Substation Equipment Fault Detection Based on Multi-scale Feature Fusion

Xing Zhuang; Wen Shukai

doi:10.11959/j.issn.1000-0801.DXKX260174

您当前的位置：

首页 >

文章列表页 >

Substation Equipment Fault Detection Based on Multi-scale Feature Fusion

更新时间：2026-06-24

- Substation Equipment Fault Detection Based on Multi-scale Feature Fusion
- Telecommunications Science (2026)
- 作者机构：
  
  1.西安大麦网络科技有限公司，陕西西安 710016
  2.西安交通大学，电气工程学院，陕西西安 710049
- 作者简介：
- 基金信息：
- DOI：10.11959/j.issn.1000-0801.DXKX260174
  CLC： TP391.4;TM63
- Received：17 March 2026，
  
  Revised：2026-05-13，
  
  Accepted：01 June 2026，
- 稿件说明：
移动端阅览
Xing Zhuang, Wen Shukai. Substation Equipment Fault Detection Based on Multi-scale Feature Fusion[J/OL]. Telecommunications Science, 2026.
DOI：

Xing Zhuang, Wen Shukai. Substation Equipment Fault Detection Based on Multi-scale Feature Fusion[J/OL]. Telecommunications Science, 2026. DOI： 10.11959/j.issn.1000-0801.DXKX260174.

摘要

针对变电站设备缺陷目标尺度跨度大及背景干扰严重等问题，提出一种基于多尺度特征融合的变电站设备缺陷检测模型（Substation Equipment Defect Detection Transformer

SED-DETR）。模型在编码阶段设计分组级联自注意力模块（Grouped Cascaded Self-Attention

GCSA），增强对关键区域的聚焦能力；在颈部阶段构建跨阶段多分支特征融合网络（Cross-Scale Multi-Branch Fusion Feed-Forward Network

CSMB-FFN），通过改进单位步长融合卷积（Unit Step Fusion Convolution

US-Conv）实现多尺度特征精准对齐，引入跨阶段多分支建模模块（Cross-Stage Partial Omni-Kernel Block

CSP-OKB）加强全局感知与局部纹理信息融合；训练阶段采用MPDIoU（Minimum Pairwise Distance Intersection over Union）替换边界框回归损失中的GIoU（Generalized Intersection over Union），增强高重叠样本条件下的回归监督敏感性。SED-DETR在自建数据集和IGV Dataset上均取得优于当前主流检测模型的检测结果，mAP@0.5分别为94.1%和95.4%，mAP@0.5:0.95分别达到77.6%和85.2%。端侧部署结果说明，该模型在Jetson Orin Nano 4GB平台上具备良好的实时推理能力，经FP16加速后，推理时间缩短至23.9ms，且mAP@0.5达到92.5%。结果表明，SED-DETR在保证较高检测精度的同时兼顾计算效率，为变电站设备缺陷检测任务提供可行方案。

Abstract

To address the large-scale variation of defect targets and severe background interference in substation equipment inspection images

this paper proposes a substation equipment defect detection model

termed the Substation Equipment Defect Detection Transformer (SED-DETR)

based on multi-scale feature fusion. In the encoder

a Grouped Cascaded Self-Attention (GCSA) module is introduced to enhance focus on critical regions and improve cross-region contextual modeling. In the neck

a Cross-Scale Multi-Branch Fusion Feed-Forward Network (CSMB-FFN) is constructed

where an improved Unit Step Fusion Convolution (US-Conv) is employed for precise multi-scale feature alignment

and a Cross-Stage Partial Omni-Kernel Block (CSP-OKB) module is incorporated to strengthen the fusion of global context and local texture information. During training

the Generalized Intersection over Union (GIoU) term in the bounding-box regression loss is replaced with the Minimum Pairwise Distance Intersection over Union (MPDIoU) to improve the sensitivity of regression supervision under highly overlapped samples. Experimental results show that SED-DETR achieves mAP@0.5 values of 94.1% and 95.4%

and mAP@0.5:0.95 values of 77.6% and 85.2%

on the self-collected dataset and the public IGV Dataset

respectively

outperforming current mainstream detection models. Edge-side deployment results further demonstrate that the proposed model maintains favorable real-time inference capability on the Jetson Orin Nano 4GB platform; after FP16 acceleration

the inference time is reduced to 23.9 ms while the mAP@0.5 remains 92.5%. These results indicate that SED-DETR achieves a favorable balance between detection accuracy and computational efficiency

providing a practical solution for substation equipment defect detection.

关键词

Keywords

references

于景飞 , 刘俊俊 . 基于动态贝叶斯网络的变电站运维风险评估 [J ] . 科学技术与工程 , 2025 , 25 ( 33 ): 14507 - 14516 .

Yu J F , Liu J J . Risk assessment of substation operation and maintenance based on dynamic Bayesian network [J ] . Science Technology and Engineering , 2025 , 25 ( 33 ): 14507 - 14516 . [维普 ]

张宗包 , 韩茂林 , 王业震 , 等 . 基于模糊生成对抗网络的变电站数字孪生自动化建模方法 [J ] . 电力自动化设备 , 2025 , 45 ( 11 ): 52 - 58 .

Zhang Z B , Han M L , Wang Y Z , et al . Automation modeling method for substation digital twin based on fuzzy generative adversarial networks [J ] . Electric Power Automation Equipment , 2025 , 45 ( 11 ): 52 - 58 . [万方 ]

叶远波 , 王吉文 , 孙振兴 , 等 . 分区协同下适应新一代智能变电站的通信链路故障定位方法 [J ] . 中国电力 , 2025 , 58 ( 9 ): 79 - 87 .

Ye Y B , Wang J W , Sun Z X , et al . Communication link fault localization method adapted to new smart substations under partition collaboration [J ] . Electric Power , 2025 , 58 ( 9 ): 79 - 87 . [维普 ]

强仁 , 龚钢军 , 王大为 , 等 . 基于变电站信息物理耦合特征的扰动建模、韧性评估和提升策略[J/OL]. 中国电机工程学报, 2025: 1-22. Qiang R, Gong G J, Wang D W, et al. Substation Disturbance Modeling, Resilience Evaluation Method and Enhancement Strategy Based on the Characteristics of Cyber Physical Coupling[J ] . Proceedings of the CSEE , 2025 : 1 - 22 . [知网 ]

Yang Q , Ma S , Guo D Q , et al . A small object detection method for oil leakage defects in substations based on improved faster-RCNN [J ] . Sensors , 2023 , 23 ( 17 ): 7390 . [ https://doi.org/10.3390/s23177390 https://doi.org/10.3390/s23177390 ]

张铭泉 , 邢福德 , 刘冬 . 基于改进Faster R-CNN的变电站设备外部缺陷检测 [J ] . 智能系统学报 , 2024 , 19 ( 2 ): 290 - 298 .

Zhang M Q , Xing F D , Liu D . External defect detection of transformer substation equipment based on improved Faster R-CNN [J ] . CAAI Transactions on Intelligent Systems , 2024 , 19 ( 2 ): 290 - 298 . [万方 ]

Xin M Y , Xu C B , Gao J P , et al . High-precision recognition algorithm for equipment defects based on mask R-CNN algorithm framework in power system [J ] . Processes , 2024 , 12 ( 12 ): 2940 . [ https://doi.org/10.3390/pr12122940 https://doi.org/10.3390/pr12122940 ]

Jiang J J , Zhang Y Q , Wan A P , et al . Multi-target detection for safety monitoring in complex substation environments using YOLO-DySE [J ] . Signal, Image and Video Processing , 2025 , 19 ( 10 ): 804 . [ https://doi.org/10.1007/s11760-025-04425-9 https://doi.org/10.1007/s11760-025-04425-9 ]

Wang Q , Yang L X , Zhou B , et al . YOLO-SS-large: a lightweight and high-performance model for defect detection in substations [J ] . Sensors , 2023 , 23 ( 19 ): 8080 . [ https://doi.org/10.3390/s23198080 https://doi.org/10.3390/s23198080 ]

Wang J Q , Sun Y W , Lin Y , et al . Lightweight substation equipment defect detection algorithm for small targets [J ] . Sensors , 2024 , 24 ( 18 ): 5914 . [ https://doi.org/10.3390/s24185914 https://doi.org/10.3390/s24185914 ]

Sun Y W , Sun X R , Lin Y , et al . Substation equipment defect detection based on improved YOLOv8 [J ] . Sensors , 2025 , 25 ( 11 ): 3410 . [ https://doi.org/10.3390/s25113410 https://doi.org/10.3390/s25113410 ]

Yu L , Tang L , Mu L S . A review of detection transformer: From basic architecture to advanced developments and visual perception applications [J ] . Sensors , 2025 , 25 ( 13 ): 3952 . [ https://doi.org/10.3390/s25133952 https://doi.org/10.3390/s25133952 ]

Carion N , Massa F , Synnaeve G , et al . End-to-end object detection with transformers [C ] // Computer Vision – ECCV 2020 . Cham : Springer , 2020 : 213 - 229 . [ https://doi.org/10.1007/978-3-030-58452-8_13 https://doi.org/10.1007/978-3-030-58452-8_13 ]

Zhao Y A , Lv W Y , Xu S L , et al . DETRs beat YOLOs on real-time object detection [C ] // Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Piscataway : IEEE Press , 2024 : 16965 - 16974 . [ https://doi.org/10.1109/CVPR52733.2024.01605 https://doi.org/10.1109/CVPR52733.2024.01605 ]

徐程 , 郑春厚 , 胡祥 , 等 . 基于TLI-DETR的输电线路巡检图像小目标检测方法 [J ] . 智慧电力 , 2025 , 53 ( 9 ): 117 - 125 .

Xu C , Zheng C H , Hu X , et al . A small object detection method for transmission line inspection images based on TLI-DETR [J ] . Smart Power , 2025 , 53 ( 9 ): 117 - 125 . [维普 ]

Fang H , Lin S , Hu J W , et al . CPF-DETR: an end-to-end DETR model for detecting complex patterned fabric defects [J ] . Fibers and Polymers , 2025 , 26 ( 1 ): 369 - 382 . [ https://doi.org/10.1007/s12221-024-00809-9 https://doi.org/10.1007/s12221-024-00809-9 ]

Lan A R , He W , Wang T . Research on substation equipment defect detection technology based on an improved RT-DETR model [C ] // Proceedings of the 2024 6th International Conference on Robotics, Intelligent Control and Artificial Intelligence (RICAI) . Piscataway : IEEE Press , 2024 : 680 - 686 . [ https://doi.org/10.1109/RICAI64321.2024.10911708 https://doi.org/10.1109/RICAI64321.2024.10911708 ]

Xiong C Q , Zayed T , Jiang X Y , et al . A novel model for instance segmentation and quantification of bridge surface cracks: the YOLOv8-AFPN-MPD-IoU [J ] . Sensors , 2024 , 24 ( 13 ): 4288 . [ https://doi.org/10.3390/s24134288 https://doi.org/10.3390/s24134288 ]

Hu M T , Zhuang Y , Wang J H , et al . PBZGNet: a novel defect detection network for substation equipment based on gradual parallel branch architecture [J ] . Sensors , 2026 , 26 ( 1 ): 300 . [ https://doi.org/10.3390/s26010300 https://doi.org/10.3390/s26010300 ]

Du S J , Zhang B F , Zhang P , et al . An improved bounding box regression loss function based on CIOU loss for multi-scale object detection [C ] // Proceedings of the 2021 IEEE 2nd International Conference on Pattern Recognition and Machine Learning (PRML) . Piscataway : IEEE Press , 2021 : 92 - 98 . [ https://doi.org/10.1109/PRML52754.2021.9520717 https://doi.org/10.1109/PRML52754.2021.9520717 ]

Su K K , Cao L H , Zhao B T , et al . N-IoU: better IoU-based bounding box regression loss for object detection [J ] . Neural Computing and Applications , 2024 , 36 ( 6 ): 3049 - 3063 . [ https://doi.org/10.1007/s00521-023-09133-4 https://doi.org/10.1007/s00521-023-09133-4 ]

Liu H , Watanabe H . InterpIoU: Rethinking Bounding Box Regression with Interpolation-Based IoU Optimization [EB/OL ] . arXiv : 2507 . 12420 , 2025-07-16 [ 2026-05-13 ] . [ https://doi.org/10.48550/arXiv.2507.12420 https://doi.org/10.48550/arXiv.2507.12420 ]

Wang H C , Zhang Y G , Wu W H . OHEH-RTDETR: an improved RT-DETR detection model based on frequency layered processing and advanced feature selection [J ] . Measurement Science and Technology , 2025 , 36 ( 4 ): 046211 . [ https://doi.org/10.1088/1361-6501/adc1f3 https://doi.org/10.1088/1361-6501/adc1f3 ]

Xie H , Zhou H B , Chen R L , et al . RAP-DETR: enhancing RT-DETR for railway track defect detection [J ] . Measurement , 2026 , 258 : 119058 . [ https://doi.org/10.1016/j.measurement.2025.119058 https://doi.org/10.1016/j.measurement.2025.119058 ]

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

GhostMamba-SAN: an efficient DDoS attack detection model

Influential nodes recognition of diverse complex network based on deep learning

Self-attention mechanism-based CSI eigenvector feedback for massive MIMO

Optimization method of deep time series prediction model based on meta-heuristic RIME algorithm

DGSF-AOT: dynamic gating and self-attention fusion enhancement for face image restoration

Related Author

Bao Xiaoan

Yang Fenghao

Fan Yunlong

Tu Xiaomei

Hu Tianbin

Wu Biao

MA Yulei

GUO Shasha

Related Institution

School of Computer Science and Technology, Zhejiang Sci-Tech University

School of Science, Zhejiang Sci-Tech University

School of Artificial Intelligence and Automation, Hohai University

School of Urban and Rural Construction, Zhejiang Guangsha Vocational and Technical University of Construction

Department of Computer and Information Engineering, Xinxiang University

⁰