【作者】 徐志强； 刘明光；

【Author】 Xu Zhiqiang, Liu Mingguang(School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044)

【机构】 北京交通大学电气工程学院；

【摘要】 铁路交通运输安全的重要保障设施之一是信号灯,一旦信号灯电源中断供电,将导致信号指挥系统混乱、失灵、引起机车停滞、阻塞,甚至撞车等重大事故。铁路信号供电线路中单相接地是常见的故障、准确诊断故障对保证铁路运输安全尤为重要。本文提出将单相接地故障信息看作一个多维随机变量,对这类信息处理可转化为一个模式识别问题。铁路信号电源的一次故障信息是小样本数据,为了快速、准确识别故障,本文应用一种新理论——支持向量机(Suppor Vector Machines,SVM),来进行故障诊断和识别。与传统统计学相比,基于统计学习理论(Statisti-cal Learning Theory)的支持向量机是在结构风险最小化原则基础上,发展起来的专门研究小样本情况下机器学习规律的一种新理论。在故障状态下对故障线路和非故障线路持征参量构造样本序列,使用SVM分类器进行训练,仿真分析的结果表明该方法对单相接地故障信息的识别分类可靠,其具有较好的分类推广性,方法简单、直观。更多还原

【Abstract】 Signal light is one of important support facilities for railway transportation security. Once power source of signal light is broke off, it will result in commanding system confusion, out of order and cause locomotive stagnation, jamming and knock down. Single- phase- to-ground fault is common in the power supply circuit of railway signal. Accurate diagnosing fault is special importance to ensure railway transportation. Here fault information for Single - phase - to - ground is considered as many dimensional random variables, so which can convert pattern recognition problem to treat. The fault information for railway signal power source in a fault is little sample data. In order to quick and accurate recognizing fault, a new theory was introduced, which is support Vector Machines (SVM) . It is applied to diagnose and recognize fault. Comparing with conventional statistics, SVM based on statistical learning theory develop on the basis of structural risk minimization, research specially machine learning rule under the little sample. By way of structuring sample sequences for characteristic variables of fault and non - fault circuit, and using SVM classifier to train them, the emulating results indicate that this method classify credibility for single - phase - to - ground fault infor- mation, and has good classifying popularization, simplicity and direct - viewing.更多还原