Building thermal management accounts for a significant share of global energy consumption. The application of Model Predictive Control (MPC) in this context remains limited by the complexity of modeling. Data-enabled Predictive Control (DeePC), which replaces explicit models with data-driven trajectory predictions, offers a promising alternative. However, DeePC’s reliance on quasi-linear dynamics limits its direct control of nonlinear HVAC actuators. This work proposes a hierarchical framework: DeePC optimizes thermal power allocation at the supervisory level, while decentralized PI controllers, tuned via the model-free Virtual Reference Feedback Tuning method, regulate power. Validated on the Building Optimization Testing Framework (BOPTEST), this approach eliminates explicit modeling and achieves superior energy efficiency and comfort management compared to conventional resistance-capacitance model MPC.
Liu et al. (Wed,) studied this question.