Multi-objective optimization for robust optimal green retrofit design of public buildings under multiple SSP scenarios

Climate change is fundamentally altering building energy demand patterns, with the IPCC projecting a shift from heating-dominant to cooling-dominant consumption. However, most green retrofit studies optimize designs for single climate scenarios, risking substantial performance degradation when future conditions deviate from initial assumptions. This study develops an integrated framework combining energy simulation, life cycle cost analysis, multi-objective optimization, and regret analysis to identify robust optimal green retrofit designs across diverse future climates. Eight Shared Socioeconomic Pathway (SSP) scenarios were analyzed using a 28-year-old community service center in South Korea, evaluating 1680 green retrofit designs of wall, roof, window insulation, and lighting across energy consumption, thermal comfort, and economic performance. Results revealed that extreme warming scenarios dramatically increased summer cooling loads while reducing winter heating loads, fundamentally reshaping building energy structures. Optimization based on individual SSP scenarios yielded green retrofit designs with limited climate resilience. Conversely, the robust optimal green retrofit design identified through regret analysis maintained exceptional stability, with mean regret of 0.0003 and maximum regret of 0.0022 across all scenarios, matching scenario-specific optima in six of eight cases. However, all retrofit scenarios were financially unfeasible within the 25-year horizon, yielding negative NPV and SIR<1. Critically, maximizing insulation performance did not ensure optimal outcomes. Green retrofit designs balancing energy efficiency and thermal comfort while managing cost trade-offs demonstrated superior robustness under diverse climate futures. This framework provides essential guidance for climate-resilient green retrofit strategies, demonstrating that multi-criteria robustness-oriented approaches substantially outperform conventional scenario-specific optimization under deep climate uncertainty.