Electric Power Systems Research 77 (2007) 851–858 Discriminating changes introduced in the model for the winding of a transformer based on measurements Subrat K. Sahoo ∗ , L. Satish Department of H.V. Engineering, Indian Institute of Science, Bangalore 560012, India Received 28 March 2006; received in revised form 14 June 2006; accepted 27 July 2006 Available online 30 August 2006 Abstract A systematic procedure is evolved that enables identification of the type-of-changes introduced in the model winding of a transformer, based on a comparison of measured neutral or line current and transfer-function with those corresponding to the reference case. Further, it was also possible to ascertain, whether the pertinent circuit element viz. series capacitance, shunt capacitance and self-inductance has increased or decreased (compared to initial value). The proposal was formulated based on analytical calculation and circuit simulation. Thereafter, the approach was experimentally verified using two model windings, and the results were found to be encouraging. Thus, it demonstrates one possible way, how interpretation of monitored data can lead to meaningful inferences. © 2006 Elsevier B.V. All rights reserved. Keywords: Diagnostic testing; Interpretation of monitored data; LVI testing; FRA/TF method 1. Introduction Power utilities perform condition monitoring on large power transformers to assess its status. Many monitoring methods have evolved over the years and each one of them is intended to ascer- tain some specific aspect of dielectric, thermal and mechanical integrity of the transformer. Amongst these low voltage impulse (LVI) and frequency response analysis (FRA) or transfer func- tion (TF) method have been found to be useful in detecting winding deformations. Such deformities occur due to short- circuit forces and sometimes may be caused due to unskilled handling during transportation [1]. Although, minor and incip- ient deformations/damages are generally not immediately per- ceivable, they are in fact potential sites from where major faults can develop subsequently. The existing detection philosophy of these methods is based on a comparison of two subsequently gathered records, and if a mismatch were to be observed, then a possible fault is implied. It is generally observed that, beyond a visual comparison, moni- tored data does not seem to be post-processed to draw inferences. Strictly speaking, diagnosis or interpretation of the monitored ∗ Corresponding author. Tel.: +91 80 22933179; fax: +91 80 22932373. E-mail addresses: subrat@hve.iisc.ernet.in, subrat.sahoo@gmail.com (S.K. Sahoo), satish@hve.iisc.ernet.in (L. Satish). data has, as yet, not happened. It is desirable that every moni- toring method is adequately supported by powerful diagnosis or interpretation, so that its true potential can be harnessed. In a practical scenario (e.g. when fault is implied), a diag- nostic method is expected to provide assistance to the operation engineer regarding, the extent of fault, its location, and so on, such that timely action can be initiated. Instead of resorting to expensive and time-consuming disassembly of the transformer, it would be certainly advantageous to seek answers to these questions by interpretation of monitored data, as far as possi- ble. Therefore, exploration of newer ways of interpretation is desirable. A survey of pertinent literature reveals that most publications on this topic can be classified into: • Introduce a deformation (axial, radial, etc.) on an actual wind- ing and then examine ability of a method to detect the smallest possible change [2–8]. • Attempt to model or translate the winding deformation (i.e. physical changes) into a corresponding change in the equiv- alent circuit, so that, measured and computed TF are similar [4–6]. • Preliminary efforts attempting to link observed changes in TF, to the nature and location of fault, was examined via simula- tion studies [9]. 0378-7796/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.epsr.2006.07.007