基于改进YOLOv8模型的配电网设备检测方法

摘要
图/表
参考文献
相关文章 (11)

全文: PDF (3219 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要针对配电网目标检测中目标形态多样、遮挡严重和环境复杂,以及现有目标检测算法存在检测速度慢和精度不足等问题,本文提出一种基于改进YOLOv8的低压配电网设备检测方法。首先,构建一个多层次注意力融合块（MAFB）,通过层次化的特征处理路径对输入特征进行精细化融合,从而提升特征表达的鲁棒性和任务适配性;其次,对原有的检测头做轻量化处理,设计共享卷积轻量检测头,以大幅减少参数量,解决尺度不一致问题;最后,融合MPDIoU、Inner-IoU和Wise-IoU多指标加权损失函数,对边界框回归损失函数进行优化。在Roboflow数据集实验中,改进后的模型较YOLOv8m的参数量降低了约7.0%,mAP₅₀和mAP_50:95分别提高了9.7个百分点和8.1个百分点,验证了模型的有效性。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	叶建盈
	林波
	刘磊
	陈颖婷
	袁国发

关键词 ：低压配电网, YOLOv8, 目标检测, 多维度特征融合, 轻量化检测

Abstract：To address the challenges posed by diverse target shapes, significant occlusion, and complex environments in distribution network target detection, along with the slow detection speed and inadequate accuracy of existing target detection algorithms, an improved you only look once (YOLO) v8 algorithm for detecting low-voltage distribution network equipment is proposed. A multi-attention fusion block (MAFB) is constructed to refine feature fusion through a hierarchical processing pipeline, enhancing robustness and task adaptability. The original detection head undergoes lightweight processing, resulting in the design of a shared convolutional lightweight detection head that substantially reduces the number of parameters while resolving scale inconsistency issues. The bounding box regression loss function is optimized by integrating a multi-metric weighted loss function combining MPDIoU, Inner-IoU, and Wise-IoU. Experiments on the Roboflow dataset demonstrate that the improved model achieves a 7.0% reduction in parameters compared to YOLOv8m, while boosting mAP₅₀ and mAP_50:95 by 9.7 percentage points and 8.1 percentage points, respectively, validating the model’s effectiveness.

Key words： low-voltage distribution network you only look once (YOLO) v8 object detection multi-dimensional feature fusion lightweight detection

收稿日期: 2025-10-16

基金资助:国网新疆电力有限公司科技项目（GY-H-24133）

作者简介: 叶建盈（1980—）,男,福建省福州市人,博士,副教授,主要从事智能配电网技术、电力电子与新能源技术、电力电子高频磁技术等研究工作。

引用本文:

叶建盈, 林波, 刘磊, 陈颖婷, 袁国发. 基于改进YOLOv8模型的配电网设备检测方法[J]. 电气技术, 2026, 27(4): 24-32. YE Jianying, LIN Bo, LIU Lei, CHEN Yingting, YUAN Guofa. A method for detecting equipment in distribution networks based on the improved YOLOv8 model. Electrical Engineering, 2026, 27(4): 24-32.

链接本文:

https://dqjs.cesmedia.cn/CN/Y2026/V27/I4/24

[1] 齐宁, 程林, 田立亭, 等. 考虑柔性负荷接入的配电网规划研究综述与展望[J]. 电力系统自动化, 2020, 44(10): 193-207.
[2] Ali Siddiqui Z, Park U.A drone based transmission line components inspection system with deep learning technique[J]. Energies, 2020, 13(13): 3348.
[3] Yang Lei, Fan Junfeng, Liu Yanhong, et al.A review on state-of-the-art power line inspection techniques[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(12): 9350-9365.
[4] 常安, 陈振辉, 付明, 等. 无人机电力巡检航线智能规划及自主巡检研究[J]. 电网与清洁能源, 2023, 39(7): 61-66.
[5] 陆煜锌, 赵云, 肖勇, 等. 面向数字孪生低压配电网的基于Gabor-YOLO算法的架空线高效识别方法研究[J]. 电测与仪表, 2023, 60(3): 40-46.
[6] 律方成, 牛雷雷, 王胜辉, 等. 基于优化YOLOv4的主要电气设备智能检测及调参策略[J]. 电工技术学报, 2021, 36(22): 4837-4848.
[7] 尤振飞, 赵健, 王小宇, 等. 基于语义分割与连通区域标记的隔离开关状态识别方法[J]. 电力系统自动化, 2021, 45(20): 157-165.
[8] 张友鹏, 王文豪, 赵珊鹏, 等. 基于RBCT算法的复杂背景下铁路接触网电力线自动提取研究[J]. 高电压技术, 2022, 48(6): 2234-2243.
[9] 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, Columbus, OH, USA, 2014: 580-587.
[10] Ren Shaoqing, He Kaiming, Girshick R, et al.Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[11] Cai Zhaowei, Vasconcelos N.Cascade R-CNN: delving into high quality object detection[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recog-nition, Salt Lake City, UT, USA, 2018: 6154-6162.
[12] 孟子超, 杜文娟, 王海风. 基于迁移学习深度卷积神经网络的配电网故障区域定位[J]. 南方电网技术, 2019, 13(7): 25-33.
[13] 陈永, 安卓奥博, 周建宇. 基于扩散模型检测的高铁接触网绝缘子缺陷语义描述方法[J]. 电工技术学报, 2025, 40(13): 4100-4111.
[14] 米增, 连哲. 面向通用目标检测的YOLO方法研究综述[J]. 计算机工程与应用, 2024, 60(21): 38-54.
[15] Gong Wenkai.Lightweight object detection: a study based on YOLOv7 integrated with ShuffleNetv2 and Vision Transformer[PP/OL].[2026-02-04].https://arxiv.org/abs/2403.01736.
[16] 裴少通, 王玮琦, 胡晨龙, 等. 基于小目标复杂环境YOLO算法的瓷质绝缘子低值缺陷识别方法[J]. 电工技术学报, 2025, 40(23): 7793-7805.
[17] 宋智伟, 黄新波, 纪超, 等. 基于Flexible YOLOv7的输电线路绝缘子缺陷检测和故障预警方法[J]. 高电压技术, 2023, 49(12): 5084-5094.
[18] 廖方舟, 杨晓霞, 杨容浩, 等. 基于改进YOLOv8的街景图像变压器目标检测[J]. 电气技术, 2024, 25(12): 12-20, 27.
[19] 宋欣杰, 金一鸣. 基于深度学习的架空输电线路绝缘子识别方法研究[J]. 电气技术, 2025, 26(9): 62-68, 78.
[20] 聂凯, 李泽, 高文豪, 等. 基于YOLOv8-RFL模型的电力绝缘子缺陷检测方法[J]. 电网技术, 2025, 49(9): 3948-3957.
[21] Xu Shibiao, Zheng Shuchen, Xu Wenhao, et al.HCF-net: hierarchical context fusion network for infrared small object detection[C]//2024 IEEE International Conference on Multimedia and Expo (ICME), Niagara Falls, ON, Canada, 2024: 1-6.
[22] Zhong Jiaqi, Qian Huaming, Wang Huilin, et al.Improved real-time object detection method based on YOLOv8: a refined approach[J]. Journal of Real-Time Image Processing, 2025, 22: 4.
[23] Tian Zhi, Shen Chunhua, Chen Hao, et al.FCOS: fully convolutional one-stage object detection[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea, 2019: 9626-9635.
[24] Ioffe S, Szegedy C.Batch normalization: accelerating deep network training by reducing internal covariate shift[C]//The 32nd International Conference on Machine Learning, Lille, France, 2015.
[25] Wu Yuxin, He Kaiming.Group normalization[C]//The 15th European Conference on Computer Vision (ECCV), Munich, Germany, 2018: 3-19.
[26] Ba J L, Kiros J R, Hinton G E.Layer normali-zation[PP/OL].[2026-02-04].https://arxiv.org/abs/1607.06450.
[27] Ghiasi G, Lin T Y, Le Q V.NAS-FPN: learning scalable feature pyramid architecture for object detection[C]//2019 IEEE/CVF Conference on Com-puter Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 2019: 7029-7038.
[28] Zhang Hao, Xu Cong, Zhang Shuaijie.Inner-IoU: more effective intersection over union loss with auxiliary bounding box[PP/OL].[2026-02-04].https://arxiv.org/abs/2311.02877.
[29] Ma Siliang, Xu Yong.MPDIoU: a loss for efficient and accurate bounding box regression[PP/OL].[2026-02-04].https://arxiv.org/abs/2307.07662.
[30] Tong Zanjia, Chen Yuhang, Xu Zewei, et al.Wise-IoU: bounding box regression loss with dynamic focusing mechanism[PP/OL].[2026-02-04].https://arxiv.org/abs/2301.10051.