中文摘要：

基于PD信号的复小波变换模极大值随尺度s的增加而增加，而白噪声的复小波变换模极大值随尺度s的增加而减小的特点，创建一种分层复阈值算法。将分层复阈值用于复小波变换对仿真PD信号的去噪，并与全局复阈值算法进行了比较。同时在对仿真PD信号去噪后的波形畸变研究中，为弥补波形相似性参数（NCC）的不足，提出一个新的描述波形相似程度参数——变化趋势参数（VTP），并结合去噪前后的信噪比（SNR）对去噪效果进行综合评价。将分层复阈值技术用于对实验室实测的模拟GIS超高频PD信号的去噪。结果表明，采用分层复阈值技术的复小波变换抑制自噪声的效果更好，在高信噪比（SNR）或噪声抑制比（NRR）的情况下，对信号的畸变也更小。

英文摘要：

According to the characteristic of modulus maximum of partial discharge（PD） signal and white-noise, an algorithm of complex multilevel threshold is put forward. The de-noising effect of complex multilevel threshold is compared with the effect of complex globe threshold. In order to make up the shortage of NCC, a new parameter--variational trend parameter（VTP） is put forward, which describes the degree of waveforms＇ similarity. The de-noising effect can be synthetically evaluated by VTP, NCC and SNR. Finally, complex multilevel threshold is used to suppress the white-noise in ultrahigh frequency PD signal which is got from stimulant GIS in lab. The findings indicate that this threshold can suppress white-noise effectively with small distortion of PD signal.