A hybrid method for arcing faults detection in large distribution networks

This paper presents a reliable protection scheme for high impedance fault (HIF) detection in large and complex electrical distribution networks. The proposed algorithm uses dynamic features extracted by stationary wavelet transform and feeds them to a support vector machine-based decision making system. HIF occurrence is detected according to the normalized changes of the features extracted from three post disturbance data windows. To provide more security, the output of the main fault classifier is compared with the outputs of three other classifiers in a voting system for final decision making. The results of simulations of a real distribution system demonstrate that the proposed scheme is able to detect a wide range of arcing faults, from a few amperes to hundreds of amperes, with high level of security against non-HIF phenomena. (C) 2017 Elsevier Ltd. All rights reserved.