Abstract: The strong load fluctuations caused by the large-scale connection of new energy sources to the grid urgently require pumped storage units to play a role in peak-shaving and valley-filling. This study proposes an optimization scheduling method for peak-shaving and valley-filling operations of pumped-storage units based on mixed-integer linear programming (MILP). The objective is to minimize equivalent load variability and maximize power generation benefits. The complex multi-dimensional nonlinear constraints of pumped-storage units are linearized, and MILP is employed to optimize the operation of pumped-storage units. Taking a pumped storage power station in Hunan Province, Central China as the research object, simulation analysis was conducted using typical scenarios of equivalent load in 2023. The results show that compared with conventional scheduling schemes, the optimized scheduling scheme reduces equivalent load variability by more than 5% in four typical scenarios, with a maximum reduction of 34%. Daily power generation benefits reach more than 302 thousand yuan, with a maximum of 405 thousand yuan, demonstrating significant comprehensive benefits. The detailed modeling of the start/stop conditions of the units and the hydraulic characteristics between units ensures that the number of start/stopcycles for each unit is limited to less than six, thereby ensuring the economic and efficient operation of the units. The findings provide technical support for the optimized scheduling of pumped-storage power systems.