【作者】 肖桂枝；

【导师】 黄辉先；

【作者基本信息】 湘潭大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 电力变压器是电力系统中最重要的设备之一,其运行状态直接关系到整个电力系统的可靠性。在电力系统中变压器运行时出现故障的情况时有发生,对电力系统的正常运行造成了严重威胁。变压器故障诊断是根据故障特征来判断其故障类型、定位故障位置或者确定故障原因等,为变压器的检修提供智能化的决策。本文的主要工作包括:(1)对应用于电力变压器故障诊断的各种人工智能方法进行了深入研究,分析它们在电力变压器故障诊断中应用的特点以及存在的主要问题。(2)查阅了大量生物免疫学文献,总结了生物免疫系统的一些重要机理和特征,以及人工免疫系统及人工免疫算法的研究应用现状。(3)研究基于粗糙集理论的知识提取算法,包括基于可辨识矩阵和数学逻辑运算的属性约简算法和基于互信息的最佳属性约简组合提取算法。研究了基于朴素贝叶斯网络的故障诊断模型及算法。(4)提出了克隆免疫网络分类算法,并通过仿真实验和实际应用检验了该算法在变压器故障定性诊断中应用的可行性。(5)提出了一种将粗糙集理论与朴素贝叶斯网络有机结合在一起的电力变压器故障定位诊断方法,并通过实例分析证明了该方法的有效性。本文提出的优化算法——克隆免疫网络分类算法,有别于传统的免疫算法,是将基于群体免疫算法中的克隆选择原理与独特型免疫网络原理相结合,改进训练抗原、记忆抗体在形态空间的表现形式,即增加训练抗原、记忆抗体的类别信息,并且设计了“抗体选择”、“抗体克隆”、“抗体重组’、“抗体变异”和“抗体抑制”等算子,并通过在电力变压器故障定性诊断中的应用,证明了该算法对故障性质的判断具有很高的准确率。本文提出的将粗糙集理论和朴素贝叶斯网络相结合的电力变压器故障定位诊断模型,在故障诊断过程中综合利用油中溶解气体分析与电气试验结果中的有效信息,对电力变压器的故障进行了定位诊断,诊断结果证明该方法是正确和有效的,具有较好的实用价值。更多还原

【Abstract】 Power transformer is one of the most important equipment in power system, the operation status of which will be directly related to the reliability of the entire power system. In power system, transformer fault happens sometimes, which threatens the normal operation of power system. Transformer fault diagnosis is based on the fault features to judge the fault type, locate the fault position or determine the fault reason, it can provides an intelligent decision-making for daily transformer maintenance.In this paper, the work includes:(1) Carry out an in-depth study on a variety of artificial intelligence methods which are applied to power transformer fault diagnosis, and analysis these methods’features and exsiting main problems.(2) Access to a large number of biological immunology literatures; summarize some important mechanism and features of biological immune system, and some applications of research of artificial immune system and artificial immune algorithm.(3) Research the kownledge extraction algorithm which based on rough set theory, it contains the attribute reduction algorithm based on discernibility matrix and mathematical logic operations, and the optimal attribute reduction combination extraction algorithm based on information entropy.(4) Propose clone immune network classification algorithm, and test the feasibility of this algorithm when it is used in transformer fault qualitative diagnosis through simulation experiments and practical application.(5) Propose a power transformer fault positioning diagnosis method in which the rough set theory is well integrated with native Bayesian network, and the effectiveness of this method are validated by the result of practical fault diagnosis examples.The optimization algorithm called clone immune network classification algorithm which was proposed in this paper, integrate population-based clonal selection principle and idiotype immune network theory, improve the representation of antigen and antibody, namely, add type information into training antigen and memory antibody, and design "antibody selection", "antibody clone", "antibody reorganization", "antibody mutation" and "antibody suppression" operators, the high judgement of failure of this algorithm is proved by the practical application.In this paper, the power transformer fault positioning diagnosis model which the rough set theory is combine with the native Bayesian network, is using the effective information of dissolved gas analysis and electrical test result in the process of fault diagnosis, and carring out the positioning diagnosis of power transformer fault. The diagnosis results show that the method is correct and effective, and have good practical value.更多还原