| In recent years, great improvements have been made in on-line monitoring of high voltage (HV) apparatus by detection of partial discharge (PD) signals. In this method, it is essential to investigate the characteristics of the PD pulses to evaluate insulation conditions. Recognition of PD types is fundamental for insulation system diagnosis. In former researches, the traditionalφ-q-n 3-Dimentional (3-D) plots are usually applied to recognize the PD pattern, which received obvious recognition results in researches. But it is rather difficult to use in practice.Firstly, because of the complex environment of HV apparatus, signals are buried in various noises, which requires better denoised method. Secondly,φ-q-n 3-D plots contain excessive information, which need more advanced hardware and software of signal process. In some of the specific detection scene and practical facility, it might be hard to create the 3-D plot. Meanwhile, the specific PD patterns in certain type of HV apparatus, such as generators, transformers, etc, can be recognized by some indexes or 2-D plots instead of 3-D plots with less calculation.This paper creates four relevant PD models upon the peculiar PD patterns of generator stators. The four PD patterns are slot discharge, end-winding discharge, inner discharge and abrasion discharge. The Rogowski coil is designed and rejusted to test the low high frequency current in the experiment. The signals are decomposed and reconstructed using mother wavelet Dmey of wavelet packet (WP) theory. In this way, the noises are removed and polarity characteristics are abstracted. The plots of negative and positive discharge signals differ between different PD patterns. Accordingly, the feasible operation suggestions are given out to certain PD types. The results show that this method is much simpler compared to the traditionalφ-q-n 3-D plots but remains the accuracy of pattern-recognition.
|
PDF Full Text Request