Integrated Multilevel Optimization in Large-Scale Poly(Ethylene Terephthalate) Plants

Abstract

The present paper discusses the vertical integration of advanced predictive control, businesswide optimization and markets laws, based on a large-scale first-principles model. The case study is the poly(ethylene terephthalate) (PET) plant, with esterifier sections, finisher reactors, a crystallizer, and a solid-state polymerizer, and the overall mathematical model consists of more than 1500 differential and algebraic equations (DAEs). The complex integration among supply chain layers is reached using a Windows−Unix hybrid system; the methodology adopted is described, and a real-time strategy is implemented by using a secure shell connection for linking a PentiumIV, which simulates the plant, and a cluster Opteron, where the optimizer operates.