(373a) On the Design and Solution of an Online Scheduling Optimization for Open-Loop and Closed-Loop Strategies | AIChE

(373a) On the Design and Solution of an Online Scheduling Optimization for Open-Loop and Closed-Loop Strategies


Franzoi, R. E. - Presenter, University of São Paulo
Menezes, B. C., Hamad Bin Khalifa University, Qatar Foundation
Kelly, J. D., Industrial Algorithms
Gut, J. A. W., University of São Paulo
Although real-world applications in industry are highly dynamic environments in which data is uncertain and change over time, the online updates of information are usually neglected since they are hard to be synchronized in the models and to be accounted for algorithmically in a systematic way [1]. However, recent advances in flowsheet formulation [2] as well as solving algorithms and computer-aided resources [3] allow a more efficient modeling and optimization. In such a way, this work evaluates the design and solution of an online scheduling optimization considering a) uncertain data from disturbances in a chemical process; b) rescheduling cycles to update the state of the system and to properly handle the disturbances; and c) different features of open-loop and closed-loop solutions, the former without intimate integration between the systematic solutions and the latter generated by sequentially integrating the rescheduling of the problem.

To this date, the scheduling optimization problem is typically performed with delayed or non-integrated process data in the models which may lead to inconsistencies in the prescriptions of logistics and quality decision-making and operations. However, the recent expansion of information and communication technologies in industry may allow data measurement of a complete plant network in a few minutes [4]. The collected data can be reconciled for consistency as well as the needed data can be estimated to provide the capabilities for design and solution of online scheduling cycles by mainly considering process feedback that reduces the deviation between the model results and the actual plant values [5].

The traditional scheduling optimization considers feedforward or open-loop strategies. When new data is integrated to the scheduling modeling and optimization, there is a need of re-optimizing the problem (rescheduling) in order to take advantage of this new information [6] by using feedback strategies. Moreover, if rescheduling is performed on a regular basis it will lead to a rolling horizon approach in which sequential open-loop solutions are implemented to enable the closed-loop solution, whereby process feedback on parameters, variables and gradients may continually improve new scheduling determinations as well. This can be part of a systematic method running the scheduling optimization every hour based on timely measurements (online scheduling) to account for disturbances in the operating or production system such as a failure of a unit or a disruption in any link of the production-chain.

Kelly and Zyngier [7] introduced parameter feedback to a chemical production system in which the gap between the model and the actual inventory of the tank can be reduced to zero by coupling online measurements to the modeling. According to the authors, the lack of feedback imposes difficulties to discern over inconsistent data in the modeling and inadequate implementation of the optimization as the cause of the differences between the scheduled (prescribed) and the actual values of the process. Franzoi et al. [3] recently addressed a complete crude-oil refinery blend scheduling problem applying both data reconciliation techniques and parameter feedback to effectively optimize the complex process system. The authors solved the feedback strategy within an iterative mixed-integer linear and nonlinear programming (MILP-NLP) decomposition and highlighted the importance of properly integrating data to the decision automation core to cope with uncertainties, to reduce inaccuracies and to close the gap between predictions/prescriptions and productions.

In this work three main topics are discussed: a) uncertainties/disturbances in the process; b) online scheduling; and c) differences and features between open-loop and closed-loop strategies. We address the design of an online scheduling algorithm for a chemical production problem based on a dynamic rescheduling approach. The problem has been first modeled and optimized using an open-loop strategy; afterwards, rescheduling has been continuously applied in order to build a closed-loop solution. The optimization is carried out within an iterative mixed-integer linear and nonlinear programming (MILP-NLP) decomposition by updating NLP results of process-shop’s yields and properties, and blend-shops’ recipes in the next MILP solution until convergence is achieved. A parameter feedback strategy is used to handle (randomly simulated) uncertainties in the process and the solutions for both open-loop and closed-loop problems are obtained. This online environment might be integrated within the scheduling cycle and can effectively cope with the proposed disturbances, quickly generating a new solution whenever needed.

[1] Larsen R, Pranzo M. (2018). A framework for dynamic rescheduling problems. International Journal of Production Research, 1-18.

[2] Brunaud B, Amaran S, Bury S, Wassick J, Grossmann IE. (2019). Batch scheduling with quality-based changeovers. Computer and Chemical Engineering. Just Accepted.

[3] Franzoi RE, Menezes BC, Kelly JD, Gut JW. (2018). Effective Scheduling of Complex Process-shops using Online Parameter Feedback in Crude-Oil Refineries. Computer Aided Chemical Engineering, 44, 1279-1284.

[4] Menezes BC, Kelly JD, Leal AG, Le Roux GC. (2019). Predictive, prescriptive and detective analytics for smart manufacturing in the information age. In 12th IFAC Symposium on Dynamics and Control of Process Systems (DYCOPS), Florianópolis, Brazil, Apr 23-26.

[5] Menezes BC, Kelly JD, Leal AG. (2019). Identification and Design of Industry 4.0 Opportunities in Manufacturing: Examples from Mature Industries to Laboratory Level Systems. In 9th IFAC Conference on Management, Modeling and Control Systems (MIM), Berlin, Germany, Aug 28-30.

[6] Gupta D, Maravelias CT. (2016). On Deterministic Online Scheduling: Major Considerations, Paradoxes and Remedies. Computers & Chemical Engineering, 94, 312-330.

[7] Kelly JD, Zyngier D. (2008). Continuously improve the performance of planning and scheduling models with parameter feedback. In: Proceedings of the foundations of computer-aided process operations (FOCAPO).