变电站作为电网的重要组成部分，其中电气设备的实时状态监测对保障电网安全稳定运行和降低国民经济损失起着至关重要的作用。随着智能电网的发展以及故障检测方法的不断进步，红外技术凭借无接触测温、操作安全便捷等优点被广泛应用于变电站的日常设备运维中。但是现存的设备故障诊断方法对人工依赖性较大，缺乏高效处理海量数据的能力。为进一步提高变电站设备故障诊断的智能化和高效化，本文提出了一种基于电气设备红外图像的变电设备故障诊断方法。

       本文首先研究了红外图像增强方法。为了突出待诊断设备细节，便于后续学习目标特征，提出了一种基于灰狼寻优的自适应多尺度图像增强算法。将图像分解为高频和低频两部分，并根据图像特点分别来提高低频图像对比度，多尺度增强高频图像细节。在图像重构时，通过灰狼算法自适应地寻找最佳融合系数，获得最佳增强效果。实验结果表明，与其他经典算法相比较，该方法能够将峰值信噪比、结构相似性、梯度相似性以及空间频率四个客观评价指标平均提高6.208%、9.483%、0.160%和11.294%，在减少图像噪声含量的同时保留丰富边缘细节信息。

      针对传统的变电设备识别方法容易受到环境干扰，且工作效率较低的问题，本文设计了一种基于轻量结构与和注意力机制的改进YOLOv5模型。利用轻量的卷积结构来替换主干网络中的普通卷积，精简模型结构；同时，增添通道注意力模块来优化目标特征表达，增强目标显著度；然后，利用自注意力机制来改进C3模块，缓解复杂变电站背景下的设备遮挡问题；再通过优化损失函数来提高预测框的定位精确度和模型收敛速度；最终模型能够将识别精度提升至97%，同时有效缓解目标遮挡现象。相较于典型的深度学习模型，本文算法能够将识别速度平均提高75.417%。

       通过对图像中的设备进行分类识别，能够有效排除无关信息干扰，为后续设备疑似故障区域的提取以及对不同类型设备进行故障分析提供良好条件。在此基础上，针对传统分割算法精度不足的问题，本文研究了一种基于颜色信息与改进模糊聚类算法的疑似故障区域提取方法。首先，通过红外图像的颜色特点，实现基于超像素分割的预分割准备。再利用空间信息和邻域像素重新构造模糊C均值聚类算法的目标函数，并引入自适应参数来增强算法鲁棒性和分割完整度。最终，使用全局阈值分割来提取出温度异常的区域，为后续获取设备不同区域的温度信息做好准备工作。由实验数据可知，该方法在保证区域提取完整度的同时有效提升分割速度，增加预分割处理和自适应操作能够分别将算法运行时间减少9.099%和48.368%。

       基于上述研究结果，本文采用相对温差法来确定设备故障与否，并结合相关的诊断依据来判断热故障严重程度。实验结果证明，本文所提出的热故障诊断方法，能够有效提高变电站设备的自动化、智能化诊断水平。并根据设备运行状态，及时制定检修计划，为变电站的安全稳定运行奠定基础。

As an important part of the power grid, the real-time status monitoring of electrical equipment plays a vital role which ensures the safe and stable operation of the power grid and reduces the loss of the national economy. With the development of smart grid and the continuous progress of fault detection methods, infrared technology is widely used in daily operation and maintenance of substation equipment with the advantages of non-contact temperature measurement, safe and convenient operation. However, the current equipment fault diagnosis methods rely on manual work heavily and lack the ability to process massive data efficiently. The method based on infrared image of electrical equipment is proposed to further improve the intelligence and efficiency of substation equipment fault diagnosis.

The infrared image enhancement method is studied. In order to highlight the details of the equipment and facilitate the subsequent learning of target features, an adaptive multi-scale image enhancement algorithm based on grey wolf optimization is proposed. The image is decomposed into high frequency and low frequency parts. According to the characteristics of the image, the low-frequency image contrast is improved, and the high-frequency image details are enhanced by multi-scale. In image reconstruction, the grey wolf algorithm is used to adaptively find the best fusion coefficient, so as to obtain the best enhancement effect. The experimental results show that compared with other classical algorithms, this method can improve the four objective evaluation indexes of peak signal-to-noise ratio, structural similarity, gradient similarity and spatial frequency by 6.208 %, 9.483 %, 0.160 % and 11.294 % on average. The rich edge detail information is retained while noise content is reduced.

Due to environmental interference and low work efficiency by traditional recognition methods of substation equipment, this paper designs an improved YOLOv5 model based on lightweight structure and attention mechanism. The lightweight convolution structure is used to replace the ordinary convolution in the backbone network to simplify the model structure. At the same time, the channel attention module is introduced to optimize the target feature expression and enhance the target salience. Then, the self-attention mechanism is utilized to improve the C3(Concentrated-Comprehensive Convolution Module) module to alleviate the equipment occlusion in the complex substation background. Then, the loss function is optimized to improve the positioning accuracy of the prediction box and the convergence speed of the model. The final model can improve the recognition accuracy to 97 %, and effectively alleviate the target occlusion. Compared with typical deep learning models, the recognition speed is increased by 75.417 % on average.

The interference of irrelevant information can be eliminated effectively by classifying and identifying the equipment in the image. It provides good conditions for subsequent equipment suspected fault area extraction and fault analysis of different equipment. On this basis, aiming at the problem of insufficient accuracy with traditional segmentation algorithm, a method based on color information and improved fuzzy clustering algorithm is proposed for the suspected fault area extraction. Firstly, the pre-segmentation based on superpixel segmentation is realized by the color characteristics of infrared images. The objective function of fuzzy c-means clustering is reconstructed by using spatial information and neighborhood pixels, and adaptive parameters are introduced to enhance the robustness and segmentation integrity of the algorithm. Finally, the global threshold segmentation is adopted to extract the region of abnormal temperature, which is ready for the subsequent acquisition of temperature information in different regions of the equipment. The experimental results show that the method improves the segmentation speed while the integrity of the region extraction is ensured. With the operation of pre-segmentation processing and adaptive operation, the running time of the algorithm can be reduced by 9.099 % and 48.368 %, respectively.

Based on the above research results, the relative temperature difference method is adopted in this paper to determine whether the equipment fails or not, and combines the related diagnostic criteria to judge the severity of the thermal fault. The experimental results show that the proposed method can effectively improve the automation and intellect diagnosis level of substation equipment. According to the operation status of the equipment, the maintenance plan is formulated in time, which lays a foundation for the safe and stable operation of the substation.

[1]李小芳，彭敏放，罗翊君，等．变电站接地网的信息融合故障诊断[J]．电力系统及其自动化学报，2015，27(12)：36-41．

[2]Han S, Yang F, Yang G, et al. Electrical equipment identification in infrared images based on ROI-selected CNN method[J]. Electric Power Systems Research, 2020, 188: 106534.

[3]邹辉，黄福珍． 基于FAsT-Match算法的电力设备红外图像多目标定位[J]．中国电机工程学报，2017，37(02)：591-599．

[4]冯振新，许晓路，周东国，等．基于局部区域聚类的电力设备故障区域提取方法[J]．电测与仪表，2020，57(08)：45-50．

[5]陈达，何全才，迪二镇，等．自适应权重的偏微分分割模型在变电设备红外图像中的应用[J]．红外技术，2022，44(02)：179-188．

[6]Han S, Yang F, Yang G, et al. Electrical equipment identification in infrared images based on ROI-selected CNN method[J]. Electric Power Systems Research, 2020, 188(2): 106534.

[7]Gao Z W, CECATI C, DING S X. A survey of fault diagnosis and fault-tolerant techniques—part Ⅰ: fault diagnosis with model-based and signal-based approaches[J]. IEEE Transactions on Industrial Electronics , 2015, 62( 6): 3757-3767.

[8]Gou B, Ge X L, Wang S L, et al. An open-switch fault diagnosis method for single-phase pwm rectifier using a model-based approach in high-speed railway electrical traction drive system [J]. IEEE Transactions on Power Electronics, 2016, 31 (5): 3816 -3826.

[9]陈浈斐，章黄勇，马宏忠，等．基于深度学习的电力设备故障诊断方法研究综述[J]． 电气自动化，2022，44(01)：1-2+6．

[10]李鑫，崔昊杨，许永鹏，等．电力设备IR图像特征提取及故障诊断方法研究[J]．激光与红外，2018，48(05)：659-664．

[11]Dutta T, Slilc J, Chottopadhyay P. Condition monitoring of electrical equipment using thermal image processing[C]//2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI). Kolkata, India: IEEE, 2016: 311-315.

[12]胡凡奎．基于红外图像的变电站设备分类及热故障诊断[D]．黑龙江大学，2020．

[13]Zhang W B, Liu W J. IFCM: Fuzzy clustering for rule extraction of interval Type-2 fuzzy logic system[C]//IEEE Conference on Decision & Control. New Orleans, LA: IEEE, 2008: 5318-5322.

[14]曾军，王东杰，范伟，等．基于红外热成像的电气设备组件识别研究[J]．红外技术，2021，43(07)：679-687．

[15]张国治，陈康，方荣行，等．基于DGA与鲸鱼算法优化Logit Boost-决策树的变压器故障诊断方法[J]．电力系统保护与控制，2023，51(07)：63-72．

[16]Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25(2): 1097-1105.

[17]Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceeding of the IEEE Conference on Computer Vision ang Pattern Recognition. Boston, MA, USA: IEEE, 2015: 1-9.

[18]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[J]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 770-778.

[19]Liu Wei, Anguelov D, Erhan D, et al. SSD: single shot multibox detector[C]// Proceedings of the 14th European Conference on Computer Vision. Amsterdam: Springer, 2016: 21-37.

[20]Redmon J, DivvalaI S, Girshick R, et al. You only look once: unified, real-time object detection [C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 779-788．

[21]Redmon J, Farhadi A, YOLO9000: better, faster, stronger[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 6517-6525.

[22]Redmon J, Farhadi A. Yolov3: An incremental improvement[J]. arXiv e-prints, 2018.

[23]Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE, 2014: 580-587．

[24]Girshick R. Fast R-CNN[C]//Proceedings of 2015 IEEE International Conference on Computer Vision. Santiago: IEEE, 2015: 1440-1448.

[25]Ren S, He K, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]// Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal: NIPS, 2015: 91-99．

[26]朱惠玲，牛哲文，黄克灿，等．基于单阶段目标检测算法的变电设备红外图像目标识别及定位[J]．电力自动化设备，2021，41(08)：217-224．

[27]Arthur D, Vassilvitskii S. K-Means++: The Advantages of Careful Seeding[C]// Proceedings of the Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2007, New Orleans, Louisiana, USA, ACM, 2007.

[28]曹恩宇，王旭红．基于深度学习的变电设备红外热像识别[J]．电力学报， 2022，37(03)：263-271．

[29]闫建平，张辉疆，李刚，等．基于高精度图像识别技术的输变电设备巡检方法研究[J]．电子设计工程，2023，31(01)：124-128．

[30]李斌，屈璐瑶，朱新山，等．基于多尺度特征融合的绝缘子缺陷检测[J]．电工技术学报，2023，38(01)：60-70．

[31]李文璞，毛颖科，廖逍，等．基于旋转目标检测的变电设备红外图像电压致热型缺陷智能诊断方法[J]．高电压技术，2021，47(09)：3246-3253．

[32]Yang X, Liu Q, Yan J, et al. R3det: refined single-stage detector with feature refinement for rotating object[J]. arXiv preprint arXiv: 1908.05612, 2019.

[33]Dai J, Qi H, Xiong Y, et al. Deformable convolutional networks[C]∥Proceedings of the IEEE international conference on computer vision. Venice, Italy: IEEE, 2017: 764-773.

[34]Zhang P, Zhang Y, Cao H, et al. Outlier-Factor-Based Cluster Analysis for Infrared Image Segmentation of Electric Equipment[C]∥第30届中国控制与决策会议论文集. Shenyang, China: IEEE, 2018: 932-937.

[35]余成波，曾亮，张林．基于OTSU和区域生长的电气设备多点故障分割[J]．红外技术，2018，40(10)：008-1012．

[36]Qi C, Li Q, Liu Y, et al. Infrared image segmentation based on multi-information fused fuzzy clustering method for electrical equipment[J]. International Journal of Advanced Robotic Systems, 2020, 17(2): 728-732.

[37]刘赫，赵天成，刘俊博，等．基于深度残差UNet网络的电气设备红外图像分割方法[J]．红外技术，2022，44(12)：1351-1357．

[38]张秋铭，李云红，罗雪敏，等．改进Chan-Vese模型的电力设备红外图像分割算法[J]．红外技术，2023，45(02)：129-136．

[39]He K, Sun J, Tang X. Guided Image Filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013(6): 1397-1409.

[40]Ming Z, Li Y, Meng Q, et al. Improving histogram-based image contrast enhancement using gray-level information histogram with application to X-ray images[J]. Optik-International Journal for Light and Electron Optics, 2012, 123(6): 511-520.

[41]Zuiderveld K. Contrast Limited Adaptive Histogram Equalization[J]. Graphics Gems, 1994: 474-485.

[42]Ming Z, Li Y, Meng Q, et al. Improving histogram-based image contrast enhancement using gray-level information histogram with application to X-ray images[J]. Optik-International Journal for Light and Electron Optics, 2012, 123(6): 511-520.

[43]Mirjalili S, Mirjalili S M, Lewis A. Grey Wolf Optimizer[J]. Advances in Engineering Software,2014, 69(3): 46-61.

[44]樊勇，刘学雨，钱方．使用梯度相似度的激光干扰图像评估[J]．电光与控制，2014，21(08)：20-23．

[45]Han K, Wang Y, Tian Q, et al. GhostNet: More Features From Cheap Operations[C]//Conference on Computer Vision and Pattern Recognition (CVPR). Seattle: IEEE, 2020: 1580-1589.

[46]Chollet F. Xception: Deep Learning with Depthwise Separable Convolutions[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA: IEEE, 2017.

[47]Wang Q, Wu B, Zhu P, et al. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA: IEEE, 2020.

[48]Ashish V, Noam S, Niki P, et al. Attention is all you need [C]//In Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS'17). Curran Associates Inc, Red Hook, NY, USA, 2017: 6000–6010.

[49]Zheng Z, Wang P, Ren D, et al. Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation[J]. IEEE Transactions on Cybernetics，2020．

[50]Zhan Y F, Ren W Q, Zhang Z, et al. Focal and Efficient IOU Loss for Accurate Bounding Box Regression[J]. Neurocomputing, 2022, 506(9): 4321-4330.

[51]Ren X, Malik J. Learning a classification model for segmentation[C]// Proceedings. Proceedings Ninth IEEE International Conference on Computer Vision. Nice, France: IEEE, 2003: 10-17.

[52]Achanta R, Shaji A, Smith K, et al. SLIC Superpixels Compared to State-of-the-Art Superpixel Methods[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(11), 2274–2282.

[53] Krinidis S and Chatzis V. A robust fuzzy local information C-Means clustering algorithm[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2010, 19(5): 1328–1337.

[54]Hua Z, Wang Q, Shi W, et al. A Novel Adaptive Fuzzy Local Information C-Means Clustering Algorithm for Remotely Sensed Imagery Classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(9): 5057-5068.

[55]赵庆生，王雨滢，王旭平，等．基于新型阈值选择方法的变电站红外图像分割[J]．光学学报，2019，39(08)：109-116．

[56]Hui Z, Huang F. An intelligent fault diagnosis method for electrical equipment using infrared images[C]//Control Conference. Hangzhou, China: IEEE, 2015: 6372-6376.

[57]中华人民共和国电力行业标准．带电设备红外诊断应用规范(DL/T 664-2016)[S]．北京：中国电力出版社，2016．