Optimal wavelet de-noising of acoustic emission signals caused by the impact between metal particles and grounding electrode in DC GIL

In recent years, the application of high voltage direct current transmission is becoming more common, because of a lot of characteristics such as large transmission capacity, low loss, small capacitance, less area, high reliability, DC Gas Insulated Line (GIL) play an important role in power system, however, metal particles will inevitably be produced because of mechanical collision, equipment vibration and thermal expansion friction which can weaken the dielectric strength of DC GIL significantly. Metal particles with different motion state have different influence to the insulation. The more intense of the Particles' movement, the greater damage to the insulation. Acoustic emission signal caused by the impact between metal particles and grounding electrode can reflect the motion state of the metal particles. The acoustic emission signal often accompanied by noise in the process of the detection. The wavelet analysis can effectively eliminate the noise in the signal. But because of the diversity of wavelet basis and wavelet de-noising algorithms, have been searching for the optimal wavelet de-noising algorithm. Most of Scholars use a single evaluation index such as root-mean-square error and smoothness to evaluate the effect of de-noising algorithms. Through constructing the experimental platform and building Wavelet de-noising composite evaluation index, which based on root-mean-square error and smoothness of no signal period, This paper get the conclusion that Fixed form threshold (hard) algorithm under the decomposition of wavelet basis sym8 in the three layers is suitable for the de-noising of acoustic emission signal, what's more, with the increase of decomposition layers, the de-noising ability of wavelet basis sym8 get better first and then get bad. It perform worst in the one layer, and perform best in the three layers.