An industrially relevant formulation of a distributed model predictive control algorithm based on minimal process information

•A novel formulation for DMPC architecture with input–output model.•Industrial considerations given through complexity simplifications in the design.•Robustness to modelling errors caused by main dynamics approximations. Plant-wide control implies advanced supervisory algorithms to maintain desired...

Full description

Saved in:
Bibliographic Details
Published in:Journal of process control Vol. 68; pp. 240 - 253
Main Authors: Maxim, Anca, Copot, Dana, De Keyser, Robin, Ionescu, Clara M.
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.08.2018
Subjects:
Information exchange
Interacting sub-systems
Model predictive control
Model uncertainties
Robustness
Model predictive control
Interacting sub-systems
Robustness
Model uncertainties
Information exchange
ISSN:0959-1524, 1873-2771
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•A novel formulation for DMPC architecture with input–output model.•Industrial considerations given through complexity simplifications in the design.•Robustness to modelling errors caused by main dynamics approximations. Plant-wide control implies advanced supervisory algorithms to maintain desired performance in the involved coupled sub-systems. The dynamical interactions among these sub-systems can vary with the operating point, material properties and disturbances present in the process. Recirculating loops introduce additional phenomena in the dynamic response, further challenging the control tasks. Complex process dynamics may be linear parameter varying (LPV) and may be difficult, if not impossible, to identify properly. In this context, maintaining global performance is a challenge one must undertake with limited information at hand. This paper investigates the trade-off between the complexity of the implementation and achieved performance, using supervisory predictive control with limited information shared, applied on a test-bench representative for process control industry. The robustness of the proposed algorithms is tested against a nominal scenario in which the prediction model is fully identified, with complete information exchange. Experimental tests are performed on a test-bench process characterized by strong interactions, and the results illustrate the usefulness of this work.
ISSN:0959-1524
1873-2771
DOI:10.1016/j.jprocont.2018.06.004