Research on the Synergistic Optimization of Daylighting and Thermal Performance in University Teaching Buildings from the Perspective of Spatial Heterogeneity

Amid the low-carbon transition, university teaching buildings feature high occupancy and energy use, making the synergistic enhancement of their daylighting and thermal environments crucial for “dual carbon” goals. However, traditional “north–south homogenization” designs in cold regions fail to address the spatial heterogeneity of solar radiation and climate resources, intensifying the trade-off between natural daylighting and Heating Energy Use Intensity (Eh) while restricting space performance optimization. Focusing on a typical cold-region teaching building, this study proposes a “parametric modeling–multi-objective optimization–machine learning” integrated framework. Targeting spatial daylight autonomy (sDA), useful daylight illuminance (UDI), and Eh, we compared the homogeneous baseline model with the Pareto-optimal solution set, demarcated key design parameter boundaries, and developed an ensemble-based rapid prediction model. Based on the parametric simulation analysis of this representative case building in a cold region, results indicate that: (1) Compared to the baseline, the overall optimal scheme reduced Eh by 17.43% while increasing UDI and sDA by 12.0% and 10.5%, respectively. (2) The Pareto set strictly converges toward a due-south orientation and a “deep-south, shallow-north” layout (depth ratio: 0.66–0.77); thermal configurations exhibit “enhanced northern insulation and southern heat gain,” confirming heterogeneous design matches cold climates better. (3) The four constructed machine learning models (MLP, LightGBM, XGBoost, and Random Forest) uniformly achieved test recall rates exceeding 99%, enabling highly precise, rapid classification of top-performing design scenarios during early-stage design. This study overcomes climate-matching blindness in traditional design, providing a multi-objective synergistic optimization path balancing low energy and high-quality daylighting with substantial engineering and theoretical value.