Integrating local climate considerations into dormitory energy optimization: An explainable machine learning and multi-objective design approach

Low-carbon, energy-efficient, and comfortable design has become a global trend in the development of the building industry. However, traditional multi-objective optimization methods often rely on simulation, which significantly reduces decision-making efficiency during the early design stages of urban residential areas. This study proposes a multi-objective optimization framework that couples Bayesian Optimization with an ensemble learning algorithm, aiming to improve the Energy Use Intensity (EUI), Photovoltaic Generation potential (PVG), and Universal Thermal Climate Index (UTCI) of university dormitories, and employing explainable artificial intelligence (AI) methods to analyze the impact of each factor on EUI, PVG, and UTCI. The framework is validated through a case study of a university dormitory area in Tianjin, China. Results show that the BO-ensemble learning algorithm achieves high predictive accuracy, with R² values of 0.99, 0.98, and 0.92 for EUI, PVG, and UTCI. SHapley Additive exPlanations (SHAP) analysis further reveals the contribution of each factor to the prediction model. Compared with the original design, performance improvements are substantial: EUI is reduced by 12.8 %, PVG is increased by 20.68 %, and UTCI is reduced by 2.8 %. Clustering analysis identifies optimal design schemes for different optimization objectives, with Cluster 3 (C3) offering the best trade-off among EUI, PVG, and UTCI. The proposed BO-ensemble based multi objective optimization framework provides a new perspective and method for sustainable dormitory design in university campuses.