油纸绝缘老化拉曼光谱特征量提取及诊断方法  

The Raman Spectral Feature Extraction and Diagnosis of Oil-Paper Insulation Ageing

作　　者：邹经鑫[1] 陈伟根[1] 万福[1] 范舟 

机构地区：[1]输配电装备及系统安全与新技术国家重点实验室(重庆大学),重庆400044

出　　处：《电工技术学报》2018年第5期1133-1142,共10页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金青年基金(61605020);高等学校学科创新引智计划(111计划)(B08036)资助项目

摘　　要：实现油纸绝缘老化的准确诊断是预防油纸绝缘设备故障的重要技术手段,获取能准确反映油纸绝缘老化的有效特征量是关键。结合拉曼光谱分析在物质成分分析及状态诊断的良好应用,论文基于实验室油纸绝缘拉曼光谱分析平台,针对通过加速老化获得的不同老化阶段的油纸绝缘样本,开展了油纸绝缘老化拉曼光谱特征量提取及诊断研究。通过对不同样本的拉曼光谱检测,运用主成分分析法对拉曼光谱数据降维,提取与油纸绝缘老化具有明显对应特征的拉曼光谱特征量;结合粒子群优化算法和多分类支持向量机诊断方法,建立了基于拉曼光谱老化特征量的油纸绝缘老化诊断模型,实验验证了该模型的有效性。Accurate diagnosis of oil-paper insulation ageing stage serves as an important technology to prevent major accidents of oil-paper insulation equipment. The extraction of effective characteristics which reflect the insulation ageing proves to be the essential step. Raman spectroscopy has been demonstrated that it has great potential of mixture composition analysis and condition diagnosis. In this paper, the Raman spectral features of oil-paper insulation samples were researched and extracted based on the Raman spectroscopy platform. Firstly, the samples in different ageing stages were prepared by thermal accelerated ageing process. Then the principal component analysis method was employed to reduce dimensionality of the obtained Raman spectral data. Secondly, the spectral features strongly corresponding to the ageing of oil-paper insulation were extracted. Based on Raman spectral features, multi-classification support vector machine optimized by particle swarm algorithm was used to set up an oil-paper insulation ageing diagnosis model. Finally, the diagnostic capability and universality of the established algorithm were verified by the samples made in lab.

关 键 词：拉曼光谱 油纸绝缘 特征量 老化诊断 

分 类 号：TM835[电气工程—高电压与绝缘技术]