变压器局部放电监测希尔伯特分形天线优化与自适应去噪方法

发布时间：2018-06-06 02:18

本文选题：变压器 + 局部放电　；参考：《重庆大学》2014年博士论文

【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。
[Abstract]:The power transformer is an important hub equipment of power system . Its safe and stable operation is especially key to the power system . The reliability of the power transformer depends greatly on the reliability of the insulation . The partial discharge is closely related to the internal insulation defects of the transformer . In this paper , the on - line monitoring and fault diagnosis of transformer ' s internal insulation can be determined in time by means of local discharge online monitoring .

On the basis of the design principle of Hilbert fractal antenna , the design criterion of UHF monitoring sensor for partial discharge of transformer is proposed . Through simulation analysis , the influence of conductor width , conductor thickness , dielectric thickness and feed point position on the standing wave ratio , gain and directivity of fractal antenna are studied .
In this paper , a fractal antenna optimization method combining genetic algorithm and simulation is proposed in order to detect that the standing wave ratio in the frequency band is not greater than 2 as the design target , and the design of the fourth - order Hilbert fractal antenna is realized ;
This paper proves that the optimized fractal antenna can meet the requirement of partial discharge ultra - high frequency detection of transformer by detecting the partial discharge of the model of the insulation defect of the artificial oil paper .

In this paper , the principle and design method of chaotic oscillator filter are studied , and an adaptive chaotic oscillator filter for suppressing narrow - band periodic interference is proposed . A self - adaptive chaotic oscillator filter for eliminating local discharge monitoring narrow - band periodic interference is designed and realized based on Lyapunov exponent method .
The de - noising results of this method and the second - class IIR lattice trap are compared and analyzed . The results show that the adaptive chaotic oscillator filter can effectively extract the local discharge signal from the narrow - band periodic interference , and the de - noising signal distortion ratio and amplitude error are significantly lower than that of the notch filter .

( 3 ) Based on the principle of empirical mode decomposition and wavelet threshold method , an adaptive optimal wavelet de - noising method for local discharge signal is proposed . The method comprises the following steps : firstly , performing empirical mode decomposition on local discharge signals to obtain a plurality of intrinsic modes ;

The characteristic feature extraction and identification method of local discharge ultra - high frequency signal are proposed according to the signal identification problem in local discharge ultra - high frequency monitoring . Firstly , the typical insulation defect of transformer is designed , and a lot of local discharge ultra - high frequency sample data is obtained by experiment . Secondly , the partial discharge ultra - high frequency signal is subjected to empirical mode analysis to extract the fractal dimension and energy coefficient of the natural mode as the feature quantity . Finally , the possibility fuzzy C - means algorithm and the reverse propagation neural network are adopted to classify and identify the characteristic . The results show that the recognition rate of the reverse propagation neural network is higher ;
The recognition accuracy of fractal feature extraction is higher than that of wavelet coefficient .

Through the research work , the broadband optimization design of partial discharge monitoring fractal antenna is realized , the self - adaptive de - noising rate of local discharge electric signal is further reduced , the recognition accuracy of local discharge ultra - high frequency signal multi - scale characteristic parameter is obviously improved , the anti - interference performance and the detection sensitivity problem of the local discharge on - line monitoring system are solved , and the local discharge monitoring fractal antenna has strong practical value and application prospect .
【学位授予单位】：重庆大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM855

【参考文献】