Advanced Machine Learning Models for Electrical Fault Detection and Classification in Transmission Lines

With the fast development of power systems, efficient methods for fault detection and classification are needed to maintain the stability, safety and efficiency of the systems. In particular, this paper investigates advanced machine learning techniques for fault detection and classification in transmission lines using supervised learning algorithms, including Support Vector Machines (SVM), Decision Trees, Random Forests, Naïve Bayes and Linear Discriminant Analysis (LDA). A dataset consisting of three phase currents and voltages from MATLAB simulations of normal and faulty system operations are used to study the problem. These models classify faults to be single line to ground, line to line and the three phase faults. Performance evaluation is done by metrics such as accuracy, precision, recall, F1 score, and AUC with respect to ROC analysis. Decision Trees are of high reliability and have good computational efficiency for fault detection among the models. The results presented in this research point to the importance of machine learning in reducing fault diagnosis circuit in order to achieve a more intelligent and resilient power transmission systems.