Synergizing passive building envelopes and data-driven Model Predictive Control for building energy savings

Model Predictive Control (MPC) is an effective energy-saving strategy for passive buildings, but its high computational cost hinders integration with complex building systems, necessitating rapid dynamics modelling. Moreover, while passive buildings are highly sensitive to weather, current MPC mainly relies on cost function with fixed weighting factors, hard to compute optimal control actions. To overcome these challenges, this study proposes an adaptive data-driven MPC framework tailored for passive buildings to optimize their energy allocation under different weather conditions. Specifically, this framework integrates a dual-level algorithm featuring the Nonlinear Autoregressive Network with Exogenous Inputs (NARX) model and an adaptive MPC algorithm. Validation confirms the exceptional generalizability of NARX models, with R2 exceeding 0.96. Simulations reveal that the proposed framework reduces cooling energy by 19% compared to a conventional control strategy. Crucially, the adaptive MPC algorithm additionally saved 31.2% cooling energy in Hong Kong. Overall, the proposed MPC framework incorporates passive building envelopes technologies, namely radiative cooling roofs and thermochromic smart windows, effectively improving energy-saving performance while maintaining thermal comfort in buildings.