In　the　process　of　model　predictive　control(MPC)　of　off-grid　three-level　NPC　inverters,　the　operation　of　the　inverter　switch　optimization　costs　a　lot　of　time,　so　in　this　paper,　a　new　type　of　BP-MPC　inverter　control　technology　is　proposed.　the　BP-MPC　method　uses　the　BP　neural　network　to　fit　the　MPC　algorithm,　and　uses　the　trained　neural　network　to　control　the　inverter.　This　method　collects　the　control　data　of　the　MPC,　and　uses　the　data　to　train　the　BP　neural　network,　so　that　the　BP　neural　network　can　control　the　inverter　as　MPC.　The　simulation　results　show　that　the　BP-MPC　method　in　this　paper　can　effectively　control　the　NPC　inverter　and　reduce　the　time　of　calculating　the　switch　compared　with　the　MPC.　©　2021　IEEE.