Artificial Neural Network Based Open-Transistor Fault Detection in Symmetric Cascaded H-Bridge Multilevel Inverter

Cascaded H-bridge inverter is finding increased attention in the industry and is used in high power multilevel inverter (MLI) based drive system applications. The prospect of faults in MLI increases with an increase in number of power semiconductor devices used for switching. Among the numerous faults in multilevel inverter, the power semiconductor switches are more prone to short-circuit (SC) fault and open-circuit (OC) fault. Owing to the applications of multilevel inverters in high power industrial applications and considering the investment costs and the safety aspects, it’s necessary to develop rapid and accurate fault detection (FD) method. Numerous fault detection methods have been developed for symmetric cascaded H-bridge (CHB) MLI. However, the fault identification strategy is limited to the identification of pair of switches (among which one switch is faulty). OC-FD strategy for symmetric CHB multilevel inverter is presented in this paper. Switches are controlled by Modified Phase Disposition (MPD) Pulse Width Modulation scheme. Here, the output voltage of each H-bridge is used as a diagnostic variable to extract the fault features, which are used for artificial neural network (ANN) training. This fault detection method can efficiently and accurately identify the faulty switch of the symmetric CHB multilevel inverter without any misclassification.