Bivariate dependent rate analyses for fault identification in transformer windings during impulse test

Identification of winding insulation failure during lightning impulse testing of transformer has been an important issue over a long period of time. It requires a significant human expertise and knowledge for proper judgment of transformer insulation. In this paper, the bivariate dependent rate analyses are proposed to identify the failure in transformer windings. The proposed analyses enable the calculation of winding response due to any impulse voltage excitation of full wave (FW) and chopped wave impulse (CW) for typical transformers. Initially, the line of best-fit on the scatter diagram is utilized to visualize the similarities between the measured winding responses and also to identify the type of relationship with potential outliers. Further, ellipse analysis is utilized to calculate the diversification (size, shape and orientation) between the responses. The actual interdependency between winding responses is established through nonlinear interpretation coefficients. To prove the feasibility of the analyses, 66.7 MVA, 90 MVA, 150 MVA, 250 MVA transformers, 80 MVAr reactor and 22 kV interleaved winding are used.