A nonlinear mathematical model for the optimal operation of a parallel pumping station group was established with the objective of minimizing the operation costs of the station group considering the target quantity of water extraction and flow unevenness between units of each station as constraints. The original model is decomposed into several sub-models with a single station multi-objective optimization operation with the target water lifting capacity of a single pump station as the coordinating variable. This constructed model was solved using a large-scale system decomposition dynamic programming aggregation method based on sub-system multi-objective genetic algorithm optimization. Taking the Jiangdu parallel pumping station group in the Chinese East Route of the South-to-North Water Diversion Project as a case study, the results show that under the condition of 80% water lifting load of parallel stations and 7.8 m daily average lift, the unit water lifting cost of the optimal operation of each station decreases by 4.81%, 4.81%, 19.83% and 11.06% compared with the constant speed operation at the specified angle. The unevenness of the flow of each station is 2.16 m³/s, 2.16 m³/s, 0.60 m³/s and 14.10 m³/s. The erosion of the outlet pool is small. This article provides theoretical reference for the optimal operation of the same type of large-scale inter-basin water transfer parallel pumping station groups.