摘要
电力行业设备发热是设备巡检过程中最为关注的问题之一,输电线路耐张线夹作为电流传导的关键设备,其引流板发热会对输电线路稳定运行造成严重影响。红外检测方式无须直接接触设备,可在不停电的状态下发现故障类型,在系统运维的安全性、可靠性和经济性等方面具有明显优势。目前已有部分利用红外图像监测设备发热的算法研究,但普遍具有设备定位不准确、无法识别发热程度等短板。针对这些短板本文提出了一种识别耐张线夹引流板发热程度的方案,首先基于目标检测领域中的yolo-v4算法在红外图像中定位出引流板位置,然后提取红外图像中的原始热量矩阵并转化为温度信息,最后参照正常部件或者导线的温度判别是否发热以及发热等级。实验证明无论是识别效果还是识别效率都有较好的表现,为后续其他场景的设备发热识别提供了思路。
Equipment heating in power industry is one of the most concerned problems in the process of equipment inspection,transmission line tension clamp is the key equipment of current conduction,the heating of the drainage plate will have a serious impact on the stable operation of the transmission line.Infrared detection does not require direct contact with the equipment,the fault type can be found without power failure,it has obvious advantages in safety,reliability and economy of system operation and maintenance.At present,some algorithms using infrared image monitoring equipment have been studied,however,it generally has some shortcomings,such as inaccurate equipment positioning and unable to identify the degree of heating.Aiming at these short plates,a scheme to identify the heating degree of strain clamp drainage plate is proposed in this paper.Firstly,the position of the drainage plate is located in the infrared image based on the yolo-v4 algorithm in the field of target detection.Then,the original heat matrix in the infrared image is extracted and transformed into temperature information.Finally,judge whether heating and heating grade with reference to the temperature of normal parts or wires.Experiments show that both recognition effect and recognition efficiency have good performance,it provides ideas for equipment heating identification in other subsequent scenes.
作者
吴洋
陈英达
黄士超
黄文琦
梁凌宇
敖榜
WU Yang;CHEN Yingda;HUANG Shichao;HUANG Wenqi;AO Bang(Digital Grid Research Institute of China Southerm Power Grid Co.,Ltd.,Guangzhou 510700,Guangdong,China)
出处
《电力大数据》
2022年第8期12-19,共8页
Power Systems and Big Data
基金
中国南方电网有限责任公司科技项目(批准号:GZ2015-2-0047)。
关键词
红外图像
引流板
发热程度
识别算法
目标检测
infrared image
drainage plate
heat generation
recognition algorithm
object detection
作者简介
吴洋(1989),男,硕士,高级工程师,长期从事图像识别算法研发工作,主要研究方向为神经网络深度学习在电网生产域巡检领域的应用技术。
