Machine learning based condition monitoring of a DC-link capacitor in a Back-to-Back converter

The utilization of power electronic converters has increased, and the significance of continuous operation is essential in various applications. Therefore, proper condition monitoring (CM) is vital for power converters to eradicate unpredictable maintenance. However, the existing CM techniques may require additional sensors or injection of controlled voltage to the converters. The following machine learning algorithms, such as a K-nearest neighbors (KNN), Support Vector Machine (SVM), and Naive Bayes (NB), have been proposed to monitor the condition of the dc-link capacitor in a Back-to-Back (BTB) converter. The dc-link voltage is measured, and a wavelet decomposition is employed for the feature extraction. Moreover, the performance index evaluates the efficacy of the different classifiers. Further, different datasets have been considered for the evaluation of the classifiers. From this analysis, it is found that the SVM classifier performs better than others.