Estimating transformer oil parameters using polynomial networks

In this paper we use polynomial networks to model the relationship between the insulation resistance measured between distribution transformer high voltage winding, low voltage winding and the ground and the breakdown strength, interfacial tension and the water content of the transformer oil. Polynomial networks have the advantage of effectively capturing nonlinearities in the modeled data even when the dataset is relatively small. The proposed model has been applied to real data collected from thirty four distribution transformers. The modeling results showed that the proposed model can predict values of the interfacial tension of the transformer oil with an impressive accuracy of 93% while it can predict the breakdown down with an accuracy of 84%. On the other hand, the study showed that this technique canpsilat be used to predict the transformer oil water content.