Novel modeling for the prediction of aged transformer oil characteristics

Abstract

The effect of aging on transformer oil physical, chemical and electrical properties has been studied using the international testing methods for the evaluation of transformer oil quality. The study has been carried out on twelve transformers in the field and for monitoring periods up to 8 years. The properties which are strongly time dependent have been specified and those which have a great impact on the transformer oil breakdown voltage have been defined. Mathematical models for the breakdown voltage, total acidity and water content as a function of service periods have been given. The validity and applicability of these models for future prediction of these properties have been verified by the good agreement between the measured end predicted values. A multiple linear regression model for each transformer oil breakdown voltage as a function of its water content, total acidity and service period has been introduced and its adequacy has been illustrated by statistical analysis. Another multiple linear regression model has been developed by combining the results of a group of transformers into that of a single equivalent transformer. This model has been validated by predicting the properties of some other transformers and comparing them with the measured values. The comparison showed a good agreement for the results of transformers which have either been used or not in the derivation of the model.