In order to improve the accuracy and stability of sensorless control system of surface-mounted permanent magnet synchronous motor (SPMSM), a fast super twisting sliding mode model reference adaptive system observer (FSTA-SM-MRASO) with linear correction is proposed to solve the problem of insufficient robustness of traditional PI control. The proposed observer combines the fast super twisting algorithm (FSTA) with the model reference adaptive system (MRAS) observer to construct the sensorless vector control system of SPMSM based on FSTA-SM-MRASO. The effectiveness of the proposed strategy is verified by simulation on the MATLAB/Simulink platform.