Adaptive Model Predictive Controller for Building Thermal Dynamics

Model predictive controllers are becoming widespread in building thermal dynamic control and energy management systems. Decreasing building energy consumption, load shifting, cost reduction, and indoor air quality improvement are some of the topics that these controllers have been shown to be efficient. However, they rely on accurate models that are hard to develop and can be expensive. Additionally, the model should be time-varying to represent the thermal dynamics in a building. To address this issue, this paper proposes an adaptive model predictive controller for thermal dynamic control in buildings. It includes an adaptive parameter identification algorithm that updates the model parameters and guarantees that the estimated parameters converge to the actual values. Moreover, a model predictive controller with an additional constraint to ensure the boundedness of the system trajectories is introduced. The proposed framework uses a simple linear grey-box model of the thermal dynamics, as a nominal model, and the adaptive parameter identification updates the model. This eliminates the need for an accurate model and an enormous bank of data, while the benefits of the model predictive controller and adaptive controller are retained. Simulation results are also provided to demonstrate the capability to identify the deviations and the efficiency of using the updated model in the model predictive controller design.