Bilevel Model Predictive Control of a Semi-Batch Emulsion Copolymerisation Reactor
MetadataShow full item record
The objective of this thesis is to study the implementation of an on-line bilevel model predictive control (MPC) application to an emulsion copolymerisation process in a semi-batch reactor. Previous work has been performed on a regular MPC layer as well as an off-line open-loop study of the process. Using MPC in the process industry is becoming more common. This allows for the optimisation of chemical processes. However, the implementation of MPC requires a process model that accurately depicts the behaviour of the process plant. This is often time-consuming and cumbersome work. The nonlinear model used for this thesis was developed during the COOPOL EU-project, and had been implemented on a pilot plant. Some differences on behaviour between the model predictions and the plant behaviour were observed. The thesis gives an introduction to polymers, emulsion polymerisation and the most important kinetic mechanisms in free-radical emulsion polymerisation. This provides the equations in the model of the process, which is implemented in the programming language C. Using a standard template, it can be easily accessed by Cybernetica AS' specialised software for nonlinear real-time optimisation. The concept behind MPC is presented, as well as the concerns of the control of polymerisation processes. This provides a sufficient base for the understanding of the main purpose, implementation of a bilevel MPC control of the process. All process plants contain a control hierarchy, from the long-term economic objectives to the fast PID controllers that ensure safe operation of the plant. This thesis presents a hierarchy of two MPC-controllers, one with a long time horizon that provides the trajectories, and the lower which strives to follow these trajectories. This aim is to discover if there is a benefit of bilevel control compared to one single MPC controller. The main objective is to optimise the polymerisation time of the process, by allowing the temperature of the reactor and the flow rate of monomer and initiator feed to vary. However, the final product quality must remain as before. This thesis explores the possibility of controlling the process using two-level MPC control. The results are promising. This thesis shows that the implementation of two control layers is possible, though it requires extensive tuning by the engineer. Under the original conditions of the pilot plant the polymerisation time can be decreased with around five minutes using the bilevel control structure compared to the off-line calculations in the preliminary project.