Detecção de Faltas de Alta Impedância Baseada em Boosting de Árvores de Decisão

Bernardo A. V. de Souza; Gustavo S. R. Araújo; Jonathan N. Gois; Thiago A. Brasil

doi:10.20906/sbse.v2i1.3021

Bernardo A. V. de Souza Centro Federal de Educação Tecnológica Celso Suckow da Fonseca, RJ
Gustavo S. R. Araújo Centro Federal de Educação Tecnológica Celso Suckow da Fonseca, RJ
Jonathan N. Gois Centro Federal de Educação Tecnológica Celso Suckow da Fonseca, RJ
Thiago A. Brasil Centro Federal de Educação Tecnológica Celso Suckow da Fonseca, RJ

DOI: https://doi.org/10.20906/sbse.v2i1.3021

Keywords: High impedance fault, Classification, Decision trees, Boosting, Machine Learning

Abstract

The occurrence of high impedance faults is an event that is difficult to detect in distribution lines and may cause various hazards to the population. With the conductor falling to the ground, the fault current does not sensibilize the protection systems and can cause several accidents. Frequently, this type of fault is confused with other line events, further delaying the detection. The present work aims to insert simulated data of high impedance faults and several other events common to the network, to detect them using a boosting of decision trees, the XGBoost classifier. The obtained results demonstrate the accuracy of the 97.67% classification model.