Infinite horizon MPC applied to batch processes. Part I

GONZÁLEZ, ALEJANDRO HERNÁN; ADAM, EDUARDO; ODLOAK, DARCI; MARCHETTI, JACINTO LUIS

Rosario, Argentina

Congreso; XIIIº Reunión de Trabajo en Procesamiento de la Información y Control (RPIC 09); 2009

Several scenarios could arise when a batch process is attempting to be controlled. Here it is considered the case where the process parameters do not change substantially from one batch to next, for large periods of time. In this context, the so-called Iterative Learning Predictive Control (ILPC) - a Model Predictive Control that “learns” a control strategy from past batches - is a suitable strategy to control the process, since it can take several iterations to learn the way the process behaves. On the other hand, there are cases in which the process parameters substantially changes from one batch to the next. In this context, the ILPC must be able to achieve a good performance at the first iterations after the changes (it should learns quickly), since otherwise the controller would keep searching the best behavior from one change to the other. In this work we propose an IHMPC controller, under closed-loop paradigm, that explicitly takes into accounts the second case. Several simulation examples show that the proposed algorithm presents a good performance in contrast with existing algorithms. In addition, the proposed methodology allows a direct extension to a new MPC with learning properties that accounts for the first case (i.e. for batch processes with almost constant parameters). This last procedure is presented in a second stage of this work (González et al., 2009).