摘要
复合绝缘子随着运行年限的增加会逐渐老化,且各项性能指标降低。为研究在线运行复合绝缘子老化程度定量分析的检测方法,引入高光谱遥感技术通过非接触方式获取复合绝缘子的图谱特征,采用主成分分析法构建了特征空间。结果表明:不同老化程度复合绝缘子的光谱曲线不同;同一老化程度复合绝缘子样本点在特征空间集中在1个点簇,不同老化程度的点簇基本无交叉,可用于定量判断其老化程度。进一步建立了老化神经网络模型,该模型预测结果与实际值比较接近,交叉验证均方根误差为0.014 1。研究结果为在线非接触定量评估复合绝缘子老化程度提供了依据。
Composite insulator ages with operation time, which deteriorates its functional performance. In order to de- velop a detection method for quantitatively evaluating the aging of working composite insulators, we introduced hyperspectral remote sensing technology to acquire image and spectra characteristics of polymeric insulators in a non-contact way, and constructed feature spaces using a principal analysis method. The results show that different aging degrees of polymeric insulators vary evidently in the spectral feature spaces, the data points of the same aging degrees concentrate as a cluster, and the clusters of different aging degrees has no overlapping with each other. This fact could be used for quantitatively determining aging degrees of polymeric insulators. Furthermore, we have established an aging neural network model, which can provide accurate predicting results that have root mean square errors of cross validation of 0.014 1, compared with actual values. The results pave the way for quantitatively evaluating aging degrees of polymer- ic insulators in a non-contact manner.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2014年第3期861-867,共7页
High Voltage Engineering
基金
国家自然科学基金(51277167)~~
关键词
复合绝缘子
老化程度
高光谱遥感
图谱特征
神经网络
非接触检测
composite insulator
aging degree
hyperspectral remote sensing
image and spectra characteristic
neuralnetwork
non-contact detection
