity and effectiveness of this method in analyzing energy storage investments.