计算机工程与设计
計算機工程與設計
계산궤공정여설계
COMPUTER ENGINEERING AND DESIGN
2015年
8期
2271-2276
,共6页
路光辉%姬波%雍明超%周钟%牧继清
路光輝%姬波%雍明超%週鐘%牧繼清
로광휘%희파%옹명초%주종%목계청
变压器%诊断算法%信息瓶颈%聚类%大卫三角形方法
變壓器%診斷算法%信息瓶頸%聚類%大衛三角形方法
변압기%진단산법%신식병경%취류%대위삼각형방법
transformer%fault diagnosis algorithm%information bottleneck%cluster%Duval method
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%。
為保證電網安全運行，解決變壓器故障早期預警問題，提齣一種基于信息瓶頸的變壓器故障診斷方法。該方法分為兩箇階段，採用信息瓶頸方法對數據進行聚類，得到簇內相似性最大的結果簇；通過簇內訓練樣本的簡單多數投票，確定待測樣本的故障類彆。實際數據集上的實驗結果錶明，該方法是可行有效的，相比大衛三角形法、貝葉斯算法和神經網絡算法，該算法的分類結果正確率分彆提高瞭14.65%、25.00%和11.23%。
위보증전망안전운행，해결변압기고장조기예경문제，제출일충기우신식병경적변압기고장진단방법。해방법분위량개계단，채용신식병경방법대수거진행취류，득도족내상사성최대적결과족；통과족내훈련양본적간단다수투표，학정대측양본적고장유별。실제수거집상적실험결과표명，해방법시가행유효적，상비대위삼각형법、패협사산법화신경망락산법，해산법적분류결과정학솔분별제고료14.65%、25.00%화11.23%。
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 Duval method,Bayes algo-rithm and BPNN algorithm,that obtained using the proposed algorithm increases 14.65%,25.00% and 11.23% respectively.
