Center for Research in Energy Systems Transformation (CREST)

A framework for the design and simulation of a building envelope and an HVAC system is used to compare advanced control algorithms in terms of energy efficiency, thermal comfort, and computational complexity. Building models are first captured in Modelica to leverage Modelica’s rich building component library and then imported into Simulink [15] to exploit Simulink’s strong control design environment. Four controllers with different computational complexity are considered and compared: a proportional (P) controller with time varying temperature bounds, a tracking linear quadratic regulator (LQR) controller with time varying tuning parameters, a tracking disturbance-aware linear quadratic regulator (d- LQR) controller with time varying tuning parameters which incorporates predictive disturbance information and a model predictive controller (MPC). We assess the performance of these controllers using two defined criteria, i.e. energy  and discomfort  measured with appropriate metrics. We show that the d-LQR and MPC, when compared to the P controller, manage to reduce energy by 41. 2%and 46%respectively, and discomfort from 3.8 to 0. While d-LQR and MPC have similar performance with respect to energy and discomfort, simulation time in the case of d-LQR is significantly less than the one of MPC.