International High Performance Buildings Conference
International High Performance Buildings Conference edition:4 location:West Lafayette date:11-14 July 2016
Model Predictive Control (MPC) is a promising control method to reduce the energy use of buildings. Its commercialization is, however, hampered by the difficulty of obtaining a reliable controller model. This paper compares two approaches to obtain such controller model: (1) a white-box model approach for which a detailed first-principles building model is linearized, and (2) a system identification method using a grey-box model approach. The MPC performance using both model approaches is evaluated on a validated 12 zones model of an existing office building.
The results indicate that the MPC performance is very sensitive to the prediction accuracy of the controller model. This paper shows that both approaches can lead to an efficient MPC as long as very accurate identification data sets are available. For the considered simulation case, the white-box MPC resulted in a better thermal comfort and used only 50% of the energy used by the best grey-box MPC.