Decision Tree Regressor-Based Approach for DC-Link Electrolytic Capacitors Health Monitoring

Power electronic converters depend on capacitors, which are more prone to failure than other electronic components. One of the most commonly used capacitors in power converters is the Aluminum Electrolytic Capacitor (AEC). As AEC ages, its quality and capacitance gradually deteriorate, and its corresponding value of the equivalent series resistance (ESR) rises. ESR and capacitance of the capacitor are frequently used to determine the operating health condition. Several advanced intelligent techniques are adopted to estimate the state of health of an AEC, many of which require the use of a current sensor on the capacitor branch. The current sensor imposes practical constraints, namely it requires space and introduces an unwanted resistive and inductive effect.In this paper, a Decision Tree Regressor (DTR)-based model, combined with the Short Time Least Squares Prony’s (STLSP) approach, is used to detect the fault. Initially, the STLSP algorithm is used to estimate the ESR, using both the current and voltage across the capacitor. The data provided by the STLSP algorithm, together with the converter output current, provide enough information to train the DTR model to estimate the ESR without the need to use a current sensor inside the converter.