During the operation of permanent magnet synchronous motor (PMSM), parameter mismatches caused by time-varying motor parameters between the real motor and the model of predictive current controller lead to the degradation of the control performance. An improved predictive current control method based on the model reference adaptive system (MRAS) is presented in this paper. D, q-axis current equations in the rotating reference frame are utilized as a reference model in this method. An online identification system of inductance and flux for PMSM based on Popov’s hyper-stability theory is established. The obtained identification parameters are used in the predictive current control model, thus online updating of the control model parameters is achieved. Analysis and experimental results demonstrate that this method improved the performance of predictive current control under the motor parameters variation efficiently.