Analysis of Optimization Algorithms for Real-Time Optimization Applied on the Model of a Fluid Catalytic Cracking Unit

Abstract

Abstract On-line optimization of Fluid Catalytic Cracking Units (FCCU) is essential in modern refineries, because of high commercial value of cracking products. Within Real-Time Optimization (RTO) the on-line optimization module plays a fundamental role since operational conditions must be frequently adjusted. The performance of this module must guarantee good solutions in a sufficiently short time to allow on-line implementation. This work presents an assessment of some deterministic and stochastic optimization algorithms into a three-layer RTO strategy for a virtual FCCU. The implementation begins with definition of an economic objective function, a sensitivity analysis to evaluate how profit is affected by variations on operational variables and concludes with a performance comparison of some algorithms. Results from deterministic algorithms confirm that there are multiple local solutions in FCCU optimal operation problem, and deterministic search methods could be trapped in these solutions. On the other hand, GA allowed to obtain satisfactory solution, but with the default values for stopping parameters, it doesn´t generate the solution in the available time for an RTO implementation. As alternative to obtain good solution in a reasonable time, some GA parameters were adjusted and a hybrid strategy was assessed. With these approaches, execution time was satisfactorily reduced, and the hybrid strategy was confirmed as an interesting option for optimization modules in on-line applications.