A two-level ANN-based method using synchronized measurements to locate high-impedance fault in distribution systems

Abstract

This paper describes a new method for locating and identifying high-impedance fault in medium-voltage networks of unbalanced distribution systems, considering: load variation, reconfiguration, distributed generation, fault resistance variation, inaccuracy in feeders data, and protection devices operations. The method is based on artificial neural networks (ANN) and considers data from synchronized measurement units. Two sets of ANNs are created. The first set has the objective to identify the faulty area and the second set has the objective to indicate the distance between the fault point and the first meter of the area. The IEEE 123 Node Test System is used to validate the proposed method.