Studies on three feature extraction methods for the location and classification of dynamic fault patterns during impulse testing of transformer winding
by P. Rajamani; Sivaji Chakravorti
International Journal of Power and Energy Conversion (IJPEC), Vol. 3, No. 3/4, 2012

Abstract: Many computer aided classifiers have been employed for identification of impulse insulation failure parameters, viz. type, condition of insulation and location using significant features extracted from winding current. Classification accuracy of these classifiers is dependent on the ability of extracted features. In this present approach, an attempt has been made to identify suitable feature extractor for accurate classification of insulation failure using simple classifier. The suitability of three feature extraction methods, viz. cross-wavelet transform (XWT), discrete wavelet transform (DWT) and cross-correlation (CCL) are assessed for identification of failure using differential evolution (DE) classifier. The required winding currents for feature extraction are acquired by emulating different insulation failure in an analogue model of 33 kV winding of 3 MVA transformer. Result of developed DE classifier shows that XWT features identified the insulation failure more accurately than DWT and CCL.

Online publication date: Thu, 07-Aug-2014

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Power and Energy Conversion (IJPEC):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com