In this paper, a new power control strategy based on a model predictive power control for the grid-tie three-level neutral point clamped inverter is presented. A dynamical model based on the orientation of grid voltage is used to predict the performance of control variables required for the system. A cost function that includes the tracking power ability, the neutral-point voltage balancing and switching frequency reduction is used to achieve the optimal switching state. A proposed selection scheme of control input is introduced to comprehend the stability of the closed-loop system and reduce the computational cost. At each sampling time, only candidate switching state inputs that guarantee the stability condition through a control Lyapunov function is evaluated in the cost function of the MPC algorithm. Thus, the execution time is remarkably decreased by 26% in comparison with traditional model predictive control. Moreover, the dead-time effect is compensated by incorporating its influence in the proposed prediction model. Simulation and experimental results compared with the traditional model predictive control are used to confirm the effectiveness of the proposed scheme.