Semantic building energy modeling: Analysis across geospatial scales

Abstract

Rapid decarbonization of the building sector is critical for mitigating climate change. While simulation-based stock level approaches such as urban building energy modeling (UBEM) help develop carbon reduction plans, they have not reached their full potential convincing individual building owners to act: by relying on archetypes averaged across multiple buildings, UBEM saving predictions can be unreliable at the building-level. Meanwhile, at larger scales, heterogeneity in building stocks requires excessive efforts to accommodate the growing number of archetypes and handle patchworks of geographic information system (GIS) datasets. This paper introduces Semantic Building Energy Modeling (SBEM), a novel framework evolved from UBEMs. It replaces UBEM's static templates with problem-specific semantic building descriptions which are decoupled from model translation layers. By decoupling high-level, human-readable building features from computational representations, SBEM accommodates incomplete or probabilistic data and facilitates coordination between teams, including GIS experts, stock-modeling experts, and software engineers. UBEMs can be seen as a special case of SBEMs appropriate for urban-scale analysis, where SBEMs represent a complementary, augmented set of capabilities. To illustrate the flexibility offered by SBEM, a case study was conducted modeling 2.5 million residential buildings in Massachusetts to assess the economic viability of heat pump adoption. The SBEM approach enables detailed, building-specific analyses, revealing significant variations in economic outcomes based on heating systems and regional characteristics. These insights underscore the importance of semantic granularity for individual homeowner decision-making. By providing a scalable and adaptable framework, SBEM can resolve some existing challenges with UBEMs by allowing consistent model use across scales.