Optimisation of Grade Transitions in an Industrial Gas-phase Olefin Polymerization Fluidized Bed Reactor via Nco Tracking

Gas-phase olefin polymerization fluidized-bed reactors are complex processes that are characterized by large time constants and slow transitions. Optimal grade changes can be computed off-line on the basis of a detailed process model and applied to the process in an open-loop mode. However, the presence of uncertainties in the form of process-model mismatch and disturbances make this approach clearly non-optimal. Fortunately, measurements can be used to improve the situation; this can be done either explicitly via process model refinement and repeated optimisation, or implicitly via direct input adjustment. This work considers the optimisation of grade-transition changes by enforcing on-line the necessary conditions of optimality (NCO tracking) using available measurements. The key steps are the generation of an adjustable model of the optimal solution and its adaptation using appropriate measurements or estimates of the controlled variables. A realistic simulation process model is employed to compare the performance of NCO tracking to that obtained via the open-loop application of nominal optimal trajectories. Copyright © 2005 IFAC