Advanced Condition Monitoring of Metal Oxide Varistors (MOVs) in DC Circuit Breakers Using Unsupervised Machine Learning Methods

Metal Oxide Varistors (MOVs) are essential components in DC circuit breakers, providing critical voltage clamping and surge protection. However, MOVs degrade and age over time due to repetitive energy surges and thermal stress, eventually leading to catastrophic failure. This paper presents an advanced online machine learning framework for real-time MOV condition monitoring, leveraging unsupervised learning techniques. Utilizing Gaussian Mixture Models (GMM), the proposed approach achieves 95.2% failure detection accuracy and precise remaining useful life (RUL) estimation. By combining MATLAB simulations with experimental MOV degradation data, this framework demonstrates its viability as a robust solution for early failure prognostics and proactive maintenance, significantly enhancing the reliability of DC circuit breakers and the power systems in general.