基于主成分分析与支持向量机的汽轮机故障诊断  

Fault Diagnosis of Turbine Generator Based on Pca and Svm

作　　者：司娟宁[1] 刘金园[2] 董泽[1] 廖薇[1] 

机构地区：[1]华北电力大学控制理论与工程学院,保定071003 [2]华能沾化热电厂,沾化256800

出　　处：《汽轮机技术》2011年第2期139-142,共4页Turbine Technology

摘　　要：汽轮机故障诊断的一大难题是故障样本的缺乏,由于支持向量机针对小样本情况能取得很好的效果,为此,提出基于主成分分析与支持向量机的故障诊断方法,首先采用主成分分析方法对汽轮机故障数据进行故障特征提取,将特征向量作为支持向量分类器的输入,按照汽轮机的故障类型训练分类函数。对于支持向量机参数的选取,提出了基于错分样本数的蚁群优化算法。在小样本情况下对汽轮发电机组故障诊断进行了仿真研究。结果表明,应用该算法可以正确且有效地诊断多类汽轮机故障。One of the problems that hamper the process of fault diagnosis of turbine generator is a lack of fault samples, moreover,SVM will achieve a very good resuhs in the small sample situation, therefore, a new approach based on Support Vector Machine and Principal Component Analysis is proposed in this paper. First of all, PCA was used to extraet the feature vectors of the fault samples, and then use these feature vectors as the input vectors of SVM classifier, moreover, the classification functions were trained according to the types of turbine faults. Based on number of error samples, a new continuous ant colony optimization algorithm called MG-CACO is used to optimize the parameters of SVM. This method was applied to diagnosis the fault for turbine generator in small sample eases. The results demonstrate that it can accurately and efficiently diagnosis many types of faults for turbine generator.

关 键 词：主成分分析 支持向量机 蚁群算法 汽轮机 故障诊断 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]