Distributed Convolutional Neural Network-Based Online Capacitor Aging Detection of DC-Link Capacitors in Power Electronic Converters

The mechanisms of aging in electronic components are crucial to rely on the power electronic converters. and understanding them is crucial, with capacitors being a major contributor to component failures for up to 30%. This paper presents a different approach for condition monitoring which uses a distributed convolutional neural network (CNN). The lifetime of capacitors in the DC-link is assessed through a degradation data model that factors in ESR (Equivalent Series Resistance), capacitance and dissipation factor. In addition, a three-level power electronic converter model is constructed to simulate various levels of capacitance deprivation. This paper provides a thorough summary of the degradation parameters associated with capacitor aging, along with the development of a converter model that incorporates the above-mentioned indicators. Furthermore, distributed CNN-based method is explained, emphasizing its ability to detect the aging of capacitors. The outcomes of the simulations validate the effectiveness of the distributed CNN approach in detecting aging and distinguishing those from other anomalies.