(329e) Economically Optimizing Model Predictive Control

The overall objective of a chemical process is to convert raw materials into desired products using available resources, while satisfying operational constraints. Since there are multiple ways of satisfying these objectives, the primary goal of a process plant is to maximize its profit, i.e. operate the process in the most economical way possible at all times subject to uncertainty and fluctuations in operating conditions, which are inevitable in all processes. Hence, in any modern chemical manufacturing facility, automatic feedback control becomes indispensable. Model Predictive Control (MPC) is the most widely used advanced feedback control technology in the process industries [1, 2]. Its ability to handle constraints and satisfy some optimal performance criterion, by solving an online optimization problem, is its major strength [3]. Currently, MPC is implemented as a two layer strategy, which is based on fundamental ideas on hierarchical control introduced in the 1960's [4, 5, 6]. The first layer, known as the Real Time Optimization (RTO) layer [7], solves an economic optimization problem to estimate the best operating steady state of the process subject to the given constraints. These are the set-points. The second layer, which is the dynamic optimization layer, then manipulates the inputs in such a fashion as to ?track? these setpoints.

Backx et al. [8] stresses the need for dynamic operations in process industries in an increasingly market-driven economy. Engell [9] reviews the various answers to this need, most of which focus on ways to improve the structure of the RTO layer. Hence there is an opportunity to improve the economic performance of MPC framework. The set-points are essentially the economic objectives translated into process control objectives. This translation results in loss of economic information as the dynamic regulation layer has no information about the original plant economics except for a point or a trajectory which it tries to track. Reducing this loss of information can improve the economic performance of the controller and hence this provides an opportunity to rethink the distribution of functionality between the two layers of MPC hierarchy.

This presentation introduces a new strategy for MPC to reduce this loss of economic information. This strategy uses plant economics directly in the dynamic regulation layer, avoiding the conversion of economic objective to set-points. In such a scenario the function of the RTO layer is combined with the dynamic regulation, allowing a single layer. We consider a simple process, and compare the performance of the new methodology with the traditional two layer technique, demonstrating the potential of the new strategy. First, the economics for the process are adequately modeled taking into account the operating and utility cost. Next, the performances of both the strategies are compared and the economically better performance of the new strategy is shown. The traditional two layer strategy is implemented using the above economic model in the RTO layer. In the new strategy, the dynamic regulator is made to optimize the modeled economics directly. Finally, the example is used to bring out economic modeling issues that arise when implementing the new strategy, and possible solutions are proposed.

References

[1] S. Joe Qin and Thomas A. Badgwell. A survey of industrial model predictive control technology. Control Eng. Prac., 11(7):733?764, 2003.

[2] Robert E. Young, R. Donald Bartusiak, and Robert W. Fontaine. Evolution of an industrial nonlinear model predictive controller. In James B. Rawlings, Babatunde A. Ogunnaike, and John W. Eaton, editors, Chemical Process Control?VI: Sixth International Conference on Chemical Process Control, pages 342?351, Tucson, Arizona, January 2001. AIChE Symposium Series, Volume 98, Number 326.

[3] David Q. Mayne, James B. Rawlings, Christopher V. Rao, and Pierre O. M. Scokaert. Constrained model predictive control: Stability and optimality. Automatica, 36(6):789?814, 2000.

[4] I. Lefkowitz. Multilevel Approach Applied to Control System Design. Trans. ASME, 88:2?10, 1966.

[5] MD Mesarovic, D. Macko, and Y. Takahara. Theory of hierarchical, multilevel systems. Academic Press, New York, 1970.

[6] W Findeisen, F.N. Bailey, M. Brdys, K. Malinowski, P. Wozniak, and A. Woznaik. Control and Coordination in Hierarchical Systems. John Wiley & Sons, New York, N.Y., 1980.

[7] Thomas E. Marlin and Andrew N. Hrymak. Real-time operations optimization of continuous processes. In Jeffrey C. Kantor, Carlos E. Garcia and Brice Carnahan, editors, Chemical Process Control?V, pages 156?164. CACHE, AIChE, 1997.

[8] T. Backx, O. Bosagra, and W. Marquardt. Integration of model predictive control and optimization of processes. In Advanced Control of Chemical Processes, June 2000.

[9] S. Engell. Feedback control for optimal process operation. J. Proc. Cont., 17:203?219, 2007.