A production optimization problem for continuously operated processes is presented and a solution strategy is proposed. The strategy consists of decoupling the complete problem into a mixed-integer linear programming (MILP) scheduling problem, including sequencing and allocation, and a multi-stage dynamic optimization (DO) problem, including the determination of optimal trajectories and setpoints. The scheduling problem is treated as a master problem and the DO problem as a primal problem, and the complete problem is solved through iteration between the two. The approach is similar in nature to standard methods for solving mixed-integer nonlinear (MINLP) problems, such as Outer Approximation and Benders Decomposition, but more adapted to the specific problem, by permitting more freedom in choosing the binary representation. The decomposition strategy implies flexibility in choosing the optimization tools required, and enables the treatment of larger problems. The approach is a generalization of a previously reported one, where only the single-unit case was discussed. Splitting the primal problem into smaller DO subproblems, parameter estimation from the DO subproblems, termination criteria and other topics are discussed. The target process for demonstration of the method is an industrial polymerization process.
- Dynamic model-based scheduling
- Multi-stage optimization
- Multiple-unit production optimization
- Optimal manufacturing
- Product sequencing