摘要
为修复全天空成像仪拍摄的地基云图中的大面积遮挡区域,提出了一种基于局部卷积神经网络的地基云图修复方法。通过在传统卷积神经网络的基础上增加网络结果整体跨层传输结构,连接对需修复图像的编码和解码两部分,实现了以含遮挡图像的输入到修复完图像的输出全过程。采用美国国家新能源实验室网上公开数据,从定性和定量的角度分析不同天气情况下所提出修复方法的性能,实验结果表明,基于局部卷积网络的地基云图修复方法可以较为真实地还原天空情况,取得优于现有方法的修复效果,为光伏电站输出功率预测提供了重要的数据。
In order to restore the large-area occlusion in the ground-based cloud image taken by a total sky imager,a inpainting method was proposed based on partial convolutional network.The cross-layer transmission of the whole output of a convolutional layer was added to connect the process of the coding and decoding of a ground-based cloud image.Therefore,it was realized that the occluded image was set as input and the repaired image was set as output directly.The open dataset of the National Renewable Energy Library in the USA was used to evaluate qualitatively and quantitatively the performance of the proposed method under different sky conditions.Experimental results show that the proposed method can do well in restoring the sky condition and performed better than other published works.The inpainting ground-based cloud image can provide important data for predicting the output power of photovoltaic power station.
作者
朱婷婷
过奕任
李元哲
ZHU Tingting;GUO Yiren;LI Yuanzhe(College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China;Key Laboratory of Measurement and Control of Complex Systems of Engineering of Ministry of Education, Southeast University, Nanjing 210096, China)
出处
《中国科技论文》
CAS
北大核心
2022年第3期269-273,共5页
China Sciencepaper
基金
国家自然科学基金资助项目(62006120)
东南大学复杂工程系统测量与控制教育部重点实验室开放课题基金资助项目(MCCSE2020A02)。
关键词
模式识别
图像处理
局部卷积网络
地基云图
pattern recognition
image processing
partial convolutional network
ground-based cloud image
