基于粗糙集理论与贪心算法的变压器故障诊断方法研究

发布时间：2018-02-23 18:07

本文关键词： 变压器故障诊断粗糙集贪心算法离散化　出处：《东北大学》2014年硕士论文　论文类型：学位论文

【摘要】：电力变压器是电力系统重要的运行设备之一,由于变压器内部结构复杂,电场分布不均匀,且随着电压等级增高,事故率成上升趋势。其故障不仅会给自身带来重大损失,还会对电力系统的安全造成重大影响。能否有效地检测和诊断变压器故障,准确判断变压器的潜伏性故障或故障隐患具有重要意义。本文围绕变压器的故障诊断这个主题,着眼于研究一种新型实用的变压器故障诊断方法,对现有变压器故障诊断方法存在的不足进行了改进,设计了变压器状态分析系统的故障诊断子系统并论述了电力变压器的常规试验和试运行。首先,本文论述了变压器故障诊断的要求和意义,在对当前采用的几种主要的变压器故障诊断方法进行比较后,提出利用对变压器油中溶解气体的分析来检测变压器内部故障。考虑到现有的变压器油中溶解气体的分析法的独立性差、局限性强、对造成变压器故障的各种因素考虑不够周全等问题,本文提出了一种新型的基于变压器油中溶解气体含量的变压器故障诊断方法。针对变压器油中溶解气体的采集存在某种程度的不确定性这一问题,本文利用的粗糙集理论具有在处理不精确问题时能在保留关键信息的前提下对数据进行化简并求得知识的最小表达的特点,应用粗糙集理论对海量的历史数据进行处理,求取最终约简规则,从而将变压器油中溶解气体的采集所造成的不确定性降到最低。其次本文考虑求取全局最优的变压器故障诊断规则,采用了基于粗糙集和布尔逻辑相结合的连续属性离散化方法,并采用可辨识矩阵算法进行属性约简和值约简,从而保证得到全局最优结果。但是在实际约简过程中,由于算法的时间复杂度和空间复杂度是以规则数和属性数乘积的立方形式增长,因此,可辨识矩阵算法在处理海量数据时是NP-hard问题,实际中无法求出最小约简。由此我们提出了一种实用的变压器故障诊断方法,以变压器油中溶解的所有气体的百分含量及重要比值作为条件属性建立决策表,利用贪心算法来对此决策表进行连续属性离散化、属性约简以及值约简。整个约简过程采用统一的约简算法,大大减少了计算时间及计算量。通过对此实用的新型变压器故障诊断方法进行仿真验证,得到了较好的效果。最后,指出了变压器故障诊断方法在理论研究和实际应用中所存在的问题,并对下一步研究工作进行了展望。
[Abstract]:Power transformer is one of the important operating equipments in power system. Because of its complex internal structure, uneven distribution of electric field, and increasing voltage grade, the accident rate increases. It is of great significance to detect and diagnose transformer faults effectively and accurately judge the latent faults or hidden troubles of transformers. This paper focuses on the subject of transformer fault diagnosis. Based on the research of a new practical transformer fault diagnosis method, the shortcomings of the existing transformer fault diagnosis methods are improved. The fault diagnosis subsystem of transformer state analysis system is designed, and the routine test and trial operation of power transformer are discussed. Firstly, the requirements and significance of transformer fault diagnosis are discussed in this paper. After comparing several main methods of transformer fault diagnosis, It is proposed that the analysis of dissolved gas in transformer oil be used to detect the internal fault of transformer. Considering the poor independence of the analysis method of dissolved gas in transformer oil, the limitation is strong. Problems such as inadequate consideration of various factors causing transformer failures, In this paper, a new transformer fault diagnosis method based on dissolved gas content in transformer oil is proposed. The rough set theory, which is used in this paper, can simplify the data and obtain the minimum representation of knowledge on the premise of keeping the key information while dealing with imprecise problems. The rough set theory is used to process the massive historical data. The final reduction rule is obtained to minimize the uncertainty caused by the acquisition of dissolved gas in transformer oil. Secondly, the global optimal rule of transformer fault diagnosis is considered in this paper. The continuous attribute discretization method based on rough set and Boolean logic is adopted, and the discernible matrix algorithm is used to reduce the attribute and the value, so as to obtain the global optimal result. Because the time complexity and space complexity of the algorithm increase in cubic form of the product of the rule number and the attribute number, the identifiable matrix algorithm is a NP-hard problem when dealing with the massive data. A practical method for transformer fault diagnosis is proposed, in which the percentage content and important ratio of all gases dissolved in transformer oil are taken as conditional attributes to establish a decision table. The continuous attribute discretization, attribute reduction and value reduction are used to discretize the decision table by greedy algorithm. The calculation time and calculation amount are greatly reduced. The simulation results of this practical new transformer fault diagnosis method are verified and good results are obtained. The problems of transformer fault diagnosis in theory research and practical application are pointed out, and the future research work is prospected.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP18;TM41

【相似文献】