Active sequential identification and self-contained disturbance reconstruction for building predictive control

The widespread adoption of Model Predictive Control (MPC) in buildings remains limited by the cost and complexity of model development. This paper presents a unified, practically deployable framework for control-oriented thermal modelling, disturbance reconstruction, and predictive control of multi-zone commercial buildings. The framework combines active HVAC excitation with a sequential gray-box identification procedure to estimate parameters of a second-order resistance-capacitance (2R2C) thermal model, requiring no additional sensors. Reconstructed disturbance signals from operational data are used to train forecasting models, eliminating dependency on external data sources. The identified model and forecasts are embedded in a convex MPC formulation to minimize cost while maintaining thermal comfort. Validated on a detailed EnergyPlus simulation of a seven-zone airport terminal, the framework converged in 12 days and reduced monthly peak demand by more than 50% and total electricity costs by approximately 34% relative to a rule-based baseline, with a modest increase in total energy consumption of 2–4%.