Prediction of winding faults between non-stationary signals: Granger causality analysis for lightning impulse testing of transformers

摘要

One of the important research areas in transient phenomena is sensitive recognition of fault in the complicated insulation system of transformer to make a decision about either go (acceptance) or no-go criterion after impulse test at factory. The existing methods are, in principle, able to indicate the delay inrncoupling between the signals, but inferring causality from the time delay is not always straightforward.rnIn this paper, the time delay between simultaneously measured signals is calculated and further extendedrnto identify the impulse fault in transformers using Granger causality analysis. The principle of Bayesianrninformation criterion is used in Granger causality analysis to estimate the minimum number of lags.rnGranger causality analysis enables the calculation of neutral current signal for any input excitation of fullrnwave and chopped wave impulse for typical power transformers. The experimental studies are performedrnon 60/90 MVA (132/33 kV) and 150/200/250 MVA (500/275/33 kV) transformers to prove the feasibilityrnof the analysis. Copyright © 2015 John Wiley & Sons, Ltd.