Optimization of heating and cooling energy demand and indoor thermal comfort under oceanic climate conditions using machine learning and NSGA-II Algorithm: A case study of Prague, Czech Republic

Optimizing building energy performance and indoor thermal comfort is essential for enhancing sustainability and climate resilience. This study develops a multi-objective optimization (MOO) framework integrating machine learning algorithms with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimize a typical residential building in Prague, Czech Republic. The optimization, conducted using jEPlus+EA software, minimizes annual heating and cooling energy loads while improving occupant thermal comfort, measured by the Predicted Percentage Dissatisfied (PPD) index. Sensitivity Analysis (SA) using SHapley Additive Explanations (SHAP) identifies the most influential 18 passive and 2 active retrofit strategies. The results indicate that optimal retrofit configurations can reduce heating and cooling energy demand by up to 89%, and 82% respectively while maintaining indoor thermal comfort.

By integrating machine learning-driven optimization with SHAP-based sensitivity analysis, this study provides a computationally efficient, data-driven methodology for sustainable building retrofits. The proposed framework enables evidence-based decision-making, offering practical insights for balancing energy efficiency, thermal comfort, and climate adaptation in the Czech building sector.