An effectual approach to detect the winding faults from frequency response using artificial neural networking techniques

The most efficient approach for identifying transformer winding deformation is Frequency Response Analysis (FRA). Transformers frequently experience the effect of short-circuit current in power systems, the mechanical design and windings are subjected to high mechanical stress. These stresses have the potential to severely distort the winding. However, winding deformation is difficult to measure using traditional methods. It also has the potential for an accident, which will result in unforeseen mishaps. The artificial neural network approach is used to the FRA method for the identification of transformer winding deformation. In this study, a Transfer function and its related parameters were calculated using a subspace-based identification technique for various forms of transformer failures. An effort was made to train an Artificial Neural Network (ANN) using these parameters. The outcomes of an investigation on the use of an ANN method for detecting and identifying transformer winding problems from FRA data results are presented.