Multi-objective optimization for energy-efficient building design considering urban heat island effects

Building energy performance (BEP) associated with climate change and urban heat island effects (UHI) play an important role in urban sustainable development. To predict and optimize BEP under various socioeconomic scenarios, a new framework combining the physical simulation modeling integrated explainable machine learning and multi-objective optimization is proposed in this study. A Grasshopper-based simulation model incorporates BO-LGBM (Bayesian optimization-LightGBM) is developed to construct a solid prediction system, which tends to tune the hyperparameters accurately and explain more details with the aid of SHapley Additive explanation (SHAP). Two major aspects, including the building energy use intensity and indoor thermal comfort, are modeled by considering the different Shared Socioeconomic Pathways (SSPs) climate change scenarios in the near and far future. A multi-objective optimization method is employed to find an optimal solution for energyefficient building design under constraints or uncertainties. Key findings include a 54% improvement in the Pareto front for building energy optimization and a significant impact of SSP585 scenarios on future energy consumption. The main novelty lies in the incorporation of machine learning into a physical model to achieve energy-efficient building design in urban contexts by considering UHI effects and climate change, offering actionable strategies for BEP assessment and promoting sustainable city planning.