Optimal Grade Transition for Polymerization Using a Novel Sequential Evolution Algorithm

Abstract

In order to reduce off-specification polymer products, it is important to obtain the optimal trajectories of manipulated variables during the grade transition process. A real-coded genetic algorithm based sequential evolution algorithm is proposed, wherein the population evolves in each dimension gradually based on the introduced evolution points and a grouped evolution strategy is used in each subpopulation. The Lee-Ramirez bioreactor case test showed that the sequential evolution algorithm is high-effective and is quite suitable for solving dynamic optimization problem. Then the sequential evolution algorithm is applied to solve the grade transition optimization problem in a gas phase fluidized-bed polyethylene reactor, where the different situations of multiple grades change over, various constraints, model mismatch and process disturbance are considered respectively. The results demonstrated that the optimal profiles could be determined precisely with short computing time.