LVQ neural network for identification of abnormal conditions within transformers

Simulation and discrimination of several different types of insulation failure has been proposed. In the present paper, five types of insulation failures that are apt to occur in power transformers are simulated using PSIM. Input-output voltage as well as input current of each insulation failure type is monitored and hence constructing the (ΔV- Iin) locus diagram which is used for providing the state of the transformer. A discrimination process utilizing neural networks is developed to distinguish any deviations of the locus with respect to the reference one.