Hybrid Nonlinear Model-Predictive Control of a Supermarket Refrigeration System

Abstract

Supermarket refrigeration systems are of hybrid nature due to forced and autonomous switching of the continuous dynamics and the existence of discretely switched actuators such as valves and compressors. One of the main control goals for these systems is to keep temperature and pressure levels within tight bounds while minimizing the wear of the compressors. Since traditional control schemes leave much room for improvement, this paper presents a new nonlinear model-predictive control approach that is tailored specifically to supermarket refrigeration systems. To increase the computation time available to the nonlinear optimization step in every NMPC iteration, a low-level controller is designed which regulates the process on a short time scale. The high-level optimization task is reduced to the determination of optimal parameters for this controller. Simulation results for a hybrid process model demonstrate that the control scheme is capable of reducing the wear of the compressors compared to existing control schemes.