Global optimization for the cyclic scheduling and operation of multistage continuous plants

This work addresses the global optimization of the simultaneous problem of the cyclic scheduling and operation of multistage continuous plants. In this problem, production rates and yields are additional optimization variables for plant scheduling. The representation proposed for this problem is a mixed-integer nonlinear programming (MINLP) model that has a nonconvex feasible region and a nonconvex objective function. To address nonconvexity, a spatial branch-and-bound global optimization algorithm is developed to solve the model. An illustrative example shows that the global approach is effectively able to yield a more profitable solution than a local optimization algorithm. Moreover, it is shown that modifications in the steps of the global optimization algorithm, such as preferential branching at a variable, can significantly improve its performance. Results also show that local optimization can provide very good estimates for the global solution when processing conditions have narrow variability ranges and plants operate at nearly full capacity.