高电压技术
高電壓技術
고전압기술
HIGH VOLTAGE ENGINEERING
2012年
11期
2964-2971
,共8页
故障诊断%电力变压器%溶解气体分析(DGA)%主元分析(PCA)%相对重构贡献(rRBC)%灰关联
故障診斷%電力變壓器%溶解氣體分析(DGA)%主元分析(PCA)%相對重構貢獻(rRBC)%灰關聯
고장진단%전력변압기%용해기체분석(DGA)%주원분석(PCA)%상대중구공헌(rRBC)%회관련
fault diagnosis%power transformer%dissolved gas analysis(DGA)%principal component analysis(PCA)%relative reconstruction-based contribution(rRBC)%grey relation entropy(GRE)
为了充分挖掘油中溶解气体分析(DGA)数据隐藏的故障特征信息,提出了一种基于相对重构贡献(rRBC)的变压器故障诊断新方法。该方法首先利用DGA数据建立主元分析(PCA)模型,基于故障重构的思想,计算样本各变量重构贡献率(RBC);考虑各变量重构贡献率之间的可比性,计算其相对重构贡献率并作为特征量,通过归一化处理来提取故障特征;然后,建立变压器分层故障诊断模型,利用灰关联熵(GRE)信息利用率高等优点,求出待诊模式与各标准模式的综合灰熵关联序,实现故障诊断。实例研究结果表明,所提出的相对重构贡献灰关联熵方法与重构贡献灰关联熵、灰关联熵方法相比,使特征样本集的可分性变大,提高了分类正确率。
為瞭充分挖掘油中溶解氣體分析(DGA)數據隱藏的故障特徵信息,提齣瞭一種基于相對重構貢獻(rRBC)的變壓器故障診斷新方法。該方法首先利用DGA數據建立主元分析(PCA)模型,基于故障重構的思想,計算樣本各變量重構貢獻率(RBC);攷慮各變量重構貢獻率之間的可比性,計算其相對重構貢獻率併作為特徵量,通過歸一化處理來提取故障特徵;然後,建立變壓器分層故障診斷模型,利用灰關聯熵(GRE)信息利用率高等優點,求齣待診模式與各標準模式的綜閤灰熵關聯序,實現故障診斷。實例研究結果錶明,所提齣的相對重構貢獻灰關聯熵方法與重構貢獻灰關聯熵、灰關聯熵方法相比,使特徵樣本集的可分性變大,提高瞭分類正確率。
위료충분알굴유중용해기체분석(DGA)수거은장적고장특정신식,제출료일충기우상대중구공헌(rRBC)적변압기고장진단신방법。해방법수선이용DGA수거건립주원분석(PCA)모형,기우고장중구적사상,계산양본각변량중구공헌솔(RBC);고필각변량중구공헌솔지간적가비성,계산기상대중구공헌솔병작위특정량,통과귀일화처리래제취고장특정;연후,건립변압기분층고장진단모형,이용회관련적(GRE)신식이용솔고등우점,구출대진모식여각표준모식적종합회적관련서,실현고장진단。실례연구결과표명,소제출적상대중구공헌회관련적방법여중구공헌회관련적、회관련적방법상비,사특정양본집적가분성변대,제고료분류정학솔。
In order to fully get the hidden fault feature information in the dissolved gas analysis(DGA)data,a new method of the transformer fault diagnosis based on relative reconstruction-based contribution(rRBC)is proposed.The reconstruction-based contribution(RBC)of variables are calculated based on the idea of fault reconstruction after establishing aprincipal component analysis(PCA)model of dissolved gas-in-oil.Then,considering the RBCs' comparability of each variable,the feature information of oil dissolved gas is extracted by calculating the relative RBCs and normalization.Finally,a hierarchical structure of transformer fault diagnosis model is built up,and due to the high information utilizing rate of the grey relation entropy(GRE),the synthesis GRE incidence order of the diagnosing pattern with standard patterns is computed to realize the fault analysis.The experimental results show that,compared with the features extracted by RBC-GRE and GRE,the proposed rRBC method increases the separability of data set and the classification accuracy.