Abstract: Finite control set model predictive control (FCS-MPC) has drawn a lot of attention in medium-voltage high-power traction drives due to its multi-objective and multi-constraint control capability. The speed of traction motors in railway vehicle varies widely, necessitating maximum DC-link voltage utilization in the high-speed flux weakening region. However, conventional FCS-MPC presents very limited overmodulation capability, resulting in reduced current tracking performance in the overmodulation speed region and inability to achieve six-step operation. To solve this problem, a dual-mode model predictive control method is proposed in this paper. The conventional one-step FCS-MPC is adopted for the linear speed region, while an improved multi-step MPC is proposed for the high-speed flux weakening region, which realizes the traction motor's full speed range operation. First, the current prediction model and the objective function of permanent magnet synchronous motor were established for linear speed range. Then, the current trajectory characteristic of motor in the six-step operation are analyzed, and the multi-step prediction is introduced to calculate average value of the current trajectory. An objective function based on the average value of current is established to determine the optimal commutation time. Furthermore, a control mode switching method with double hysteresis structure is proposed to realize seamless switching between linear control mode and six-step control mode. Next, a calculation model of feedforward flux weakening control is established to derive current references in the high-speed flux weakening region. Finally, the steady-state and dynamic experimental results have demonstrated the effectiveness of the proposed method.