(136c) Integration of Planning, Scheduling and Control Using Feasibility Analysis and Surrogate Models

Optimal decision-making in a process industry is fundamental in order to improve performance and guarantee efficient operation in highly competitive and global markets. Process operations decision-making problems range across different scales, and include control, operational, tactical and strategic decisions. Traditionally, these decisions are considered in a hierarchical fashion and addressed individually by different areas within an organization. However, remarkable benefits can be achieved though the proper integration of the decision-making process. The academic community has made significant progress towards this goal [1, 2]. Nevertheless, the available methodologies and tools do not fully meet industrial needs, and more practical and efficient methods to coordinate the different levels of decisions are needed.

In this work, a novel framework for the integration of planning, scheduling and control problems is presented. The first step is to design the control level simulation of the system in consideration. The simulation consists of iterations between the solution of a Model Predictive Control problem and the implementation of control actions in a detailed dynamic model of the system. The systems of interest commonly involve highly nonlinear models, and the sample time of the MPC problem is usually in the order of minutes or seconds. Therefore, transmitting the closed-loop behavior of the simulation in a time scale that is relevant to the scheduling problem is challenging, and it is achieved through the use of surrogate models. Once surrogates are built to effectively capture the closed-loop behavior of the control-level simulations, they can be incorporated as constraints in the scheduling problem.

To enable the complete vertical integration of the decision making process, we then investigate how the scheduling problem can be posed as a feasibility problem [3, 4], where feasible production targets are defined as a function of the initial state of the system, the scheduling horizon, and the given network. By reformulating the scheduling problem as a feasibility problem, surrogate models and classification methods can be used to define the feasible space, following the work by Wang and Ierapetritou [5]. We discuss how the feasible space of the integrated scheduling and control problem is expected to change when compared to the individual scheduling problem. An explicit function that provides the feasible production targets and associated production costs is obtained, and the explicit models can be incorporated as constraints in the planning problem.

Finally, the planning problem is posed as a nonlinear problem and solved in a moving-horizon fashion in order to address uncertainties and disruptions associated to the scheduling and control levels. Several case studies are presented to illustrate the proposed approach and provide comparisons with the hierarchical decision-making process.

1. Dias, L.S. and M.G. Ierapetritou, From process control to supply chain management: An overview of integrated decision making strategies. Computers & Chemical Engineering, 2017. 106(Supplement C): p. 826-835.
2. Grossmann, I., Enterprise‐wide optimization: A new frontier in process systems engineering. AIChE Journal, 2005. 51(7): p. 1846-1857.
3. Sung, C. and C.T. Maravelias, An attainable region approach for production planning of multiproduct processes. AIChE Journal, 2007. 53(5): p. 1298-1315.
4. Sung, C. and C.T. Maravelias, A projection-based method for production planning of multiproduct facilities. AIChE Journal, 2009. 55(10): p. 2614-2630.
5. Wang, Z. and M. Ierapetritou, A novel feasibility analysis method for black-box processes using a radial basis function adaptive sampling approach. AIChE Journal, 2017. 63(2): p. 532-550.