Factorisation-based transfer function estimation technique for deformation diagnostics of windings in transformers

Mechanical deformations of transformer windings are identified through variations observed in their frequency response (FR). This study proposes a new way of determining a transfer function (TF) from measured FR data using factorisation principles as against established ways of finding it through rational fitting algorithms or least-square estimation. It is observed that some characteristics of FR of a transformer winding are similar to FR of first- and second-order systems, and this has been the basic motivation for the reported work. In the proposed method, TF is expressed as a combination of first- and second-order polynomials. The algorithm is applicable over a wide-frequency range and for high-order systems. Initial guesstimates for locations of poles/zeros, orders of TFs, and enforcement of poles to the left half of the s-plane are not required. The algorithm is validated through three case studies and its accuracy is checked through commonly used statistical indicators. The application of the algorithm is also demonstrated on a 500 kVA distribution transformer. The method is simple and accurate in estimating TFs for their further use in the deformation diagnostics of transformer windings.