Abstract: Since model predictive control (MPC) works based on a motor model, the mismatch between the actual parameters of the motor and the parameters of the predicted model will lead to a decrease in the control performance of the system. Aiming at this problem, this paper proposes a least square parameter identification method by using adaptive forgetting factors. This method adjusts the forgetting degree of the old data in the identification process through the changing forgetting factors so that the identification results have fast convergence and can stably follow the changes of the motor parameters. The parameters of the prediction model are corrected in real time by identifying the motor parameters, which can effectively reduce the current and torque fluctuations caused by changes in motor parameters. The control performance and the parameter robustness of the MPC algorithm are both improved. Finally, Matlab/Simulink is used for simulation analysis, and its results verify the effectiveness of the method.