Journal of Frontiers of Computer Science and Technology ›› 2022, Vol. 16 ›› Issue (2): 458-467.DOI: 10.3778/j.issn.1673-9418.2111036

• Graphics and Image • Previous Articles     Next Articles

Micro-cracks Detection of Solar Cells Based on Few Shot Samples with Multi-loss

NA Zhixiong1,+(), FAN Tao2, SUN Tao1, XIE Xiangying3,1, LAI Guangzhi1   

  1. 1. State Grid Electronic Commerce Co., Ltd., Beijing 100053, China
    2. State Grid Corporation of China, Beijing 100031, China
    3. Beihang University, Beijing 100191, China
  • Received:2021-11-05 Revised:2022-01-12 Online:2022-02-01 Published:2022-01-17
  • About author:NA Zhixiong, born in 1974, M.S., engineer. His research interests include renewable energy, new power system, etc.
    FAN Tao, born in 1971, M.S., professor of engineering. His research interests include enterprise informatization, power system automation, energy Internet, etc.
    SUN Tao, born in 1968, senior engineer. His research interests include power system automation, energy industry Internet, etc.
    XIE Xiangying, born in 1979, Ph.D. candidate, senior engineer. His research interest is industry Internet.
    LAI Guangzhi, born in 1969, M.S., professor of economist. His research interests include new energy development and consumption management, etc.
  • Supported by:
    National Key Research and Development Program of China(2018YFB1500800);Technology Project of State Grid Corporation of China(SGTJDK00DYJS2000148)

多损失融合的小样本光伏组件隐裂检测算法

那峙雄1,+(), 樊涛2, 孙涛1, 谢祥颖3,1, 来广志1   

  1. 1.国网电子商务有限公司,北京 100053
    2.国家电网有限公司,北京 100031
    3.北京航空航天大学,北京 100191
  • 通讯作者: + E-mail: hellonazx@163.com
  • 作者简介:那峙雄(1974—),男,北京人,硕士,工程师,主要研究方向为可再生能源与新型电力系统研究应用等。
    樊涛(1971—),男,山东人,硕士,教授级高级工程师,主要研究方向为企业信息化、电力系统自动化、能源互联网等。
    孙涛(1968—),男,新疆人,高级工程师,主要研究方向为电力系统自动化、能源工业互联网等。
    谢祥颖(1979—),男,湖北人,博士研究生,高级工程师,主要研究方向为工业互联网。
    来广志(1969—), 男, 陕西人, 硕士, 教授级高级经济师, 主要研究方向为新能源发展与消纳管理等。
  • 基金资助:
    国家重点研发计划(2018YFB1500800);国家电网有限公司科技项目(SGTJDK00DYJS2000148)

Abstract:

Aiming at the problem of micro-cracks detection of photovoltaic modules in industrial production line, in order to reduce labor cost, improve detection efficiency and quickly adapt to the micro-cracks detection of new products with the support of a few number of samples, a micro-cracks detection algorithm of solar cells based on few shot samples with multi-loss is proposed. Firstly, in order to enrich the semantic information extracted by convolutional neural network, Transformer’s multi-head attention mechanism is introduced to alleviate the impact of the distribution difference of each batch of products on crack detection, and promote the model to focus on crack information from diversified products. Secondly, the strategy of combining multiple loss functions to constrain the model training is used to optimize feature extraction. On the basis of direct classification loss, the triplet loss is used to shorten the feature distance between cracked samples. In addition, the implicit classification loss is designed to adapt to the characteristics of type differences between the two types of cells with or without cracks, and fully learn the diversity of historical component data. This algorithm can use a small number of samples to quickly extract the features of new components and detect micro-crack defects of new products accurately. The experimental results on the actual industrial production data sets show that the recall of the algorithm can be improved by 10 percentage points compared with other baseline models. This algorithm can effectively alleviate the problem of scarce samples with hidden cracks and greatly reduce the cost of frequent data labeling and model training for each batch of new products.

Key words: solar cells, micro-cracks detection, few shot samples, deep learning, feature extraction

摘要:

针对工业生产线光伏组件隐性纹检测问题,为了降低人力成本,提高检测效率,并快速适应新型产品的隐裂检测,提出了一种多损失融合的小样本光伏组件隐裂检测算法。首先,为丰富卷积神经网络提取的语义信息,引入了Transformer的多头注意力机制,缓解各批次产品的分布差异对隐裂检测的影响,促使模型从多样化产品中关注于隐裂信息;其次,利用多损失结合约束模型训练的策略优化特征提取,在直接分类损失的基础上,利用三元组损失拉近含隐裂样本间特征距离;此外,设计了隐式分类损失以适应有无隐裂两类电池片内部也存在类型差异的特点,充分学习历史组件数据的多样性。该算法能够快速提取新型组件特征,利用少量的样本特征对新产品隐裂缺陷进行准确检测。在实际工业生产数据集上的实验结果表明,该算法对新型组件的隐裂检测的召回率相较于其他基线模型可提高10个百分点,能够有效缓解含隐裂样本数量不足的问题,极大地降低了频繁对每批新产品进行数据标记和训练的开销。

关键词: 光伏组件, 隐裂检测, 小样本, 深度学习, 特征提取

CLC Number: