Due to the deficient-rank of the permanent magnet synchronous motor (PMSM) identification equations, the parameter identification results are not unique and cannot be guaranteed to converge to the true values. In this study, we propose a step-by-step model reference adaptive identification method to identify the stator resistance, d-q axis inductances and permanent magnet flux linkage, effectively addressing the deficient-rank problem. The accuracy and stability of the identification results are enhanced by using the parameter loop iteration, and the influence of initial parameter selection on the identification results is analyzed. The simulation results show that the method can accurately identify multi-parameter of the motor, with good rapidity and steady-state accuracy of the identification results, and reduce the dependence of the identification on the initial values of the parameters.