【作者】 刘辉；

【导师】 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2012， 博士

【摘要】 随着电网规模越来越大,电力电缆线路规模也随之变大,确保电缆线路安全十分关键。因此,电力电缆线路在线状态监测和故障诊断成为热门研究领域。由于实际电缆故障环境状态复杂多变,影响环境变化和故障定位精度的因素很多,造成电缆故障信息的多样性,故障信息获取困难,故障定位精度不高。因此研究如何对故障信息进行分析、处理以及如何提高故障定位的精度具有迫切的现实意义。本文分析了电力电缆线路故障产生的机理,在综述了电力电缆故障诊断在国内外的研究现状的基础上,研究了小波分析、DNA遗传算法、粗糙集和Petri网等技术在电力电缆线路故障诊断方向中的实际应用,深入分析了多智能体技术在电缆线路故障的应用,并根据本论文的研究目的设计电力电缆线路故障系统。论文的主要工作与创新有：1.针对故障特征信号的提取问题,基于小波去噪的基本原理,给出了常用小波阈值去噪方法与步骤。通过小波去噪仿真比较,验证了3σ去噪方法在信号去噪方面有很好的优越性。2.在已有的遗传属性约简算法的基础上,提出在算法编码过程中引入分子DNA遗传算法的思想,采取特有的两位二进制编码方式,得到一种改进属性约简算法。通过实例验证了该方法用于故障诊断是可行的,并有较高的应用价值。3.针对搜索空间随着问题复杂程度的增加成指数倍增加,而经典的优化算法不能很好地解决高维空间的优化问题,提出了将引力搜索算法与BP神经网络结合的一种混合算法(GSA-BP),并给出了实现步骤。通过仿真实验验证了该方法在网络训练的前期收敛速度是比较快的,基本能满足诊断要求。4.针对电缆故障测距时,高压脉冲注入法的高压脉冲信号在电力电缆线路传播时会不可避免的产生一些噪声,导致在接收故障信号的时候不能够直观的找出故障点所对应的波头这一问题,提出通过希尔伯特-黄变换算法对接收的故障信号进行去噪声处理。通过ATP仿真验证了这种方法能得到接近理想状态的波形并且可以直观地找出故障点所对应的波头。5.根据对故障信号去噪、故障特征约简以及故障分类的处理结果,设计了一种可以根据故障征兆与故障间联系建立模糊推理的电力电缆线路故障诊断多智能体模型。模型中设计了多种功能Agent。其中多Agent组采用结合合同网模型与黑板模型的结构来协调组中各故障检测Agent工作。根据整理出的电缆故障征兆与故障对应关系,利用模糊Petri网建立故障诊断模块。模糊Petri网故障诊断模块按照故障种类建立诊断推理结构,能够有效对“多因多果”故障进行诊断。算例分析证明了该诊断模块的有效性和可靠性。更多还原

【Abstract】 With the increasing scale of Power Grid and power cables of power system, it is crucial to ensure the security of power cables. Therefore, the research of online status monitoring and fault diagnosing of power cables becomes a popular research field. However, due to the complicated and changeable environment of practical cable faults, various factors which influence the change of environment and accuracy of faults location, it leads to the diversity of the cable fault information, the difficulty of obtaining fault information and the fault location accuracy not high enough. Therefore, it is with urgent practical significance to research how to analysis, process and improve the accuracy of fault location.Generation mechanism of power cable fault is analyzed in this paper. Based on the summary of domestic and foreign research of power cable fault diagnosis, the application of Wavelet Analysis, DNA Genetic Algorithm, Rough Set and Petri nets in the field of power cable fault diagnosis is studied in this paper. Furthermore, the application of Multi-Agent technology in the power cable faults is analyzed, and faults diagnosis system of power cables is designed based on this paper’s research purposes. The main work and innovation of the paper are listed as below:1. Against the problem of fault feature signal extraction, the commonly used wavelet threshold de-noising methods and steps are put forward based on the basic principles of wavelet de-noising. Being compared with wavelet de-noising simulation, the 3σde-noising methods are verified good superiority in terms of signal de-noising.2. Based on previous Genetic Algorithm for attribute reduction, the idea of introducing molecular DNA algorithm into the coding process, taking the approach of applying the unique two bit binary encoding and obtaining an improved algorithm for attribute reduction is proposed. This new method is verified feasible in fault diagnosis and more valuable based on examples.3. Against the problem that classic optimization algorithm cannot solve the optimization problem of the high-dimensional space and search space increases exponentially with the increasing degree of problem complexity, a hybrid algorithm, the combination of the Gravity Search Algorithm and BP neural network, and feasible steps are proposed in this paper. Simulation results show that this method has relatively fast convergence speed in the early stage of network training, and can basically meet the diagnosing requirements..4 Against the problem that when locating cable fault, the high-voltage pulse signal produced by high-voltage pulse injection method propagating in the cable will inevitably generate noise, causing the received signal not be direct to identify the faults point position corresponding to the first wave, the problem can be solved by de-noising the signal wave by Hilbert-Huang transform algorithm. ATP simulation verified that the waveform is getting close to the ideal state obtained from this method and, in addition, faults point position corresponding to the wavehead can be identified visually.5. According to processing results of faults signal de-noising, fault feature reduction and fault classification, a Multi-Agent model of power cable fault diagnosis, which can build fuzzy reasoning on the basis of the relationship between fault symptoms and faults, is designed in this paper. Multiple functional Agents are also designed. Multi-Agent groups take the approach of combining contract net model with blackboard model to coordinate the work of fault detection among the groups. Based on the relationship between fault symptoms and faults, fault diagnosing model is established by applying fuzzy Petri Nets. Then fault diagnosing model of fuzzy Petri Nets be built to diagnosis reasoning structure through the category of faults, which can diagnose faults of more reasons leading to more results effectively. Calculation analysis proved the effectiveness and reliability of this diagnosing model.更多还原