In view of the traditional flower pollination algorithm for parameter identification in permanent magnet synchronous motors, the late iteration often results in easily plunging into a local optimum, leading to slow convergence speed and the defect of low optimization accuracy. An improved flower pollination algorithm based on t-distributed perturbation and Gaussian perturbation (tGFPA) is proposed to realize high-precision identification of permanent magnet synchronous motors. Firstly, the individual position of flowers is initialized by chaotic logistic mapping, and then t-distribution perturbation is introduced in the global pollination process to improve the diversity of search space. Gaussian perturbation is added during local pollination to enhance the ability to jump out of the local optimal solution. Finally, comparing the simulation results show that the flower pollinate algorithm based on double perturbation strategy to improve convergence speed is faster and have higher precision, which has important significance for improving control performance of permanent magnet synchronous motor.