Risk-resilient retrofit strategies for complex buildings: a probabilistic, spatially differentiated optimization framework

Retrofitting complex buildings is complicated by long-term uncertainty and divergent stakeholder priorities, which make it difficult for traditional single-package strategies to balance Life-cycle Cost (LCC), operational carbon emissions, and thermal comfort across different functional zones within complexes. This study aims to develop a robust, preference-aware optimization framework capable of identifying spatially differentiated retrofit strategies that remain effective under uncertain inflation trajectories, evolving grid carbon intensity, and variable stakeholder preferences. The framework integrates Bayesian Structural Time Series (BSTS) forecasting, joint Monte Carlo scenario generation with Monte Carlo Standard Error (MCSE) controlled convergence, a two-stage optimization with full-ensemble evaluation, and Stochastic Multicriteria Acceptability Analysis (SMAA)-based decision analysis with Dirichlet-sampled preference uncertainty. Applied to a complex building in Suzhou, the spatially differentiated retrofit packages achieved a 17% reduction in the 20-year tail risk of operational carbon emissions and a 10% reduction in LCC, outperforming the holistic single-package alternative (12% and 4% reductions, respectively). Robustness diagnostics confirmed that zone-specific decisions remain stable across a wide spectrum of preference and scenario uncertainties. Results show that spatially differentiated packages achieve higher acceptability probabilities and visibility, and more favourable expected trade-offs than holistic packages, although they require more implementation complexity. The proposed framework offers a transparent and reproducible approach to uncertainty-robust retrofit planning, and spatial differentiation provides clear benefits when functional zones exhibit conflicting priorities. Future work will extend the validation to additional climates and building prototypes.