Lightweight Real-time Detection Network for Insulator Self-explosion Defects Based on Model Compression and Knowledge Distillation

YANG Hongkuan; LI Zuoqing; YE Shengyan; WAN Shunchang

doi:10.13296/j.1001-1609.hva.2026.10.013

您当前的位置：

首页 >

文章列表页 >

Lightweight Real-time Detection Network for Insulator Self-explosion Defects Based on Model Compression and Knowledge Distillation

Research & Analysis | 更新时间：2026-10-14

- Lightweight Real-time Detection Network for Insulator Self-explosion Defects Based on Model Compression and Knowledge Distillation
- High Voltage Apparatus Vol. 62, Issue 10, Pages: 108-116(2026)
- 作者机构：
  
  云南电网有限责任公司德宏供电局，云南芒市 678400
- 作者简介：
- 基金信息：
- DOI：10.13296/j.1001-1609.hva.2026.10.013
  CLC：
- Received：15 May 2026，
  
  Revised：2026-07-28，
  
  Published：16 October 2026
- 稿件说明：
移动端阅览
YANG Hongkuan, LI Zuoqing, YE Shengyan, et al. Lightweight Real-time Detection Network for Insulator Self-explosion Defects Based on Model Compression and Knowledge Distillation[J]. High Voltage Apparatus, 2026, 62(10): 108-116.
DOI：

YANG Hongkuan, LI Zuoqing, YE Shengyan, et al. Lightweight Real-time Detection Network for Insulator Self-explosion Defects Based on Model Compression and Knowledge Distillation[J]. High Voltage Apparatus, 2026, 62(10): 108-116. DOI： 10.13296/j.1001-1609.hva.2026.10.013.

摘要

针对目前绝缘子缺陷检测网络架构复杂、推理计算量大导致占用大量的储存空间、检测效率低下的问题，设计出一种绝缘子自爆缺陷轻量化实时检测网络（insulator defects net，IDNet）。首先，通过空间尺度变换的方法将YOLOv4-tiny基础网络中用于小目标检测任务的26像素×26像素的YOLOhead进行删除，降低网络复杂程度，提升检测速度；然后，根据预设的剪枝比例阈值进行卷积层通道剪枝来实现模型压缩，以减小网络权重，进一步提升网络检测速度；最后，采用知识蒸馏的方法将模型压缩时损失的精度做微量调控，恢复网络的准确性。实验结果表明，文中设计的IDNet绝缘子自爆缺陷轻量化实时检测网络较YOLOv4-tiny基础网络能保证在检测精度损失较少的前提下，网络权重压缩85.76%，降为3.19 MB；平均检测速度提升11.27%，达到44.02帧/s。文中为研制绝缘子自爆缺陷实时检测设备提供了检测速度快、准确性高且占用储存空间小的有效绝缘子自爆缺陷实时检测算法。

Abstract

In view of issues of present insulator such as complex network architecture and significant storage occupation as well as low detection efficiency due to large inference computation in defect detection

a lightweight realtime detection network for insulator selfe-explosion defects(insulator defects net

IDNet)is designed. First

the 26 pixel × 26 pixel YOLOhead in the YOLOv4-tiny base network

which is used for small object detection

is removed through spatial scale transformation

thereby reducingnetwork complexity and improving detection speed. Then

the channel pruning on the convolutional layer is performed in accordance with the preset pruning ratio threshold

thereby reducing network weights and further improving the detection speed. Finally

the knowledge distillation method is used to finely adjust the accuracy incurred during model compression

thereby restoring the network's accuracy.Experimental results show that

compared with the YOLOv4-tiny baseline network

the proposed IDNet lightweight real-time detection network for insulator self-explosion defects can achieve 85.76% compression in network weights(reduced to 3.19 MB)while incurring only a minor loss in detection accuracy. Moreover

the average detection speed is improved by 11.27%

reaching 44.02 frames per second. This paper provides an effective real-time detection algorithm for insulator self-explosion defects

which offers fast detection speed

high accuracy

and low storage space occupation

thereby faciliating the development of real-time detection equipment for insulator selfexplosion defects.

关键词

Keywords

references

张秋敏,马思维,孟然.改进YOLOv5s的绝缘子缺陷检测研究[J].电子设计工程,2025,33(10):181-185.

Zhang Qiumin,Ma Siwei,Meng Ran.Research on improving YOLOv5s for insulator defect detection[J].Electronic Design Engineering,2025,33(10):181-185.

王敩青,王红星,张瑞晨,等. 500 kV紧凑型输电线路复合绝缘子均压环优化设计[J].电瓷避雷器,2024(3):114-122.

Wang Xiaoqing,Wang Hongxing,Zhang Ruichen,et al.The optimization of grading ring for composite insulator in 500 kV compact transmission line[J].Insulators and Surge Arresters,2024 (3):114-122.

李进,赵仁勇,杜伯学,等.电工环氧绝缘件缺陷无损检测方法研究进展[J].电工技术学报,2021,36(21):4598-4607.

Li Jin,Zhao Renyong,Du Boxue,et al. Research progress of nondestructive detection methods for defects of electrical epoxy insulators[J].Transactions of China Electrotechnical Society,2021, 36(21):4598-4607.

吕岩婷,蒋菲,张素慧,等.一起GIS用盆式绝缘子原始缺陷导致开裂故障分析[J].电瓷避雷器,2024(3):180-185.

Lyu Yanting,Jiang Fei,Zhang Suhui,et al.Cracking fault caused by original defect of gis basin insulator[J].Insulators and Surge Arresters,2024(3):180-185.

缪金,秦军,陈峻宇,等.基于高光谱成像的复合绝缘子表面污秽成分识别方法[J].浙江电力,2025,44(1):124-132.

Miao Jin,Qin Jun,Chen Junyu,et al.A method for identifying surface contamination components of composite insulators based on hyperspectral imaging[J].Zhejiang Electric Power,2025,44(1):124-132.

常波,张启哲,王胜辉,等.动车组车顶隔离开关用支柱绝缘子污秽状态下的电场分布特性研究[J].电测与仪表,2024,61 (4):81-85.

Chang Bo,Zhang Qizhe,Wang Shenghui,et al. Research on the electric field distribution characteristics under polluted conditionsof EMU roof isolation switch post insulators[J].Electrical Measurement & Instrumentation,2024,61(4):81-85.

贺春,何金,朱旭亮,等.35 kV C-GIS环氧树脂外锥套管残余应力检测研究[J].高压电器,2023,59(4):38-43.

He Chun,He Jin,Zhu Xuliang,et al. Study on residual stress detection of external conical epoxy bushing of 35 kV C-GIS[J].High Voltage Apparatus,2023,59(4):38-43.

谢金鹏,张素慧,李军,等.某变电站高压跌落式熔断器瓷绝缘子断裂原因分析[J].电瓷避雷器,2024(6):153-159.

Xie Jinpeng,Zhang Suhui,Li Jun,et al.Cause of porcelain insulator fracture of high-voltage drop fuse in a certain substation[J].Insulators and Surge Arresters,2024(6):153-159.

尹向雷,解永芳,屈少鹏,等.基于动态蛇形卷积和非跨步卷积的绝缘子缺陷检测[J].电力系统保护与控制,2024,52(20):177-187.

Yin Xianlei,Xie Yongfang,Qu Shaopeng,et al.Insulator defect detection based on dynamic snake convolution and space-to-depth convolution[J].Power System Protection and Control,2024,52(20):177-187.

李玄,尹奕淳,师伟,等.三支柱绝缘子微缺陷超声临界纵波探测方法与应用[J].电工技术学报,2025,40(21):6905-6921.

Li Xuan,Yin Yichun,Shi Wei,et al.Ultrasonic critical longitudinal wave detection method and application for microdefects in tri-post insulators[J].Transactions of China Electrotechnical Society,2025, 40(21):6905-6921.

冯世凯,陈千.复杂环境下玻璃绝缘子自爆的实时检测方法[J].电瓷避雷器,2024(5):135-144.

Feng Shikai,Chen Qian.Real-time detection method for selfdetonation of glass insulator in complex environment[J].Insulators and Surge Arresters,2024(5):135-144.

孙喆,张静,李建兴,等.基于轻量化YOLOv7-DenseBlock的绝缘子检测算法[J].电网与清洁能源,2025,41(5):68-74.

Sun Zhe,Zhang Jing,Li Jianxing,et al.An insulator detection algorithm based on lightweight YOLOv7-DenseBlock[J].Power System and Clean Energy,2025,41(5):68-74.

梁纲,马斌,饶宇飞,等.基于ADFE-Net的航拍图像绝缘子缺陷检测[J].智慧电力,2025,53(3):107-116.

Liang Gang,Ma Bin,Rao Yufei,et al.Insulator defect detection in aerial images based on ADFE-Net[J].Smart Power,2025,53(3):107-116.

刘开培,李博强,秦亮,等.深度学习目标检测算法在架空输电线路绝缘子缺陷检测中的应用研究综述[J].高电压技术, 2023,49(9):3584-3595.

Liu Kaipei,Li Boqiang,Qin Liang,et al.Review of application research of deep learning object detection algorithms in insulator defect detection of overhead transmission lines[J].High Voltage Engineering,2023,49(9):3584-3595.

Girshick R,Donahue J,Darrell T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014:580-587.

Girshick R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision(ICCV),2015:1440-1448.

Ren Shaoqing,He Kaiming,Girshick R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems,2015:91-99.

Liu Wei,Anguelov D,Erhan D,et al. SSD:single shot MultiBox detector[C]//European Conference on Computer Vision(ECCV), 2016:21-37.

Redmon J,Divvala S,Girshick R,et al. You only look once:unified,real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:779-788.

Redmon J,Farhadi A. YOLO9000:better,faster,stronger[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR),2017:6517-6525.

Redmon J,Farhadi A. YOLOv3:an incremental improvement[EB/OL].[2026-05-15].https://arXiv:1804.02767.

Bochkovskiy A,Wang C Y,Liao H Y M. YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2026-05-15].https://arxiv.org/abs/2004.10934.

Zhou Fangbo,Zhao Huailin,Nie Zhen. Safety helmet detection based on YOLOv5[C]//2021 IEEE International Conference on Power Electronics,Computer Applications(ICPECA),2021:6-11.

李俊,任景,王晔琳,等.基于改进U-net和CNN的绝缘子自爆检测方法研究[J].智慧电力,2021,49(8):98-103.

LI Jun,REN Jing,WANG Yelin,et al. Insulator self-explosion detection method based on improved U-net and CNN[J].Smart Power,2021,49(8):98-103.

李斌,曾筠婷,朱新山,等.基于多尺度上下文感知的绝缘子缺陷检测网络[J].高电压技术,2022,48(8):2905-2914.

Li Bin,Zeng Junting,Zhu Xinshan,et al. Detection network for insulator defects based on multi-scale context awareness[J].High Voltage Engineering,2022,48(8):2905-2914.

罗潇,於锋,彭勇.基于深度学习的无人机电网巡检缺陷检测研究[J].电力系统保护与控制,2022,50(10):132-139.

Luo Xiao,Yu Feng,Peng Yong. UAV power grid inspection defect detection based on deep learning[J].Power System Protection and Control,2022,50(10):132-139.

马进,白雨生.应用于绝缘子缺陷检测的轻量化YOLOv4研究[J].电子测量技术,2022,45(14):123-130.

Ma Jin,Bai Yusheng. Research on lightweight YOLOv4 applied to insulator defect detection[J].Electronic Measurement Technology, 2022,45(14):123-130.

李斌,屈璐瑶,朱新山,等.基于多尺度特征融合的绝缘子缺陷检测[J].电工技术学报,2023,38(1):60-70.

Li Bin,Qu Luyao,Zhu Xinshan,et al. Insulator defect detection based on multi-scale feature fusion[J].Transactions of China Electrotechnical Society,2023,38(1):60-70.

Wang C Y,Bochkovskiy A,Liao Hongyuan. Scaled-YOLOv4:scaling cross stage partial network[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR),2021:13024-13033.

Ioffe S,Szegedy C. Batch normalization:accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on International Conference on Machine Learning,2015:448-456.

张印辉,杨宏宽,朱守业,等.基于轻型调控网络的下茧机器视觉实时检测[J].农业机械学报,2022,53(4):261-270.

Zhang Yinhui,Yang Hongkuan,Zhu Shouye,et al. Machine vision real time detection of inferior cocoons based on lightweight manipulation network[J].Transactions of the Chinese Society for Agricultural Machinery,2022,53(4):261-270.

Liu Zhuang,Li Jianguo,Shen Zhiqiang,et al. Learning efficient convolutional networks through network slimming[C]//2017 IEEE International Conference on Computer Vision(ICCV),2017:2755-2763.

Hioton G,Vinyals O,Dean J. Distilling the knowledge in neural network[J].Computer Science,2015,14(7):38-39.

Anon.Insulator data set - chinese power line insulator dataset (CPLID)[EB/OL][2026-05-15].https://github.com/InsulatorData/InsulatorDataSet.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Prediction of Transient Magnetic Field of Transformer Inrush Current Fusing Multi-surce Data and CNN-Transformer

Influence of Geological Activity on Deformation Characteristics of GIS/GIL with Long-distance and Large Elevation Difference

Multi-physics Field Coupling Analysis and Optimal Design of Spring Finger Structure in 550 kV GIL

Distribution Characteristics of Electromagnetic Forces in Windings of Thyristor-controlled Phase Shifters Under Typical Fault Conditions

Design of High-voltage Sinusoidal Power Supply with Superimposed High-frequency Oscillations for Air Dielectric Barrier Discharge

Related Author

No data

Related Institution

No data

⁰