Power distribution systems have undergone several transformations in recent years, including the integration of new load profiles, distributed generation, and the grid expansion. This article presents an approach based on traveling wave theory and artificial neural networks for fault classification and location, which assists the utility by reducing response time and directly impacting system quality and reliability indicators. The proposed method utilizes three- phase voltage and current signals acquired by smart meters installed at the system endpoints. Voltage signals were used to determine the distance from fault using traveling wave theory. In addition, features were extracted from voltage and current signals, which were used as inputs to artificial neural networks, resposible for classifying the fault. The faulty scenarios were simulated using the PSCAD/EMTP software, considering the CIGRE medium voltage test system. The results were very promising. They presented: (i) detection accuracy rate of 100%; (ii) classification accuracy of nearly 100%; and (iii) average errors of less than 1% in fault distance estimation with a mean deviation of 0.4 and mitigation multiple estimation problem of more than 90%.
