Application of Machine Learning in Fault Detection and Predictive Maintenance of Power Transformers

Le résumé fourni par la source

Transformers in which power is converted into and out of power systems are critical in achieving stability, efficiency, and reliability of modern power systems. Being essential resources of the transmission and distribution system, their breakdown may result in considerable economic losses, delays, and even safety risks. Conventional fault detection and maintenance approaches, though valuable, can hardly manage the requirements of growingly complicated smart grid settings where quick choice-making and correct diagnostics are crucial. The faults, which include degradation of insulation, oil contamination, thermal stress, and partial discharges, are still hard to predict through the traditional monitoring systems and, as such, result in reactive but not proactive maintenance processes. Machine learning (ML) has become a disruptive technology to address those challenges, providing high-tech methods of fault detection, condition evaluation, and predictive maintenance of power transformers. With access to historical and real-time data, the power of early fault detection, trend detection, and the prediction of failures is offered by the ML models of support vector machines, artificial neural networks, and various deep learning structures. Also, when paired with digital twins, Internet of Things (IoT) technologies, and explainable artificial intelligence (XAI) frameworks, the interpretability, scalability, and trustworthiness of predictive models can be increased by combining ML and digital twins. This paper will discuss the concept of ML as applied in transformer fault detection and predictive maintenance, with the focus on the types of faults, machine learning techniques, proactive maintenance approaches, difficulties, and perspectives. The conclusions highlight the value of ML-based methods in enhancing asset reliability, lowering operational expenses, and enhancing the digitalization of the power industry.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.