Detecção e classificação da corrente de inrush em transformadores de potência, usando as técnicas de estatística de ordem superior, discriminante linear de Fisher e Redes Neurais

Romão O. K. Polyana; Almeida R. Aryfrance; Carvalho S. G. José; Araújo G. Cléia; Sobreira O. Danillo; Junior F. S. Bartolomeu

doi:10.20906/CBA2022/3348

Romão O. K. Polyana Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI
Almeida R. Aryfrance Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI
Carvalho S. G. José Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI
Araújo G. Cléia Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI
Sobreira O. Danillo Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI
Junior F. S. Bartolomeu Departamento de Engenharia Elétrica, Universidade Federal do Piauí, PI

DOI: https://doi.org/10.20906/CBA2022/3348

Keywords: Inrush Current, Higher Order Statistics, Fisher's Linear Discriminant, Neural networks, Transients

Abstract

This article presents the application of Higher Order Statistics for the extraction of characteristics of the inrush current phenomenon, together with Fisher's Linear Discriminant for the selection of characteristics with the greatest potential for distinction, in relation to steady-state and short- term events. circuit, as well as the application of the Neural Networks technique for the classification of these types of events. The modeling process of these events was carried out in the software APTDraw (Alternative Transient Program), having as an initial step the process of validating the simulation data, in relation to the real data measured in a 69/13.8kV Substation. The application of the techniques together resulted in a satisfactory distinction and classification of events, giving credibility to the efficiency of the application.