Support Vector Machines no suporte à detecção de faltas de alta impedância. Caso base: LT 500 kV Xingu / Tucuruí

Vittorio T. Missagia1; Thiago A. do Brasil1; Rodrigo A. das Neves; João A. Moor

doi:10.20906/sbse.v2i1.2885

Vittorio T. Missagia1 Centro Federal de Educação Tecnológica Celso Suckow da Fonseca Rio de Janeiro
Thiago A. do Brasil1 Centro Federal de Educação Tecnológica Celso Suckow da Fonseca Rio de Janeiro
Rodrigo A. das Neves Operador Nacional do Sistema Elétrico, Rio de Janeiro
João A. Moor Centro Federal de Educação Tecnológica Celso Suckow da Fonseca Rio de Janeiro

DOI: https://doi.org/10.20906/sbse.v2i1.2885

Keywords: high impedance fault, FAI, identification, costs, LT, supervised learning

Abstract

The equipments of an electrical system is subject to several types of failures, especially those present in the transmission that are built in an unsheltered way, thus being exposed to bad weather, which invariably results in wear and, therefore, equipment failure. Of these, transmission lines (LT) are the most susceptible to failure, due to their large extensions. From this feature in particular, the construction of LT in dense tropical forest regions has led to the occurrence of high impedance faults (FAI), which are difficult to identify by current protection systems. In this sense, identifying this phenomenon quickly before it evolves to more severe short circuits are essential, protecting equipment, reducing operation and maintenance costs, protecting the environment from possible fires and reducing the energy consequences of transmission outages. Therefore, getting the LT 500 kV Xingu / Tucuruí as an example, this work aims to analyze and propose a strategy for identifying FAI based on the use of supervised learning classifiers of the Support Vector Machines type.