A Generalized Switch Fault Diagnosis for Cascaded H-Bridge Multilevel Inverters Using Mean Voltage Prediction

Abstract

A cascaded H-bridge multilevel inverter is an indispensable part of high-power industrial drives. The multilevel inverters have an increased chance of switch fault as a consequence of a considerably large number of semiconductor switches associated in the power conversion. A quick and precise fault detection–isolation strategy improves the system reliability as well as reduces the shutdown time of the drive. This article presents a novel generalized open-switch fault-diagnostic approach for an $N$ -level cascaded H-bridge multilevel inverter. In this detection technique, half-cycle mean values of bridge voltages, which are calculated for positive and negative half cycles individually, are used as fault identification features. These means under open-switch fault are predicted from the reference half-cycle means and compared with those measured values to locate the open-switch fault. This quick detection scheme can identify the faulty switch within one fundamental period of output voltage for the cascaded inverters having different number of voltage levels. The computation requirement is minimum, since this approach does not need complex calculations or domain transformations. This strategy can efficiently locate the faulty switch of the cascaded inverter working with various level-shifted pulsewidth modulation schemes, different loading conditions, modulation indexes, and switching frequencies. The potency of the fault classification algorithm is verified through simulation and experimentation.