Distributed control of process networks : a dissipativity based approach

Download files
Access & Terms of Use
open access
Copyright: Tippett, Michael James
Altmetric
Abstract
Process networks are characterised by their large-scale and strong interaction effects. As such, a scalable control design approach, which can account for interactions, is required. This thesis addresses the problem of control design for process networks by developing a distributed approach, whereby local controllers which communicate with one another are used. This approach has the ability to address both the issues of scalability and interactions. The system is viewed as two interacting networks, a process network and controller network. Dissipative systems theory is adopted to facilitate a two step design approach. In the first step, the required dissipativity of the controllers is determined, such that the process network is stable, and satisfies specified H-infinity performance bounds. These bounds are determined from the dissipativity of the closed-loop process network. In the second step, the controllers are synthesised. Design approaches for both closed-form controllers, and MPC are developed. In both cases, once the first step is completed, the controllers may be synthesised independently in the second step. This approach is computationally tractable as, depending on the specific case, the calculations require either the solution of linear matrix inequalities or algebraic Riccati equations. Quadratic differential/difference forms are used as supply rates in the dissipativity based analysis. These capture more system information than traditional supply rates, thus providing sharper stability results. They also allow for weighted H-infinity norm bounds to be represented as closed-loop dissipativity conditions. The quadratic differential/difference form framework provides a convenient algebra for manipulations of the process and controller supply rates through polynomial matrix operations. In addition to the framework for closed-form and MPC design, several important extensions are developed. Including: multi-rate control, design for bounded unknown time delays, design for uncertainty, and design for changing network topology. Multi-rate control may facilitate decreased capital costs. In addition, there is a reduction in controller communication network traffic. Design for delays, model uncertainty and changing network topology provides robustness to changes in operating conditions, model error and failures of the controller communication network. In the case that the changing operating conditions are known, the controllers reconfigure to compensate for these changes.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
Tippett, Michael James
Supervisor(s)
Bao, Jie
Ydstie, B Erik
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2014
Resource Type
Thesis
Degree Type
PhD Doctorate
UNSW Faculty
Files
download public copy.pdf 1.62 MB Adobe Portable Document Format
Related dataset(s)