基于主成分分析与支持向量回归的MOV劣化状态诊断研究  

Diagnosis of MOV Deterioration State Based on Principal Component Analysis and Support Vector Regression

作　　者：贺敬安 陶世银 罗少辉 金欣 HE Jingan;TAO Shiyin;LUO Shaohui;JIN Xin(Qinghai Meteorological Disaster Prevention Center,Xining 810001,China)

机构地区：[1]青海省气象灾害防御技术中心,西宁810001

出　　处：《电瓷避雷器》2023年第2期41-48,共8页Insulators and Surge Arresters

基　　金：青海省气象局科研项目(编号:2019-ZJ-T033,2019-ZJ-7098)。

摘　　要：基于主成分分析(PCA)和支持向量回归(SVR)构建诊断模型,探索性的预测MOV器件在遭受浪涌后的劣化状态。研究过程:前期分别对数个MOV进行多脉冲冲击实验,记录器件在每次冲击前后的各项电参数,随后通过主成分分析得到每个MOV的劣化数值变化曲线,并结合对实际器件损伤情况的判断来确定完全损坏阈值点,最后基于支持向量回归算法在实验训练集上构建关于MOV劣化状态的诊断模型并使用测试数据进行检验。检验结果表明,构建的模型在接收一组遭受冲击后的MOV电参数数据作为输入后,可以较为精准的输出表征其劣化状态的数值量(0~1范围内),与实际劣化值相比,计算得到均方根误差为0.081。为在实际电气系统中对MOV器件的劣化监测预警提供了一种新的方法思路。The author builds a diagnostic model based on principal component analysis(PCA)and sup-port vector regression(SVR)to predict the degradation state of MOV devices after being subjected to sur-ges.Research process:In the early stage,a multi-pulse impact experiment was performed on several MOVs,and the electrical parameters of the device before and after each impact were recorded,and then the degradation value change curve of each MOV was obtained through principal component analysis,and combined with the actual device damage to determine the complete damage threshold point,finally based on the support vector regression algorithm to construct a diagnosis model about the deterioration state of MOV on the experimental training set and use the test data to test it.The test results show that the con-structed model receives a set of MOV electrical parameter data after the impact as input,and can more accurately output the numerical value(in the range of O to 1)that characterizes its degradation state.Compared with the actual degradation value,the root mean square error is calculated to be 0.081.It pro-vides a new method for monitoring and early warning of the deterioration of MOV devices in actual electri-cal systems.

关 键 词：MOV 劣化监测 多脉冲 主成分分析 支持向量回归 

分 类 号：TM54[电气工程—电器] TP181[自动化与计算机技术—控制理论与控制工程]