A UAV-Based Framework for Community-Scale Residential Building Energy Modelling in China

With increasing emphasis on energy efficiency and carbon emission reduction in the building sector, rapid and scalable energy modelling of existing buildings is critical for retrofit projects and policy development. Conventional surveys, data collection and energy modelling processes remain labour-intensive and impractical for large-scale applications, and energy modelling often dominates project time and cost. Unmanned Aerial Vehicles (UAVs) equipped with high-resolution and infrared cameras now offer fast and safe exterior surveys, showing great potential in building energy assessments. Yet, a systematic framework for UAV integration, from image capture through processing to inputs for building energy simulation, has not been established. This research addresses these challenges by developing a structured framework for community-scale building energy modelling, which integrates UAV photogrammetry to accelerate data acquisition, reduce labour, and shorten energy modelling time for retrofit projects. The framework is designed to minimise prior data requirements while maximising automation and retaining sufficient accuracy and reliability. It also addressed systemic issues, such as the lack of data necessary to fully exploit the potential of available technologies in the context of this study. To achieve this research objective, a mixed-methodology approach, combining desk studies and fieldwork, was adopted. First, a comprehensive review of UAV-based building energy studies mapped the current applications and available tools, proposing a preliminary framework. Fieldwork then encompassed a pilot study and a detailed case study. In the pilot test, a single building was surveyed to identify an optimal toolchain for flight planning, image capture, photogrammetric reconstruction, model generation, and selection of a simulation platform. The resulting refined framework was applied to a case study of a 39-building community, and it was demonstrated that further automation was necessary. The potential for automation was then further explored. Geomatic algorithms, machine learning, and deep learning techniques were compared, and the combination that best balanced accuracy and processing time was selected for 3D model generation and window-to-wall ratio (WWR) extraction. In parallel, an archetype library of residential buildings in the Hot Summer Cold Winter (HSCW) zone in China was compiled to facilitate the assignment of thermal parameters. Finally, a reusable community-scale building energy modelling (CBEM) workflow was assembled in Grasshopper using Ladybug Tools components to process the aforementioned inputs. In conclusion, this study integrates UAV photogrammetry, GIS-based geomatic algorithms, machine learning, and graphical programming to develop an UAV-CBEM framework. The outcomes establish a foundation for scalable UAV-based building energy assessments, enhancing applicability in data-limited and large-scale settings worldwide.