Diagnóstico de faltas incipientes em transformadores de potência baseado na análise de gases dissolvidos no óleo isolante empregando redes neurais artificiais
Data
2022-07-11
Autores
Batista, Adrian Felipe Nogueira
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A análise de gases dissolvidos no óleo de transformadores é uma metodologia empregada nos
dias atuais para diagnosticar faltas em transformadores de potência. O objetivo deste trabalho
é utilizar redes neurais artificiais para criar um método alternativo de análise de gases
dissolvidos no óleo do transformador utilizando como base a norma técnica IEC 60559. Propõese, assim, analisar os dados obtidos utilizando-se diversas formas de treinamento da rede para
verificar qual método de treinamento e qual configuração de rede apresenta melhor eficiência
quando os resultados são comparados aos diagnósticos esperados. A partir dos resultados
obtidos foi possível atingir uma eficiência de até 91% indicando que a metodologia empregada
apresenta potencial para ser utilizada no diagnóstico de faltas em transformadores de potência.
The dissolved gasses analysis in transformer’s oil is a methodology used in present days to diagnostic faults in power transformers. This work’s objective is to use artificial neural networks to create an alternative method of dissolved gasses analysis in transformer’s oil with technical standard IEC 60559 as base. Thereby, it is proposed to analyze the data obtained with several types of network training methods verify which method and which network configuration has better efficiency with the expected results. In that way, it was possible to achieve a 91% efficiency indicating that the used method has potential to be used in dissolved gasses analysis in power transformers.
The dissolved gasses analysis in transformer’s oil is a methodology used in present days to diagnostic faults in power transformers. This work’s objective is to use artificial neural networks to create an alternative method of dissolved gasses analysis in transformer’s oil with technical standard IEC 60559 as base. Thereby, it is proposed to analyze the data obtained with several types of network training methods verify which method and which network configuration has better efficiency with the expected results. In that way, it was possible to achieve a 91% efficiency indicating that the used method has potential to be used in dissolved gasses analysis in power transformers.
Descrição
Palavras-chave
Transformadores, Redes neurais artificiais, Análise de gases dissolvidos no óleo, Transformers, Artificial neural networks, Dissolved gasses analysis