Comparison of MPC strategies for building control

Abstract

We propose three different Model Predictive Control (MPC) strategies for controlling temperatures in buildings. We show that maximization of thermal comfort, along with minimization of energy consumption, can be cast in various ways, each having their pros and cons. The three strategies include tracking of setpoint temperatures, tracking of a comfort zone, and minimization of number of violations of such a zone. Even though the latter formulation is the most demanding from a computational point of view, it indeed provides best thermal comfort. All proposed methods are compared with respect to two qualitative criteria on simulations.