变电站过电压监测及其波形的研究

发布时间：2018-01-27 22:47

本文关键词： 电压传感器波形分析支持向量机参数优化 GUI　出处：《西华大学》2015年硕士论文　论文类型：学位论文

【摘要】：如今我国经济社会正处于持续高速发展阶段,电力需求长期保持快速增长,电力系统规模日益扩大。相应的电网运行方式及系统控制的复杂程度愈来愈高,如何降低事故风险保持电能的安全可靠输送成为了目前建设坚强智能电网的一项重大挑战。研究表明,各种设备的绝缘水平及运行方式是影响电力系统可靠性的重要因素,而造成系统中各种设备绝缘故障的主要原因则是内部过电压及外部过电压。对变电站电压信号进行实时监测及故障信号的处理分析有助于工作人员及时掌握系统运行情况,查找故障原因,及时发现绝缘薄弱处以便提出应对方案快速排除故障减少经济损失。并且通过过电压数据对故障发生原因的分析也有助于为电网设计提供更确实可靠的科学依据,为电网中各设备绝缘水平的确定及各类保护元件的合理选取配置提供帮助。本文首先介绍了一种氧化锌避雷器电压传感器,分别在工作特性,安全防护等方面对其可行性进行了分析。另外为了验证其可靠性及准确性,本文又加以仿真计算、试验测试及现场对比测量等手段。通过最终结果分析,氧化锌避雷器电压传感器在频率响应及线性度方面具有优良表现,可用其实现对暂态过电压的采集监测,所获信号数据准确度高。为有针对的研究所采集的暂态过电压数据,本文对各暂态过电压产生机理进行了详细分析。利用ATPDraw电磁暂态仿真软件对现场进行建模,通过仿真波形与实际测量波形相结合的方式研究各过电压特点,并基于树形结构对暂态过电压进行了分层归类,为智能识别系统的建立打下基础。特征量的选取在识别系统中起着至关重要的作用,本文在时频域分析的基础上利用小波变换分析、奇异值理论等手段对过电压数据进行处理,使信号局部特性得以凸显,提取了多种特征参量。根据各类暂态过电压特点,选取能表征相应过电压的特征参量建立了各层分类器。在暂态过电压识别系统实现的过程中,本文将传统二分类支持向量机理念与多分类支持向量机相结合采取了多层递进归类方式开发程序。为进一步提高识别系统的准确率及性能,首先采用主成分分析及归一化方法对数据进行预处理,然后综合使用交叉验证、遗传算法、粒子群优化算法三种手段对各层支持向量机参数进行优化,选取最优方式确定支持向量机参数。最后本文制作了暂态过电压识别系统图形操作界面,进一步提升了其交互性、可视性及操作性。
[Abstract]:Nowadays, the economy and society of our country are in the stage of sustained and high-speed development, the power demand has been growing rapidly for a long time, the scale of the power system is expanding day by day, and the complexity of the corresponding operation mode and system control of the power network is becoming more and more high. How to reduce the risk of accidents to maintain the safe and reliable transmission of electricity has become a major challenge in building a strong smart grid. The insulation level and operation mode of various equipments are important factors that affect the reliability of power system. The main causes of insulation failure of various equipments in the system are internal overvoltage and external overvoltage. The real-time monitoring of substation voltage signal and the processing and analysis of fault signal are helpful for the staff to master the system in time. Operational status. Find the cause of the failure. When the insulation weakness is discovered in time, the solution is put forward to eliminate the fault quickly and reduce the economic loss. The analysis of the cause of the fault through the overvoltage data is also helpful to provide a more reliable scientific basis for the power network design. . It can help to determine the insulation level of each equipment and the reasonable selection and configuration of various protective elements. Firstly, a voltage sensor of zinc oxide arrester is introduced in this paper. In addition, in order to verify its reliability and accuracy, this paper carries on the simulation calculation, the test test and the field contrast measurement and so on, through the final result analysis. The voltage sensor of zinc oxide arrester has excellent performance in frequency response and linearity, which can be used to collect and monitor transient overvoltage. The obtained signal data has high accuracy. It is the transient overvoltage data collected by the research institute. In this paper, the generation mechanism of transient overvoltage is analyzed in detail, and the field modeling is carried out by using ATPDraw electromagnetic transient simulation software. By combining the simulation waveform with the actual measurement waveform, the characteristics of each overvoltage are studied, and the transient overvoltage is classified in layers based on the tree structure. It lays the foundation for the establishment of intelligent recognition system. The selection of feature plays an important role in the recognition system. In this paper, wavelet transform is used on the basis of time-frequency domain analysis. Singular value theory and other means to process the over-voltage data, so that the local characteristics of the signal can be highlighted, extracted a variety of characteristic parameters, according to the characteristics of various transient overvoltage. Selecting the characteristic parameters which can represent the corresponding overvoltage, the classifier of each layer is established, and the realization of transient overvoltage identification system is carried out. In this paper, the traditional two-classification support vector machine and multi-classification support vector machine are combined to develop a multi-layer progressive classification program to further improve the accuracy and performance of the recognition system. First, the data are preprocessed by principal component analysis and normalization, then the parameters of support vector machine are optimized by three methods: crossover verification, genetic algorithm and particle swarm optimization. Finally, the graphical operation interface of transient overvoltage recognition system is made, which further improves its interaction, visibility and maneuverability.
【学位授予单位】：西华大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM63

【参考文献】