| 杨阳,杨帅,闫敏,胡晨龙,裴少通.基于UDD-YOLO的边缘端绝缘子
放电严重程度评估算法[J].电子测量与仪器学报,2024,38(1):219-227 |
| 基于UDD-YOLO的边缘端绝缘子
放电严重程度评估算法 |
| UDD-YOLO based edge-end insulator discharge severity assessment algorithm |
| |
| DOI： |
| 中文关键词: 紫外成像 放电评估 目标识别 YOLOv8 |
| 英文关键词:ultraviolet imaging discharge assessment target recognition YOLOv8 |
| 基金项目:国网河北省电力有限公司科技项目（kj2022-052）资助 |
|
|
| Author | Institution |
| Yang Yang | State Grid Hebei Electric Power Co., Ltd., Ultra High Voltage Branch, Shijiazhuang 050000, China |
| Yang Shuai | State Grid Hebei Electric Power Co., Ltd., Ultra High Voltage Branch, Shijiazhuang 050000, China |
| Yan Min | State Grid Hebei Electric Power Co., Ltd., Ultra High Voltage Branch, Shijiazhuang 050000, China |
| Hu Chenlong | North China Electric Power University (Baoding), Baoding 071000, China |
| Pei Shaotong | North China Electric Power University (Baoding), Baoding 071000, China |
|
| 摘要点击次数: 477 |
| 全文下载次数: 696 |
| 中文摘要: |
| 绝缘子是输电线路的重要组成部分，其放电问题是导致输电线路故障的主要原因之一，故需要可以准确对绝缘子放电严重程度进行快速评估且可在边缘端实时监测的算法方法。本文针对以上问题，首先对YOLOv8目标检测算法进行轻量化改进，首先引入Mosaic-9数据增强方法改进输入端，提高了算法的鲁棒性及泛用能力；而后引入了GhostNet网络替换主干网络，实现了对模型的轻量化；再引入GeLU激活函数替换ReLU激活函数，提高算法的收敛速度和检测精度；最后，引入了SIoU损失函数，对网络进行了优化，最终形成了UDD-YOLO边缘端绝缘子放电严重程度评估算法。经实验验证，其在边缘端设备实现了87.6% mAP及58 fps的推理速度，满足了在边缘端对绝缘子放电严重程度进行评估的要求，且通过消融、对比试验证明了本文提出的算法的有效性及优越性。 |
| 英文摘要: |
| Insulators are an important part of transmission lines, and their discharge problem is one of the main causes of transmission line faults, so there is a need for algorithms that can accurately and quickly assess the severity of insulator discharge and can be monitored in real time at the edge. In this paper, in order to address the above problems, the YOLOv8 target detection algorithm is firstly lightweighted and improved. Firstly, Mosaic-9 data enhancement method is introduced to improve the input, which improves the robustness and generalization ability of the algorithm; then GhostNet network is introduced to replace the backbone network, which realizes the lightweighting of the model; then the GeLU activation function is introduced to replace the ReLU activation function, which improves the convergence speed and detection accuracy of the algorithm; then the GELU activation function is introduced to replace the ReLU activation function. The GeLU activation function is introduced to replace the RELU activation function to improve the convergence speed and detection accuracy of the algorithm; finally, the SIoU loss function is introduced to optimize the network, and the UDD-YOLO edge-end insulator discharge severity assessment algorithm is finally formed. Experimentally verified, it achieves 87.6% mAP and 58 frames/s inference speed in the edge-end device, which meets the requirement of evaluating the severity of insulator discharge in the edge-end, and the effectiveness and superiority of the algorithm proposed in this paper is proved by ablation and comparison tests. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
