The surrogate model was used to replace the computational fluid dynamic (CFD) method to optimize and design rotor airfoil considering dynamic stall characteristics. The unsteady aerodynamics calculation of rotor airfoil based on moving grid technology was established to obtain the lift, drag and torque coefficients under different airfoil shapes. The class shape transformation (CST) airfoil parameterization method was used to conduct fitting and reconstruction of the initial airfoil, and 12 design parameters were selected. The global optimal differential evolution algorithm based on natural heuristic was used to reduce the airfoil’s torque and drag coefficients. The main limiting condition was to ensure that the lift characteristics were not reduced and the airfoil thickness increase was not obvious. The optimization results of the method were compared with those of the adjoint and CFD method. Results show that the optimization method based on Kriging model has better search performance and better aerodynamic performance than the adjoint method in the two-dimensional airfoil optimization. The possibility of prematurely falling into the local best was reduced under the advantage of using the global optimization algorithm compared with CFD method. The comparison optimization results show that the lift characteristics are better when the torque and resistance characteristics are almost the same.