Model-based predictive control of a multi-evaporator vapor compression cooling cycle

This paper presents a decentralized control architecture for multiple evaporator vapor compression systems using model-based predictive control. Vapor compression systems are widely used for heating, air-conditioning and refrigeration, and constitute a major part of total US energy use. Advanced control strategies have the potential to significantly increase energy efficiency, while delivering the necessary amount of cooling capacity. This paper proposes a decentralized control approach based on a study of interacting dynamics, wherein the cooling capacity of each evaporator is controlled by a multi-input, multi-output MPC controller and standard PI controllers are used to regulate system pressures by modulating compressor speed and discharge valve opening. This is in contrast with traditional single-input, single-output control approaches, which can result in undesired dynamic behavior. The efficacy of the proposed control architecture is demonstrated on an experimental system.