基于模糊TOPSIS和BP网络的电力变压器故障诊断方法

针对传统故障诊断技术中存在诊断模型结构复杂以及收集故障样本数据非常繁琐的问题,将TOPSIS方法在Vague集下进行扩展。介绍了Vague集的基本概念及其相似度量方法,以及使用Vague集表达的语义变量集,并据此对原始样本集进行优劣排序和聚类,从而缩减了样本集的容量,使得故障特征信息量和映射空间复杂度的问题在一定程度上得以平衡。在此基础上构建了适应于变压器故障诊断的BP网络诊断模型,实现对不同类型故障的诊断。算例分析表明,此方法与传统的变压器故障诊断的方法相比较具有明显的优越性。

Fuzzy TOPSIS and BPANN Based Fault Diagnosis Method for Power Transformer

Aiming at the problem of diagnosing the structure of model and collecting the data of failure sample being very complicated in the traditional diagnosis technology, this paper extendes the method of TOPSIS on vague sets. The basic concepts of vague sets, similarity measurement and the usage of semanteme variable set expressed by vague value are introduced. Hereby the original sample data can be distinguished and clustered so to reduce the quantity of sample data enough to somehow balance the fault feature information and the mapping space complexity. Furthermore, a BP neural diagnosing network is constructed for diagnosis of all kinds of transformer faults. The cases analysis indicate that the proposed method has obvious advantage than the conventional diagnosis methods for transformer faults.