A simulation-informed data-driven thermal zoning automation framework for urban building energy modeling with applications to community-scale and archetype-based modeling

초록

Urban Building Energy Modeling (UBEM) faces persistent challenges in acquiring detailed building information, such as internal loads, HVAC systems, and occupancy conditions, at the urban scale. Consequently, thermal zoning, despite its strong influence on building energy use, is often oversimplified to reduce computational burden, with single-zone and core/perimeter approaches widely adopted. However, such conventional zoning methods frequently lead to inaccurate energy predictions. This study proposes a simulation-informed data-driven thermal zoning automation method based on simulated intrinsic indoor air temperature time-series data, applying K-medoids and Hierarchical clustering techniques to identify thermally homogeneous spaces. The proposed approach was implemented within a UBEM framework and evaluated against measured energy-use data. Results show that the Hierarchical zoning model achieved the highest prediction accuracy, with errors within ± 5.2%, showing consistent underestimation. In contrast, conventional and K-medoids zoning models exhibited substantially higher errors ranging from 15.6% to 30%. In addition, the Hierarchical model reduced natural gas prediction errors to 13.7%, compared to 42.4–56.4% for other models. These results demonstrate that thermally informed zoning can significantly improve the predictive performance of building energy simulation. The proposed framework provides a systematic approach for developing thermally meaningful zoning structures, which can support archetype development and calibration processes in urban building energy modeling. Furthermore, the proposed method provides quantitative insights into the influence of zoning resolution in urban-scale energy analysis. © 2026 Elsevier B.V.

키워드