(523d) Optimal Scheduling for Ethylene Cracking Furnace System

Industrial ethylene cracking process employs multiple cracking furnaces in parallel to convert various hydrocarbon feedstocks to smaller hydrocarbon molecules, where mostly are ethylene and propylene. Ethylene cracking operation is also a typical semi-continuous dynamic process, where coking phenomena always accompanies with cracking reactions. Because of coking, a cracking furnace has to be periodically shut down for maintance to restore its performance. Given data of multiple feed characteristics, different furnace performances, various product prices and manufacturing costs, the trade-off between the productivity and cleanup cost gives rise to the optimization opportunities to the entire cracking system for achieving the best economic performance through operation scheduling. To improve the ethylene furnace performance, it is also necessary to consider multiple products, such as propylene and ethylene productions simultaneously.

In this paper, a multi-product considered MINLP model has been developed to address a cyclic scheduling strategy for the best performance of ethylene cracking furnace system. Compared to previous studies, the new developed model has more capabilities to address operation profitability of multiple feeds cracked in multiple furnaces with multiple products. The identified schedule considers the total productivity, maintenance cost, production loss due to the unit shutdown and re-startup, and production yields of major products. In this MINLP model, the batch processing time of a feed processed in the same unit may be different during one cycle operation to maximize the profitability. Meanwhile, the scheduling model can inherently avoid simultaneous cleanups for multiple furnaces, and could maintain the downstream ethylene and propylene production ratio within a specified range, thus results in a more applicable and more stable production schedule for the plant. In this paper, the inherent disturbance due to furnace shutdown will be evaluated. The conceptual design and additional scheduling strategy to completely iron out such disturbances are also proposed. With production yield models from real plant data, a case study has demonstrated the efficacy of the developed methodology.