(8g) Upset-Conscious Scheduling of Cracking Furnace System for Olefin Plants

Facing the challenge of volatile raw material and product market, modern olefin plants employ multi-type cracking furnaces in parallel to convert various hydrocarbon feeds into products such as ethylene and propylene. However, different feeds have different yields for different products. For example, the ethylene yield is above 50% and the propylene yield is just 3% when the feed is ethane. While, the ethylene and propylene yield are about 35% and 15% respectively when the feed is naphtha. Thus, the final yield of each product will change significantly in the total cracked gas which is further processed in the downstream units, if different feeds are cracked for a feed flexible plant. The cracking process of each furnace is a performance-decaying batch operation. The ethylene yield decreases and propylene yield increases with processing time increasing. This yield variation makes the final product yield in the cracked gas unstable and generates inherently frequent upsets to the downstream process. Moreover, the furnace needs periodic shutdowns for decoking operations or maintenance due to the deposition of coke on the tube internal surface. This kind of shutdowns further aggravates the variation of each product yield in the cracked gas. Therefore, the furnace feed allocation, batch processing time and decoking sequence of the entire furnace system must be optimally scheduled to maximize the plant profitability; meanwhile, the induced upset of the main product yields also needs to be restricted within an appropriate range at any time for the sake of the plant operability. Facing this challenge, a new MILP (mixed-integer linear programming) model has been developed in this paper for the upset-conscious scheduling of cracking furnace systems. The model can restrict the final product yields within a certain range according to the plant’s request. The study can not only benefit both profitability and operability for the entire olefin plant operation, but also be applied to other continuous parallel-process and performance-decaying unit systems. The efficacy of the developed scheduling model has been demonstrated by various case studies. Optimal furnace schedules are solved under different product yield constraints. The chart of profit loss versus product yield variation can be used by plant decision makers to fine tune their production schedules.