Vertical centrifugal pump is a high specific speed centrifugal pump, which is usually with radial vane diffuser structure. As the core power equipment for large-scale irrigation projects and long-distance water transfer, the matching motor power for vertical centrifugal pump is huge and can reach 40MW,and the efficiency directly determines its operating energy consumption. In order to reduce the energy consumption of vertical centrifugal pumps, an optimization on multi-components was proposed based on back propagation neural network (BPNN) and multi-island genetic algorithm (MIGA) . The matching of the hydraulic components was taken into account and the Plackett-Burman test design was used to screen out the optimal design variables from the 10 design parameters of the vane diffuser and the volute. Then, totally 106 sets of cases were sampled by using optimal Latin hypercube sampling (OLHS), and an automatic numerical simulation optimization platform for the vertical centrifugal pump was built to quickly obtain the optimization objective values corresponding to each set of case. The BPNN was used to construct the high-precision nonlinear relationship between the optimization variables and the optimization objective. Finally, the optimal parameter combination of vane diffuser and volute was obtained through MIGA. The results showed that the performance parameters of vertical centrifugal pump could be more accurately predicted by using SST k-ωturbulence model. BPNN was an effective method to construct high-precision nonlinear relationship between pump design parameters and performance parameters. The efficiency of the optimized model under design condition reached 90.21%, which was 3.61 percentage points higher than that of the original model. The optimized vane diffuser and volute had a more obvious influence on the performance of vertical centrifugal pumps under design condition and part-load conditions. The matching between the vane diffuser and other hydraulic components was better, and the internal flow pattern of the vane diffuser was significantly improved after optimization. The optimization method proposed could provide a certain reference for the optimization design of centrifugal pumps.