Aiming at the buffeting problem of traditional integral SMC of PMSM, a predictive time-varying sliding mode control of permanent magnet synchronous motor with improved sand cat swarm algorithm was proposed. Firstly, an improved double power reaching law is proposed, which solves the contradiction between buffeting and rapidity effectively. The idea of prediction is introduced into the SMC, and the optimal time-varying sliding mode surface at the current time is selected online by predictive control, which improves the convergence of the global stage of sliding mode control. Then the extended state observer is used to observe the unknown interference of the system and to compensate it by feed forward, which improves the robustness of the system. Finally, in order to reduce the difficulty of manual parameter adjustment, an improved sand cat swarm algorithm is proposed to optimize each parameter of the controller off-line, which effectively improves the parameter accuracy. Simulation and experimental results show that the control strategy has strong robustness and excellent dynamic and steady state performance.