ChemE Spring Seminar Series

Dynamics and Control of Integrated Networks: A Multi-Scale Perspective

Prodromos Daoutidis
Department of Chemical Engineering & Materials Science
University of Minnesota

Friday, April 18, 2008
3:00 p.m., 66-110
Refreshments will be served at 2:45 p.m.

Tight integration is the rule rather than the exception in modern plants. Integration leads to significant reduction in capital and operating costs, but also results in complex designs which are difficult to operate and control due to the strong coupling induced by the feedback interconnections in the plant.

Extensive research has focused on steady state design and optimization of integrated networks over the last two or three decades. In the area of control past research has focused mostly on linear decentralized control, with very few attempts to compensate systematically for the core nonlinear dynamics induced by integration. Yet, the efficient transient operation of such networks is becoming increasingly important, as the current economic environment dictates frequent changes in operating states and a tight coordination between the optimization and supervisory control levels. This in turn dictates going beyond the linear regulatory control paradigm and developing strategies that compensate for the core dynamics of integrated networks and enable effective transitions between different operating states.

In this talk, recent results are presented on the dynamic analysis and control of broad classes of networks with material and energy integration. It is shown that tight integration, achieved through large material and / or energy recycle, leads to multi-time-scale dynamics, with individual units evolving in a fast time scale with weak connections, which become significant over slower time scales giving rise to a slow evolution of the entire network. A class of (non-standard) singularly perturbed systems is identified as a common underlying model structure of such integrated networks, and a model reduction framework is described which enables obtaining a hierarchy of low-order nonlinear  models valid in the different time scales. The analysis lends itself naturally to a hierarchical control framework, which reconciles systematically regulatory and supervisory control objectives, and allows for the development of robust nonlinear supervisory control strategies for effective network transitions.

The singular perturbation modeling and analysis framework developed can also be used as the basis for nonlinear model reduction in chemical and biological reaction networks with fast and slow reactions. Recent results on these problems will also be briefly discussed.