A novel deep learning and GIS integrated method for accurate city-scale assessment of building facade solar energy potential

Accurately assessing building solar potential is becoming increasingly important for sustainable urban development. However, evaluating the solar energy potential of building facades in urban areas poses significant challenges due to complex shading from surrounding structures and a lack of detailed facade information. This study proposes a comprehensive framework for assessing the solar PV potential of urban facades by integrating deep learning and geographic information systems (GIS). GIS was used to extract information about the layouts and heights of buildings, while a deep learning-based approach was developed to identify the window-to-wall ratio (WWR) of various building facades from street view images. To validate the proposed methodology, a region in Wuhan with a diverse range of architectural features was selected. The solar energy potential was estimated by combining facade information with shadow analysis. Additionally, a solar irradiance measurement experiment was conducted to verify the findings. The results revealed that a lack of WWR information for building facades can lead to significant overestimations of their solar energy potential, with errors ranging from 15 % to 50 %. Moreover, using standardized WWRs in the assessment can still result in errors between 3 % and 20 %. These discrepancies primarily stem from differences between actual and assumed WWRs used in the calculations. Further analysis shows that accurately assessing the solar energy potential of facades in various orientations requires considering both WWR data and shading effects. This comprehensive approach can be employed to more accurately characterize the solar energy potential of building facades in urban areas, facilitating the broader adoption of solar energy at the city scale.