A novel approach towards predicting faults in power systems using machine learning

Transmission lines are one of the most important components of a Power System. Overhead transmission lines run for hundreds of kilometres to supply power to the cities. At the same time, these lines are very susceptible to damage due to myriad reasons which include natural causes like particle contamination or technical reasons such as insulation failures, which do not immediately lead to a fault but lead to a phenomenon known as partial discharge (PD). Partial discharge is a localized dielectric breakdown which does not completely bridge the space between the two conductors in a small portion of a solid or liquid electrical insulation system under high voltage stress. Partial discharges slowly deteriorate the condition of the power line and, if left unrepaired, can lead to potential faults. Partial discharges are also a proven way of diagnosing the condition of power cables. This paper describes an approach towards detecting partial discharge signal patterns using machine learning algorithms and hence, predicting potential faults. The model described in this paper was able to detect PD patterns with 97% accuracy.