Article detail · 2026
Gaussian Process Regression Based Condition Monitoring of DC-Link Capacitors in Three-Phase AC/DC/AC Converters
- Year
- 2026
- Type
- article
Data source split
- YÖKSİS YÖKSİS article record
- YÖKSİS venue IEEE Transactions on Power Electronics
- Catalog match (ISSN) IEEE Transactions on Power Electronics
- OpenAlex OpenAlex enrichment (abstract, citations, topics)
Abstract
English (OpenAlex)
While dc-link capacitors (DLCs) are one of the core components of power electronic converters, their failure can cause unexpected downtime, costly maintenance, and even safety hazards. Therefore, monitoring their health condition is essential. This article proposes a Gaussian process regression (GPR) based condition monitoring (CM) method for DLCs in three-phase AC/DC/AC converters. The proposed method has three merits, which make it more suitable for practical applications. 1) Unlike traditional approaches, an additional hardware or a signal injection is not required. 2) It can be applied during the normal operation of the converter under different load levels and output frequencies. 3) It is possible to quantify the uncertainty associated with each prediction, which is often overlooked in similar machine learning based methods. Moreover, the experimental results on a laboratory prototype showed that the proposed method can achieve an average prediction error less than 2% and remains effective even at low sampling rates.
Topics
- Machine Fault Diagnosis Techniques
- Advanced Battery Technologies Research
- Multilevel Inverters and Converters
Primary topic Machine Fault Diagnosis Techniques