【作者】 程昌奎；

【导师】 李剑；

【作者基本信息】 重庆大学 ， 电气工程， 2014， 博士

【摘要】 电力变压器是电力系统重要的枢纽设备，其安全稳定运行对于电力系统尤为关键。电力变压器的运行可靠性很大程度上取决于其绝缘的可靠性。局部放电与变压器内部绝缘缺陷具有紧密联系，通过局部放电在线监测能够及时判断变压器内部绝缘状态，对防止电力变压器事故发生，保障电力系统安全稳定运行具有重要意义。本文针对传感器技术、抗干扰、模式识别等变压器局部放电在线监测与故障诊断的三个主要问题，总结分析了变压器局部放电在线监测研究现状，对局部放电监测的分形天线优化、自适应去噪与识别等问题进行了深入的研究。论文主要包括以下内容：①根据希尔伯特分形天线的设计原理，提出变压器局部放电超高频监测传感器设计准则。通过仿真分析，研究了导体宽度、导体厚度、介质厚度、馈电点位置对分形天线的驻波比、增益和方向性的影响规律；提出了结合遗传算法和仿真计算的分形天线优化方法，以检测频带内驻波比不大于2为设计目标，实现了四阶希尔伯特分形天线的设计；通过对人工油纸绝缘缺陷模型的局部放电检测，证明了优化后的分形天线能够满足变压器局部放电超高频检测的要求。②研究了混沌振子滤波器的原理与设计方法，提出了抑制窄带周期干扰的自适应混沌振子滤波器。提出了基于Lyapunov指数法判别混沌振子的运动状态，设计并实现了消除局部放电监测窄带周期性干扰的自适应混沌振子滤波器；通过局部放电信号去噪实例，对比分析了该方法与二阶级联IIR格型陷波器的去噪结果。结果表明，自适应混沌振子滤波器能够从窄带周期干扰中有效提取局部放电信号，去噪信号畸变率与幅值误差显著低于陷波器去噪结果。③基于经验模式分解及小波阈值法原理，提出了局部放电信号固有模态自适应最优小波去噪方法。该方法首先对局部放电信号进行经验模式分解得到多个固有模态，对每个固有模态采取自适应最优小波去噪并相加重构得到去噪后信号。固有模态小波分解时基于尺度系数能量最大原则自适应选择最优小波。通过对染噪局部放电信号的去噪试验，证明了固有模态自适应最优小波去噪对染噪局部放电信号造成的畸变更小。④根据局部放电超高频监测中的信号识别问题，提出了局部放电超高频信号固有模态特征提取及识别方法。首先，设计了变压器典型绝缘缺陷，通过试验获得大量局部放电超高频样本数据。其次，对局部放电超高频信号进行经验模态分析，提取固有模态的分形维数和能量系数作为特征量。最后，采取可能性模糊C-均值算法和反向传播神经网络进行分类识别，结果表明：反向传播神经网络识别率更高；固有模态提取的分形特征识别正确率高于小波系数。通过上述研究工作，本文实现了局部放电监测分形天线的宽频带优化设计，进一步降低了局部放电信号自适应去噪畸变率，显著提升了局部放电超高频信号多尺度特征参数识别正确率，解决了局部放电在线监测系统的抗干扰性和检测灵敏度难题，具有很强的实用价值和应用前景。更多还原

【Abstract】 Power transformer is one of the most important equipment in power system, theinternal fault of transformer is usually due to insulation defects. There is a tight linkbetween the insulation defects of transformer and partial discharge (PD). The PDmonitoring can be used to faults diagnosis for power transformer, and it is valuable toprevent accident faults in power transformer and insure the stable performance of powersystem. The PD approach has already become present research focuses in recent years.However, It is still need further studies on sensing technology, anti-interferencetechnique, PD pattern recognition before the PD on-line monitoring system can beapplied in actual projects. Based on summarize and analyze of the research status of PDon-line monitoring for transformers, this thesis made a systematic and thorough studyon the Hilbert fractal antenna optimization, the noise suppression based on adaptiveoptimal wavelet de-nosing and the chaotic oscillator filter etc., The main contents areshown as follows:①Firstly, thorough research on basic principle of fractal antenna and the practicalsituation of power transformer, the design criterion of UHF antenna and theoptimization design method of Hilbert fractal antenna was put forward. The influence ofwidth, conductor thickness, medium thickness, feed position of the four order Hilbertfractal curve to the performance of the Hilbert fractal antenna, such as voltage standingwave ratio(VSWR), gain, directivity have been analyzed. The optimization designmethod for antenna conductor by internal slotted method was put forward. The optionalfour-order Hilbert fractal slotting antenna was designed through the PD experimentsresults, the bandwidth, VSWR were verified to meet the requirement of PD UHFon-line monitoring.②The adaptive chaotic oscillator filter which can remove the communicationinterference in the PD signals was proposed. The basic principles of chaotic oscillatorinterference filter in suppressing PD signal narrow-band was studied. The method basedon the characteristic that the moving state of chaotic oscillators which was determinedbased on the chaotic phase diagram and Lyapunov exponent. Chaos oscillator filter wasput forward to inhibition the narrow-band interference in UHF PD detections. Thede-nosing results of simulation PD high-frequency signals and PD UHF signals showthat the chaotic oscillator filter denoising effect was better than the two order II R-rated filter.③Based on empirical mode decomposition (EMD) decomposition theory andwavelet threshold denoising method, The denoising methods based on EMDdecomposition and adaptive optimal wavelet method was proposed. The method basedon EMD decomposition, then multiple IMF components can obtained, for each IMFcomponent adopt adaptive optimal wavelet denoising, and then the denoising signal canobtained by adding reconstruction. For the adaptive optimal wavelet method, Accordingto the principle of maximum energy for scale coefficients, the optimal wavelet can beselected adaptively in each scale. The de-nosing results of simulation PDhigh-frequency signals and PD UHF signals show that the EMD with adaptive optimalwavelet de-nosing method presented in this thesis is superior to the standard waveletthreshold method.④According to the problems of PD signals recognition in PD UHF monitoring,the fracal dimension extraction method of IMF component of PD UHF signals wasproposed. The method based on the EMD of PD UHF signals, it can be got more IMFs.The fracal dimension and energy of each IMF calculated as as waveform features forPD UHF signals. Then the FPCM, BPNN were used for the artificial defect dischargeanalyzing. The results show that：the method of BPNN haved a better recognition ratethat the FPCM, and the feature of fracal was better that the energy feature.Through the above research work, this paper realizes broadband optimizationfractal antenna for partial discharge monitoring, further reducing the adaptive partialdischarge signal denoising distortion rate, significantly enhance the UHF partialdischarge signal multiscale feature parameter recognition correct rate, it has a strongpractical value and application prospect.更多还原