(739h) A Novel Back-Off Algorithm for Integrating Dynamic Optimization and Scheduling of Batch Processes Under Uncertainty | AIChE

(739h) A Novel Back-Off Algorithm for Integrating Dynamic Optimization and Scheduling of Batch Processes Under Uncertainty


Ricardez-Sandoval, L. A. - Presenter, University of Waterloo
Valdez-Navarro, Y. I., University of Waterloo
As industries become more competitive, the need to meet highly restrictive constraints arises. Batch processes are ideal in this context, as their flexibility is an attractive property that can be adjusted depending on the volatility of the changing markets. Smarter, comprehensive manufacturing decisions have to made in order to meet the increasing demand of specialty products. This motivates a rethinking of the traditional sequential-based decision making that regularly takes place in plant operations (i.e. supply chain management, planning, scheduling and process control)[1]. Hence, an integrated decision approach has received interest from the process systems engineering (PSE) community since it has been shown to provide better quality solutions through the development of algorithms that consider interactions between layers, previously ignored by the sequential methods [2]–[4].

The proposed integrated methods are generally casted as mixed-integer dynamic optimization problems (MIDO), which may become computationally intractable for large-scale applications [5]. Moreover, many of the proposed approaches assume process model parameters to be deterministic in nature. This assumption may not hold when applied to real processes, as it has been shown that the presence of uncertainty can cause deviation from nominal plant operation [6]. Consequently, sub-optimal or infeasible manufacturing operation can be expected. Efforts from the PSE community aimed to tackle this challenging problem; however, there is still a gap in the literature for integrating methods between the scheduling and control layer that account for parameter uncertainty in batch processes.

This study aims to develop a decomposition algorithm that employs a new back-off methodology to account for stochastic model parameter uncertainty. The purpose of the novel algorithm is the integration of dynamic optimization and scheduling for multi-unit batch plants. A series of optimization problems involving scheduling and control decisions are solved. Consequently, uncertainty is propagated into the system by iteratively solving Monte Carlo-based simulations. These simulations provide sufficient statistical information, which is captured and used to determine back-off terms for the process operational constraints. At each step in the algorithm, the back-off terms are updated such that the system moves away from a nominal deterministic solution until a convergence criterion is met. Convergence of the algorithm results in an optimal schedule and control profiles that remain dynamically feasible, up to a user-defined criterion, in the presence of stochastic-based uncertainty. This algorithm has been successfully applied to a multi-product multi-unit batch plant and compared in performance against a fully integrating multi-scenario-based uncertain MIDO, casted as an MINLP. Preliminary results show the proposed decomposition algorithm is, when compared against the integrating MINLP, computationally tractable. This is achieved without compromising on solution quality. Moreover, the back-off approach has been shown to account for a greater range of uncertainty while still maintaining its computational performance.


[1] I. Grossmann, “Enterprise-wide optimization: A new frontier in process systems engineering,” AIChE J., vol. 51, no. 7, pp. 1846–1857, 2005.

[2] R. W. Koller and L. A. Ricardez-Sandoval, “A Dynamic Optimization Framework for Integration of Design, Control and Scheduling of Multi-product Chemical Processes under Disturbance and Uncertainty,” Comput. Chem. Eng., 2017.

[3] B. Burnak, N. A. Diangelakis, J. Katz, and E. N. Pistikopoulos, “Integrated process design, scheduling, and control using multiparametric programming,” Comput. Chem. Eng., 2019.

[4] M. Baldea, J. Du, J. Park, and I. Harjunkoski, “Integrated production scheduling and model predictive control of continuous processes,” AIChE J., vol. 61, no. 12, pp. 4179–4190, Dec. 2015.

[5] L. S. Dias and M. G. Ierapetritou, “From process control to supply chain management: An overview of integrated decision making strategies,” Comput. Chem. Eng., vol. 106, pp. 826–835, 2017.

[6] B. P. Patil, E. Maia, and L. A. Ricardez-Sandoval, “Integration of scheduling, design, and control of multiproduct chemical processes under uncertainty,” AIChE J., vol. 61, no. 8, pp. 2456–2470, Aug. 2015.