Leveraging online housing data for large-scale building energy modeling

Buildings are significant contributors to global energy demand, and their energy demand continues to increase annually. Addressing this growing demand requires effective and scalable solutions supported by adaptable tools. Large-scale Building Energy Modeling (BEM) approaches based on survey data can inform decisions to enhance building energy efficiency. However, the scalability of BEM is often limited by the sparsity of the survey data for studies requiring higher granularity. To address this limitation, a novel framework was proposed for Online housing data-informed Building Energy Modeling (OBEM), which leverages online housing data to enable large-scale BEM at smaller geographic levels, such as census tracts. The framework involved extracting individual building information from online housing data, creating survey-based building energy models using a Bayesian prediction network, and performing model-based imputation through a weighted cosine similarity evaluation to address missing inputs. The model developed through OBEM was calibrated using actual utility data and applied to two communities in Baltimore City, MD – one disadvantaged and one non-disadvantaged—encompassing 954 housing units. To demonstrate the application of OBEM, the impacts of future climate changes and window improvements on building energy performance for two communities were evaluated. Results revealed disparities in energy consumption between the two communities, which are both expected to experience decreased natural gas and increased electricity usage under future climate trends. Additionally, window replacement resulted in greater energy demand reduction in the disadvantaged community. This work provides a robust tool for policymakers and stakeholders to make informed decisions on building performance improvements.