Comparison of different neural networks algorithms used in the diagnosis and thermal ageing prediction of transformer oil

In this paper supervised and unsupervised neural networks are applied. To help the inexperienced transformer oil analyst to make good diagnosis, a Levenberg-Marquardt net and a Bayesian network are applied in the diagnosis of the transformer oil. The last net presents the best generalization. A Kohonen net is applied also to classify the diagnosis. An RBFG (Radial Basis Function Gaussian) net is used to predict thermal ageing of the same oil.