Application of numerical evaluation techniques for interpreting frequency response measurements in power transformers

Frequency response analysis (FRA) is an emerging, powerful non-intrusive condition monitoring and diagnostic tool for verifying the mechanical integrity of power transformers. FRA results are graphical in nature and require trained experts to interpret test results. The work reported discusses numerical-criteria-based evaluation techniques. Persons not familiar with interpreting the FRA results can apply the evaluation criteria. The various criteria help in deriving proper conclusions. By evaluating correlation coefficient (CC), standard deviation and absolute sum of logarithmic error (ASLE) techniques, it is possible to discriminate between defective and non-defective windings. Experimental studies were conducted on two test transformers for axial and radial displacements, and additionally two sets of identical substation transformers. The techniques mentioned above are useful for interpreting frequency responses even in situations when a reference fingerprint was not available. However, it was concluded that if original fingerprints are available, the method gives very reliable indication for diagnosing the faulty winding. In addition, the severity of displacement/deformation can also be concluded from the amount of variation of the parameters from the suggested critical values.