Multi-performance coupled optimization drives low-carbon retrofitting of site museums

Abstract

As typical public buildings, site museums face the challenge of achieving low-carbon renovation while simultaneously addressing the dual objectives of preserving immovable cultural relics and providing a comfortable indoor thermal environment for visitors. This study proposes a multi-performance coupled prediction and optimization method tailored to the unique operational conditions of site museums and applies it two representative case studies of different scales (M1 and M2). Among the several predictive models examined, ANN and LGB are found to be better suited for this study, achieving R² values of 0.878 and 0.899 for M1, and 0.914 and 0.925 for M2. The optimal solution identified by the entropy weighting method led to a 24.23% improvement in indoor Daylighting Index (DLI) for M1, while the proportion of space where glare is effectively mitigated was increased by 63.35%. Simultaneously, the Thermal Comfort Hours (TCH) increased by 9.98%, and the Carbon Emission Intensity (ECI) per unit area decreased by 12.76%. For M2, the optimization solution resulted in a 17.76% improvement in TCH and a 13.63% reduction in ECI. Although the improvement in DLI was marginal at 0.21%, the space for enhancing Spatial Glare Autonomy (sGA) increased by 6.53%. The SHapley Additive exPlanations (SHAP) method was employed for interpretability analysis, quantifying the interactions between renovation parameters. Sensitivity analysis revealed significant variations in the impact of design parameters on performance indicators, with results consistent with the SHAP analysis, thereby confirming the reliability of the findings. The approach proposed in this study can promote environmental enhancement in heritage preservation and contribute to achieving sustainable urban and social development goals.