Abstract: The river-reservoir connection project in “Mengkaige Area” is a key project for the optimal allocation and unified dispatching of water resources in central cities in southern Yunnan including Nandong Primary Pumping Station, Nandong Secondary Pumping Station and Changqiaohai Pumping Station. Studying the unit commitment of pumping stations will help to reduce the operation cost of the project. Taking the minimum energy consumption of pumping stations as the goal, this paper constructs the optimal unit commitment model of three pump stations, and solves the model by genetic algorithm（GA） and differential evolution algorithm（DE/rand, DE/best）. The results show that the DE/rand algorithm has better solution effect, larger search space and strong stability and is suitable for solving the optimal unit commitment models of river-reservoir connection project in “Mengkaige Area”. The optimal operation process will maintain the minimum number of startup units. When the total flow is at the start-up critical point, the total power will increase sharply. The optimized unit commitment modes improve the operation efficiency of the pump stations and reduce the operation cost, which can provide a reference for the pumping station economic operation of the river-reservoir connection project in “Mengkaige Area”.