Three-phase NPC-type grid-connected inverter has been widely used in the renewable energy generation systems with the advantages of high efficiency, high power and low cost. Finite control set model predictive control (FCS-MPC) has strong robustness and fast dynamic response, so it is widely used in the control system of grid-connected inverter. However, the traditional FCS-MPC bears with irregular switching state, which will reduce the current tracking accuracy, accidentally produce high current ripple and electromagnetic noise. Aiming at the above shortcomings, an efficient and fixed operation frequency model predictive based on the discrete space vector modulation (MPC-DSVM) is presented in this paper. Firstly, the mathematical model and voltage vector of NPC-type converter are given according to the principle of deadbeat control. By applying the virtual vectors which are linear synthesized with the real vectors, the MPC-DSVM method can output more voltage vectors with fixed switching frequency. The results obtained show that the current tracking accuracy is improved, the current distortion and imbalance of neutral point potential are reduced.