Regression-Based Machine Learning Approaches for Capacitor Condition Monitoring in Boost PFC Converter

Condition monitoring is essential for assessing aging indicators and identifying the end-of-life stages of critical components in power electronic converters from a reliability perspective. This study focuses on estimating the aging indicators of DC-link capacitors with a balance of accuracy and cost-effectiveness by leveraging readily available converter data. The objective is to enhance system reliability, reduce maintenance time, and streamline troubleshooting processes. Various methodologies exist for monitoring capacitor conditions; however, this article employs machine learning techniques, specifically regression algorithms, to estimate capacitor aging indicators with high precision while maintaining explainability. In this study, a simulation of the DC-link capacitor in a Boost PFC converter was conducted using the Simulink environment in MATLAB. The resulting simulation data were utilized to train a regression model, which achieved a high level of accuracy in estimating the electrical aging indicators of capacitors.