Non linear predictive control of a batch refrigeration process
Résumé
A predictive optimal control algorithm for a fruit refngeration process is described. The future inlet air temperature profile is optimized on line, taking into account the most recent process measurements, and thus introducing feedback. The optimality criterion is directly based on economic costs. Technological and product quality constraints are included. The future control profile optimization requires long term predictions of process variables evolution, achieved using a physical model of the plant. The model is adjusted on line in order to account for unmeasured disturbances and poorly known product properties. Both the optimality criterion and the plant model are nonlinear. The approach is experimentally validated on a pilot scale refngera-
tion plant. Economic benefits from using an optimal, rather than a traditional constant temperature profile are
discussed, revealing the importance of the process configuration. Robustness of the control algorithm with
respect to modeling errors, measured and unmeasured disturbances, and temporary plant failures is demonstrated.