基于语义信息距离解耦的变电运维多类别缺陷图像检测

doi:10.11930/j.issn.1004-9649.202208117

中国电力 ›› 2023, Vol. 56 ›› Issue (6): 209-218.DOI: 10.11930/j.issn.1004-9649.202208117

• 信息与通信 • 上一篇

基于语义信息距离解耦的变电运维多类别缺陷图像检测

张鑫^1,2, 叶俊杰³, 崔瑶^1,2, 黄鑫^1,2, 仲林林³

1. 南瑞集团有限公司（国网电力科学研究院有限公司），江苏南京 211106;
2. 智能电网保护和运行控制国家重点实验室，江苏南京 211106;
3. 东南大学电气工程学院，江苏南京 210096

收稿日期:2022-08-30 修回日期:2022-11-18 发布日期:2023-07-04
作者简介:张鑫（1987—），男，硕士，高级工程师，从事电力系统通信、电力智能运维技术及应用研究，E-mail：zhangxin4@sgepri.sgcc.com.cn;仲林林（1990—），男，通信作者，博士，副研究员，从事高电压技术、放电等离子体技术、人工智能技术研究，E-mail：linlin@seu.edu.cn
基金资助:
国网电力科学研究院有限公司科技项目（524606210002）。

Decoupled Sematic Distance Based Multi-class Defect Scene Detecting for Substations

ZHANG Xin^1,2, YE Junjie³, CUI Yao^1,2, HUANG Xin^1,2, ZHONG Linlin³

1. NARI Group Corporation (State Grid Electric Power Research Institute), Nanjing 211106, China;
2. State Key Laboratory of Smart Grid Protection and Control, Nanjing 211106, China;
3. School of Electrical Engineering, Southeast University, Nanjing 210096, China

Received:2022-08-30 Revised:2022-11-18 Published:2023-07-04
Supported by:
This work is supported by Science and Technology Project of State Grid Electric Power Research Institute (No.524606210002).

摘要/Abstract

摘要： 变电站设备种类繁多、缺陷类型复杂、特征差异大，传统的基于深度学习的缺陷图像检测模型难以同时有效处理不同设备的多种缺陷。为此，提出了一种基于语义信息距离解耦的缺陷图像检测模型（sematic-distance based decoupling detection model，SDB-DDM）。首先对缺陷类别进行语义信息聚簇，构建解耦式网络结构，然后对网络输出进行加权锚框融合，并在损失函数中加入局部预测损失以提升预测能力，同时提出解耦式非极大值抑制策略以加快模型推理速度。该模型可根据缺陷类别进行自适应调整，以适用变电运维多类别缺陷图像检测的应用场景。实验结果显示，该模型的平均精度均值达到了69.68%。同平台下相较于目前性能最佳的目标检测模型（YOLOX），精度提升了1.36个百分点，参数量下降了5%，推理速度提升了34%。

关键词: 变电运维场景, 缺陷检测, 深度学习, 语义信息距离, 解耦式模型

Abstract: Due to the complexity and differences of defect types in substations, traditional deep learning models for defects detection lack comprehensive response ability. It proposes a sematic distance based decoupling detection model. Firstly, the decoupled model structure is determined by clustering defect classes according to the semantic information distance between each other. Then, the weighted anchor fusion and local prediction loss techniques are used to improve the model performance. Meanwhile, the decoupled non-maximum suppression strategy is proposed to accelerate the model inference process. The experiment results show that the mean average precision of the model reaches 69.68%. Compared with YOLOX, which has been recognized as the best real-time object detection model, the accuracy of proposed model is improved by 1.36 percentage points, the parameter quantity is reduced by 5%, and the inference speed is improved by 34%.

Key words: substation maintain and operation scene, defect detection, deep learning, semantic information distance, decoupled model

张鑫, 叶俊杰, 崔瑶, 黄鑫, 仲林林. 基于语义信息距离解耦的变电运维多类别缺陷图像检测[J]. 中国电力, 2023, 56(6): 209-218.

ZHANG Xin, YE Junjie, CUI Yao, HUANG Xin, ZHONG Linlin. Decoupled Sematic Distance Based Multi-class Defect Scene Detecting for Substations[J]. Electric Power, 2023, 56(6): 209-218.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202208117

https://www.electricpower.com.cn/CN/Y2023/V56/I6/209

参考文献

[1] 罗琨, 时永肖, 李正新, 等. 智能变电站继电保护装置寿命模型及其辨识方法[J]. 智慧电力, 2021, 49(1): 96–101
LUO Kun, SHI Yongxiao, LI Zhengxin, et al. Life model and identification method of relay protection device in smart substation[J]. Smart Power, 2021, 49(1): 96–101
[2] 李肖博, 于杨, 姚浩, 等. 新一代智能变电站采控装置[J]. 中国电力, 2022, 55(4): 85–92
LI Xiaobo, YU Yang, YAO Hao, et al. Sample-control-device of smart substation[J]. Electric Power, 2022, 55(4): 85–92
[3] 李泽文, 王志刚, 穆利智, 等. 变电站智能安监可穿戴设备设计[J]. 电力科学与技术学报, 2022, 37(4): 217–226
LI Zewen, WANG Zhigang, MU Lizhi, et al. Design of wearable equipment for substation intelligent safety supervision[J]. Journal of Electric Power Science and Technology, 2022, 37(4): 217–226
[4] 李自若, 沈曦, 张亦兵, 等. 基于深度强化学习的智慧变电站网络异常检测方法[J]. 南方电网技术, 2021, 15(6): 98–105
LI Ziruo, SHEN Xi, ZHANG Yibing, et al. Network anomaly detection method for smart substation based on deep reinforcement learning[J]. Southern Power System Technology, 2021, 15(6): 98–105
[5] 高熠, 田联房, 杜启亮. 基于Mask R-CNN的复合绝缘子过热缺陷检测[J]. 中国电力, 2021, 54(1): 135–141
GAO Yi, TIAN Lianfang, DU Qiliang. Overheating defect detection of composite insulator based on mask R-CNN[J]. Electric Power, 2021, 54(1): 135–141
[6] 王帅, 姜敏, 李江林, 等. 全维度智能变电站设备状态监测关键技术研究[J]. 电测与仪表, 2020, 57(7): 82–86
WANG Shuai, JIANG Min, LI Jianglin, et al. Research on key technologies of condition monitoring of full-dimensional intelligent substation equipment[J]. Electrical Measurement & Instrumentation, 2020, 57(7): 82–86
[7] 刘颖, 胡楠, 杨壮观, 等. 基于深度学习的电网监控视频中工作人员检测与识别[J]. 沈阳工业大学学报, 2019, 41(5): 544–548
LIU Ying, HU Nan, YANG Zhuangguan, et al. Detection and identification of staff in power grid monitoring video based on deep learning[J]. Journal of Shenyang University of Technology, 2019, 41(5): 544–548
[8] 王旭红, 李浩, 樊绍胜, 等. 基于改进SSD的电力设备红外图像异常自动检测方法[J]. 电工技术学报, 2020, 35(增刊1): 302–310
WANG Xuhong, LI Hao, FAN Shaosheng, et al. Infrared image anomaly automatic detection method for power equipment based on improved single shot multi box detection[J]. Transactions of China Electrotechnical Society, 2020, 35(S1): 302–310
[9] 王凯, 王健, 刘刚, 等. 基于RetinaNet和类别平衡采样方法的销钉缺陷检测[J]. 电力工程技术, 2019, 38(4): 80–85
WANG Kai, WANG Jian, LIU Gang, et al. Defect detection of pins based on RetinaNet and class balanced sampling methods[J]. Electric Power Engineering Technology, 2019, 38(4): 80–85
[10] 黄锐勇, 戴美胜, 郑跃斌, 等. 电力设备红外图像缺陷检测[J]. 中国电力, 2021, 54(2): 147–155
HUANG Ruiyong, DAI Meisheng, ZHENG Yuebin, et al. Defect detection of power equipment by infrared image[J]. Electric Power, 2021, 54(2): 147–155
[11] 位一鸣, 童力, 罗麟, 等. 基于卷积神经网络的主变压器外观缺陷检测方法[J]. 浙江电力, 2019, 38(4): 61–68
WEI Yiming, TONG Li, LUO Lin, et al. An exterior defects detecting method of main transformer based on convolutional neural networks[J]. Zhejiang Electric Power, 2019, 38(4): 61–68
[12] 张晶焯, 佘楚云, 伍国兴, 等. 基于增强特征金字塔和可变形卷积的绝缘子缺陷检测[J]. 电力工程技术, 2021, 40(4): 155–160
ZHANG Jingzhuo, SHE Chuyun, WU Guoxing, et al. Insulator defect detection based on enhanced feature pyramid and deformable convolution[J]. Electric Power Engineering Technology, 2021, 40(4): 155–160
[13] 罗鹏, 王波, 马恒瑞, 等. 基于组合式目标检测框架的低漏报率缺陷识别方法[J]. 高电压技术, 2021, 47(2): 454–464
LUO Peng, WANG Bo, MA Hengrui, et al. Defect recognition method with low false negative rate based on combined target detection framework[J]. High Voltage Engineering, 2021, 47(2): 454–464
[14] 顾晓东, 唐丹宏, 黄晓华. 基于深度学习的电网巡检图像缺陷检测与识别[J]. 电力系统保护与控制, 2021, 49(5): 91–97
GU Xiaodong, TANG Danhong, HUANG Xiaohua. Deep learning-based defect detection and recognition of a power grid inspection image[J]. Power System Protection and Control, 2021, 49(5): 91–97
[15] 应樱. 基于改进Faster R-CNN的变电站设备缺陷检测算法研究[D]. 杭州: 浙江大学, 2021.
YING Ying. Research on defect detection algorithm of substation equipment based on improved Faster R-CNN[D]. Hangzhou: Zhejiang University, 2021.
[16] 尹思宇. 深度学习在变电站电力设备缺陷检测中的应用研究[D]. 成都: 电子科技大学, 2020.
YIN Siyu. Research on application of deep learning in defect detection of substation power equipment[D]. Chengdu: University of Electronic Science and Technology of China, 2020.
[17] 徐海洋. 变电站设备表面缺陷图像识别关键技术研究[D]. 南宁: 广西大学, 2021.
XU Haiyang. Research on key technologies of image recognition for surface defects of substation equipment[D]. Nanning: Guangxi University, 2021.
[18] GE Z, LIU S, WANG F, et al. YOLOX: exceeding YOLO series in 2021[EB/OL]. 2021: arXiv: 2107.08430.https://arxiv.org/abs/2107.08430.
[19] WANG C Y, LIAO H Y M, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Seattle, WA, USA. IEEE, 2020: 1571–1580.
[20] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. 2020: arXiv: 2004.10934.https://arxiv.org/abs/2004.10934.
[21] REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. 2018: arXiv: 1804.02767.https://arxiv.org/abs/1804.02767.
[22] SALEH S, PAUL M, GREG F. “How many images do I need?” Understanding how sample size per class affects deep learning model performance metrics for balanced designs in autonomous wildlife monitoring[J]. Ecological Informatics, 2020, 57: 101085.
[23] HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904–1916.
[24] BODLA N, SINGH B, CHELLAPPA R, et al. Soft-NMS—improving object detection with one line of code[C]//2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy. IEEE, 2017: 5562–5570.
[25] DUAN K W, BAI S, XIE L X, et al. CenterNet: keypoint triplets for object detection[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, Korea (South). IEEE, 2020: 6568–6577.

基于语义信息距离解耦的变电运维多类别缺陷图像检测

Decoupled Sematic Distance Based Multi-class Defect Scene Detecting for Substations

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	姜通海, 王峰, 刘子琪, 单帅杰. 基于改进生成对抗网络的风光气象资源联合场景生成方法[J]. 中国电力, 2025, 58(3): 183-192.
[2]	刘东, 张越, 皮俊波, 单连飞, 刘赫, 宋鹏程, 姜涛. 面向电网调控信息智能检索的知识图谱构建及应用[J]. 中国电力, 2023, 56(7): 78-84.
[3]	周震震, 宋云海, 何宇浩, 王黎伟, 黄和燕, 何珏, 朱志航, 闫云凤. 基于分组查询注意力的可扩展电力人员行为分类方法[J]. 中国电力, 2023, 56(11): 77-85.
[4]	李冰, 白云山, 赵宽, 郭聪彬, 翟永杰. 基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法[J]. 中国电力, 2023, 56(10): 43-52.
[5]	李丰君, 王磊, 赵健, 张建宾, 张世尧, 田杨阳. 基于天气融合和LSTM网络的分布式光伏短期功率预测方法[J]. 中国电力, 2022, 55(11): 149-154.
[6]	赵永良, 付鑫, 郭阳, 边迎迎, 王思宁. 基于深度学习和图像识别的电力配件智能出入库[J]. 中国电力, 2021, 54(3): 55-60.
[7]	马静怡, 崔昊杨, 张明达, 孙益辉, 许永鹏. 基于改进Faster RCNN的小尺度入侵目标识别及定位[J]. 中国电力, 2021, 54(3): 38-44.
[8]	刘逸凡, 王淑青, 庆毅辉, 王晨曦, 兰天泽, 要若天. 基于EfficientDet和双目摄像头的绝缘子缺陷检测[J]. 中国电力, 2021, 54(2): 156-163,196.
[9]	黄锐勇, 戴美胜, 郑跃斌, 黄勤琴, 康立烨, 苟先太, 周维超. 电力设备红外图像缺陷检测[J]. 中国电力, 2021, 54(2): 147-155.
[10]	曹培, 徐鹏, 贺建明, 高凯, 田昊洋, 季怡萍. 基于多源感知的开关柜绝缘缺陷检测技术[J]. 中国电力, 2021, 54(10): 117-124,133.
[11]	周俊煌, 黄廷城, 谢小瑜, 范纹郡, 易婷婷, 张勇军. 视频图像智能识别技术在输变电系统中的应用研究综述[J]. 中国电力, 2021, 54(1): 124-134,166.
[12]	吴云亮, 张建新, 李豹, 李鹏, 李智勇, 周鑫, 杨燕, 赖晓文. 深度学习辅助约束辨识的电力市场快速出清方法[J]. 中国电力, 2020, 53(9): 90-97,207.
[13]	赵会茹, 赵一航, 郭森. 基于互补集合经验模态分解和长短期记忆神经网络的短期电力负荷预测[J]. 中国电力, 2020, 53(6): 48-55.
[14]	杨晓楠, 孙博, 郎燕生. 基于深度学习的特高压直流闭锁故障智能调度决策[J]. 中国电力, 2020, 53(6): 8-17.
[15]	周艳真, 查显煜, 兰健, 郭庆来, 孙宏斌, 薛峰, 王胜明. 基于数据增强和深度残差网络的电力系统暂态稳定预测[J]. 中国电力, 2020, 53(1): 22-31.

模态框（Modal）标题

基于语义信息距离解耦的变电运维多类别缺陷图像检测

Decoupled Sematic Distance Based Multi-class Defect Scene Detecting for Substations

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics