Closed-loop Scheduling with Process Faults: Framwork and an Air Separation Unit Example

The strong increase of the contribution of renewable sources (wind, solar) to the power generation portfolio has increased uncertainty in the operation of the grid and has motivated efforts in demand-side management. Of particular interest is the demand response (DR) operation of large industrial electricity consumers, which is encouraged via (real-)time-varying electricity pricing schemes. For industrial users, taking advantage of such price structures calls for far more frequent changes in the production schedule than ever before, and requires a close coordination between scheduling decisions, the actions of the control system, and the dynamic capabilities of a plant. Crucial to this end are, i) embedding representations of the process dynamics in the scheduling calculations and, ii) developing a framework for recomputing the scheduling decisions upon the occurrence of process events and faults. In this work, we address the latter challenge by proposing a novel closed-loop, moving-horizon approach to production scheduling, whereby process dynamics are represented using previously-developed scale-bridging models, and schedule updates are triggered by both scheduling-relevant market events and by process events that do not have an explicit impact on the schedule execution. We use the DR operation of an industrial-scale air separation unit model to demonstrate the successful implementation of these concepts.