针对光伏组件样本不均衡及缺陷识别精度低问题,提出一种基于LS-DCGAN的GCSE-DenseNet光伏组件缺陷识别方法。首先,针对光伏组件样本的不均衡问题,构建最小二乘深度卷积生成对抗网络（LS-DCGAN）,进行样本数据增强,以扩充数据集。其次,在传统DenseNet网络基础上引入分组卷积和注意力机制,提出一种基于分组卷积和注意力机制的改进GCSE-DenseNet网络模型。改进模型仍采用密集连接机制,实现特征重用防止梯度消失;同时,采用分组卷积优化模型密集模块结构,以降低模型参数量;引入注意力机制加强有效特征、削弱无效特征,以增强模型特征学习能力。最后,对所提模型的有效性进行实验验证。实验表明,改进的GCSE-DenseNet网络模型能有效提高光伏组件缺陷识别精度。
Abstract
Aiming at the problem of unbalanced of photovoltaic module samples and low defect identification accuracy, a defect identification method of GCSE-DenseNet photovoltaic modules based on LS-DCGAN is proposed. Firstly, aiming at the imbalance problem of PV module samples, a least squares deep convolutional generative adversarial network （LS-DCGAN） is constructed to enhance the sample data to expand the dataset. Secondly, an improved GCSE-DenseNet network model is proposed by introducing group convolution and attention mechanism on the basis of the traditional DenseNet network. The improved model still adopts the dense connection mechanism to realize feature reuse and prevent the gradient from disappearing. Simultaneously, group convolution is employed to optimize the dense module structure and reduce the number of model parameters. The attention mechanism is introduced to strengthen effective features and weaken invalid features to enhance model feature learning ability. Finally, the effectiveness of the proposed model is verified through experiments. Experimental results show that the improved GCSE-DenseNet network model can effectively improve the defect recognition accuracy of photovoltaic modules.
关键词
光伏组件 /
深度学习 /
图像分类 /
数据增强 /
分组卷积
Key words
photovoltaic modules /
deep learning /
image classification /
data enhancement /
group convolution
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 国家能源局. 2023年上半年光伏发电建设运行情况[EB/OL]. [2022-04-15]. http://www.nea.gov.cn/2023-07/27/c_1310734298.htm.
National Energy Administration. Photovoltaic power generation construction and operation in the first half of 2023[EB/OL].[2022-04-15]. http://www.nea.gov.cn/2023-07/27/c_1310734298.htm.
[2] 孙荣富, 王隆扬, 王玉林, 等. 基于数字孪生的光伏发电功率超短期预测[J]. 电网技术, 2021, 45(4): 1258-1264.
SUN R F, WANG L Y, WANG Y L, et al.Ultra-short-term prediction of photovoltaic power generation based on digital twins[J]. Power system technology, 2021, 45(4): 1258-1264.
[3] 张雪莉, 刘其辉, 马会萌, 等. 光伏电站输出功率影响因素分析[J]. 电网与清洁能源, 2012, 28(5) : 75-81.
ZHANG X L, LIU Q H, MA H M, et al.Analysis of influencing factors of output power of photovoltaic power plant[J]. Power system and clean energy, 2012, 28(5): 75-81.
[4] 王元章, 李智华, 吴春华. 光伏系统故障诊断方法综述[J]. 电源技术, 2013, 37(9): 1700-1705.
WANG Y Z, LI Z H, WU C H.Fault diagnosis technologies for photovoltaic system[J]. Chinese journal of power sources, 2013, 37(9): 1700-1705.
[5] CHEN H Y, ZHAO H F, HAN D, et al.Accurate and robust crack detection using steerable evidence filtering in electroluminescence images of solar cells[J]. Optics and lasers in engineering, 2019, 118: 22-33.
[6] 蒋琳, 苏建徽, 施永, 等. 基于红外热图像处理的光伏阵列热斑检测方法[J]. 太阳能学报, 2020, 41(8): 180-184.
JIANG L, SU J H, SHI Y, et al.Hot spots detection of operating PV arrays through IR thermal image[J]. Acta energiae solaris sinica, 2020, 41(8): 180-184.
[7] TSAI D M, WU S C, CHIU W Y.Defect detection in solar modules using ICA basis images[J]. IEEE transactions on industrial informatics, 2013, 9(1): 122-131.
[8] 周颖, 毛立, 张燕, 等. 改进CNN的太阳电池缺陷识别方法研究[J]. 太阳能学报, 2020, 41(12): 69-76.
ZHOU Y, MAO L, ZHANG Y, et al.Research on defect detection and classification for solar cells based on improved convolutional neural network[J]. Acta energiae solaris sinica, 2020, 41(12): 69-76.
[9] DEMIRCI M Y, BESLI N, GÜMÜŞÇÜ A. Defective PV cell detection using deep transfer learning and EL imaging[J]. Proceedings book, 2019: 311.
[10] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Honolulu, HI, USA, 2017: 2261-2269.
[11] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al.Generative adversarial nets[C] //Proceedings of the International Conference on Neural Information Processing Systems. Montreal, Quebec, Canada, 2014: 2672-2680.
[12] MAO X D, LI Q, XIE H R, et al.Least squares generative adversarial networks[C]//2017 IEEE International Conference on Computer Vision(ICCV). Venice, Italy, 2017: 2813-2821.
[13] RADFORD A, METZ L, CHINTALA S.Unsupervised representation learning with deep convolutional generative adversarial networks[C]//4th International Conference or Learning Representations, ICLR 2016 Conference Track Proceedings, 2016.
[14] KRIZHEVSKY A, SUTSKEVER I, HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th International Conference on Neural Information Processing Systems-Volume 1. Lake Tahoe, Nevada, 2012: 1097-1105.
[15] HU J, SHEN L, ALBANIE S, et al.Squeeze-and-excitation networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2020, 42(8): 2011-2023.
[16] BUERHOP C, DEITSCH S, MAIER A F, et al.A benchmark for visual identification of defective solar cells in electroluminescence imagery[C]//35th European Photovolataic Solar Energy Conference and Exhibition. Brussels, Belyium, 2018.
[17] KÖNTGES M, KURTZ S, PACKARD C, et al. IEA-PVPS T13-012014 review of failures of photovoltaic modules final[R]. IEA-PVPs TB-01, 2014.
[18] BATISTA G E A P A, PRATI R C, MONARD M C. A study of the behavior of several methods for balancing machine learning training data[J]. ACM SIGKDD explorations newsletter, 2004, 6(1): 20-29.
[19] CHAWLA N V, BOWYER K W, HALL L O, et al.SMOTE: synthetic minority over-sampling technique[J]. Journal of artificial intelligence research, 2002, 16: 321-357.
基金
国家自然科学基金（51807063）; 中央高校基本科研业务费专项资金（2021MS065）
PDF(1972 KB)
