基于LS-SVM和IMF能量矩的配电网故障区段定位方法
发布时间：2018-11-12 14:12
【摘要】：目前我国正处于建设智能电网的关键阶段,而配电网保护智能化正是其重要的研究方向。由于配电网接地故障时故障电流小,电流信号容易受到外界环境的干扰,因此故障特征量变得难以检测,而且接地故障时往往伴随着非故障相电压的升高,使得馈线绝缘薄弱处易于击穿,故障更有可能发展为多相、多点短路,扩大事故范围。因此有必要找到一种能在各种复杂信号中识别出故障特征量,并快速定位故障区段的方法。论文首先讨论了现有的小电流故障定位的理论,从主动式和被动式两方面介绍了目前不同定位方法的优点和缺点,并在被动式保护方法中着重论述了经验模态分解(EMD)和支持向量机在电力系统上的应用;提出采用集合经验模态分解(EEMD)算法对信号进行处理,它除了继承小波分解等算法的优点外,还解决了EMD信号分解时容易产生的模态混叠现象,并通过三组对比算例,就不同算法在信号提取方面的性能进行了对比;简要介绍了最小二乘支持向量机(LS-SVM)在特征信号分类方面的特点,并与BP神经网络和支持向量机就分类的快速性和准确性上进行了比较;提出了基于本征模态函数(IMF)能量矩和LS-SVM的故障区段定位方法,利用EEMD分解电流信号得到IMF,进而将IMF与时间积分获得能量矩,最后将能量矩作为特征向量输入到LS-SVM,训练得到故障区段定位模型并用于未知故障的定位;然后基于数字化采样技术,将新的故障区段定位方法应用到智能电网中,定位采用低功率互感器测量电流,并结合IEEE1588时钟同步原理与电力云,使得故障区段定位精度有了进一步提高;最后采用MATLAB GUI平台编写了图形用户界面,实现了与用户间的交互式操作。将不同的信号提取与分类算法在信号处理性能方面分别进行比较,通过对比算例表明EEMD更能反映原信号各自分量的特点,而LS-SVM的分类准确率更高,速度更快。然后结合这两种方法提出了新的故障区段定位理论,通过10kV的配电网故障仿真,表明基于IMF能量矩的LS-SVM故障定位新方法能有效定位不同区段配电网接地故障,具备在不同接地电阻下的区段定位的能力。
[Abstract]:At present, China is in the key stage of smart grid construction, and intelligent distribution network protection is an important research direction. Because the fault current is small and the current signal is easily disturbed by the external environment, the fault characteristic change is difficult to detect, and the non-fault phase voltage increases when the grounding fault occurs. The fault is more likely to develop into multi-phase, multi-point short circuit and extend the range of accidents. Therefore, it is necessary to find a method that can identify the fault characteristic quantity in various complex signals and locate the fault section quickly. This paper first discusses the existing theory of small current fault location, and introduces the advantages and disadvantages of different localization methods from the active and passive aspects. In the passive protection method, the application of empirical mode decomposition (EMD) and support vector machine (SVM) in power system is discussed. A set empirical mode decomposition (EEMD) algorithm is proposed to process signals, which not only inherits the advantages of wavelet decomposition, but also resolves the phenomenon of modal aliasing which is easy to occur in the decomposition of EMD signals. The performance of different algorithms in signal extraction is compared. This paper briefly introduces the features of least squares support vector machine (LS-SVM) in feature signal classification, and compares with BP neural network and support vector machine on the rapidity and accuracy of classification. A fault zone location method based on intrinsic mode function (IMF) energy moment and LS-SVM is proposed. The IMF, is obtained by decomposing the current signal by EEMD, and then the energy moment is obtained by integrating IMF with time. Finally, the energy moment is input into the LS-SVM, as the eigenvector to obtain the location model of the fault section and be used to locate the unknown fault. Then, based on the digital sampling technology, the new fault section location method is applied to smart grid. The low power transformer is used to measure the current, and combined with the principle of IEEE1588 clock synchronization and the power cloud. The location accuracy of the fault section has been further improved; Finally, the graphical user interface is written on MATLAB GUI platform, which realizes the interactive operation with users. The different signal extraction and classification algorithms are compared in signal processing performance. The results show that EEMD can better reflect the characteristics of each component of the original signal, while the classification accuracy of LS-SVM is higher and the speed is faster. Then, combining these two methods, a new fault section location theory is proposed. Through the fault simulation of 10kV distribution network, it is shown that the new method of LS-SVM fault location based on IMF energy moment can effectively locate the grounding fault of distribution network in different sections. Ability to position sections at different ground resistances.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM862
[Abstract]:At present, China is in the key stage of smart grid construction, and intelligent distribution network protection is an important research direction. Because the fault current is small and the current signal is easily disturbed by the external environment, the fault characteristic change is difficult to detect, and the non-fault phase voltage increases when the grounding fault occurs. The fault is more likely to develop into multi-phase, multi-point short circuit and extend the range of accidents. Therefore, it is necessary to find a method that can identify the fault characteristic quantity in various complex signals and locate the fault section quickly. This paper first discusses the existing theory of small current fault location, and introduces the advantages and disadvantages of different localization methods from the active and passive aspects. In the passive protection method, the application of empirical mode decomposition (EMD) and support vector machine (SVM) in power system is discussed. A set empirical mode decomposition (EEMD) algorithm is proposed to process signals, which not only inherits the advantages of wavelet decomposition, but also resolves the phenomenon of modal aliasing which is easy to occur in the decomposition of EMD signals. The performance of different algorithms in signal extraction is compared. This paper briefly introduces the features of least squares support vector machine (LS-SVM) in feature signal classification, and compares with BP neural network and support vector machine on the rapidity and accuracy of classification. A fault zone location method based on intrinsic mode function (IMF) energy moment and LS-SVM is proposed. The IMF, is obtained by decomposing the current signal by EEMD, and then the energy moment is obtained by integrating IMF with time. Finally, the energy moment is input into the LS-SVM, as the eigenvector to obtain the location model of the fault section and be used to locate the unknown fault. Then, based on the digital sampling technology, the new fault section location method is applied to smart grid. The low power transformer is used to measure the current, and combined with the principle of IEEE1588 clock synchronization and the power cloud. The location accuracy of the fault section has been further improved; Finally, the graphical user interface is written on MATLAB GUI platform, which realizes the interactive operation with users. The different signal extraction and classification algorithms are compared in signal processing performance. The results show that EEMD can better reflect the characteristics of each component of the original signal, while the classification accuracy of LS-SVM is higher and the speed is faster. Then, combining these two methods, a new fault section location theory is proposed. Through the fault simulation of 10kV distribution network, it is shown that the new method of LS-SVM fault location based on IMF energy moment can effectively locate the grounding fault of distribution network in different sections. Ability to position sections at different ground resistances.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM862
【相似文献】
相关期刊论文 前10条
1 卫志农;孙国强;于峰;;配电网故障区段定位[J];重庆理工大学学报(自然科学版);2010年01期
2 张愿章;余艳伟;;配电网故障区段判别与隔离改进算法[J];河北工业大学学报;2009年04期
3 张颖;周韧;钟凯;;改进蚁群算法在复杂配电网故障区段定位中的应用[J];电网技术;2011年01期
4 宗剑;陈静;;基于概率条件下配电网故障区段定位方法[J];上海应用技术学院学报(自然科学版);2007年02期
5 朱江;;利用信息化技术快速确定配网电缆故障区段研究[J];上海电力;2007年05期
6 史燕琨;肖湘宁;邹积岩;;基于边界保护的配电网故障区段无通信定位方法[J];电网技术;2009年04期
7 刘健,倪建立,杜宇;配电网故障区段判断和隔离的统一矩阵算法[J];电力系统自动化;1999年01期
8 林秋金;苏燕红;;判别中低压配电网故障区段的方法[J];农村电气化;2012年03期
9 刘耀湘;乐秀t，
本文编号：2327344
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2327344.html
教材专著
