基于改进YOLOv11n模型的变电站设备及生产行为异常检测  

Anomaly Detection of Substation Equipment and Operational Behavior Based on the Improved YOLOv11n Model

在线阅读下载全文

作  者:魏迎澳 徐建 李英豪 胡浩特 WEI Yingao;XU Jian;LI Yinghao;HU Haote(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi 445000,China;Liupanshui Power Supply Bureau of Guizhou Power Grid Co.,Ltd.,Liupanshui 553000,China)

机构地区:[1]湖北民族大学智能科学与工程学院,湖北恩施445000 [2]贵州电网有限责任公司六盘水供电局,贵州六盘水553000

出  处:《湖北民族大学学报(自然科学版)》2025年第1期41-46,共6页Journal of Hubei Minzu University:Natural Science Edition

基  金:湖北民族大学研究生科研创新项目(MYK2024090)。

摘  要:针对目前人工巡视导致的变电站设备及生产行为异常检测效率低、人工风险高等问题,提出改进你只看一次11纳米版(you only look once version 11 nano, YOLOv11n)模型。首先,通过设计基于自注意力机制的3尺度卷积双路径可变核(convolutional three-scale kernel-adaptive dual-path self-attention mechanism, C3k2-SA)模块在较小特征图衔接特征融合部分,优化了网络结构,增强了全局特征提取能力。然后,在主干网络末层引入了基于注意力机制的特征增强(feature enhancement, FEN)模块,动态调整不同区域的特征权重,实现自适应的特征增强,缓解深层网络中的梯度消失问题。最后,对拼接(concatenate, Concat)模块进行优化,通过卷积层调整通道数,采用池化和sigmoid激活函数进行特征的精细处理,提高了模型对不同类型特征的自适应,增强了特征融合效果,同时抑制了无关或冗余特征,防止过拟合。结果表明,与原始YOLOv11n模型相比,改进YOLOv11n模型的精确率、召回率、平均精确率均值分别上升了1.7、6.6、3.6个百分点。改进YOLOv11n模型能够提高变电站异常状态检测的准确性,为智能变电站的异常检测工作提供一定参考。To address the current issues of low detection efficiency and high manual risks in detecting abnormal equipment and operational behavior in substations due to manual inspections,an improved you only look once version 11 nano(YOLOv11n)model was proposed.Firstly,a convolutional three-scale kernel-adaptive dual-path self-attention(C3k2-SA)module was designed,which was used to optimize the network structure and enhance the global feature extraction capability in the small feature map fusion section.Subsequently,a feature enhancement(FEN)module based on the attention mechanism was added to the final layer of the backbone network.This module dynamically adjusted the feature weights of different regions,enabling adaptive feature enhancement and alleviating the gradient vanishing problem in deep networks.Finally,the concatenate(Concat)module was optimized by adjusting the channel numbers through convolution layers,with pooling and sigmoid activation function applied for fine-grained feature processing.The model′s adaptability to different types of features and the feature fusion effect were strengthened and irrelevant or redundant features were suppressed to prevent overfitting.The results showed the precision,recall,and mean average precision of the improved YOLOv11n model increased by 1.7,6.6 and 3.6 percentage point respectively,compared to the original YOLOv11n model.The improved YOLOv11n model could be used to enhance the accuracy of abnormal state detection in substations and provide valuable insights for abnormal detection tasks in intelligent substations.

关 键 词:目标检测 YOLOv11n 激活函数 自注意力机制 C3k2 自适应平均池化 

分 类 号:TP391.41[自动化与计算机技术—计算机应用技术] TM63[自动化与计算机技术—计算机科学与技术]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象