Renewable energy plays an increasingly significant role in the power generation process. Wind energy, in particular, has brilliant prospects, especially high-power wind energy. The three-level neutral-point-clamped (3L-NPC) converter for power transformed into the grid has proven to be a promising and effective configuration. For this topology, finite control set model predictive control (FCS-MPC) is an effective control option. However, the performance of FCS-MPC, like most model based control schemes, will deteriorate when system parameter (in particular, the inductance of grid side filters) varies. In this work, we propose a robust FCS-MPC scheme with revised predictions. This proposed robust FCS-MPC outperforms the classic scheme at both normal and parameter mismatched conditions, which has been validated experimentally for such systems with a fully FPGA based real-time hardware. Besides, the proposed method enlarges the range of the stability of the control method. The stability of the proposed scheme is derived theoretically and proven to be better in comparison with classic FCS-MPC.