Predição do Grau de Polimerização Utilizando Machine Learning: Nova Metodologia de Avaliação da Vida Útil de Transformadores de Potência

Rafael Prux Fehlberg; Daniel Carrijo Polonio Araujo; Gabriel de Souza Pereira Gomes; Sofia Moreira de Andrade Lopes; Rogério Andrade Flauzino; Renan Ferreira Santa Rosa; Iony Patriota de Siqueira

doi:10.20906/SBAI-SBSE-2023/3880

Rafael Prux Fehlberg Esc. Eng. de S. Carlos - EESC, Univ. de S. Paulo - USP, SP
Daniel Carrijo Polonio Araujo Esc. Eng. de S. Carlos - EESC, Univ. de S. Paulo - USP, SP
Gabriel de Souza Pereira Gomes Esc. Eng. de S. Carlos - EESC, Univ. de S. Paulo - USP, SP
Sofia Moreira de Andrade Lopes Esc. Eng. de S. Carlos - EESC, Univ. de S. Paulo - USP, SP
Rogério Andrade Flauzino Esc. Eng. de S. Carlos - EESC, Univ. de S. Paulo - USP, SP
Renan Ferreira Santa Rosa Treetech Tecnologia, Atibaia, SP
Iony Patriota de Siqueira Tecnix Engenharia e Arquitetura Ltda. Recife, PE

DOI: https://doi.org/10.20906/SBAI-SBSE-2023/3880

Keywords: Aging, Power Transformer, Assets, Degree of Polymerization

Abstract

The main indicator used to assess the condition of solid insulation in power equipment currently is the degree of polymerization (DP). This work presents a systematic study where machine learning techniques are used to estimate DP from 2-fal and other indicators. The results are promising, indicating that 2-fal, CO2/CO, the Chendong formula, and equipment power can be used together to better predict the current value of its service life.