摘要
针对油浸式变压器的故障诊断问题,提出一种利用差异化的粗糙集属性约简与有向无环图-支持向量机(DAG-SVM)相结合的方法实现变压器故障类型的快速诊断。该方法首先通过油浸液中溶解气体量的历史数据及其对应的故障类型建立原始故障决策表;然后利用等频率间隔划分法对条件属性数据作离散化处理,利用差分矩阵对决策表信息进行差异化属性约简,建立每两类故障间的诊断规则,消除对决策结果区分度较低的冗余属性;最后根据约简属性的对应数据作为特征向量输入,构建有向无环图-支持向量机多分类诊断网络,进而实现对故障类型的判断。仿真实例表明,该方法在系统检测中提高了故障诊断的准确率。
To solve the fault diagnosis of oil-filled transformer,an approach based on attribute reduction of differential rough set with directed acyclic graph-support vector machine（DAG-SVM）is proposed to rapidly identify fault reasons.Fault decision table of volume of dissolved gas in oil-filled transformer is firstly established according to historical data and corresponding fault type.Then data of condition attributes are discretized by means of equal frequency division method,differentiated attribute reduction is conducted by means of discernibility matrix in rough set theory for decision table,and diagnosis rules between every two kinds of faults are set up,so the redundant attributes of low identification are removed.Finally,multi-classified diagnosis classifier DAG-SVM is constructed by SVMs,in which the data of reduced attributes are character vectors.Case analysis indicates that this method improves the accuracy of fault diagnosis in system detection.
出处
《南京航空航天大学学报》
EI
CAS
CSCD
北大核心
2017年第4期504-510,共7页
Journal of Nanjing University of Aeronautics & Astronautics
基金
国家自然科学基金(61673209
71471087)资助项目
南京航空航天大学研究生创新基地(实验室)开放基金(kfjj20160318)资助项目
关键词
故障诊断
粗糙集
差异化属性约简
有向无环图-支持向量机
fault diagnosis
rough set
differentiated attribute reduction
directed acyclic graph-support vector machine
作者简介
通信作者：徐海燕，女，教授，博士生导师，E-mail：xuhaiyan@nuaa．edu．cn
