RST和NBN用于电力变压器故障诊断  

Power Transformer Fault Diagnosis Using RST and NBN

作　　者：黄辉先[1] 肖桂枝[1] 阳敏[1] 

机构地区：[1]湘潭大学信息工程学院,湘潭411105

出　　处：《高电压技术》2009年第7期1589-1594,共6页High Voltage Engineering

基　　金：湖南省自然科学基金(06JJ50112)~~

摘　　要：电力变压器发生故障后,当故障信息存在不完整或不确定性,甚至关键信息丢失时,会导致故障诊断难以得出正确结论。针对此问题,提出了一种将粗糙集理论(RST)与朴素贝叶斯网络(NBN)结合的电力变压器故障诊断新方法。首先将油中溶解气体分析(DGA)结果和其他电气试验结果作为条件属性,故障区域作为决策属性,考察各种故障与征兆间的连接关系并建立决策表,接着利用基于可辨识矩阵和信息熵的属性约简算法实现对专家知识的简化与故障特征的压缩,提取最佳属性约简组合,然后以最佳属性约简组合形成的约简决策表建立朴素贝叶斯网络模型,利用贝叶斯网络实现概率推理,便于描述故障特征的变化及对变压器故障原因的快速分析。最后对变压器故障进行实例分析,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。If the fault information of power transformer is incomplete or indeterminate or even the key information is lost when power transformer goes wrong, correct conclusion could not be given by fault diagnosis. To settle this problem, we proposed a new power transformer fault diagnosis method in which the rough set （RS） theory is well integrated with Native Bayesian Network （NBN）. First, the results of dissolved gas-in-oil analysis （DGA） and conventional electrical tests were taken as conditional attributes and the faulty region was taken as decision attributes. Various connection relations between fault andsymptom were investigated and decision table was established. Next, the optimal attribute reduction combination can be obtained by using attribute reducing method based on cognizable matrix and information entropy to simplify expert knowledge and to reduce fault symptoms. Then, the Native Bayesian Networks model was established according to the reduction decision table formed by the optimal attribute reduction combination, and the nodal probability is trained. Finally, the correctness and effectiveness of this method were validated by the results of practical fault diagnosis examples.

关 键 词：电力变压器 故障诊断 粗糙集 朴素贝叶斯网络 约简 信息熵 

分 类 号：TM41[电气工程—电器]