分类号 TM7 密级 UDC
硕 士 学 位 论 文
支持向量机在输电线路故障识别中
的应用研究
学 位 申 请 人： 马新明
学 科 专 业： 电力系统及其自动化 指 导 教 师： 王成江 副教授
二○一○年五月
A Dissertation Submitted in Partial Fulfillment of the Requirements for
the Degree of Master of Science in Engineering
Research on Application of Support Vector Machine in Fault
Identification for Transmission lines
Graduate Student:
Ma xinming
Major: Power System and its Automation Supervisor: Associate prof. Wang chengjiang
China Three Gorges University Yichang, 443002, P.R.China
May, 2010
三峡大学学位论文原创性声明
本人郑重声明：所呈交的学位论文，是本人在导师的指导下，独立进行研究工作所取得的成果，除文中已经注明引用的内容外，本论文不含任何其他个人或集体已经发表或撰写过的作品成果。对本文的研究做出重要贡献的个人和集体均已在文中以明确方式标明，本人完全意识到本声明的法律后果由本人承担。
学位论文作者签名：
日 期：
I
内 容 摘 要
随着电力系统电网规模的不断发展，对输电线路的安全运行和供电可靠性的要求亦越来越高。输电线路作为电力系统传输电力的关键设备，在电力系统的安全运行中起着至关重要的作用，其发生的各种故障将会直接威胁着电力系统的安全运行，这些故障若不及时排除就将会对电力系统产生严重的甚至是灾难性的后果，而且其引起的经济损失也将会越来越大。输电线路故障的迅速排除是电力系统安全、可靠和经济运行的重要保证。当输电线路一旦发生故障时，将会在较短的时间段内给出大量的故障信息，如果单凭运行工作人员的工作经验进行故障分析，其正确性和快速性将会很难得到保证。因此及时、准确的判断并识别出这些故障就显得尤为重要，并且对电力系统管理人员恢复决策的制定和电力系统的安全运行具有重要意义。
支持向量机(Support Vector Machine)是基于统计学习理论基础上提出的一种新的模式识别方法。它基于结构风险最小化原则（SRM准则），并且兼顾训练误差和泛化能力，在解决小样本、非线性等模式识别问题中表现出其独有的优势。
本论文在研究支持向量机的二叉树分类的基础上，结合输电线路故障的特点，考虑其不同故障发生的优先级后，设计了基于SVM的改进二叉树输电线路故障多类分类器的模型，通过仿真实验选择了最小二乘支持向量机（LS-SVM）算法和线性函数转换表达式的归一化算法，并利用电力系统分析综合程序（PSASP）软件仿真小样本短路电流数据和短路电压数据，结合MATLAB7.0软件编写分类程序，对所建立的故障分类模型进行训练，得到了九个二分类SVM的判别函数系数矩阵和判别函数，再利用这九个判别函数系数矩阵重新编写故障分类器程序，最后利用测试样本对故障分类器进行测试，并且和其他SVM算法与BP神经网络算法等几种算法分类器的测试结果相比较。结果表明，在输电线路所有故障类型情况下，利用LS-SVM算法设计的改进二叉树输电线路故障分类器都具有很高的分类正确率，尤其是对两相接地和不接地短路电流分类的效果尤为显著，另外，本文所设计的故障分类器的数据预处理过程简单，分类步骤少，可以实现输电线路故障的快速分类。 关键词：支持向量机 二叉树 输电线路 故障识别
II
Abstract
With the development of power system, The requirements for the transmission lines safe operation and the reliability of power supply are also getting higher and higher.Transmission lines as the key equipment of power system, it plays a vital role in the operation of power system, a variety of its fault would be a direct threat to the safe operation of power system，if these fault would not be immediately removed then will produce serious and even catastrophic consequences for the power system, and its resulting economic loss will also be more greater. The rapid transmission line fault exclusion is an important guarantee for the power system security, reliable and economic operation. When the transmission line occurs fault, woule have a large number of fault information in a short time, if only have the experience of the workers to analysis, its speed and accuracy would be difficult to be guaranteed. Therefore, timely and accurate judgments, and identify these failures is particularly important. It have a great significance for power system management to making the decision of the restoration and the safe operation of. power systems.
Support Vector Machine(SVM) is a new pattern recognition method based on statistical learning theory. It is based on the principle of structural risk minimization (SRM guidelines), and the balance between training error and generalization capabilities, It shown its unique advantage in addressing the small sample, nonlinear problems. In this paper, based on the study of binary tree SVM multi-classification and combination of the characteristics of transmission line fault, considering the priority of the different faults, the design improvements based on binary tree SVM classifier model of transmission line fault, by simulation experiments to chosen least squares support vector machine (LS-SVM) algorithm and a linear function of converting an expression of the normalization algorithm, and use the power system analysis software package (PSASP) software simulation small samples of short-circuit voltage short-circuit current data, combination of MATLAB7.0 to write the classification procedures and to training the fault classification model, and gain nine SVM discriminant function coefficient matrix, then take it to re-write classification program, finally, use obtained SVM classifier and other algorithm and BP neural network classifier algorithms to compare. The results showed that all the fault type in the transmission line cases, the use of LS-SVM algorithm is designed to improve the binary tree transmission line fault
III