Application of cross-correlation denoising algorithm for UHF partial discharge detection

Abstract

For insulation condition assessment of HV power apparatus, partial discharge (PD) monitoring is one of the most effective techniques. However, on-line PD measurements are affected by high levels of electromagnetic interference (EMI) that makes sensitive PD detection very difficult. On the basis of crosscorrelation algorithm, this paper described a new PD extraction method which can reject noise from weak UHF signal. According to the analysis of PD feature, a double exponential model was setup. Then a damped oscillatory pulse (DOP) based on this model was applied for cross-correlation calculation with the test signal. In order to stimulate the noisy on-site testing environment, narrow band interference and Gaussian white noise was used in the laboratory tests. The noises have been successfully eliminated because the signal and noise was incorrelate with each other. The denoising performance of cross-correlation algorithm was analyzed by Signal-to-Noise (SNR) and correlation coefficient. The results showed that the method was effective in rejecting noise in high levels of interference environment. In addition, PD pulse integrity has been greatly improved. The reference signalx(t) T y (t)=s(t)+n(t) y 0 (t)=s 0 (t)+n 0 (t)