Abstract: In order to improve the operation efficiency and head of centrifugal pumps under the design conditions, a centrifugal pump impeller optimization design method based on high-dimensional hybrid model was proposed. A single-stage centrifugal pump with specific speed of 157 was selected as the research object for this study. The optimization variables were parameterized in CFturbo software, and then combined with numerical simulation to obtain the training set of the high-dimensional hybrid model. By using MATLAB machine learning on the obtained training set, a high-dimensional support vector regression model between efficiency, head and optimization parameters was obtained, and a genetic algorithm was used for the optimization. Under the design working conditions, the predicted efficiency and head values of the fitted high-dimensional hybrid model are 1.5% and 3.2 m higher than the original model, respectively. The efficiency and head values of the optimized scheme verified by numerical simulation are 0.9% and 2.1 m higher than the original model, respectively. This case study shows that the application of a high-dimensional hybrid model in the optimal design of a centrifugal pump impeller can achieve rapid optimization and improve the hydraulic performance of the centrifugal pump.