For the problem that current in single-vector model predictive current control (MPCC) of permanent magnet synchronous motor (PMSM) fluctuates greatly, an improved two-vector model predictive current control (TV-MPCC) algorithm is proposed. A zero voltage vector and an effective voltage vector or two effective voltage vectors in each sector are used to synthesize a desired voltage vector, and the optimal voltage vector is selected by using the principle of minimization of value function. The action time of the voltage vector is calculated by the dq-axis current dead-beat control method, which can realize dead-beat control of the current. Finally, the model is built in MATLAB/Simulink. The simulation results show that, compared with the single-vector MPCC, the optimized control strategy reduces the current fluctuation and improves the system stability.