基于信息瓶颈的变压器故障诊断算法  

Transformer fault diagnosis algorithm based on information bottleneck

作　　者：路光辉[1] 姬波[2] 雍明超[1] 周钟[1] 牧继清[1] 

机构地区：[1]许继集团有限公司,河南许昌461000 [2]郑州大学信息工程学院,河南郑州450001

出　　处：《计算机工程与设计》2015年第8期2271-2276,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61170223);国家自然科学基金河南人才培养联合基金项目(U1204610);河南省科技攻关计划基金项目(132102210404)

摘　　要：为保证电网安全运行,解决变压器故障早期预警问题,提出一种基于信息瓶颈的变压器故障诊断方法。该方法分为两个阶段,采用信息瓶颈方法对数据进行聚类,得到簇内相似性最大的结果簇;通过簇内训练样本的简单多数投票,确定待测样本的故障类别。实际数据集上的实验结果表明,该方法是可行有效的,相比大卫三角形法、贝叶斯算法和神经网络算法,该算法的分类结果正确率分别提高了14.65%、25.00%和11.23%。To ensure the safe operation of power grid, a transformer fault diagnosis algorithm based on information bottleneck was proposed for transformer fault early-warning. The proposed method had two steps. Firstly, data were clustered and result clusters were gained using information bottleneck algorithm. Secondly, the fault type of the test sample was determined by sim- ply majority voting among training samples in the cluster which test sample ＇belongs to. Results of experiments on real datasets show that the proposed method is feasible and effective. Compared with the accuracies obtained using Dural method, Bayes algo- rithm and BPNN algorithm, that obtained using the proposed algorithm increases 14.65%, 25.00% and 11.23% respectively.

关 键 词：变压器 诊断算法 信息瓶颈 聚类 大卫三角形方法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]