Model Predictive Control of Vapor Compression Systems

Model Predictive Control (MPC) of vapor compression systems (VCSs) offers several advantages over conventional control methods (such as multivariable process control with selector logic) in terms of 1) the resulting closed-loop performance and 2) the control engineering design process. VCSs are multivariable systems and feature constraints on system variables and actuators that must be enforced during steady-state and transient operation. We present the design and validation of an MPC for a split ductless VCS. The design regulates room temperature with zero steady state error for unknown changes in the thermal load and enforces constraints on system variables such as compressor discharge temperature and actuator ranges and rates. We show how the MPC design can evolve during the engineering process by adding and modifying constraints and process variables. The design methodology provides guarantees in terms of closed loop stability and convergence. Importantly, in contrast to other published results on MPC for VCSs, our design makes use of only available temperature measurements and does not require pressure or mass flow measurements which are typically not available in production VCSs.