Abstract: Aiming at the problems of chattering, slow speed response and large overshoot in the existing traditionalPID control and conventional sliding mode control of permanent magnet synchronous motor (PMSM), in order to improve the dynamic performance of the system, this paper improves a non-singular fast terminal sliding mode control (NFTSMC) strategy based on crow search algorithm optimization (CSA). At the same time, in order to reduce the speed drop caused by interference such as load mutation and so on, linear extended observer (LESO) is used to dynamically identify interference and perform online estimation compensation. The simulation experiment is carried out by building a vector control simulation model based on PMSM rotor flux orientation. The results show that the dynamic performance of PMSM is improved under the control strategy, and the robustness and steady-state accuracy of the system are obviously improved.