Improving urban building energy modeling (UBEM) accuracy through shape and energy use pattern calibration factors based on representative buildings

Rapid urbanization and population concentration have increased urban energy demand, intensifying stress on urban power grids. Urban building energy modeling (UBEM) is widely used to predict urban-scale energy demand; however, its accuracy strongly depends on input data availability and model resolution. Because detailed building-level energy data are rarely available at scale, many UBEM studies rely on representative buildings to reduce data and computational burdens, which can oversimplify geometry and temporal load patterns and reduce reliability for decision-making. To address these limitations, this study proposes a two-step calibration framework for a representative building-based UBEM that accounts for differences in building geometry and monthly energy use patterns. First, we derive shape calibration factors using shape indices calculated from building height, floor area, and volume. Second, we divide energy use data into (i) heating and cooling and (ii) baseload energy use intensity (EUI) [kWh/m2] and derive energy use pattern calibration factors that adjust the monthly distribution. The framework is demonstrated using UBEM simulations based on the representative buildings and a target dataset of Seoul office buildings. Compared to the uncalibrated representative building-based UBEM, the shape calibration significantly reduced end-use CVRMSE (e.g., heating: 964.60% — 427.92%, cooling: 210.97% — 103.63%, baseload: 58.69% — 35.02%). Energy use pattern calibration further improved the consistency of monthly patterns, increasing the Pearson correlation coefficient from 0.57 to 0.99 and reducing the average monthly total EUI error rate from 21.66% (before calibration) to 7.79% (after pattern calibration). This approach improves accuracy while maintaining the efficiency benefits of representative buildings, and it provides a practical pathway to more reliable UBEM for urban-scale applications such as load forecasting and retrofit prioritization.