A knowledge graph-based performance evaluation framework for sustainable residential block design

Incorporating sustainable performance considerations during the early design stage plays a pivotal role in the whole life cycle of buildings. However, the existing performance-based generative design (PGD) framework heavily relies on computational resources while overlooking the potential benefits of incorporating domain knowledge for sustainable performance evaluation.

This study proposes a knowledge graph (KG)-based performance evaluation framework for sustainable residential block design, which includes rule-based KG reasoning and data-driven KG-based surrogate modeling. The framework is implemented in a residential block design project.

Results show that, compared to the conventional PGD optimization framework, the KG reasoning method is able to identify a substantial number of Pareto-optimal solutions while significantly reducing computational time. In addition, KG-based surrogate modeling reduces the evaluation time from approximately 4.93 days to 25.27 seconds, with CV(RMSE)s remaining below the maximum acceptance threshold of 25% set by ASHRAE.