基于极性鉴别与暂态特征分布特性的电网过电压识别研究
发布时间：2018-11-09 14:21
【摘要】：随着电网规模的不断扩大、输送容量和电压等级的不断提高，电力系统过电压对输电线路和电气设备绝缘造成的危害越来越严重，因此研究电力系统过电压是保障电网安全可靠运行的重要课题。电力系统过电压种类较多，各类型过电压的产生机理不同，防护措施也不尽相同。实时监测电力系统出现的各种过电压信号，快速准确的判别故障类型，对处理故障和改善绝缘配合是十分必要的。 本文分别对输电线路雷电过电压与变电站过电压进行了研究。文章首先分析雷电过电压的产生机理，采用ATP-EMTP电磁暂态仿真软件搭建输电线路绕击和反击模型，对绕击和反击进行仿真，基于理论分析和仿真结果，比较三种雷电过电压之间的差异，基于时域分析和小波模极大值原理提取绝缘子电压极性、杆塔入地电流极性与杆塔入地电流突变极性特征量，建立了基于极性鉴别的雷电过电压识别方法。该方法同时引入绝缘子串电位差和杆塔入地电流两组物理量，更为完整地描述雷击物理过程，反映绕击与反击的本质差异。与以往方法只引入线路电压或线路电流单一物理量相比，方法的物理意义更为清晰直观。同时，本文所提方法只提取信号的极性特征，不依赖雷电过电压波头的细节特征，不易受到冲击电晕和行波传输过程中折反射的干扰，算法简便直观，，可靠性高。 基于对变电站常见过电压的产生机理和波形特征的分析归纳，得到不同类型过电压的暂态特征在时域的分布特点，提出各类型过电压的暂态特征提取区间，使所提取的特征量具有更好的可分性，放大各类型过电压间的差异。提出基于时域理论、频域理论、小波理论和奇异值分解理论的特征提取方法，在过电压相应的特征提取区间提取特征量。本文对变电站内常见混合过电压进行了归类，针对不同类型过电压的产生机理和时序上的因果关系，基于小波模极大值的分布特点对时域相连的不同类型过电压进行分解，实现混合过电压的分解及特征提取。 本文对过电压分类识别方法进行了讨论，提出基于暂态特征时域分布特性的编码分类方法，基于时域特征量对过电压进行了初步识别。基于小波时频特征量和奇异值统计特征量，提出基于最小二乘支持向量机的过电压识别方法。最后将本文所提分类识别方法有机整合，建立了变电站过电压分层识别系统。经实测数据验证，本文提出的特征提取及分类识别方法能对电力系统过电压进行有效辨识。
[Abstract]:With the expansion of power network scale and the continuous improvement of transmission capacity and voltage grade, the overvoltage of power system has caused more and more serious harm to the insulation of transmission lines and electrical equipment. Therefore, the study of power system overvoltage is an important issue to ensure the safe and reliable operation of power system. There are many kinds of overvoltages in power system. It is necessary to monitor all kinds of overvoltage signals in power system and identify fault types quickly and accurately to deal with faults and improve insulation coordination. In this paper, lightning overvoltage and substation overvoltage on transmission line are studied. In this paper, firstly, the mechanism of lightning overvoltage is analyzed, and the ATP-EMTP electromagnetic transient simulation software is used to build the transmission line wound failure and counterattack model, and the simulation results are based on the theoretical analysis and simulation results. By comparing the differences of three kinds of lightning overvoltages, the polarity of insulator voltage, the polarity of tower ground current and the abrupt polarity of tower ground current are extracted based on time domain analysis and wavelet modulus maximum principle. A method of lightning overvoltage recognition based on polarity discrimination is established. This method also introduces two sets of physical quantities: the potential difference of insulator string and the ground current of tower. The method describes the physical process of lightning stroke more completely and reflects the essential difference between round strike and counterattack. Compared with the single physical quantity of line voltage or line current, the physical meaning of the method is clearer and more intuitive. At the same time, the method proposed in this paper only extracts the polar characteristics of the signal, does not depend on the detailed characteristics of the lightning overvoltage wave head, and is not easy to be interfered by the shock corona and the refractive reflection during the traveling wave transmission. The algorithm is simple and intuitive and has high reliability. Based on the analysis of generation mechanism and waveform characteristics of common overvoltages in substations, the distribution characteristics of transient characteristics of different types of overvoltages in time domain are obtained, and the extraction interval of transient characteristics of different types of overvoltages is proposed. The extracted features have better separability and amplify the differences between different types of overvoltages. A feature extraction method based on time domain theory, frequency domain theory, wavelet theory and singular value decomposition theory is proposed. In this paper, the common mixed overvoltages in substations are classified. According to the generation mechanism of different types of overvoltages and the causality in time series, different types of overvoltages connected in time domain are decomposed based on the distribution characteristics of wavelet modulus maximums. The decomposition and feature extraction of hybrid overvoltage are realized. In this paper, the recognition method of overvoltage classification is discussed, and the coding classification method based on the time-domain distribution characteristics of transient features is proposed, and the primary recognition of overvoltage is carried out based on the time-domain characteristic quantity. Based on wavelet time-frequency characteristic and singular value statistical feature, an overvoltage recognition method based on least squares support vector machine (LS-SVM) is proposed. Finally, a hierarchical recognition system for substation overvoltage is established by integrating the classification and recognition methods proposed in this paper. The experimental results show that the proposed feature extraction and classification recognition method can effectively identify the overvoltage in power system.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM863
[Abstract]:With the expansion of power network scale and the continuous improvement of transmission capacity and voltage grade, the overvoltage of power system has caused more and more serious harm to the insulation of transmission lines and electrical equipment. Therefore, the study of power system overvoltage is an important issue to ensure the safe and reliable operation of power system. There are many kinds of overvoltages in power system. It is necessary to monitor all kinds of overvoltage signals in power system and identify fault types quickly and accurately to deal with faults and improve insulation coordination. In this paper, lightning overvoltage and substation overvoltage on transmission line are studied. In this paper, firstly, the mechanism of lightning overvoltage is analyzed, and the ATP-EMTP electromagnetic transient simulation software is used to build the transmission line wound failure and counterattack model, and the simulation results are based on the theoretical analysis and simulation results. By comparing the differences of three kinds of lightning overvoltages, the polarity of insulator voltage, the polarity of tower ground current and the abrupt polarity of tower ground current are extracted based on time domain analysis and wavelet modulus maximum principle. A method of lightning overvoltage recognition based on polarity discrimination is established. This method also introduces two sets of physical quantities: the potential difference of insulator string and the ground current of tower. The method describes the physical process of lightning stroke more completely and reflects the essential difference between round strike and counterattack. Compared with the single physical quantity of line voltage or line current, the physical meaning of the method is clearer and more intuitive. At the same time, the method proposed in this paper only extracts the polar characteristics of the signal, does not depend on the detailed characteristics of the lightning overvoltage wave head, and is not easy to be interfered by the shock corona and the refractive reflection during the traveling wave transmission. The algorithm is simple and intuitive and has high reliability. Based on the analysis of generation mechanism and waveform characteristics of common overvoltages in substations, the distribution characteristics of transient characteristics of different types of overvoltages in time domain are obtained, and the extraction interval of transient characteristics of different types of overvoltages is proposed. The extracted features have better separability and amplify the differences between different types of overvoltages. A feature extraction method based on time domain theory, frequency domain theory, wavelet theory and singular value decomposition theory is proposed. In this paper, the common mixed overvoltages in substations are classified. According to the generation mechanism of different types of overvoltages and the causality in time series, different types of overvoltages connected in time domain are decomposed based on the distribution characteristics of wavelet modulus maximums. The decomposition and feature extraction of hybrid overvoltage are realized. In this paper, the recognition method of overvoltage classification is discussed, and the coding classification method based on the time-domain distribution characteristics of transient features is proposed, and the primary recognition of overvoltage is carried out based on the time-domain characteristic quantity. Based on wavelet time-frequency characteristic and singular value statistical feature, an overvoltage recognition method based on least squares support vector machine (LS-SVM) is proposed. Finally, a hierarchical recognition system for substation overvoltage is established by integrating the classification and recognition methods proposed in this paper. The experimental results show that the proposed feature extraction and classification recognition method can effectively identify the overvoltage in power system.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM863
【参考文献】
相关期刊论文 前10条
1 李天云,赵妍,季小慧,李楠;HHT方法在电力系统故障信号分析中的应用[J];电工技术学报;2005年06期
2 杨丽君,廖瑞金,孙才新,梁帅伟,程涣超;基于Fisher判别法的油纸绝缘老化阶段识别[J];电工技术学报;2005年08期
3 杜林;戴斌;陆国俊;孙才新;王有元;;基于S变换局部奇异值分解的过电压特征提取[J];电工技术学报;2010年12期
4 王梦云;薛辰东;;1995～1999年全国变压器类设备事故统计与分析[J];电力设备;2001年01期
5 何正友,钱清泉;电力系统暂态信号分析中小波基的选择原则[J];电力系统自动化;2003年10期
6 李天云,赵妍,李楠;基于EMD的Hilbert变换应用于暂态信号分析[J];电力系统自动化;2005年04期
7 周凯;张涛;董秀成;张彼德;王军;席世友;;基于电容分压的配电网过电压在线监测[J];电力系统自动化;2007年21期
8 张峰;梁军;张利;
本文编号：2320665
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2320665.html
教材专著
