Applications of ANFIS to Estimate the Degree of Polymerization Using Transformer Dissolve Gas Analysis and Oil Characteristics

Abstract

Degree of Polymerization is a reliable parameter that can identify the health condition and remaining operational life of the paper insulation within power transformers. Measuring this parameter is a challenging process due to the difficulty in acquiring paper samples from hot spot locations of operating transformers. Insulation oil characteristics and dissolved gas analysis (DGA) of power transformers have become powerful tools to overcome this limitation, because these tests are regularly performed by utilities in Indonesia, unlike furfural and degree of polymerization measurements. This paper presents a novel Adaptive Neuro Fuzzy Inference System (ANFIS) approach to estimate the paper degree of polymerization based on dissolved gas analysis and oil characteristics. In this context, data of oil insulation characteristics, DGA, and furan compounds of 200 150/20 kV and 42 500/150 kV operating transformers are collected from PT. PLN (Persero). The Calculated degree of polymerization from furan data of each investigated transformer is analyzed using the corresponding oil insulation characteristics and DGA data. Results show that carbon oxides compounds, acidity number, interfacial tension, and colour of the oil are statistically correlated with the degree of polymerization of the insulation paper. Parameters that highly correlated with the degree of polymerization are used as input parameters to the proposed ANFIS model to estimate the degree of polymerization value. The results of the developed ANFIS model reveal an overall accuracy more than 86% for the two investigated transformers’ data. The accuracy of the proposed ANFIS model can be enhanced through tuning the model using future obtained results.