基于深度学习的武汉市屋顶光伏地理潜力评估

doi:10.13474/j.cnki.11-2246.2026.0725

摘要/Abstract

摘要： [目的] 本文旨在精准评估武汉市屋顶光伏装机潜力以支撑城市分布式光伏规划。[方法] 本文融合武汉市手工标注数据集与公开数据集训练模型,对比多种深度学习语义分割模型以确定最优屋顶识别方案,结合城市功能区规划完成屋顶分类,并结合可利用系数计算光伏装机容量。[结果] 结果表明,最优模型屋顶分割精度达 96%以上,识别了武汉市5类屋顶资源,全市屋顶面积为522 km²,屋顶光伏装机容量38 554.64 MW,呈现“核心区密度高、远城区总量大”的格局。[结论] 本文研究通过适配武汉本地的数据集构建、模型筛选及分类潜力量化,实现了武汉市屋顶光伏地理潜力的精准量化,为武汉市分布式光伏的布局与开发提供了科学依据。

关键词: 屋顶提取, 深度学习, 高分辨率影像, 光伏潜力评估, 遥感影像识别

Abstract: [Purposes] This study aims to accurately assess the rooftop photovoltaic (PV)installation potential in Wuhan and support urban distributed PV planning. [Methods] This study integrates Wuhan's manually labeled dataset and public datasets for model training,systematically compares multiple deep learning semantic segmentation models to determine the optimal solution,completes rooftop classification based on urban functional zone planning,and calculates the installed capacity with availability coefficients. [Findings] The results show that the optimal model achieves a rooftop segmentation accuracy of over 96%,identifies five types of rooftop resources in Wuhan.The city's total rooftop area is 522 km²,and the rooftop PV installed capacity reaches 38 554.64 MW,presenting a pattern of “high density in core areas and large total capacity in outer suburbs”. [Conclusions] This study realizes the accurate quantification of Wuhan's rooftop PV geographic potential through the construction of a dataset adapted to local conditions in Wuhan,model selection,and quantitative evaluation of classified potential,providing a scientific basis for the layout and development of distributed PV in Wuhan.

Key words: roof extraction, deep learning, high-resolution imagery, photovoltaic potential assessment, remote sensing image recognition

中图分类号:

P23
P208

刘伟丽, 田志勇, 罗勇强. 基于深度学习的武汉市屋顶光伏地理潜力评估[J]. 测绘通报, 2026, 0(7): 167-172.

Liu Weili, Tian Zhiyong, Luo Yongqiang. Assessment of Wuhan's rooftop photovoltaic geographic potential based on deep learning[J]. Bulletin of Surveying and Mapping, 2026, 0(7): 167-172.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://tb.chinasmp.com/CN/10.13474/j.cnki.11-2246.2026.0725

http://tb.chinasmp.com/CN/Y2026/V0/I7/167

参考文献

[1] 郭晓琳.基于屋顶面积的徐州市屋顶太阳能光伏潜力评估[D].徐州:中国矿业大学,2015.
[2] Gutschner M,Nowak S,Ruoss D,et al.Potential for building integrated photovoltaics,IEA-PVPS Task 7[EB/OL].[2025-09-16].https://www.iea-pvps.org/wp-content/uploads/2020/01/rep7_04.pdf.
[3] Peng Jinqing,Lu Lin.Investigation on the development potential of rooftop PV system in Hong Kong and its environmental benefits[J].Renewable and Sustainable Energy Reviews,2013,27:149-162.
[4] Wiginton L K,Nguyen H T,Pearce J M.Quantifying rooftop solar photovoltaic potential for regional renewable energy policy[J].Computers,Environment and Urban Systems,2010,34(4):345-357.
[5] 齐中琨,沈朝,王芳.基于ArcGIS的城市屋顶太阳能应用潜力评估[J].低温建筑技术,2023,45(6):96-101.
[6] 张华,王立雄,李卓.城市建筑屋顶光伏利用潜力评估方法及其应用[J].城市问题,2017(2):33-39.
[7] Ko L,Wang J C,Chen C Y,et al.Evaluation of the development potential of rooftop solar photovoltaic in Taiwan[J].Renewable Energy,2015,76:582-595.
[8] 刘梦月,彭佳强,肖睿,等.基于深度学习的城市屋顶光伏面积识别及光伏发电潜力研究[J].建筑科学,2023,39(6):151-158.
[9] 倪浩展,赵文卓,蒋沃林,等.基于深度学习的低碳城市建筑屋顶太阳能利用潜力评估——以上海市虹口区为例[C]//2022全国建筑院系建筑数字技术教学与研究学术研讨会.厦门:[出版者不详],2022.
[10] Zhong Teng,Zhang Zhixin,Chen Min,et al.A city-scale estimation of rooftop solar photovoltaic potential based on deep learning[J].Applied Energy,2021,298:117132.
[11] 殷吉崇,武芳,翟仁健,等.面向建筑物轮廓规则化的双路径边界约束与相对论生成对抗网络[J].测绘学报, 2024,53(7):1444-1457.
[12] Chen L C,Zhu Yukun,Papandreou G,et al.Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of 2018 European conference on computer vision (ECCV).Cham:Springer,2018:833-851.
[13] Sandler M,Howard A,Zhu Menglong,et al.MobileNetV2:inverted residuals and linear bottlenecks[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:4510-4520.
[14] Chollet F.Xception:deep learning with depthwise separable convolutions[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu:IEEE,2017:1800-1807.
[15] Ronneberger O,Fischer P,Brox T.U-Net:convolutional networks for biomedical image segmentation[C]//Proceedings of 2015 Medical Image Computing and Computer-Assisted Intervention-MICCAI.Cham:Springer,2015:234-241.
[16] Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[EB/OL].[2025-09-22].https://arxiv.org/abs/1409.1556.
[17] He Kaiming,Zhang Xiangyu,Ren Shaoqing,et al.Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas:IEEE,2016:770-778.
[18] Hussain Z K,Jiang Congshi,Adrees M,et al.A novel architecture for building rooftop extraction using remote sensing and deep learning[J].Remote Sensing Applications:Society and Environment,2025,38:101551.
[19] 彭博,司东雨,谢志伟,等.基于U-Net算法的城市屋顶太阳能光伏潜力估算[C]//第二十届沈阳科学学术年会.沈阳:[出版者不详],2023.
[20] 武汉市自然资源和规划局.《武汉市主体功能区规划》今起征集意见！落实总规有了新抓手,武汉每个区片将明确功能定位[EB/OL].2021-10-28[2025-09-22].https://zrzyhgh.wuhan.gov.cn/zwdt/gzdt/202110/t20211028_1821816.shtml.
[21] Charlie Dou,左冲,贾彦,等.中国农村民居屋顶分布式光伏发电的发展潜力分析[J].太阳能,2023(1):5-16.
[22] 徐福圆.基于遥感图像的屋顶面积识别及屋顶光伏容量估计[D].杭州:杭州电子科技大学,2016.
[23] Mao Hongzhi,Liu Weili,Li Chongzheng,et al.An accurate quantification study on the rooftop PV potential based UAV field photography in dense urban environments[J].Applied Energy,2025,399:126499.