978-1-5386-0517-2/18/$31.00 ©2018 IEEE High Impedance Fault Detection and Location in Distribution Networks Using Smart Meters Francinei L. Vieira, José M. C. Filho, Paulo M. Silveira and Carlos A. V. Guerrero Universidade Federal de Itajubá - UNIFEI Itajubá, Minas Gerais, Brazil {nei, jmaria, pmsilveira}@unifei.edu.br carlosvillegasguerrero@unifei.edu.br Marino P. Leite TSE Tecnologia LTDA Itajubá, Brazil marino.piazza@tse-tecnologia.com.br Abstract — High impedance faults (HIFs) present great difficulty of identification and location in distribution networks (DN) due to their characteristics of low current magnitude. Advances in smart grids and distribution automation allow the detection of disturbances that were previously unnoticeable in DN. This work aims to present a new method for detection and location of HIFs from smart meters placed at strategic points in the feeder, using a voltage unbalance based approach. The methodology was evaluated through simulations in MATLAB / Simulink, focusing on the detection of high impedance series faults. The results showed that the algorithm effectively identifies broken conductors, with or without ground faults, located either at the load or source side. Once technical and economic feasibility is proven, these methods can assist energy distribution utilities in restoring the normal operating conditions of the distribution network. Index Terms -- distribution network, high impedance faults, smart meters, voltage unbalance. I. INTRODUCTION There is a constant requirement for electrical equipment manufacturers, transmission and distribution utilities to adopt solutions aimed to minimize the frequency and duration of in- terruptions, as well as reducing costs due to faults. However, despite the investments made in the electrical systems, the dis- tribution network (DN) is susceptible to some faults hard to de- tect. Among these faults, the high impedance fault (HIF) is a specific type of abnormality that traditional overcurrent protec- tion devices usually do not detect. This happens because the current of this kind of fault is confused with the normal opera- tive current on the system [1]. A HIF occurs when an energized conductor meets a surface of considerable resistive value, such as tree branches, civil structures, asphalt, soil, among others. With the difficulty of having effective contact with the ground, the fault current is limited to values lower than those normally detectable by over- current protection. Therefore, detection and identification of these faults is crucial and challenging for protection engineers. The problem of the detection of HIFs is known by the in- dustries of the electric field and several techniques have been proposed in the literature since 1970, for instance statistical hy- pothesis tests, neural networks, 3rd harmonic fault current, de- cision trees, wavelet decomposition, fuzzy logic, among other methods [2]. In [3], the performance of four different algorithms for HIFs detection were evaluated by means of a failure test by stages, using proportional relays, second and third harmonic current re- lays and earth-fault relays. The results showed that some algo- rithms had better performance than others under certain condi- tions. In [4], several measurements of harmonic currents were obtained in situations of high impedance faults in sandy soil of an energy substation, during a week. The objective was to de- termine the harmonic order of the current that could be used for detection of a HIF. The authors concluded that monitoring the second harmonic current could contribute to this objective. In [5], an algorithm for the detection of electric arcs was developed using bus voltages and high order odd harmonics. On the other hand, an algorithm based on discrete wavelet trans- forms, artificial neural networks and statistical analysis was presented in [6]. Such algorithm is used in an Asea-Brown Bo- veri (ABB) relay for detection of HIFs in medium voltage net- works. A HIF detection method was also presented in [7], which measures the phase angle of the third harmonic current in rela- tion to the fundamental voltage. The sampling rate of 32 sam- ples/cycle was used to obtain the frequency response required to detect HIFs. This method was incorporated into a set of al- gorithms based on the analysis of abrupt changes of high-fre- quency energies, signal randomness, load behavior, and arc pat- terns in the neutral current, installed in General Electric (GE) relays. An algorithm proposed by Schweitzer Engineering Labora- tories (SEL) based on harmonic and interharmonic components was presented in [8], where artificial intelligence techniques were replaced by counters and compilers, thus simplifying the implementation of protection functions. In [9], the discrete wavelet transform was used to analyze the reignition of electric