A centrifugal-gear combined fuel pump is a new type of engine fuel pump, and the performance of the centrifugal stage is crucial for the reliable operation of the combined fuel pump. This paper presents a multi-objective optimization of the centrifugal impeller of a combined fuel pump using the higher-order response surface method and NSGA-II algorithm. First, a turbulence model used in numerical simulation is validated and established through a combination of physical test and simulation calculation. Second, the complexity of design parameters is reduced by the significance analysis of full factorial screening model, and the surrogate model of efficiency and head of centrifugal stage of combined fuel pump is constructed using Latin hypercube sampling and high-order response surface method. Finally, multi-objective optimization is performed via NSGA-II algorithm, and the optimization results are validated by numerical simulation. The research findings show that the blade angle at leading edge, blade angle at trailing edge and wrap angle determined by the significance analysis of the full factorial screening model have a significant influence on the efficiency response value and are therefore the key sensitive parameters for optimization. The surrogate model of head and efficiency constructed based on higher-order response surface method demonstrates high calculation accuracy, with sample standard deviations of 0.132 and 0.189 respectively. At the design point, the head and efficiency of the optimized centrifugal stage of the pump are increased by 2.25% and 2.54% respectively, and the internal flow becomes more even, and the adverse flow is significantly alleviated.