基于朴素贝叶斯的电力变压器故障诊断

发布时间：2018-01-02 03:07

本文关键词：基于朴素贝叶斯的电力变压器故障诊断　出处：《上海交通大学》2015年硕士论文　论文类型：学位论文

【摘要】：针对电力变压器在电力系统中的特殊地位和重要作用,对其日常运维中的常见故障及其机理进行了深入研究,确定了电力变压器各种故障的排除方法。电力变压器发生故障后,当其故障信息不完整或不确定时,会导致处理故障的难度大大增加,有时甚至会拖延返厂维修时机。通过对目前较为常用的几种故障诊断方法的特点进行对比,凸显了贝叶斯网络的在这方面优势,确定了基于贝叶斯网络的变压器故障诊断方法。由于TAN模型属性变量关联边的不确定性以及因果关系模型的主观性,本文确定了朴素贝叶斯网络作为故障诊断方法,并通过简单实例分析,证明了模型具有很强的可靠性和良好的准确率。在完备数据分析的基础上,简要概括了在属性缺失情况下的EM及SEM算法。给出了利用朴素贝叶斯网络进行故障诊断的基本方法。同项目结合,在对一定规模的数据样本进行分析处理的基础上,对网络模型进行了仿真计算。通过对结果的分析,说明了贝叶斯网络在故障诊断领域的出色能力和广阔前景。论文最后探讨了应用贝叶斯网络进行故障诊断工作的重点以及研究方向。
[Abstract]:In view of the special position and important role of power transformer in power system, the common faults and their mechanism in daily operation and maintenance are deeply studied. The troubleshooting methods of power transformer are determined. When the fault information of power transformer is incomplete or uncertain, the difficulty of dealing with the fault will be greatly increased. By comparing the characteristics of several commonly used fault diagnosis methods, it highlights the advantages of Bayesian network in this respect. The method of transformer fault diagnosis based on Bayesian network is determined. Because of the uncertainty of attribute variable correlation edge of TAN model and the subjectivity of causality model. In this paper, the naive Bayesian network is determined as the fault diagnosis method, and through a simple example analysis, it is proved that the model has strong reliability and good accuracy, on the basis of complete data analysis. The EM and SEM algorithms in the absence of attributes are briefly summarized, and the basic method of fault diagnosis using naive Bayesian network is given. On the basis of analyzing and processing the data samples of a certain scale, the network model is simulated and calculated. The excellent capability and broad prospect of Bayesian network in the field of fault diagnosis are explained. Finally, the emphasis and research direction of the application of Bayesian network in fault diagnosis are discussed.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM407

【参考文献】