Optimal operation of adsorption chillers: First implementation and experimental evaluation of a nonlinear model-predictive-control strategy

The control strategy strongly influences the performance of adsorption chillers: efficiency and power density depend on phase times for adsorption and desorption. The optimal phase times depend on system characteristics, inlet conditions, and user's preferences regarding the trade-off between efficiency and power density. In principle, these optimal phase times can be determined during operation using nonlinear model predictive control (NMPC), but implementation is complex and, therefore, still missing. In this paper, we propose and implement a NMPC strategy at an adsorption-chiller test stand and experimentally evaluate the performance. The NMPC strategy combines a nonlinear process model, state estimation, phase time optimisation, and prediction of future inlet conditions. The NMPC is experimentally tested for two scenarios: (1) a step in desorption inlet temperature and (2) a typical solar-cooling application. Comparison with a typical state-based control shows that NMPC can increase the specific cooling power by 31.1%.