【作者】 唐炬；

【导师】 孙才新；

【作者基本信息】 重庆大学 ， 电气工程， 2004， 博士

【摘要】 由于气体绝缘组合电器(Gas Insulated Switchgear,简称GIS)内部产生局部放电(Partial Discharge,简称PD)是引发事故的重要因素之一,因此,为了保证GIS的安全可靠运行,国内外对GIS局部放电在线监测技术开展了大量的研究,但到目前为止,对在线监测中的PD模式识别还是一个尚待深入研究的前沿课题,其难点主要是从强烈的电磁干扰环境中准确获取PD信号。为此,本文主要研究用于GIS局部放电在线监测的超高频外置传感器和前置滤波放大器,并用研制的模拟GIS内部缺陷的人工物理模型在实验室里获取大量不同缺陷下产生的PD信号数据;同时,构建用于强电磁干扰环境下GIS中PD信号处理的复小波,并阐述用复小波变换的简单信息和复合信息处理PD信号的原理和特点;通过仿真研究与实验室应用,验证用所构造的复小波变换抑制白噪声和周期窄带干扰的能力,并用软硬联合去噪法抑制GIS中PD信号干扰。取得主要成果有:采用并矢格林函数对PD信号电流陡脉冲激发瞬变电场进行分析计算,推导出瞬变电场的简化计算公式,证明了各次模波对计算瞬变电场的贡献大小和在仿真研究中采用高斯分布电流脉冲来模拟GIS中PD电流陡脉冲的可行性;同时,研制出了用于GIS中PD试验的3种人工模拟缺陷物理模型,为获取大量的人工试验数据和进一步开展混合缺陷下的局部放电机理研究奠定了基础。采用独特而巧妙的附加电抗等结构设计,使传统DA成为超宽频带天线,相对带宽比达到0.73;设计出屏蔽谐振式环天线,使得驻波比更加优良,抑制干扰能力更强;同时,研制出工作频带为300 MHz～1000MHz、增益达到20dB的JBF-1000型微带线滤波放大器,当信号经过滤波放大预处理后,信噪比大为提高,满足了超高频微弱信号的滤波与放大预处理要求。利用复小波变换独有的相频特性,通过对简单信息的不同组合,构建了多种复合信息,为监测GIS中的PD信号提供了更多而有效的去噪方法和手段;同时,以db小波为基础,成功地构建出db系列复小波,并详细阐明了构造复小波的思想、方法和步骤,给出了db3和db4滤波器组系数;以db4复小波为例进行了去噪应用,分解和重构表明所构建的复小波及其算法具有较高的准确性。在对影响复小波去噪的阈值选取、分解层数、噪声水平、干扰频率等主要因素进行深入研究的基础上,首次提出一种用于复小波变换的有效小波阈值法;从仿真和实验室去噪研究结果表明:有效小波阈值法比史坦得无偏似然估计法和基于极大<WP=6>极小原理法更优越;同时,在用众多复合信息对PD信号的去噪处理中,以实部、虚部和相位(WTRIPH)复合的信息为最佳去噪信息。更多还原

【Abstract】 Partial discharge (PD) in Gas Insulated Switchgear (GIS) is one of the important factors for arising the fault, and therefore the on-line monitoring PD in GIS has been being researched at home and abroad in order to guarantee the safe operation of GIS. At the same time, PD pattern recognition is still an advanced problem to be studied, and it is difficult to correctly obtain the PD signal from the intense electromagnetic interferences. So this disquisition designs an outer ultra high frequency (UHF) sensor and a pre-positive filter amplifier utilized to monitor the PD signal in GIS, and a plenty of PD data are sampled with artificial physical models of PD defect in laboratory. Meanwhile, this paper introduces how to construct the complex wavelet applied to suppress the interferences, of which the characteristics of simple and multiplex information are introduced. The simulation results and practical application prove the complex wavelet transformation can excellently suppress the white noise and narrow band interference. Finally, associated method of software and hardware is adapted to extract PD signals. The main conclusions are shown as following:The paper uses Greens function to calculate the transient electric field excited by the PD, and develops the simplified expression of the transient electric field, and shows clearly modes’ contribution to the transient electric field and Gauss distributed current can be used to simulate the PD pulse. Three artificial physical models of PD defects establish the foundation for sampling plenty of PD data and developing the research of PD mechanism under multiplex defects.Annexed impedance adopted to optimize dipole antenna (DA) makes DA become ultra broadband antenna, which relative bandwidth ratio reaches 0.73. The designed shield resonance loop antenna (SRLA) holds better standing wave ratio (SWR), stronger ability to suppress interferences. On the other hand, the designed microstrip line filter amplifier JBF-1000, which bandwidth is from 300MHz to 1000MHz and the gain is 20 dB, and can effectively improve the signal-to-noise ratio and satisfy the practical application.The phase-frequency characteristic of the complex wavelet has been used to compose the multiplex information of the simple information, and the multiplex information can provide more effective means to monitor the PD signals in GIS and better effect to restrain the noise. This paper introduces the idea and the process of <WP=8>constructing complex wavelet. The complex wavelets based on db wavelet have been expressed according to the way, and the complex wavelet filters are shown based on db3 and db4. Furthermore, the complex wavelet based on db4 is applied to restrain the noise, and the results of the wavelet decomposition and reconstruction suggest that constructed complex wavelet and its algorithms can greatly meet the needs of engineering application.Based on analyzing the main factor influenced the effect of complex wavelet compressing noise, which are threshold selection, decomposition scales, noise level and interference frequency, the paper firstly puts forward the effective wavelet coefficient (EWC) threshold. The results of the simulation and application indicate that the effect of EWC suppressing noise is better than that of maximal theory and Stein unbiased risk estimate theory. Finally, the results show clearly that the multiplex information WTRIPH is the best one, which is composed of real part, imaginary part and phase of complex wavelet transformation.更多还原