微型电脑应用2025，Vol.41Issue(3)：5-9,14,6.
应用DenseNet识别电网绝缘子故障的研究
Research on the Application of DenseNet in Identifying Insulator Faults in Power Grids
摘要
Abstract
The detection of power grid insulators based on unmanned aerial vehicle images is a significant research topic of the power grid industry applying modern technology.This paper employs DenseNet-121 and DenseNet-169 models,utilizing nine sample sizes and seven evaluation metrics to comprehensively analyze the performance of 2 models in recognizing and classifying insulators and their defects.The results indicate that both DenseNet-121 and DenseNet-169 can effectively identify and classify transmission line insulators and their defects,achieving a maximum accuracy of 99.1％.With a sample size of 400,the accura-cies of DenseNet-121 and DenseNet-169 are 96.4％and 97.2％,respectively,outperforming existing models such as Faster RCNN,YOLO.The accuracy of DenseNet models exceeds 95.0％when the sample size is less than 500,and their detection accuracy does not significantly improve with an increase in sample size,and DenseNet-121 outperforms DenseNet-169.For sample sizes between 500～800,DenseNet-169 performs better.For sample sizes exceeding 1000,both models exhibit similar performance.This study provides new methodological support and insights for the application of unmanned aerial vehicle re-mote sensing and deep learning in intelligent inspection of power grid,and also offers new ideas for interpreting high-resolution unmanned aerial vehicle images,which has important application prospects.关键词
DenseNet/无人机应用/故障诊断/绝缘子/小样本Key words
DenseNet/application of unmanned aerial vehicle/fault diagnosis/insulator/small sample分类
计算机与自动化引用本文复制引用
尘昌华,李文波,桂元苗,王亦凡..应用DenseNet识别电网绝缘子故障的研究[J].微型电脑应用,2025,41(3):5-9,14,6.基金项目
国家自然科学基金(41871302) （41871302）
国家重点研发计划(2018YFB1004600) （2018YFB1004600）
