Application of artificial neural network to the detection of the transformer winding deformation

The application of the artificial neural network technique to the frequency response analysis method (FRA) for the detection of transformer winding deformation is presented. A set of simulation experiments is performed in order to obtain the information of the deteriorated winding. The fingerprints of the state of transformer windings obtained from simulation tests of deformation with different types and positions are learned by a multilayer feedforward neural network using backpropagation algorithm. Results show that after being trained, the neural network could well discriminate the state of transformer windings.