Machine Learning-Based Classification of Electrical Low Voltage Cable Degradation

Low voltage distribution networks have not been traditionally designed to accommodate the large-scale integration of decentralized photovoltaic (PV) generations. The bidirectional power flows in existing resulting from load demand and PV generation changes as well influence ambient temperature led variations increased leakage current through cable insulation. In this paper, a machine learning-based framework is implemented for identification degradation by using data deployed smart meter (SM) measurements. Nodal are supposed be related conditions (reduction insulation thickness due wear) client net changes. Various learning techniques applied classification nodal voltages according conditions. Once trained comprehensive generated datasets, can classify new network operating points into healthy or degraded condition with high accuracy their predictions. simulation results reveal that logistic regression decision tree algorithms lead better prediction (with 97.9% 99.9% accuracy, respectively) result than k-nearest neighbors (which reach only 76.7%). proposed offers promising perspectives early LV SM