Abstract: The viscoplastic model considering dynamic recrystallization describes the coupling process of macroscopic deformation and microstructure evolution during hot working. It is difficult to measure the material parameters accurately by means of traditional testing methods. A hybrid global optimization algorithm is designed, which combines the strengths of genetic algorithm, Levenberg-Marquardt algorithm, augmented Gauss-Newton method and flexible tolerance method. The square sum of the norm of the difference between the experimental values obtained from upsetting experiment and the calculated values obtained from finite element simulation is defined as an objective function. Taking 26Cr2Ni4MoV as an example, the material parameters are identified by the designed algorithm. The comparison between simulated and experimental results shows that the calculated results are well with the experimental. This indicates that the constructed algorithm can effectively identify the material parameters of the model and the model can describe accurately the evolution of microstructure during hot working.