Neuro-fuzzy computing for large power transformers monitoring and diagnostics

Abstract

There are number of parameters and methods available for condition monitoring and diagnosis of power transformer insulation. We concentrate on a few of them. One of the parameters, which is becoming more and more vital for diagnostics of the integrity of a power transformer, is the moisture content of the insulation system. Migration of moisture between oil and paper insulation is a very complex, nonlinear process with many uncertainties. An adaptive neuro-fuzzy system identification is applied to predict the moisture content of solid insulation from on-line measurements of moisture characteristics of oil. The comparison of the measured and predicted values of the average moisture content in a paper-oil insulation system is presented. Development of a neuro-fuzzy thermal model of a power transformer, intelligent sensor technology, fuzzy signal conditioning and evaluation of partial discharges in power transformers are also discussed.