Stochastic multi-objective optimization for office energy saving with occupant uncertainty

Purpose To address the insufficient consideration of occupant behavior uncertainty in multi-objective optimization for office building energy efficiency by proposing a stochastic optimization framework that incorporates multiple occupant behaviors into the optimization process. Design/methodology/approach Field data from typical office buildings in Ningbo were used to develop four occupant behavior models (occupancy, air conditioning, shading and window operation). These models were integrated into EnergyPlus simulations within a stochastic multi-objective optimization framework based on NSGA-II and sample average approximation (SAA). Findings The proposed framework improved the robustness of the optimization results under occupant behavior uncertainty. The balanced solution reduced fluctuation ranges compared to the original model, achieving significant energy savings while revealing a trade-off with thermal comfort. Among the investigated design alternatives, a window-to-wall ratio of 0.25 provided the best overall performance. Originality/value This study presents a practical stochastic multi-objective optimization framework that integrates empirically derived occupant behavior models with SAA and NSGA-II. By embedding occupant behavior uncertainty directly into the optimization process rather than relying on post-hoc robustness analysis, the proposed framework provides a more reliable approach for energy-efficient office building design.