Study on Hydrophobicity Diagnosis Method of Composite Insulator Based on Lightweight Convolution Neural Network

MA Ziru; LIU Yunpeng; PEI Shaotong; ZHANG Gonghao; YANG Jiajun

doi:10.13296/j.1001-1609.hva.2024.03.012

您当前的位置：

首页 >

文章列表页 >

Study on Hydrophobicity Diagnosis Method of Composite Insulator Based on Lightweight Convolution Neural Network

Research & Analysis | 更新时间：2025-04-26

- Study on Hydrophobicity Diagnosis Method of Composite Insulator Based on Lightweight Convolution Neural Network
- High Voltage Apparatus Vol. 60, Issue 3, Pages: 93-100(2024)
- 作者机构：
  
  1.华北电力大学河北省输变电设备安全防御重点实验室，河北　保定　 071003
  2.华北电力大学河北省分布式储能与微网重点实验室，河北　保定　 071003
  3.华北电力大学数理系，河北　保定　 071003
- 作者简介：
- 基金信息：
  
  Hebei Natural Science Foundation(E2019502150)
- DOI：10.13296/j.1001-1609.hva.2024.03.012
  CLC：
- Received：16 October 2023，
  
  Revised：17 December 2023，
  
  Published：16 March 2024
- 稿件说明：
移动端阅览
MA Ziru, LIU Yunpeng, PEI Shaotong, et al. Study on Hydrophobicity Diagnosis Method of Composite Insulator Based on Lightweight Convolution Neural Network[J]. High voltage apparatus, 2024, 60(3): 93-100.
DOI：

MA Ziru, LIU Yunpeng, PEI Shaotong, et al. Study on Hydrophobicity Diagnosis Method of Composite Insulator Based on Lightweight Convolution Neural Network[J]. High voltage apparatus, 2024, 60(3): 93-100. DOI： 10.13296/j.1001-1609.hva.2024.03.012.

摘要

复合绝缘子因其良好的性能在电网中普遍使用，为提高复合绝缘子带电憎水性检测的智能化和自动化水平，文中提出一种基于轻量级卷积神经网络的复合绝缘子憎水性诊断方法。文中使用EfficientNet算法对复合绝缘子不同憎水性等级对应的水迹图像数据进行特征提取训练，通过对模型准确率和参数量等方面的比较，并结合模型在边缘计算设备上的正向推理速度，确定EfficientNet-b

模型为本应用场景下的最优模型。该模型分类准确率为96.43%，在Jetson Xavier NX上的正向推理速度为57.16 FPS，在精度和速度方面均优于其他经典卷积神经网络对比算法。后接触角的测量结果验证了该模型在实际应用中的分类有效性。该模型搭载无人机可进行现场运行复合绝缘子的憎水性测量，具有一定的创新价值和良好的应用前景。

Abstract

Composite insulator is used widely in power grid due to its good performance. For improving both intelligent and automated level of live hydrophobicity detection of composite insulator

a kind of hydrophobicity diagnosis method of composite insulat

or based on lightweight convolution neural network is proposed in this paper. The EfficientNet algorithms is used in this paper to perform feature extraction and training on the water trace image data corresponding to the different hydrophobicity levels of the composite insulator. It is defined through comparison in such aspects as accuracy of model and the amount of parameters and also in combination with the model’s forward inference speed on the edge computing device that the EfficientNet-b

model is the optimal model in this application scenario. The classification accuracy rate of this model is 96.43%

and the forward inference speed on Jetson Xavier NX is 57.16 FPS

which is superior to other classic convolution neural network comparison algorithms in terms of accuracy and speed. The classification effectiveness of the model in practical applications is verified by the the measurement results of the back contact angle. This model equipped with an unmanned aerial vehicle can measure the hydrophobicity of composite insulator in operation

and has certain innovative value and good application prospects.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Thermal Aging Characteristics Analysis of HTV Silicone Rubber Composite Insulator Shed

Influence of Corona Discharge on Aging Characteristics of HTV Silicone Rubber Material

Identification of Insulator Hydrophobicity Based on Image Processing and RBF Neural Network

Design on Operation Mechanism of Hydrophobicity Detector for EHV and UHV Composite Insulators with Live Detection

Study on the Effect of Environmental Factors on Hydrophobicity of Composite Insulator

Related Author

LYU Li

XIA Lingzhi

HUANG Hongfei

GUO Kaihang

CHENG Yang

REN Chengyan

SHEN Wen-wei1

SONG Wei1

Related Institution

State Grid Anhui Electric Power Research Institute

Beijing International S & T Cooperation Base for Plasma Science and Energy Conversion, Institute of Electrical Engineering Chinese Academy of Sciences

University of Chinese Academy of Sciences

State Key Laboratory of Electrical Insulation and Power Equipment, Xi’an Jiaotong University

National Engineering Laboratory for Ultra High Voltage Engineering Technology

⁰