摘要
通过对变压器油中溶解气体进行分析,可以及早的发现变压器的故障。为了全面地反映变压器内部故障与特征气体之间的关系,提出采用5种特征气体浓度比值共计15组作为特征预输入量,并采用基因选择算法对15个特征量进行筛选,将筛选后特征量作为支持向量机模型输入。在SVM模型中,采用模拟退火算法对SVM的参数进行优化,给出其GUI界面。最后,通过数据验证基于RFE-SA-SVM模型故障诊断率要高于单一模型。
The transformer fault will be discovered earlier by the analysis on the gases dissolved in transformer oil.Five characteristic gas concentration ratios with total 15 sets have been adopted to reflect the relationship betweenthe transformer inner fault and the characteristic gas. The RFE algorithm has been used to filtrate in 15characteristic quantities, with the obtained quantities acted as the input of the SVM model. In the SVM model, theSVM parameters will be optimized by the SA algorithm, and the GUI interface will be provided. Finally, the RFE-SA-SVM model fault diagnosis rate will be better than that of single model through data validation.
出处
《电测与仪表》
北大核心
2014年第12期50-55,共6页
Electrical Measurement & Instrumentation
关键词
特征选择
基因选择算法
支持向量机
故障诊断
feature selection, recursive feature elimination, support vector machine, fault diagnosis
