Performance analysis of Model Predictive Control for the thermal control of buildings

The objective of this work is to set up a method to control the thermal behavior of existing buildings, by means of the model based predictive control approach. In the past years, several works about the predictive control problem for buildings have been produced in literature, but so far a method that describes how to approach this problem and how to optimize its control parameters has not been defined, yet. This work suggests a method supporting the customization of a model based predictive control system, in order to optimize its performance according to the real characteristics of the building. Also, the method suggests how to fine-tune control parameters according to the control policy (e.g. maximization of comfort or energy saving). The proposed method has been validated on a real case study, proving its effectiveness even in the presence of limited accuracy of the reduced model of the building. As for the selected case study, assuming a 50% error on the model, we proved that, thanks to the use of this method, the energy saving increases 8% as compared to the PID control, while thermal comfort increases by 33%.