中文摘要：

建立了基于自组织特征映射（SOM）神经网络和粗糙集理论的变压器故障诊断模型,并将SOM各种改进算法应用到厦门市110 kV变压器故障的智能诊断中。模型使用粗糙集方法对原始数据进行约简,将得到约简后的决策表作为SOM网络的训练样本集,采用包括传统SOM、静态自组织映射树（TS-SOM）、自组织树（SOTA）、改进的自组织二叉树（DBTSONN2）、动态多叉树（DMTSONN）等各种算法对电力变压器运行中的潜在故障进行诊断。试验表明,DBTSONN2、DMTSONN能有效降低SOM网络的复杂性,减少SOM网络训练时间,对于提高变压器故障诊断精度具有一定的实际意义。

英文摘要：

The fault diagnosis model for transformer was developed based on self-organizing feature map（SOM） neural network and rough set theory.Various improved SOM algorithms were applied to the fault intelligent diagnosis for the 110 kV transformers in Xiamen.Rough set theory was used by the model to simplify the original data,which was taking as training sample set of the SOM network.The potential fault was diagnosed by using various algorithms including traditional SOM,tree structure self-organizing map（TS-SOM）,self-organizing tree（SOTA）,improved self-organizing binary tree（DBTSONN2） and dynamic multiple tree（DMTSONN）.The tests indicated that DBTSONN2 and DMTSONN can efficiently reduce the complexity and training time of SOM network,which had practical significances in improving the accuracy of fault diagnosis for transformer.