Intelligent Classifiers in Distinguishing Transformer Faults Using Frequency Response Analysis

With the expansion of use frequency response analysis (FRA) as a reliable tool for fault detection in transformers, more capabilities this method are discovered every day. So that today number transformer faults can be identified by FRA has also increased. One most critical steps with is to distinguish and classify them different classes. In paper, well-known intelligent classifiers (probabilistic neural network, decision tree, support vector machine, k-nearest neighbors) used faults. For purpose, necessary measurements performed on model transformers under healthy condition conditions (axial displacement, radial deformation, disc space variation, short-circuits, core deformation). Then, dividing ranges measured transfer functions transformer, new feature based numerical statistical indices training validation proposed. After completing process, performance evaluated compared applying data obtained from real transformers.