以屋顶航拍图像为研究对象,通过图像识别和软件模拟的方法,对区域建筑屋顶的光伏发电潜力进行评估。针对建筑物轮廓提取不精确的问题,设计逐步优化的多尺度特征融合网络SOV-net,兼顾高层特征中的全局语义和低层特征中的局部细节,提升屋顶边界提取的准确性和轮廓的完整性。随后,筛选屋顶可铺设区域,将屋顶按照空间结构进行四分类,实现光伏阵列铺设方案的可视化,并评估区域的最大光伏发电潜力。结果表明,该区域可铺设光伏组件51754块,年发电总量为18.65 GWh。
Abstract
Using aerial rooftop images as the research object, the photovoltaic power generation potential of regional building roofs was evaluated by image recognition and software simulation. To address the problem of inaccurate extraction of building contours, a gradually optimized multi-scale feature fusion network SOV-net was designed, which takes into account both the global semantics in high-level features and the local details in low-level features, improving the accuracy of roof boundary extraction and the integrity of the contour. Subsequently, available areas were screened, and the roof was classified into folur categories according to spatial structure. This study visualizes the photovoltaic arrays laying scheme and evaluates the maximum photovoltaic power generation potential of the region. The results show that 51754 photovoltaic modules can be laid in this region, with a total annual power generation of 18.65 GWh.
关键词
光伏发电 /
图像识别 /
深度学习 /
建筑物提取 /
光伏潜力
Key words
photovoltaic power /
image recognition /
deep learning /
building extraction /
photovoltaic potential
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 于瑛, 姚星, 丑锦帅, 等. 城镇典型住宅建筑屋顶分布式光伏系统潜能分析[J]. 太阳能学报, 2023, 44(7): 182-190.
YU Y, YAO X, CHOU J S, et al.Potential analysis of distributed PV systems on roof of typical residential building in urban area[J]. Acta energiae solaris sinica, 2023, 44(7): 182-190.
[2] DEFAIX P R, VAN SARK W G J H M,WORRELL E, et al. Technical potential for photovoltaics on buildings in the EU-27[J]. Solar energy, 2012, 86(9): 2644-2653.
[3] 李泞吕, 赵方凯, 陈利顶. 城市建筑屋顶光伏发电潜力评估方法和模型[J]. 生态学报, 2023, 43(10): 4284-4293.
LI N L, ZHAO F K, CHEN L D.Review of rooftop solar photovoltaic electrical potential estimation: approaches and models[J]. Acta ecologica sinica, 2023, 43(10): 4284-4293.
[4] YILDIRIM D, BÜYÜKSALIH G, SAHIN A D. Rooftop photovoltaic potential in Istanbul: calculations based on LiDAR data, measurements and verifications[J]. Applied energy, 2021, 304: 117743.
[5] YU J F, WANG Z C, MAJUMDAR A, et al.DeepSolar: a machine learning framework to efficiently construct a solar deployment database in the United States[J]. Joule, 2018, 2(12): 2605-2617.
[6] WEI S Q, JI S P, LU M.Toward automatic building footprint delineation from aerial images using CNN and regularization[J]. IEEE transactions on geoscience and remote sensing, 2020, 58(3): 2178-2189.
[7] ZHU Q, LIAO C, HU H, et al.MAP-net: multiple attending path neural network for building footprint extraction from remote sensed imagery[J]. IEEE transactions on geoscience and remote sensing, 2021, 59(7): 6169-6181.
[8] 徐孝彬, 张好杰, 白建波, 等. 基于改进Unet的分布式光伏建筑物高精度分割方法[J]. 太阳能学报, 2023, 44(11): 82-90.
XU X B, ZHANG H J, BAI J B, et al.High-precision segmentation method of distributed photovoltaic buildings based on improved Unet[J]. Acta energiae solaris sinica, 2023, 44(11): 82-90.
[9] THAI C, BROUWER J.Challenges estimating distributed solar potential with utilization factors: California universities case study[J]. Applied energy, 2021, 282: 116209.
[10] SIMONYAN K, ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[C]//3rd International Conference on Learning Representation, 2015, 1409.1556.
[11] LIANG X J, ZHANG Y, ZHANG J P.Attention symbiotic neural network for hyperspectral image refined classification based on relative water content retrieval[J]. IEEE transactions on geoscience and remote sensing, 2021, 59(7): 5998-6016.
[12] BADRINARAYANAN V, KENDALL A, CIPOLLA R.SegNet: a deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481-2495.
[13] VARGHESE R, M S YOLOv8: a novel object detection algorithm with enhanced performance and robustness[C]//2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS), Chennai, India, 2024: 1-6.
[14] 杨瑛, 胡楼君, 高青, 等. 基于光伏发电预测的居住小区建筑高度组合寻优[J]. 太阳能学报, 2024, 45(5): 500-507.
YANG Y, HU L J, GAO Q, et al.Optimization of residential building height combination based on photovoltaic power generation prediction[J]. Acta energiae solaris sinica, 2024, 45(5): 500-507.
[15] RUAN J C, XIE M Y, GAO J S, et al.EGE-UNet: an efficient group enhanced UNet for skin lesion segmentation[M]. Cham: Springer Nature Switzerland, 2023: 481-490.
[16] SÁNCHEZ-APARICIO M, MARTÍN-JIMÉNEZ J, DEL POZO S, et al. Ener3DMap-SolarWeb roofs: a geospatial web-based platform to compute photovoltaic potential[J]. Renewable and sustainable energy reviews, 2021, 135: 110203.
[17] BATZELIS E I, PAPATHANASSIOU S A.A method for the analytical extraction of the single-diode PV model parameters[J]. IEEE transactions on sustainable energy, 2016, 7(2): 504-512.
[18] MÜLLER M. Potenzialatlas erneuerbare energien Baden-Württemberg[C]//Proceedings of the 27th Conference on Environmental Informatics-Informatics for Environmental Protection, Sustainable Development and Risk Management, Shaker Verlag, 2013.
基金
安徽省自然科学基金（2108085UD03）
PDF(4208 KB)
