基于SVM的变压器故障诊断研究

发布时间：2018-03-31 14:58

本文选题：变压器　切入点：参数寻优　出处：《安徽理工大学》2017年硕士论文

【摘要】：伴随着世界工业水平的快速提升,全国电网的互联已大致形成,对于电力系统安全稳定运行的要求越来越严格。变压器作为电能传送过程当中的核心装备之一,对于整个电力系统而言,起到了电能输送以及电压等级变换的作用。然而,其也是导致电力系统网络发生故障的电气装备之一。能否准确地预判出变压器的潜伏性故障类型,对于变压器的安全稳定运行是至关重要的。因此,针对于变压器的运行状态进行监控是必要的。本文将围绕变压器故障诊断方面的内容进行相关阐述。就变压器的诊断方式而言,当其是建立在分析变压器油中溶解气体的故障诊断方法时,相对比较稳妥。首先,本文将从油中溶解气体的来源、溶解、耗损等方面内容的介绍之后提出变压器故障的类型及传统意义上的诊断方法(三比值法)。由于三比值法在诊断速率、准确率等方面存在缺陷,进而提出了结合支持向量机(SVM)算法的模型诊断方式。不过,仅靠单一的诊断方式进行故障的判定,其效果必然是不尽人意的。为了进一步提高诊断的准确率以及收敛速度,提出了交叉验证的方式以及粒子群算法对支持向量机的核函数进行参数c/g的改良,从而得到更加合适的分类效果及精确率。对变压器的故障类型进行相应的编码之后,建立变压器的故障诊断仿真模型,得出基于交叉验证的SVM分类结果是84%,粒子群算法优化后得到的SVM分类结果是85.3333%。从而验证了优化核函数的参数c/g的可行性,进而验证了选取以SVM为核心的变压器故障诊断方法是可行的。
[Abstract]:With the rapid improvement of the world industrial level, the interconnection of the national power grid has been formed, and the requirements for the safe and stable operation of the power system are becoming more and more strict.As one of the core equipments in the process of power transmission, transformer plays the role of power transmission and voltage grade conversion for the whole power system.However, it is also one of the electrical equipments that lead to the failure of power system network.It is very important for the safe and stable operation of transformers to determine accurately the latent fault types of transformers.Therefore, it is necessary to monitor the operation state of transformer.This paper will focus on transformer fault diagnosis related to the content of the elaboration.As far as the diagnosis mode of transformer is concerned, it is relatively safe to establish a fault diagnosis method based on the analysis of dissolved gas in transformer oil.First of all, after introducing the source, dissolution and consumption of dissolved gas in oil, this paper puts forward the types of transformer faults and the traditional diagnosis method (three-ratio method).Due to the shortcomings of the three-ratio method in diagnosis rate and accuracy, a model diagnosis method combining support vector machine (SVM) algorithm is proposed.However, only by a single diagnosis method to determine the fault, its effect must not be satisfactory.In order to further improve the accuracy and convergence rate of diagnosis, the cross-validation method and particle swarm optimization (PSO) algorithm are proposed to improve the kernel function parameters of support vector machine (SVM), c / g, so as to obtain a more appropriate classification effect and accuracy rate.After the corresponding coding of transformer fault type, the simulation model of transformer fault diagnosis is established, and the result of SVM classification based on cross verification is 84 and the SVM classification result obtained by particle swarm optimization is 85.3333.The feasibility of optimizing the parameter c / g of the kernel function is verified, and the feasibility of selecting the transformer fault diagnosis method with SVM as the core is verified.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM407

【参考文献】