Machine Learning-Based Condition Monitoring of DC-Link Capacitors in Drive Inverters Using Case Temperature

The proliferation of the electric vehicle (EV) sector has increased the demand for reliability in EV power converters. Therefore, condition monitoring technology for power converters is gaining importance to ensure the safety and availability of EVs. For EV applications, film capacitors are preferred over aluminium electrolytic capacitors due to their higher reliability. However, aluminium electrolytic capacitors are still used in EVs and motor drive applications where cost is a significant factor. Given that dc-link electrolytic capacitors in three-phase inverters suffer from high failure rates, this paper proposes a machine learning-based condition monitoring technology using case temperature measurements. The proposed methodology utilizes low-bandwidth voltage and current measurements, which are already available in inverters for control purposes, in addition to case temperature measurements. The proposed method was simulated in MATLAB and the results are presented.