Diagnosis and identification of transformer winding faults from frequency response data by the application of ANN technique

In the present work, model transformer windings were used to obtain Sweep frequency Response Analyzer measurement data for various experimentally simulated faults. Transfer function and its corresponding parameters were computed for different types of transformer faults by subspace based identification method. An attempt was made to use these parameters for training Artificial Neural Network. The result of study on application of Artificial Neural Network algorithm for diagnosing and identifying the transformer winding faults from the Frequency Response Analysis data is presented. It was observed that, predicted fault closely follows the created fault suggesting the effectiveness of Artificial Neural Network technique for distinguishing between the normal and failed state quite satisfactorily.