Abstract: A large amount of energy is consumed during the pumping station operation, especially for the pumping station that water level of the forebay varied frequently in a wide range, pump units often run deviating from the high efficient area of pump, this results in energy waste. In order to lift water, some energy dissipation devices and flow facilities are used such as main pump unit, inlet and outlet passages, forebay and outlet sump, auxiliary equipments, power transmission and substation facilities, etc. Usually, the source water should be transported in a certain distance by the river or channel to the destination. Hydraulic loss and water loss of the river are integrated into power losses of the water. Further more, the operation performance of the pumping station is influenced. In this paper, hydraulic loss, evaporation loss and leakage loss were all considered. Based on variations of source water level, under certain water level of water transfer destination, the unknown water level, flow rate and river parameters could be obtained by the given parameter data based on energy conservation equation in fluid mechanics. Then, the relationships were determined respectively between demanding discharge of destination and pumping head, and between demanding discharge of destination and pumping discharge. The results showed that the relationships were nearly linear. Applying the linear relationships to the later optimization process in the algorithm, the repeated water level iteration was avoided. Also, the calculating amount and calculating time were reduced greatly. Taking three typical tidal level differences of maximum, mean, and minimum at Sanjisngying Intake of Yangtze River as a case, under certain water level and demanding discharge of Huai'an pumping station downstream, the model was built based on the system concluding main pump units, auxiliary equipments, power transmission and substation facilities, and water transferring facilities. The optimizing goal was the least daily operation cost, and the constraints included source water level, water level and flow rate of destination and each cross section of the river, single machine flow rate, the number of running pumps, and the balance of hydraulic loss and water loss. Also, the model was solved by Simulated Annealing-Particle Swarm Optimization Algorithm (SA-PSO), which had better global and local searching ability. Pump blade angles and the number of running pumps were defined as variables, and the objective function was chosen as the fitness function in the algorithm. The pump assembly performance at some blade angles could be got by fitting or interpolation. The results indicated that when the daily varied range of Sangjiangying water level was small, the operation cost was low because of high Sangjiangying water level and low head of Jiangdu pumping stations. When the daily varied range of Sangjiangying water level was large, it turned out the opposite. The cost could be saved by 0.62%~2.26%, 0.33%~3.26% and 0.22%~0.83%, respectively by adjusting pump blade angles than that of design blade angles. If the water-conveyance performance was considered, the optimal operation schemes were more reasonable. The methods could be applied in optimal operation for other pumping station system.