Aiming at the problem that mechanical sensors of electric vehicle were easy to fail in complex working environment, the speed sensorless technology based on model reference adaptive system (MRAS) was applied to electric vehicle. In order to solve the problem of large phase delay of rotor position estimation and large speed estimation error in traditional MRAS speed sensorless control, the model predictive control algorithm was applied to MRAS. Permanent magnet synchronous motor (PMSM) current flux linkage equation was selected as the reference model, and voltage flux linkage equation was selected as the adjustable model. The cost function was the difference of flux linkage, and the rotor position was selected as the estimated parameters. Compared with the traditional MRAS speed sensorless control algorithm, the proposed algorithm had more accurate speed and rotor position estimation, less estimation error, and excellent dynamic and steady state performance. The feasibility and effectiveness of the algorithm were verified by simulation and experiment.