基于免疫深度网络的电力设备检测算法  

Power equipment detection algorithm based on immune deep network

作　　者：于晓[1] 薛浩 YU Xiao;XUE Hao(School of Electrical Engineering and Automation,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学电气工程与自动化学院,天津300384

出　　处：《河南科技学院学报（自然科学版）》2024年第2期48-55,共8页Journal of Henan Institute of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金青年项目(61502340);天津市科技计划项目(18JCQNJCO1000)。

摘　　要：随着电力系统的快速发展,保障电力设备的稳定安全运行成为一项重要工作.针对绝缘子和变压器套管的工作状态检测问题,研究基于生物免疫智能理论与当前深度学习模型,提出了一种新的免疫深度网络,采用深度残差网络提取特征、SIoU计算检测锚框损失、3块SPP结构进行多尺度特征融合.通过实验对比分析,免疫深度网络MAP可以达到80.97%,相较于YOLOv3模型提升4.55%,并通过与Prewitt、Kmeans、Otsu、YOLOv3检测效果对比,免疫深度网络不仅未出现漏检、误检现象,且检测准确率优于其他模型.With the rapid development of the power system,ensuring the stable and safe operation of power equipment has become an important task.Aiming at the problem of detection of working state of insulators and transformer bushings,based on the theory of biological immune intelligence and the current deep learning model,a new immune depth network is proposed,and the deep residual network is used to extract features,SIoU calculation is used to detect anchor frame loss,and three SPP structures are used for multi-scale feature fusion.Through experimental comparative analysis,the MAP of the immune deep network can reach 80.97%,which is 4.55%higher than that of the YOLOV3 model,and by comparing with the detection effect of Prewitt,Kmeans,Otsu,and YOLOv3,the immune deep network not only has no missed detection and false detection,but also has a better detection accuracy than other models.

关 键 词：生物免疫 数据处理 绝缘子 套管 

分 类 号：TM642[电气工程—电力系统及自动化]