This paper proposes a model predictive fault tolerant control (MPFTC) for five-phase permanent magnet synchronous motor (PMSM) using an improved vector selection. Compared to traditional MPFTC methods, this approach avoids reliance on deadbeat control, effectively reducing computational burden and improving motor drive performance. It enhances current control accuracy by introducing a virtual vector control set and employs an improved vector selection method, using a lookup table to directly obtain reference vectors without complex derivations. This method offers significant advantages in computational load and parameter robustness. Additionally, a two-step optimization strategy is used to determine the optimal vector combination and calculate the duty cycle, with constraints introduced to prevent overflow. The optimization process also avoids reliance on deadbeat control. The feasibility and superiority of the proposed MPFTC method are confirmed through theoretical analysis and validation.