Differentiated Design of Dormitory Buildings Based on Orientation, Focusing on Thermal and Daylighting Performance

In university dormitory buildings composed of standardized units, units of different orientations are typically designed with identical parameters. Using a typical dormitory as the baseline, XGBoost-based models with Bayesian hyperparameter optimization were coupled with NSGA-II for multi-objective optimization. Sensitivity analysis was performed to reveal the orientation-dependent influence mechanisms of design variables and to identify the key variables affecting useful daylight illuminance (UDI), vertical illuminance at eye level (VCI), predicted percentage of dissatisfied (PPD), and energy use intensity (EUI) in south- and north-facing dormitories, with TOPSIS used for decision making. Results show that: (1) For both orientations, the window-to-wall ratio and window type are the key variables. South-facing units are more sensitive to the window solar heat gain coefficient, while north-facing units are more sensitive to the window thermal transmittance. (2) Window Types 8 and 5 are recommended for north- and south-facing rooms, with window-to-wall ratios of 0.30 and 0.39, respectively. (3) Comparing three daylighting optimization scenarios reveals that high correlation does not necessarily imply substitutability: the maximum difference in all performance indicators reached 6.75%, indicating that indicator selection affects the multi-objective optimization results. (4) The effect of wall reflectance on vertical illuminance depends on orientation, window parameters, and floor level, and is more pronounced in higher-floor rooms.