| Abstract | The Modular Multilevel Converter is gaining wide acceptance in both industrial and research communities for medium and high power applications. However, the increasing need for a higher number of active power switches raised more reliability issues. In this study, detection and localization of open circuit faults are studied in depth. Two multivariate statistical techniques, namely the principal component analysis and the kernel principal component analysis are proposed for fault detection and localization. To study the effectiveness of the proposed methodologies, two converters with different sizes and control techniques are considered in the simulations. The results show that both techniques are capable of accurately detecting the anomalies that can disrupt the normal behavior of the converter. However, only the kernel principal component analysis, can efficiently localize the faulty cells. |
