Fault Detection in Modular Multilevel Converter using Principle Component Analysis

Modular multilevel converter is becoming a great candidate for high power applications due to its valuable features. However, its higher performance comes at the expenses of higher complexity, which increases the risks of failures. Power switches can be accounted among the most vulnerable components in this converter and their failure can lead to the regular functioning destabilization and eventual shutdowns. Failures in power switches can be induced by either open or short circuits faults. This paper presents the principle component analysis, which is a data-driven approach, to tackle the open circuit faults detection in modular multilevel converters. The results were compared to the performance of a model-based technique, the sliding mode observer, and showed a better performance of the adopted approach.