Multiobjective Optimization of Double‐Skin Façade Performance Using XGBoost and NSGA‐II in Buildings Under Hot Climate Conditions
Abstract
Buildings account for ~40% of global energy consumption, with cooling loads dominating in hot and arid climates. Double‐skin façades (DSFs) represent promising building envelope technologies for energy efficiency enhancement, yet standardized optimization methodologies for such climatic conditions remain limited. This study aimed to develop and validate a comprehensive multiobjective optimization framework integrating machine learning algorithms with building energy simulation (BES) to optimize DSF performance in residential buildings under hot and arid climatic conditions. The research employed EnergyPlus and DesignBuilder to simulate 2500 unique DSF configurations using Latin Hypercube Sampling (LHS) across varied design parameters including cavity depth, glazing properties, and ventilation strategies. XGBoost regression models were trained to predict heating loads, cooling loads, and environmental impacts. The Nondominated Sorting Genetic Algorithm II (NSGA‐II) was implemented using XGBoost models as fitness functions for multiobjective optimization targeting energy consumption, cost, and thermal comfort in Saudi Arabian climatic conditions. XGBoost models achieved exceptional prediction accuracy with R 2 values exceeding 0.94 and mean absolute percentage errors (MAPEs) below 4.5%. Optimal DSF configurations demonstrated energy consumption reductions of 35.7–42.3% compared to conventional single‐skin façades, with corresponding and thermal comfort improvements of 58.1–78.9%. Economic analysis revealed favorable investment characteristics with payback periods of 3.1–4.2 years and net present values exceeding SR 975/m 2 . The framework achieves substantial reductions in computational time, with surrogate‐based predictions enabling near‐instant evaluation of new DSF configurations, thus making optimization feasible for practical design applications. The demonstrated economic viability and computational efficiency establish practical foundations for sustainable building design implementation in hot and arid regions.