Transformer Insulation Degree of Polymerization Estimation through Adaptive Neuro Fuzzy Inference System Approach

Abstract

The useful life of a power transformer is linked to the integrity of its solid (cellulosic) insulation. The insulation deterioration is time dependent and is influenced by varying temperatures, moisture, oxygen, oil conditions and loading profiles. Accordingly, once the solid insulation state is compromised, its dielectric and mechanical properties are not recoverable. Thus, longevity in service of a power transformer can only be achieved by incessant monitoring and assessing the credibility of its insulation system. The tensile strength of the solid insulation can be indicated by determining the degree of polymerization of the paper. Consequently, the degree of polymerization is a key indicator of the insulation status that correlates well with the transformer remnant life. Since degree of polymerization measurement is an intrusive test involving transformer disassembling, utilities have opted to the use of furans concentration in the oil as a way of determining the degree of polymerization of transformer solid insulation. However, these furans and degree of polymerization correlations are based on mathematical models. This paper introduces an adaptive neuro fuzzy model to estimate the degree of polymerization of a mineral oil-immersed power transformer based on furan content and CO2/CO gas concentration ratio in the insulation system. Practical data from numerous power transformers of different lifespans subjected to different operating regimes have been used to validate the accuracy and credibility of the established adaptive neuro fuzzy model. Compared to the conventional models, the results show that the proposed model is effective in estimating the degree of polymerization of transformer solid insulation.