Abstracts IFAC 2017 - Hierarchical MPC of Batch Reactors with Shared Resources

Published on: 
Tuesday, 22 August, 2017

S. Maxeiner, S. Engell, Hierarchical MPC of Batch Reactors with Shared Resources

Keywords: Distributed nonlinear control

Abstract: Multi-reactor semi-batch plants are widespread in the chemical and pharmaceutical industry, since they can produce more flexibly and depending on the product also more profitably than continuous plants. Such reactors usually share resources that are constrained, as for instance raw materials or cooling and heating media. In order to maximize the productivity while respecting product quality constraints, equipment limitations, as well as constraints on the utilization of the shared resources, the feeding policies and temperature profiles in such semi-batch reactors are increasingly optimized online. This can be done by a plant-wide optimizing controller, however, due to robustness, flexibility, and reduced computational effort, distributed schemes that solve the combined trajectory optimization and resource assignment problem are of high interest. In this contribution, we present a hierarchical model predictive control scheme that computes cost and resource optimal trajectories online for a set of semi-batch reactors. Each reactor maximizes its profit function locally by maximizing the amount of product while incurring a cost for the use of the shared resources. On the coordinator level, the future availability of the shared resources is taken into account and the prices are adjusted iteratively, such that the feasibility of the joint operation of all reactors is guaranteed. We show that the problem can be solved using the alternating direction method of multipliers (ADMM), modified for inequality constraints, and compare the performance of this scheme to a decentralized one. Furthermore, the need to coordinate over the whole prediction horizon is discussed and a reduced coordination horizon and its selection are investigated.