Reduction of noise by principal component analysis in on line frequency response monitoring of power transformers

Abstract

A method for the analysis of signals of On Line Frequency Response Monitoring, using Principal Components, is described in this article. The method is shown to be useful when the signal of the test has noise which could distort the original data and lead to misinterpretation of the results by giving a negative statement even when the transformer is still in good condition. It is shown that with Principal Components the noise could be reduced and the analysis could be performed closer to the actual state of the equipment. For the evaluation of signals, a method based on the Standard Deviation was used.