Seamless Fault-Tolerant Model Predictive Control for Cascaded Full-Bridge NPC Inverters With Cascaded Open-Circuit Faults

Cascaded inverters may suffer from multiple open-circuit faults (OCFs) that occur in a cascaded way, but most existing hardware-based fault-tolerant methods usually activate redundant submodules (SMs) after a long diagnosis process is finished, which leads to significant waveform distortion and power derating. This article proposes a seamless fault-tolerant model predictive control (SFTMPC) for cascaded full-bridge NPC inverters (CFNPCIs) to eliminate this problem and improve the utilization rate of remaining healthy devices. A finite state machine is designed to denote eight different operating stages of the CFNPCI with two cascaded OCFs. A redundant SM without bypass switches is directly activated as soon as an OCF is detected, and it will compensate for the current tracking error in diagnosis stages with higher control frequency than ordinary SMs. After diagnosis finishing, the allocated level and switching state selection method of the faulty SM are accordingly revised based on the diagnosis results, to provide power as much as possible utilizing the remaining healthy devices. The redundant SM is then switched to supplement the missing levels of the faulty SM with the same control frequency as ordinary SMs. Finally, experimental results are reported to validate the SFTMPC.