An investigative study on the influence of correlation of PD statistical features on PD pattern recognition

Mas'ud, Abduallhi Abubaker and Stewart, Brian G. (2018) An investigative study on the influence of correlation of PD statistical features on PD pattern recognition. In: 2nd IEEE International Conference on Dielectrics (ICD) 2018. IEEE, Piscataway, N.J..

[img]
Preview
Text (Mas-ud-Stewart-IEEE-2018-An-investigative-study-on-the-influence-of-correlation-of-PD-statistical)
Mas_ud_Stewart_IEEE_2018_An_investigative_study_on_the_influence_of_correlation_of_PD_statistical.pdf
Accepted Author Manuscript

Download (566kB)| Preview

    Abstract

    This paper investigates the influence of correlation coefficients of partial discharge (PD) statistical fingerprints on the classification performance of the ensemble neural network (ENN). PD measurements were carried out according to the IEC 60270 standard. Independent statistical parameters of skewness, kurtosis, cross-correlation, discharge factor and modified crosscorrelation were analyzed and utilized as inputs to the ENN. The ENN was applied to classify 2 PD datasets. One with PD statistical features and the other a combination of PD statistical features and their correlation coefficients. The results indicate that the ENN appears to show a statistically better performance using the statistical features mixed with their correlation coefficients as compared to the other dataset. This clearly shows that the correlation coefficients of statistical features can provide an improved classification and discrimination of PD patterns.