Abstract
This paper presents a method for remaining life assessment of oil-immersed transformers using analyzable structured neural networks. Remaining life assessment of oil-immersed transformer is very important. Furfural method is conventionally used for such an assessment. In the furfural method, the average degree of polymerization is determined by considering the nonlinear correlation between the furfural and average degree of polymerization. The remaining life is estimated by considering the average degree of polymerization. However, a range of estimates of the remaining life is obtained when the furfural method is used, making accurate estimation difficult. The proposed method can be used to estimate the remaining life of an oil-immersed transformer accurately; this method involves analyzable structured neural networks and ensemble method. In the proposed method, not only furfural but also oil temperature, operational status, cooling type, etc., are considered. Since various factors are considered as input variables and a nonlinear model, i.e., artificial neural networks are used, accurate estimation of the remaining life has been realized. The effectiveness of the proposed method is shown by numerical simulation using actual measured data.
