Application of Modified Sequential Floating Forward Feature Selection to Partial Discharge Patterns

Rania M. Sharkawy and Karim I. Mohamedeen


Signal Denoising, Feature Extraction, Wavelet transforms, Support Vector Machines


In this study, the classification of partial discharge patterns for the identification of contaminating particles in transformer oil is accomplished by utilizing sequential floating forward feature selection technique. The technique is applied to the extracted features from the measured partial discharge (PD) pulse patterns to obtain better classification results. The utilization of phase resolved in addition to time resolved partial discharge signals is undertaken to extract a numerous features using statistical and frequency analyzers. Extracted features are fed to a feature selection wrapper model utilizing the sequential floating forward selection (SFFS) as a search strategy. Finally a support vector machines (SVM) is utilized for the classification of contaminating particles identity. The presented technique in this study rendered a classification success of up to 97.5% of total tested samples.

Important Links:

Go Back