1 Department of Energy Technology, The Faculty of Engineering and Science, Aalborg University, VBN2 The Faculty of Engineering and Science (ENG), Aalborg University, VBN3 Power Electronic Systems, The Faculty of Engineering and Science, Aalborg University, VBN
As the modular multi-level converter is now developing rapidly, especially in high power applications, and the large amount of sub-modules in the modular multi-level converter will increase the probability of failures, fault detection and diagnosis of the sub-modules are of importance for continuous operation and post-fault maintenance. In this article, a fault diagnosis technique is proposed for the short circuit fault in a modular multi-level converter sub-module using the wavelet transform and adaptive neuro fuzzy inference system. The fault features are extracted from output phase voltage by employing wavelet transform under different fault conditions. Then the fuzzy logic rules are automatically trained based on the fuzzified fault features to diagnose the different faults. Neither additional sensor nor the capacitor voltages are needed in the proposed method. The high accuracy, good generalization, as well as the time-saving characteristic is verified by conducting a comparison with the neural network method.
Electric Power Components and Systems, 2015, Vol 43, Issue 8-10, p. 1080-1088