Model predictive control of a fluid catalytic cracking unit

Abstract

The dynamic model describing the FCC process is inspired from Lee and Groves model. The model uses three lump kinetics to describe the cracking reactions in the riser considered as a plug-flow system. The riser behavior is described by gasoil, gasoline, coke and energy balances as a plug flow reactor. The separator is considered as a CSTR. The regenerator mainly constituted by a dense and a dilute zone is also considered as a CSTR to represent the catalyst and its dynamic behavior is described by coke, oxygen and energy balances. This model is sufficiently complex to capture the major dynamic effects that occur in a FCC and to control the key variables which are the riser outlet temperature and the regenerator dense bed temperature. The manipulated inputs are the air inlet flow rate in the regenerator and the regenerated catalyst flow rate. Hard constraints are imposed with respect to the manipulated variables. In spite of the important nonlinearity of the FCC, Model Predictive Control is able to maintain a smooth multivariable control of the plant, while taking into account the constraints.