Defect detection method of PV panels based on multi-scale fusion and improved YOLOv8n

doi:10.3969/j.issn.2097-0706.2024.11.004

Abstract

Abstract:

Photovoltaic power generation plays a critical role in modern energy systems， and its stable operation is essential for ensuring energy supply. However， photovoltaic panels are exposed to complex environmental factors， such as ultraviolet radiation， corrosion， moisture， which can lead to defects like cracks and broken grids， thereby potentially compromising their functionality. To address the problems of high model complexity and low detection accuracy for small defects in current photovoltaic panel defect detection algorithms， a defect detection algorithm based on improved YOLOv8n was proposed. A new backbone network was designed to reduce the model's parameters and computation； The C2F-Efficient Local Attention （C2F-ELA） module was introduced to enhance the model's capability to precisely localize subtle features. Weight-Bidirectional Feature Pyramid Network （W-BiFPN） was then proposed to replace the original network structure and integrate the P2 small target detection layer. This effectively enhanced the model's ability to capture multi-scale features while leveraging shallow network information to strengthen the local feature perception， boosting the accuracy of small target recognition. Experimental results showed that， compared to the baseline YOLOv8 algorithm， this method reduced the number of parameters by 16.7% and increased the mean average precision by 3.1 percentage points， demonstrating an improvement in detection performance.

Key words: defect detection, deep learning, YOLOv8n, lightweight module, ELA, W-BiFPN

CLC Number:

TK01：TP391.41

ZHANG Wenqiang, LI Jiashu, XUAN Yang, LI Chen, QIAN Hang, ZHANG Xiaoyu. Defect detection method of PV panels based on multi-scale fusion and improved YOLOv8n[J]. Integrated Intelligent Energy, 2024, 46(11): 29-37.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Table 1

Fig.8

Table 2

Table 3

Fig.9

References 26

[1]	舒印彪, 赵勇, 赵良, 等. “双碳”目标下我国能源电力低碳转型路径[J]. 中国电机工程学报, 2023, 43(5):1663-1672.
	SHU Yinbiao, ZHAO Yong, ZHAO Liang, et al. Study on low carbon energy transition path toward carbon peak and carbon neutrality[J]. Proceedings of the CSEE, 2023, 43(5):1663-1672.
[2]	国家能源局. 2024年上半年光伏发电建设情况[J]. 电力科技与环保, 2021, 31(4):46.
	National Energy Administration. Photovoltaic power generation construction in the first half of 2024[J]. Electric Power Technology and Environmental Protection, 2021, 31(4):46.
[3]	AKRAM M W, LI G, JIN Y, et al. Failures of photovoltaic modules and their detection:A review[J]. Applied Energy, 2022, 313:118822.
[4]	KANDEAL A W, ELKADEEM M R, THAKUR A K, et al. Infrared thermography-based condition monitoring of solar photovoltaic systems:A mini review of recent advances[J]. Solar Energy, 2021, 223:33-43.
[5]	ET-TALEBY A, CHAIBI Y, BOUSSETTA M, et al. A novel fault detection technique for PV systems based on the K-means algorithm,coded wireless orthogonal frequency division multiplexing and thermal image processing techniques[J]. Solar Energy, 2022, 237:365-376.
[6]	ALI M U, KHAN H F, MASUD M, et al. A machine learning framework to identify the hotspot in photovoltaic module using infrared thermography[J]. Solar Energy, 2020, 208: 643-651.
[7]	SU B Y, CHEN H Y, CHEN P, et al. Deep learning-based solar-cell manufacturing defect detection with complementary attention network[J]. IEEE Transactions on Industrial informatics, 2020, 17(6):4084-4095.
[8]	HERRAIZ Á H MARUGÁN A P, MÁRQUEZ F P G. Photovoltaic plant condition monitoring using thermal images analysis by convolutional neural network-based structure[J]. Renewable Energy, 2020, 153:334-348.
[9]	季瑞瑞, 梅远, 杨思凡, 等. 基于改进Faster R-CNN的光伏组件红外热斑检测算法[J]. 激光与红外, 2024, 54(4):584-592.
	JI Ruirui, MEI Yuan, YANG Sifan, et al. Infrared hot spot detection in photovoltaic modules based on improved Faster R-CNN[J]. Laser & Infrared, 2024, 54(4):584-592.
[10]	WANG J, ZHANG R Z, ZHENG X. Photovoltaic panel intelligent detection method based on improved Faster-RCNN[C]// 2023 IEEE 3rd International Conference on Electronic Technology,Communication and Information (ICETCI).IEEE, 2023:1565-1569.
[11]	TRIPATHY S, SATPATHY M. SSD internal cache management policies:A survey[J]. Journal of Systems Architecture, 2022, 122:102334.
[12]	HUSSAIN M. YOLOv1 to v8:Unveiling each varianta comprehensive review of YOLO[J]. IEEE Access, 2024, 12:42816-42833.
[13]	XU M. Solar cell defect detection based on improved G-SSD network[J]. International Journal of Energy, 2023, 2(1):68-71.
[14]	艾上美, 周剑峰, 张必朝, 等. 基于改进SSD算法的光伏组件缺陷检测研究[J]. 智慧电力, 2023, 51(12):53-58.
	AI Shangme, ZHOU Jianfeng, ZHANG Bicha, et al. Defect detection of photovoltaic modules based on improved SSD algorithm[J]. Smart Power, 2023, 51(12):53-58.
[15]	DITOMMASO A, BETTI A, FONTANELLI G, et al. A multi-stage model based on YOLOv3 for defect detection in PV panels based on IR and visible imaging by unmanned aerial vehicle[J]. Renewable energy, 2022, 193:941-962.
[16]	LI L L, WANG Z F, ZHANG T T. GBH-YOLOv5:Ghost convolution with bottleneck and tiny target prediction head incorporating YOLOv5 for PV panel defect detection[J]. Electronics, 2023, 12(3):561.
[17]	ZHANG J L, YANG W, CHEN Y, et al. Fast object detection of anomaly photovoltaic (PV) cells using deep neural networks[J]. Applied Energy, 2024, 372:123759.
[18]	CAO Y K, PANG D D, ZHAO Q C, et al. Improved YOLOv8-GD deep learning model for defect detection in electroluminescence images of solar photovoltaic modules[J]. Engineering Applications of Artificial Intelligence, 2024, 131:107866.
[19]	HANG X Y, ZHU X B, GAO X X, et al. Study on crack monitoring method of wind turbine blade based on AI model:Integration of classification,detection,segmentation and fault level evaluation[J]. Renewable Energy, 2024, 224: 120152.
[20]	HAN K, WANG Y H, TIAN Q, et al. Ghostnet:More features from cheap operations[C]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020:1580-1589.
[21]	XU W, WAN Y. ELA:Efficient local attention for deep convolutional neural networks[J]. Arxiv Preprint Arxiv, 2024:2403.01123.
[22]	WANG H, LI D, ISSHIKI T. Energy-efficient implementation of YOLOv8,instance segmentation,and pose detection on RISC-V SoC[J]. IEEE Access, 2024, 12:64050-64068.
[23]	TAN M M, PANG R M, LE Q V. Efficientdet:Scalable and efficient object detection[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020:10781-10790.
[24]	任一鸣, 杜董生, 邓祥帅, 等. 基于Real-ESRGAN和改进YOLOv8n的输电线路绝缘子故障检测[J]. 综合智慧能源, 2024, 46(7):29-39. doi: 10.3969/j.issn.2097-0706.2024.07.004
	REN Yiming, DU Dongsheng, DENG Xiangshuai, et al. Optimal scheduling of virtual power plants integrating electric vehicles based on reinforcement learning[J]. Integrated Intelligent Energy, 2024, 46(7):29-39. doi: 10.3969/j.issn.2097-0706.2024.07.004
[25]	郭岚, 刘正新. 基于改进YOLOv5的光伏组件缺陷检测[J]. 激光与光电子学进展, 2023, 60(20):148-156.
	GUO Lan, LIU Zhengxin. Improved YOLOv5-based defect detection in photovoltaic modules[J]. Laser &Optoelectronics Progress, 2023, 60(20):148-156.
[26]	田浩, 周强, 贺晨龙. 基于多尺度特征融合的光伏组件缺陷检测[J]. 计算机工程与应用, 2024, 60(3):340-347. doi: 10.3778/j.issn.1002-8331.2304-0390
	TIAN Hao,ZHOU Qiang,HE Chenlong. Defect detection of photovoltaic modules based on multi-scale feature fusion[J]. Computer Engineering and Applications, 2024, 60(3):340-347. doi: 10.3778/j.issn.1002-8331.2304-0390

方法	参数量/10⁶	准确率/%	召回率/%	F₂值	mAP_0.5/%
YOLOv8n	3.0	88.4	81.8	0.850	88.8
GhostConv	2.8	82.0	83.0	0.825	88.1
C2f-ELA	3.0	86.0	85.6	0.858	89.1
W-BiFPN	2.7	89.5	84.2	0.868	89.8
GhostConv+C2f-ELA	2.8	85.7	82.8	0.842	90.1
本文	2.5	93.0	83.8	0.882	91.9

模型	F₂值	mAP_0.5/%
C2f-ELA_A	0.838	88.9
C2f-ELA_N	0.836	84.4
C2f-ELA_B	0.858	89.1

模型	参数量/10⁶	计算量/GFLOPs	准确率/%	召回率/%	F₂值	mAP_0.5/%
文献[25]	4.2	18.1	85.9	82.3	0.841	88.1
文献[26]	13.3	31.6	88.6	86.1	0.873	88.4
Faster RCNN	137.1	370.2	39.3	88.3	0.544	65.2
YOLOv3	61.5	155.3	85.6	88.6	0.871	89.7
YOLOv5m	21.1	25.2	86.4	87.2	0.868	89.3
YOLOv7	37.2	150.1	87.8	79.1	0.832	85.7
YOLOv9m	32.6	130.7	85.1	89.2	0.871	90.2
YOLOv10m	16.5	64.0	84.9	85.0	0.849	88.6
本文	2.5	11.7	93.0	83.8	0.882	91.9