基于稀疏光谱成像的线路复合绝缘子积污状态可视化评估

doi:10.11930/j.issn.1004-9649.202311070

摘要/Abstract

摘要：

对线路复合绝缘子的污秽状态进行及时、准确的在线评估，可以有效防止污闪事故的发生。提出一种线路复合绝缘子污秽状态可视化评估方法。首先，结合图像配准算法和目标区域框选构建多源反射光谱图像半自动化配准方法，解决多目相机固有的图像配准难题；其次，利用低成本、轻量化、高成像质量的多目式稀疏光谱成像设备拍摄人工染污样品并进行模型预训练，进一步通过真实自然积污样品进行迁移训练，构建复合绝缘材料表面污秽等级诊断模型；最后，借助无人机平台，在不同拍摄条件下对线路复合绝缘子污秽状态进行实测分析。结果表明，该方法对于人工染污和自然积污样品污秽等级的划分准确率分别为95.3%和87.8%，实际线路复合绝缘子污秽等级划分准确率可达90%，污秽分布区域显示清晰。通过无人机实测分析验证了基于稀疏光谱成像技术的线路绝缘子污秽等级评估和污秽分布可视化诊断的可行性，为线路绝缘子状态巡视和检修决策提供了新的技术手段。

关键词: 复合绝缘子, 稀疏光谱成像技术, 污秽状态评估, 图像配准, 迁移学习

Abstract:

The timely and accurate online estimation for composite insulators of transmission lines can effectively prevent pollution flashover accidents. A visualized estimation method of composite insulator pollution status of transmission line was proposed in this paper. Firstly, a semi-automatic registration method for multisource reflective spectral images was constructed by combining image registration algorithm and target area selection, which solved the inherent image registration problem of multi-lens camera. Secondly, model pre-training was conducted through shooting artificial contaminated samples using the low cost, lightweight, and high imaging quality multi-lens multispectral imaging equipment, and further transfer training was conducted through real natural contaminated samples to construct a pollution grade diagnosis model of composite insulation material surfaces. Finally, the actual pollution status of transmission line composite insulators was measured and analyzed under different shooting conditions using the unmanned aerial vehicle platform. The results indicate that the pollution grade classification accuracy for artificial and natural contaminated samples is 95.3% and 87.8%, respectively, and the pollution grade classification accuracy for actual transmission line composite insulators can reach up to 90% with pollution distribution area clearly displayed. The feasibility of transmission line insulator pollution grade estimation and pollution distribution visualization diagnosis based on reflective multispectral imaging technology is verified, which can provide a new technique for status inspection and maintenance decision of transmission line insulators.

Key words: composite insulator, multispectral imaging technology, estimation of pollution status, image registration, transfer learning

任明, 李乾宇, 夏昌杰, 董明. 基于稀疏光谱成像的线路复合绝缘子积污状态可视化评估[J]. 中国电力, 2025, 58(2): 203-215.

Ming REN, Qianyu LI, Changjie XIA, Ming DONG. Visualized Estimation of Composite Insulator Pollution Status of Transmission Line Based on Reflective Multispectral Imaging[J]. Electric Power, 2025, 58(2): 203-215.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202311070

https://www.electricpower.com.cn/CN/Y2025/V58/I2/203

图/表 25

图 1 稀疏光谱图像信息采集平台示意

Fig.1 Diagram of multispectral image information acquisition platform

图 2 多目式稀疏光谱图像采集设备结构示意及实物

Fig.2 Physical image and structure diagram of multi-lens multispectral image acquisition device

图 3 多目式稀疏光谱相机图像采集示意

Fig.3 Image acquisition diagram using multi-lens multispectral camera

表 1 多目式稀疏光谱图像采集设备主要参数

Table 1 Main parameters of multi-lens multispectral image acquisition device

参数名称		数值
质量/g		640
中心波长（带宽）/nm		450 (35)、555 (25)、660 (22.5)、 720 (10)、750 (10)、840 (30)
量化位数/bit		12
图像分辨率		1280×960
CMOS尺寸		1/3′′
像元尺寸/μm		3.75×3.75

图 4 污秽状态可视化评估方法的整体构建流程及应用方式

Fig.4 Overall construction process and application of visualization estimation method for pollution status

图 5 待配准线路复合绝缘子稀疏光谱图像

Fig.5 Unregistered multispectral image of transmission line composite insulators

表 2 不同配准算法的互信息和配准时间

Table 2 Mutual information and registration time of different registration algorithms

配准算法	综合互信息	配准时间/s
基于SIFT的配准算法	0.780 2	2.587 5
基于SURF的配准算法	0.775 0	0.881 3
基于ORB的配准算法	0.751 4	0.868 8

图 6 基于SURF算法的复杂背景稀疏光谱图像配准

Fig.6 Complex background multispectral image registration based on SURF algorithm

图 7 基于目标区域框选的复杂背景稀疏光谱图像配准

Fig.7 Complex background multispectral image registration based on target area selection

图 8 基于特征点对标记的复杂背景稀疏光谱图像配准

Fig.8 Complex background multispectral image registration based on feature point pair marking

表 3 不同等级污秽的NaCl和高岭土含量

Table 3 NaCl and Kaolin content of different pollution grades 单位：mg/cm2

污秽等级	NaCl	CaSO₄	高岭土
0	0	0	0
Ⅰ	0.016	0.040	0.336
Ⅱ	0.028	0.070	0.588
Ⅲ	0.070	0.175	1.470
Ⅳ	0.100	0.250	2.100

图 9 不同污秽等级复合绝缘子片样品

Fig.9 Composite insulation material samples with different pollution levels

图 10 复合绝缘材料不同等级污秽光谱曲线

Fig.10 Spectral curve of composite insulation material with different pollution grades

表 4 不同分类模型的评估结果

Table 4 Estimation result of different classification models

分类模型	超参数搜索算法	验证准确率/%	测试准确率/%
SVM	网格搜索	94.6	94.5
	随机搜索	95.2	95.0
	贝叶斯优化	95.2	95.0
KNN	网格搜索	95.0	94.8
	随机搜索	95.2	95.0
	贝叶斯优化	95.3	95.2
随机森林	网格搜索	94.9	94.6
	随机搜索	95.2	95.0
	贝叶斯优化	95.4	95.3

表 5 污秽等级诊断模型的超参数

Table 5 Hyper-parameters of pollution grade diagnosis model

参数名称		参数值
决策树数量		486
最大分离特征数		7
决策树最大深度		93
特征选取方法		基尼系数

图 11 复合绝缘子片表面污秽状态可视化结果

Fig.11 Visualized result of surface pollution status of composite insulation material

图 12 迁移训练数据构建流程示意

Fig.12 Migration training data construction process

表 6 自然积污样本数据集划分

Table 6 Data set division of natural contaminated samples

样本污秽等级	训练集数据/个	测试集数据/个
0	1 550	523
Ⅰ	1 650	546
Ⅱ	1 700	577
Ⅲ	1 750	584
Ⅳ	1 800	612

表 7 迁移训练前后自然积污数据集的评估结果

Table 7 Estimation result of natural contamination dataset before and after migration training

	训练集准确率/%	测试集准确率/%
迁移训练前	70.4	71.5
迁移训练后	90.2	87.8

图 13 实测现场实拍图

Fig.13 Actual pictures of measurement site

表 8 线路复合绝缘子拍摄计划

Table 8 Shooting plan of transmission line composite insulators

组别	环境光照	拍摄角度	绝缘子器型
a	晴天（60 000~80 000 lux）	30°俯拍	复合悬式绝缘子
b	晴天（60 000~80 000 lux）	30°仰拍	复合悬式绝缘子
c	阴天（10 000~15 000 lux）	30°仰拍	复合悬式绝缘子
d	阴天（10 000~15 000 lux）	30°俯拍	瓷复合绝缘子
e	阴天（10 000~15 000 lux）	30°仰拍	瓷复合绝缘子

图 14 本文检测方法及传统检测方法流程示意

Fig.14 Flow chart of the proposed detection method and traditional method

图 15 复合悬式绝缘子真实污秽等级与评估结果对比

Fig.15 Comparison of actual pollution grade and estimation result of composite pin insulator

图 16 瓷复合绝缘子真实污秽等级与评估结果对比

Fig.16 Comparison of actual pollution grade and estimation result of porcelain composite insulator

图 17 绝缘子稀疏光谱图像及污秽分布可视化评估结果

Fig.17 Multispectral image of insulator and visualized estimation result of pollution distribution

参考文献 27

1	王少华, 胡文堂, 何文林, 等. 自然积污绝缘子的污闪电压和泄漏电流特性试验研究[J]. 电网技术, 2013, 37 (10): 2855- 2860.
	WANG Shaohua, HU Wentang, HE Wenlin, et al. Experimental research on pollution flashover voltage and leakage current characteristics of naturally polluted insulators[J]. Power System Technology, 2013, 37 (10): 2855- 2860.
2	蒋兴良, 赵世华, 毕茂强, 等. 污秽绝缘子闪络特性与泄漏电流特性研究[J]. 中国电机工程学报, 2013, 33 (31): 220- 226, 27.
	JIANG Xingliang, ZHAO Shihua, BI Maoqiang, et al. Research on flashover performance and leakage current performance of polluted insulators[J]. Proceedings of the CSEE, 2013, 33 (31): 220- 226, 27.
3	宿志一, 李庆峰. 我国电网防污闪措施的回顾和总结[J]. 电网技术, 2010, 34 (12): 124- 130.
	SU Zhiyi, LI Qingfeng. Historical review and summary on measures against pollution flashover occurred in power grids in China[J]. Power System Technology, 2010, 34 (12): 124- 130.
4	何昱燊, 张志劲, 傅海涛, 等. 自然环境下覆冰绝缘子长串交流闪络电压及电弧发展现象[J]. 电网技术, 2021, 45 (7): 2904- 2912.
	HE Yushen, ZHANG Zhijin, FU Haitao, et al. AC flashover voltage and arc development of long insulators strings in natural icing environment[J]. Power System Technology, 2021, 45 (7): 2904- 2912.
5	姜新建, 董弘川, 王黎明, 等. 用有效盐密作为表征污秽度的新方法[J]. 高电压技术, 2017, 43 (12): 3869- 3875.
	JIANG Xinjian, DONG Hongchuan, WANG Liming, et al. New method to describe contamination degree of insulators by effective equivalent salt deposit density[J]. High Voltage Engineering, 2017, 43 (12): 3869- 3875.
6	张志劲, 卢炳宏, 傅海涛, 等. 基于XP-160污秽体积分数的绝缘子积污表征[J]. 电网技术, 2021, 45 (9): 3737- 3744.
	ZHANG Zhijin, LU Binghong, FU Haitao, et al. Characterization of insulator contamination based on XP-160 pollution volume fraction[J]. Power System Technology, 2021, 45 (9): 3737- 3744.
7	王尧平, 李特, 姜凯华, 等. 基于气象特征挖掘与AdaBoost-MEA-ELM模型的绝缘子盐密预测[J]. 中国电力, 2023, 56 (9): 157- 167.
	WANG Yaoping, LI Te, JIANG Kaihua, et al. Prediction of Insulator ESDD based on meteorological feature mining and AdaBoost-MEA-ELM model[J]. Electric Power, 2023, 56 (9): 157- 167.
8	SLAMA M E, BEROUAL A, HADI H. Influence of the linear non-uniformity of pollution layer on the insulator flashover under impulse voltage-estimation of the effective pollution thickness[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2011, 2 (18): 384- 392.
9	张志劲, 杨晟欢, 蒋兴良, 等. 迎背风不均匀污秽下支柱绝缘子直流污闪特性[J]. 电网技术, 2020, 44 (1): 354- 360.
	ZHANG Zhijin, YANG Shenghuan, JIANG Xingliang, et al. DC flashover characteristics of post insulators under non-uniform pollution between windward and leeward sides[J]. Power System Technology, 2020, 44 (1): 354- 360.
10	金立军, 田治仁, 高凯, 等. 基于红外与可见光图像信息融合的绝缘子污秽等级识别[J]. 中国电机工程学报, 2016, 36 (13): 3682- 3691, 3389.
	JIN Lijun, TIAN Zhiren, GAO Kai, et al. Discrimination of insulator contamination grades using information fusion of infrared and visible images[J]. Proceedings of the CSEE, 2016, 36 (13): 3682- 3691, 3389.
11	王丰华, 刘国坚, 张宏钊, 等. 基于改进C-V模型的外绝缘放电紫外图像特征量提取[J]. 高电压技术, 2018, 44 (8): 2525- 2532.
	WANG Fenghua, LIU Guojian, ZHANG Hongzhao, et al. Ultraviolet image parameters extraction of insulation surface discharge based on improved C-V model[J]. High Voltage Engineering, 2018, 44 (8): 2525- 2532.
12	DOUGLAS B, GAMAL E, SUN D W. Non-destructive of chemical composition intact and minced pork using near-infrared hyperspectral imaging[J]. Food Chemistry, 2013, 138, 1162- 1171.
13	KONG Y, LIU Y, GENG J, et al. Pixel-level assessment model of contamination conditions of composite insulators based on hyperspectral imaging technology and a semi-supervised ladder network[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2023, 30 (1): 326- 335. DOI
14	邱彦, 张血琴, 郭裕钧, 等. 基于高光谱技术的绝缘子污秽等级检测方法[J]. 高电压技术, 2019, 45 (11): 3587- 3594.
	QIU Yan, ZHANG Xueqin, GUO Yujun, et al. Detection method of insulator contamination grades based on hyperspectral technique[J]. High Voltage Engineering, 2019, 45 (11): 3587- 3594.
15	邱彦, 郭裕钧, 张血琴, 等. 基于高光谱技术的绝缘子污秽成分识别方法[J]. 高电压技术, 2020, 46 (11): 4023- 4030.
	QIU Yan, GUO Yujun, ZHANG Xueqin, et al. Identification method of insulator pollution components based on hyperspectral technology[J]. High Voltage Engineering, 2020, 46 (11): 4023- 4030.
16	马欢, 郭裕钧, 张血琴, 等. 基于高光谱技术的绝缘子污秽含水量检测[J]. 高电压技术, 2020, 46 (4): 1396- 1404.
	MA Huan, GUO Yujun, ZHANG Xueqin, et al. Moisture content detection of insulator contamination based on hyperspectral technology[J]. High Voltage Engineering, 2020, 46 (4): 1396- 1404.
17	MANOLAKIS D, MARDEN D, SHAW G A. Hyperspectral image processing for automatic target detection applications[J]. Lincoln Laboratory Journal, 2003, 14 (1): 79- 116.
18	XIA C J, REN M, WANG B, et al. Acquisition and analysis of hyperspectral data for surface contamination level of insulating materials[J]. Measurement, 2021, 173, 108560. DOI
19	何锦强, 李锐海, 李昊, 等. 基于Yolo v5与Grabcut的架空线路绝缘子可见光图像自动识别与分割方法[J]. 南方电网技术, 2023, 17 (6): 128- 135.
	HE Jinqiang, LI Ruihai, LI Hao, et al. Visible light image automatic recognition and segmentation method for overhead power line insulators based on Yolo v5 and Grabcut[J]. Southern Power System Technology, 2023, 17 (6): 128- 135.
20	WANG B, DONG M, REN M, et al. Automatic fault diagnosis of infrared insulator images based on image instance segmentation and temperature analysis[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69 (8): 5345- 5355. DOI
21	张二虎, 卞正中. 基于最大熵和互信息最大化的特征点配准算法[J]. 计算机研究与发展, 2004, 41 (7): 1194- 1199.
	ZHANG Erhu, BIAN Zhengzhong. A new robust point registration algorithm by maximization of entropy and mutual information[J]. Journal of Computer Research and Development, 2004, 41 (7): 1194- 1199.
22	肖明, 鲍永亮, 颜仲新. 基于点特征的图像配准方法综述[J]. 兵工学报, 2015, 36 (增刊2): 326- 340.
23	关志成, 张仁豫, 薛家麒, 等. 自然污秽可溶盐构成及其对污闪电压值的影响[J]. 电瓷避雷器, 1989, (6): 13- 18.
24	国家质量监督检验检疫总局, 中国国家标准化管理委员会. 污秽条件下使用的高压绝缘子的选择和尺寸确定第1部分: 定义、信息和一般原则: GB/T 26218.1—2010[S]. 北京: 中国标准出版社, 2011.
25	国家能源局. 高压交流系统用复合绝缘子人工污秽试验: DL/T 859—2015[S]. 北京: 中国电力出版社, 2015.
26	程乐峰, 余涛, 张孝顺, 等. 机器学习在能源与电力系统领域的应用和展望[J]. 电力系统自动化, 2019, 43 (1): 15- 31.
	CHENG Lefeng, YU Tao, ZHANG Xiaoshun, et al. Machine learning for energy and electric power systems: state of the art and prospects[J]. Automation of Electric Power Systems, 2019, 43 (1): 15- 31.
27	肖建, 于龙, 白裔峰. 支持向量回归中核函数和超参数选择方法综述[J]. 西南交通大学学报, 2008, 43 (3): 297- 303. DOI
	XIAO Jian, YU Long, BAI Yifeng. Survey of the selection of kernels and hyper-parameters in support vector regression[J]. Journal of Southwest Jiaotong University, 2008, 43 (3): 297- 303. DOI

[1]	赵晨浩, 焦在滨, 李程昊, 张迪, 张鹏辉. 基于主动迁移学习的电力系统暂态稳定自适应评估[J]. 中国电力, 2025, 58(1): 70-77.
[2]	徐浩, 刘利强, 吕超. 基于迁移学习的配电网内部过电压识别方法[J]. 中国电力, 2021, 54(8): 52-59.
[3]	王思华, 王军军, 赵磊, 陈龙. 污秽成分对复合绝缘子表面电场的影响[J]. 中国电力, 2021, 54(7): 149-157.
[4]	王思华, 陈龙, 王军军, 赵磊. 复合绝缘子伞套老化状态模糊综合评估[J]. 中国电力, 2021, 54(5): 156-165.
[5]	高熠, 田联房, 杜启亮. 基于Mask R-CNN的复合绝缘子过热缺陷检测[J]. 中国电力, 2021, 54(1): 135-141.
[6]	申巍, 高华, 张鹏, 左坤, 田毅, 杨文博, 袁正康. 高聚物复合绝缘子紫外老化性能研究[J]. 中国电力, 2019, 52(9): 73-78,85.
[7]	雷兴列, 李宇泽, 彭勇, 方玉群, 潜力群. 复合绝缘子导通性缺陷对带电作业安全距离影响[J]. 中国电力, 2019, 52(6): 134-139.
[8]	周自强, 纪扬, 苏烨, 蔡钧宇. 基于迁移学习卷积神经网络的电缆隧道锈蚀识别算法[J]. 中国电力, 2019, 52(4): 104-110.
[9]	吴高林, 周庆, 肖前波, 胡建林, 毕茂强. 复合绝缘子污闪与冰闪特性及临界闪络电流研究[J]. 中国电力, 2019, 52(4): 59-65.
[10]	王康, 朱亚洲, 周辛南, 吴照国, 胡远翔. 运行复合绝缘子界面问题分析[J]. 中国电力, 2018, 51(4): 45-52.
[11]	赵美云, 吴阳, 赵新泽, 彭文昱, 习森明. 复合绝缘子积污腐蚀表面泄漏电流分析[J]. 中国电力, 2018, 51(2): 7-12,53.
[12]	党会学, 赵均海, 吴静. V形复合绝缘子力学特性研究[J]. 中国电力, 2016, 49(7): 9-14.
[13]	闵绚，文志科，曾云飞，付晶. 脉动风作用下特高压绝缘子串的风偏特性[J]. 中国电力, 2016, 49(3): 65-71.
[14]	李孟兴，王海燕，宋新利，刘健，赵连政，侯锐锋. 具有绝缘子性能避雷器的结构设计与试验分析[J]. 中国电力, 2015, 48(9): 90-94.
[15]	郑秋玮，刘华钢，邹建明，杨帅，喻剑辉，周文俊. ±500 kV兴安直流线路复合绝缘子积污特性[J]. 中国电力, 2015, 48(8): 86-91.