Hardware-Based Novel Applications to Locate Faults in Branched Distribution Systems

Abstract

The investigations on short-circuit fault protection applications have shown that the majority of research are toward transmission networks. In the case of distribution networks, however, most studies have been focused on identifying the fault and its type or detecting the faulty bus, rather than locating the fault directly. This study presents two hardware-based applications, the first known prototype device attempts to directly predict fault location in branched distribution networks, which operate in real-time and off-line respectively. In the real-time application, a circuit is designed in Laboratory Virtual Instrument Engineering Workbench (LabVIEW) to transform 3 phase current-voltage modeling data into analog signals. The signals are then processed by a digital signal processor and the fault location is determined by the embedded algorithm in real time. The off-line application obtains results using the MATLAB Gui-based interface, which is integrated into the LattePanda development board. The algorithms inside both applications use Random Forest (RF) along with Fast Walsh Hadamard Transform (FWHT) feature to estimate the fault location. The performance of both systems is validated on a real 111-bus distribution feeder located in Turkey.