基于DGA的粗糙集与人工鱼群极限学习机的变压器故障诊断  

Transformer Fault Diagnosis by Using Rough Set and Artificial Fish Swarm Extreme Learning Machine Based on DGA

作　　者：雷帆[1] 高波[1] 袁海满 吴广宁[1] 段宗超 

机构地区：[1]西南交通大学电气工程学院,成都610031 [2]国网山东省电力公司威海供电公司,山东威海264200

出　　处：《高压电器》2017年第10期124-130,共7页High Voltage Apparatus

摘　　要：为有效克服变压器不完备故障样本数据对故障诊断结果的影响,文中构建了一种基于粗糙集的人工鱼群极限学习机变压器故障诊断方法,该方法首先运用粗糙集对决策表中的16个条件属性进行约简;其次,根据最简规则表对训练样本进行编码,利用已编码的训练样本对极限学习机进行训练,并运用人工鱼群优化方法对极限学习机的权值及阈值进行优化;最后,利用训练好的极限学习机方法对编码好的样本进行故障诊断。该方法将粗糙集在不完整数据方面所具有的优良特性与极限学习机优良的泛化能力有机融合,以有效提高故障诊断精度。经实例对比分析表明,所构建方法具有更高的诊断准确率,从而验证了该方法的有效性。In order to overcome the influence of the fault diagnosis results of the transformer fault sample data, in this paper, based on rough set, a new method of fault diagnosis for the fault diagnosis method of artificial fish swarm is constructed, in this method, the rough set is used to reduce the 16 conditional attributes in the decision table; Secondly, according to the most simple rule table, the training sample is encoding, and the training sample of en- coding is used to train the extreme learning machine. The weight and the threshold value of the extreme learning machine are optimized by artificial fish swarm optimization method; Finally, using the method of the training of good limit learning machine to the good sample of encoding fault diagnosis. This method can improve the accuracy of fault diagnosis by combining rough sets with the excellent characteristics of the incomplete data and the good generalization ability of the extreme learning machine. The comparison analysis of the case shows that the method proposed in this paper has higher diagnostic accuracy, and the validity of this method is verified.

关 键 词：变压器 故障诊断 粗糙集 极限学习机 人工鱼群 

分 类 号：TM407[电气工程—电器] TP18[自动化与计算机技术—控制理论与控制工程]