最小二乘支持向量机在绝缘子污秽放电声发射监测中的应用  

Application of Least Squares Support Vector Machine in Acoustic Emission Monitor of Polluted Insulator Discharge

作　　者：李红玲[1] 文习山[1] 舒乃秋[1] 

机构地区：[1]武汉大学电气工程学院,武汉市430072

出　　处：《电力建设》2009年第6期25-28,共4页Electric Power Construction

摘　　要：绝缘子污秽放电伴随着声发射现象,人工污秽试验表明,污秽绝缘子放电声发射信号与绝缘子污闪放电的发展存在复杂的非线性关系。为有效分析绝缘子放电声发射信号,从而判断绝缘子表面的污秽放电状况,利用最小二乘支持向量机这种新的机器学习工具,建立了以污秽绝缘子放电声发射信号中的多个变量作为输入参数,污秽程度作为输出的绝缘子运行状态评定模型。模型参数通过交叉检验的方式确定,并通过部分仿真数据验证了该模型的有效性。仿真结果表明,最小支持向量机具有很好的学习、分类和泛化能力,满足绝缘子污秽监测实际要求。Artificial contaminations tests proved that there is a complicated nonlinear corresponding relationship between the acoustic emission signals emitted by the polluted insulators and the development of contamination discharge. In order to analyze the acoustic emission signals emitted by polluted insulator discharge effectively and estimate the contamination discharge status on the insulator surface rightly, the least square support vector machine （LS-SVM） is produced, which is a new machine learning tool. By using LS-SVM, the assessment model of insulator surface contamination is built. In the model, the multiple variables come from the acoustic emission signals are chosen as the input variables and then the degree of contaminatian as the output variable. The model parameters of classifiers are tuned with cross-validation method, and the feasibility of model is proved by some data in the laboratory simulation. Simulation results show that the IS-SVM classifiers are capable of learning quite well from the raw data samples while processing good classification and generalization ability. The assessment model of insulator surface contamination can meet the actual need of insulator surface contamination monitoring.

关 键 词：绝缘子 放电 在线监测 声发射 最小二乘支持向量机 污秽程度评定 

分 类 号：TM216[一般工业技术—材料科学与工程]