IoT-enabled occupant-in-the-loop MPC for responsive built environment control and its implementation in a living lab

The uncertainty and intermittent nature of occupancy patterns pose challenges to effective built environment control, which further complicated by the need to respond to diverse personalized thermal comfort in a varying building environment. Model predictive control (MPC) integrating occupancy information has emerged as a promising approach for enhancing comfort, air quality (IAQ) and energy efficiency. However, implementing the MPC strategy in real-world HVAC systems of existing buildings with conventional control infrastructure to handle diverse comfort preferences, stochastic occupancy disturbances and dynamic thermal responses still faces significant challenges. This study proposes and experimentally implements an IoT-enabled Occupant-in-the-loop (OITL) MPC strategy that advances built environment control through several contributions. First, it incorporates dynamic reference temperature in the OITL-MPC to enable responsiveness to real-time comfort feedback and integrates stochastic occupancy prediction to account for the stochasticity and time-inhomogeneity of occupancy disturbances. Second, this study develops an IoT architecture towards adaptable implementation of the OITL-MPC in existing buildings. Third, the proposed OITL-MPC strategy is implemented in a real-world office living lab featured with highly intermittent occupancy pattern and active occupancy participation. Finally, the study comprehensively evaluates the multi-dimensional performance and benchmarks against two baseline strategies: conventional local control (FIX) and occupancy-based control (OBC). One-month experimental results show that the OITL-MPC achieves (1) 25.3% energy savings against FIX, (2) improved percentage of arrival temperature meeting Grade-I comfort from 23.1% (OBC) to 48.7%, and (3) optimized IAQ by maintaining CO₂ near 800 ppm while reducing per-person fresh air flow rates by 38% against OBC. The results demonstrate the theoretical robustness and practical viability of the proposed IoT-enabled OITL-MPC.