-
Marcelo Estevão da Silva
Engenharia Elétrica, Campus Sobral, Universidade Federal do Ceará, CE
-
Joaquim Osterwald Frota Moura Filho
Programa de Pós-Graduação em Engenharia Elétrica e Computação (PPGEEC), Campus Sobral, Universidade Federal do Ceará, CE
-
Juan Carlos Peqquenã Suni
Engenharia Elétrica, Campus Sobral, Universidade Federal do Ceará, CE
-
Márcio André Baima Amora
Programa de Pós-Graduação em Engenharia Elétrica e Computação (PPGEEC), Campus Sobral, Universidade Federal do Ceará, CE
Keywords:
Machine Learning, Fault Detection, Distribution Networks, Smart Grids, Power Systems
Abstract
Distribution systems, due to their complex topologies and configurations, present the challenge of maintaining the reliability and continuity of the energy supply. In this sense, one of the main faults in the electrical network is the emergence of multi-stage faults, which represent 20% of fault occurrences. Aiming at the context of smart grids, and considering electricity meters that will be optimally allocated, this work proposes a methodology for classifying multistage faults in primary radial and overhead distribution feeders, based on decision trees (DA), whose Input parameters are the currents of the primary distribution feeder under study, measured only at the substation. The current oscillographs were obtained from simulations with the software MATLAB/SIMULINK and the signal processing method adopted was the RMS (Root Mean Square). Therefore, the obtained results represent an accurate classification, superior to 97%, indicating efficiency of the proposed method for the classification of such defects.
