Predictive hybrid control of the supermarket refrigeration benchmark process

Predictive control is applied to a supermarket refrigeration process in which all actuators are on/off devices. This process has been proposed as a benchmark for the evaluation of hybrid control methods. It is shown that decentralized predictive control of display case temperatures, when combined with predictive control of suction manifold pressure, can greatly reduce or eliminate on/off compressor cycling. The strategy is scalable to large systems. Results are presented for an application involving 10 display cases and six compressors. Traditional hysteresis control causes synchronization of display case temperature oscillations. The proposed predictive approach includes desynchronization, which is shown to be necessary for minimization of compressor cycling. Finally, it is shown that the minimum possible number of compressor cycles can be estimated. This limit on control system performance is imposed by process design and operating conditions. It is also shown that the recommended predictive control strategy operates close to the minimum possible number of compressor cycles for the benchmark process.