MODELING, OPTIMIZATION AND CONTROL OF A FCC UNIT USING NEURAL NETWORKS AND EVOLUTIONARY METHODS

This paper presents a simulation study of the use of an artificial neural network (ANN) model for control and optimization of a Fluidized-Bed Catalytic Cracking reactor-regenerator system (FCC). This case study, whose phenomenological model was validated with industrial data, is a multivariable and nonlinear process with strong interactions among the operational variables. In order to obtain a dynamic model of the FCC system, a feedforward ANN model was identified. Genetic Algorithm (GA), and Particle Swarm Optimization (PSO) evolutionary methods were used to set optimal operating conditions for the FCC, and both algorithms presented good and consistent results for typical FCC optimization problems. The neural model was also used in the design of a Model-Based Predictive Control (MPC) for the FCC process. It was showed that the ANN-based MPC was able to reject the imposed disturbance as well as to track the proposed trajectory, while considering operational constraints of the plant.