Permanent magnet synchronous motor (PMSM) was a highly nonlinear dynamic system, which played an important role in industrial driving applications. Real time identification of moment of inertia was of great significance to the control and stability monitoring of high precision PMSM systems. However, traditional identification methods lacked accuracy. Based on the traditional discrete model reference adaptive theory, the Cauchy mutation particle swarm optimization (CMPSO) algorithm was used to replace the adaptive law design of parameters to be identified to realize the identification of PMSM moment of inertia. This method made full use of the fast and efficient convergence of CMPSO algorithm. The simulation and experimental results proved the feasibility, correctness and accuracy of the method.