(372c) Integration Of Cyclic Scheduling And Dynamic Optimization For Parallel Units Operation With Decaying Performance

Cyclic scheduling could be employed to efficiently allocate resources (feeds, units, and time) for handling multiple feeds processed on parallel units to maximize the manufacturing productivity. It aims at production optimality in the production management level, where some important information such as batch processing time, production yields, and batch size is assumed as constant. Operational optimization, on the process level, targets maximum economic performance of the manufacturing process by manipulating the setpoint for each unit. If parallel units, featuring decaying performance due to coking, fouling, catalyst deactivation or etc., have been employed in manufacturing, operational setpoints should be optimally adjusted with respect to time during each sub-cycle, which creates dynamic optimization problems in a shorter time scale. Note that the batch processing time and unit production yields may not exactly match between cyclic scheduling and unit operational optimization. Thus, the integration of these two techniques for better manufacturing performance should be investigated.

This paper introduces a general methodology to simultaneously consider the production scheduling and operational dynamic optimization for parallel units operation with decaying performance. It presents bi-level iteration scenarios. In the upper level, cyclic scheduling for multiple feeds processed by parallel units is conducted based on assumed average production yields. This task is accomplished by solving a mixed-integer nonlinear programming (MINLP) model. The identified schedule is the trade-off between the total productivity, maintenance cost, and the production loss due to the unit shutdown and re-startup. In this MINLP model, the batch processing time of a feed processed in the same unit may be different during one cycle operation. Meanwhile, two or more units can not be shut down simultaneously for the sake of production continuity and stability.

In the lower level, the information of batch processing time from upper level will be utilized to generate the optimal setpoints for each unit operation. Since the units have decaying performance, dynamic operational model should be employed, which is solved by a sequential quadratic programming (SQP) method. The average unit production yields obtained from lower level will be utilized to update the upper level assumptions. Iteratively, the bi-level optimization will result in the optimal schedule and operational setpoints, simultaneously.

In general, the methodology integrates cyclic scheduling and dynamic optimization for better performance of an important type of process, where parallel units with decaying performance are employed. The efficacy of this methodology is demonstrated by solving the scheduling and dynamic operational problem for an ethylene cracking process.