Decision Tree Technique of High Impedance Fault detection in Distribution Feeder

Volume 4 Issue 3 August - October 2016

Research Paper

Decision Tree Technique of High Impedance Fault detection in Distribution Feeder

N. Rajasekhar Varma*, Sanker Ram**

* Associate Professor & HOD, Department of Electrical and Electronics Engineering, RSS College of Engineering and Technology, Anantapur, Andhra Pradesh, India.

** Professor, Department of Electrical and Electronics Engineering, Jawaharlal Nehru Technological University, College of Engineering, Hyderabad, Telangana, India.

Varma, N. R., and Ram, B. S. (2016). Decision Tree Technique of High Impedance Fault detection in Distribution Feeder. i-manager’s Journal on Power Systems Engineering, 4(3), 12-27. https://doi.org/10.26634/jps.4.3.8268

Abstract

Distribution lines subjected to vagaries of nature are prone to metal fatigue, which results in snapping of live conductors. These broken conductors, upon contact with earthed objects or, among themselves, cause High Impedance Faults (HIF) which are difficult to be detected, based on system parameters. The faults cause arcing grounds, which are prone to fire hazards, leading to loss of life and property. This paper presents a novel method of HIF in distribution feeders, using Wavelet and Decision Tree approaches. Wavelet approach has proved to be successful in understanding and evolving solutions to many problems in Power Quality, Power System Protection, and Transient Analysis. The technique adopted, uses Wavelet Transform (WT) in the pre-processing stage for feature extraction, which is used to prepare the necessary data, to be used in the Decision Tree. The current waveform, when measured at the relaying point, yields coefficients which are used as the inputs to the decision tree. A realistically developed HIF model using a typical IEEE 13 Radial Distribution System was used to determine the performance of the technique for different types of HIF and Capacitor switching, linear faults and non linear load switching, etc. The method was found to be robust, fast, and accurate.