Distributed economic model predictive control with pseudo-steady state modifier adaptation for an industrial fluid catalytic cracking unit

Abstract

The fluid catalytic cracking (FCC) unit is a typical multi-timescale system. Its operational performance will make a significant difference in the overall economic benefits of the oil refining factory. Classical model-based optimization approaches may not always be completely satisfied due to unmeasurable disturbances and model-plant structure mismatch. A distributed economic MPC with pseudo-steady state modifier adaptation (DEMPC-MA) is proposed to cope with the difficulties. The strategy divides the plant into two parts, i.e., the slow dynamic part and pseudo-steady state part. The output modifier adaption is adopted to modify the model function according to the pseudo-steady state characteristics of the riser. Lyapunov stable constraints are added to the EMPC problem to guarantee a safe operation, while the feasibility of the algorithm is ensured by a set of auxiliary controllers, which are constructed offline based on the multi-model approach. The overall control optimization problem is solved in a distributed formulation through the alternating direction method of multipliers (ADMM) to improve the calculation performance. The proposed algorithm was applied to an industrial FCC unit model based on a refinery factory in Jiujiang, China. The results showed that the method obtained better economic benefits and finally reached the optimal steady-state operating point even if the model structure and parameter of the prediction model were mismatched.