基于图谱特征的复合绝缘子老化神经网络评估方法
发布时间：2019-01-11 14:58
【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。
[Abstract]:The composite insulator will aging gradually with the increase of service life, and the performance index will decrease. In order to study the quantitative analysis method of aging degree of composite insulators in on-line operation, hyperspectral remote sensing technology was introduced to obtain the atlas features of composite insulators by non-contact method. The feature space was constructed by principal component analysis. The results show that the spectral curves of composite insulators with different aging degrees are different. The sample points of composite insulators with the same aging degree are concentrated in one cluster in the feature space, and the clusters with different aging degrees have no crossover, which can be used to quantitatively judge the aging degree of insulators. Furthermore, the aging neural network model is established. The prediction results of the model are close to the actual values, and the root-mean-square error of cross-validation is 0.014. The results provide a basis for quantitative evaluation of aging degree of composite insulators.
【作者单位】： 中国电力科学研究院;
【基金】：国家自然科学基金(51277167)~~
【分类号】：TM216
[Abstract]:The composite insulator will aging gradually with the increase of service life, and the performance index will decrease. In order to study the quantitative analysis method of aging degree of composite insulators in on-line operation, hyperspectral remote sensing technology was introduced to obtain the atlas features of composite insulators by non-contact method. The feature space was constructed by principal component analysis. The results show that the spectral curves of composite insulators with different aging degrees are different. The sample points of composite insulators with the same aging degree are concentrated in one cluster in the feature space, and the clusters with different aging degrees have no crossover, which can be used to quantitatively judge the aging degree of insulators. Furthermore, the aging neural network model is established. The prediction results of the model are close to the actual values, and the root-mean-square error of cross-validation is 0.014. The results provide a basis for quantitative evaluation of aging degree of composite insulators.
【作者单位】： 中国电力科学研究院;
【基金】：国家自然科学基金(51277167)~~
【分类号】：TM216
【参考文献】
相关期刊论文 前10条
1 宿志一;陈刚;李庆峰;殷禹;;硅橡胶复合绝缘子伞裙护套的老化及其判据研究[J];电网技术;2006年12期
2 欧文娟;孟耀勇;张小燕;孔猛;;紫外可见吸收光谱结合主成分-反向传播人工神经网络鉴别真假蜂蜜[J];分析化学;2011年07期
3 肖猛;文曹;;一种新型绝缘子带电检测方法—紫外成像法[J];高电压技术;2006年06期
4 杨文宇;杨俊杰;杨旭英;;基于数学形态学的绝缘子在线监测[J];高电压技术;2008年11期
5 徐志钮;律方成;李Z，
本文编号：2407254
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2407254.html
教材专著
