基于DGA的粗糙集与决策信息融合变压器故障诊断

Fault diagnosis of transformer based on rough set theory and decision information fusion

针对变压器故障特征信息不确定性、冗余性及传统故障诊断手段的单一性问题,本文构建了一种粗糙集与多决策信息融合的变压器故障诊断模型。该方法首先考虑将16种特征气体比值作为故障特征参量,并利用离散化规则与粗糙集知识约简对其进行知识提取,以有效降低特征信息冗余度。其次,将降维后属性集作为BP神经网络、支持向量机以及贝叶斯网络3种单一诊断方法的特征输入,进行故障类型初步判定。最后,利用DS信息融合规则对3种初步判定结果进行决策融合,以获得更为高效的故障判断结论。实例分析表明,该方法有效削弱了冗余特征信息对诊断结果的影响,能够合理解决证据融合冲突,并切实提高了故障识别准确率,其性能明显优于单一诊断方法。

Aiming at the problem of information uncertainty and redundancy of transformer fault feature and the singularity of traditional method of fault diagnosis,a fault diagnosis model of transformer based on rough set theory and multi information fusion is proposed. Firstly,16 groups of characteristic gas ratios are considered as fault characteristic parameters,and information is extracted by discretization rules and rough set to reduce feature information redundancy effectively. Additionally,the reduced attribute set is used as the input of BP neural network（ BPNN）,support vector machine（ SVM） and Bayesian network to diagnose the fault types. Finally,the decision fusion of three kinds of preliminary judgment results is made by DS information fusion rules to obtain more efficient fault judgment conclusion. Example analysis shows that the proposed method can effectively reduce the impact of redundant feature information on the diagnosis result and solve the conflict of evidence fusion,and improve the accuracy of fault recognition. It＇s easy to see the performance of the method is superior to the single diagnosis method.