The traditional model predictive control（MPC）of permanent magnet synchronous motors（PMSM）only acts on one voltage vector in a control cycle, so a higher sampling frequency is needed to obtain better control performance. Aiming at the problem, a model predictive control algorithm based on extended vector set is proposed, which can realize the dynamic adjustment of motor control performance and switching loss without changing the sampling frequency. Because there are many voltage vectors in the extended vector set, in order to reduce the calculation, an optimization strategy of voltage vectors based on adjacent vectors is proposed, and the number of candidate voltage vectors in the control set is limited to 11. The simulation and experimental results show that the proposed PMSM model predictive control algorithm based on the extended vector set has the same dynamic performance to the traditional finite set model predictive control algorithm, and the control performance and switching loss can be adjusted by adjusting the number of voltage vectors in the extended vector set without changing the sampling frequency.