基于改进YOLOv8的输电线路故障识别方法  

Fault Identification Method for Transmission Lines Based on Improved YOLO v8

作　　者：李宁 程旭 卢景才 梁河雷 时洪刚 LI Ning;CHENG Xu;LU Jingcai;LIANG Helei;SHI Honggang(State Grid Hebei Electric Power Co.,Ltd.Hengshui Power Supply Company,Hengshui 053000,China)

机构地区：[1]国网河北省电力有限公司衡水供电分公司,河北衡水053000

出　　处：《河北电力技术》2024年第4期56-63,共8页Hebei Electric Power

基　　金：国网河北省电力有限公司科技项目(kj2022-042)。

摘　　要：针对输电线路巡检难度大且巡检信息处理可靠性不佳等问题,提出了一种基于改进YOLOv8的输电线路故障识别方法。首先,设计智能化单兵巡检装备,包括智能巡检头盔和智能信息装备服,并通过无人机获取输电线路的实时运行情况。然后,提出一种增量八叉树空间检索算法用于激光雷达等图像信息的处理,得到输电线路全景图像。最后,构建改进C2f模块、残差注意力模块以及改进损失函数优化YOLOv8模型,将其用于全景图像的学习,从而得到输电线路的故障类型。基于Pytorch平台对所提方法进行实验分析,结果表明,其识别结果的平均精度均值达到了92.03%,且识别时间仅为28ms,能够满足智能化单兵巡检装备的工作需求。A fault identification method for transmission lines based on improved YOLO v8 was proposed to address the challenges of difficult inspection of transmission lines and poor reliability of inspection information processing.Firstly,the intelligent individual inspection equipment was designed,including intelligent inspection helmets and intelligent information equipment suits,and operate drones to obtain real-time operation status of transmission lines.Then,an incremental octree spatial retrieval algorithm was proposed to process the image information such as LiDAR to obtain panoramic images of transmission lines.Finally,an improved C2f module,residual attention module,and improved loss function were constructed to optimize the YOLO v8 model,which was used for panoramic image learning to obtain the fault types of transmission lines.Based on the Pytorch platform,experimental analysis was conducted on the proposed method,and the results showed that the mPA of the recognition result reached 92.03%,and the recognition time was only 28 ms,which can meet the work requirements of intelligent individual patrol equipment.

关 键 词：智能化单兵巡检装备 增量八叉树空间检索算法 全景图像 输电线路 YOLOv8模型 故障识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TM72[自动化与计算机技术—计算机科学与技术]