Transient power grid phenomena classification based on phase diagram features and machine learning classifiers

Electrical transmission lines are the most significant part of a power systems in terms of their spread and length with respect of other components. With their huge development due to the growing demand, network losses are an issue that needs the permanent attention of power network providers and distributors. The difficulties of predictive maintenance of power grids are related to the detection of early warning indicators of weaknesses of electrical cables. These indicators might be defined by partial discharges, corona effects, electrical arcs, all these phenomena being characterized by transient signals propagating in cables. Identifying such signals can be very helpful to assess their sources usually located in the weak parts of the grid. In this paper, we present a new approach for the detection and characterization of these types of transient phenomena in power grid using the phase diagram domain. The extracted features are classified using Support Vector Machine, Naïve Bayes and k-Nearest Neighbors. The experimental results indicate that the proposed method provides interesting results in the classification of real-life power grids signals, being a potential solution for predictive maintenance of electrical cables.