【摘要】 变压器油中溶解气体分析(DGA)是电力变压器绝缘诊断的重要方法。针对模糊C均值(FCM)聚类算法用于溶解气体分析时易陷入局部极小的问题,利用全局最优化性能强的动态隧道(DT)算法,将两种算法结合,提出一种基于动态隧道的模糊C均值(DTFCM)算法。该算法首先采用FCM算法聚类得到局部最优值,再利用动态隧道算法以该局部最优值为初始值寻找更小的能量盆地,再将其值返回给FCM算法进行迭代寻优,直到找到全局最小值。通过该算法应用于变压器DGA数据分析,从而实现变压器的故障诊断。变压器油色谱样本及加噪样本故障诊断试验表明,该算法能快速、有效地对样本进行聚类,具有较高的诊断准确率。更多还原

【Abstract】 Dissolved gas analysis in transformer oil (DGA) is an important method for power transformer insulating diagnosis. Aiming at the problem that fuzzy C-means (FCM) clustering algorithm is likely to fall into local minimum point when it is used for dissolved gas analysis,dynamic tunneling (DT) algorithm was introduced for its high global optimization performance. Then a DTFCM algorithm was presented based on these two algorithms. On the basis of local minimum obtained by optimization searching of FCM algorithm,dynamic tunneling process was used to search a lower energy valley,then the value was submitted to FCM algorithm for iterative optimization until global minimum point was found by repeating the process. Through the application of this algorithm for DGA in transformer oil,transformer fault diagnosis can be achieved. These tests for fault diagnosis of chromatography transformer oil and noise samples show that,the algorithm can cluster samples quickly and effectively and with high accuracy for diagnosis.更多还原