Robust classification of partial discharges in transformer insulation based on acoustic emissions detected using fiber bragg gratings

Incipient discharges formed due to corona activity, surface discharge, and particle movement in transformer insulation are identified based on acoustic emission signals captured using fiber Bragg grating sensors and analyzed in the frequency domain. To improve the SNR of these signals, the use of an adaptive line enhancement-based technique is systematically explored through simulations, and the associated parameters are optimized. The noise-filtered spectra analyzed through ternary diagrams suggest the possibility of classifying the discharges which are further validated using appropriate classifiers. The experimental comparison of discharges generated in different oil media like mineral oil, nanoparticle-dispersed mineral oil, ester oil, and nanoparticle-dispersed ester oil reveals that the discharge characteristics are similar across multiple media, and the classification holds good.