Online integration of scheduling and control for cyclic production in CSTR

Scheduling and control problems are traditionally solved sequentially. However, integration of both problems can result in a better overall performance. A main challenge in the integration is the solution to the derived mixed-integer dynamic optimization (MIDO) problem. As a consequence, most integration methods can only be applied offline. To overcome the challenge, we present a novel integration method, which simultaneously determines the PI controller design and the scheduling decisions. The method decomposes the MIDO problem. The dynamic optimization for each transition can be solved independently offline. A set of controller candidates are generated and stored. Then the integrated problem is transformed into a mixed-integer nonlinear fractional programming problem of scheduling with controller selection. An efficient global optimization method based on the Dinkelbach's algorithm is developed, which enables the online implementation. The numerical study on a cyclic production model in a CSTR shows that the proposed integration method reduces the cost rate by 19.7% comparing with the traditional sequential method.