Fault Identification and Location System for 11 kV Distribution Line Using Magneto-Resistive Sensing

Abstract

Timely fault detection and location are of paramount importance in ensuring the reliability of the power distribution system. While there are many methodologies for detecting and locating faults, timely, accurate, and inexpensive fault location is necessary. This paper explores the use of magneto-resistive sensing to develop a system for fault detection and location on an 11kV distribution line as an improvement over systems that are already being used. To achieve this, fault data and load profile were collected from Dutse Alhaji Injection Substation and used to model a system with different fault scenarios, using Convolutional Neural Network (CNN). The modelled base case network was accurate and corresponded to the theoretical result. Eighty percent of the data was used to train the system, and twenty percent was reserved for testing. A feedback loop was implemented in real-time performance, incorporating operator feedback. The magnetic field at fault was transduced to an electrical quantity. The proposed system was able to detect fault location using the corresponding electromagnetic field, which varied inversely with the distance of the fault. Results were analysed and validated. The error was 4.98%.