Interaction-based weight optimization framework for adaptive façades: balancing cooling energy, carbon, and comfort

Adaptive facade performance depends on both window–to–wall ratio and building orientation. Yet most design methods treat WWR as fixed and overlook parameter interactions. This study presents a summer–focused optimization framework for PCM–integrated Venetian blind facades in Tehran’s semi–arid climate. The framework combines EnergyPlus simulation, Sobol sensitivity analysis, XGBoost surrogate modeling, and Bayesian Optimization. An Interaction‑Based Weight Framework derives MCDM weights directly from Sobol higher‑order interaction indices ( S H ), and simultaneously treats WWR as an active variable evaluated independently across four cardinal orientations. The results show that Slat Angle and WWR dominate both individual and interaction effects across all facades. In contrast, PCM Thickness, Slat Depth, and Slat Distance to Glass behave as purely additive parameters with negligible interactions. Optimal WWR varies sharply with orientation, reaching 78.18% for north–facing facades while converging near 20% for south, east, and west. The optimized designs achieve cooling energy reductions ranging from 37.2% to 61.9%, with comparable carbon emission savings. Adaptive thermal comfort improves to 97–99% neutral hours across all orientations. The XGBoost surrogate achieves R 2 values above 0.98, and EnergyPlus re–simulations confirm its predictions with MAPE errors of 0.02–2.27% for Cooling EUI, 0.02–0.54% for pct neutral , 0.48–3.80% for Cooling CEI, and 4.69–8.51% for deg neutral . This framework offers an interaction-informed, orientation-specific, and summer-tailored approach for semi-arid climates that moves beyond fixed-WWR conventions, providing physics-based guidance. The methodology helps prioritize facade design parameters according to their true interaction–driven influence on energy, carbon, and comfort outcomes in regions with similar climatic conditions.