摘要
光伏热斑故障对光伏组件的运行会产生严重影响,为从图像数据中进行有效的热斑检测,提出一种基于密集连接网络(DenseNet)的深度学习方法。利用数据增强、改进模型结构和迁移学习的方法,在红外光伏故障图形数据集上训练优化,并针对构建的样本数据集具有分布不平衡性的特点,选择采用Focal损失函数缓解样本的非均衡。实验结果表明,该模型网络训练构建的光伏组件红外图像热斑状态数据集,能够实现较高准确度的图像识别,与原始DenseNet模型相比,能够提升准确率。
Photovoltaic hot spot fault has a serious impact on the operation of photovoltaic modules.In order to effectively detect hot spots from image data,a deep learning method based on DenseNet network was proposed.By using the methods of data enhancement,model structure improvement and migration learning,the infrared photovoltaic fault graph data set was trained and optimized.Considering the constructed sample data set has the characteristics of unbalanced distribution,the focal loss function was selected to alleviate the sample imbalance.The experimental results show that the hot spot state data in infrared image of photovoltaic module formed by network training can achieve high accuracy image recognition and improve the accuracy compared with the original DenseNet model.
作者
贾帅康
白英君
孙海蓉
曹瑶佳
JIA Shuaikang;BAI YingJun;SUN Hairong;CAO YaoJia(School of Control Computer Engineering,North China Electric Power University,Baoding 071003,China;Hebei Technology Innovation Center of Simulation&Optimized Control for Power Generation,North China Electric Power University,Baoding 071003,China)
出处
《山东电力技术》
2021年第3期60-64,共5页
Shandong Electric Power
