Iterative Unit-based Adaptive Dynamic Programming with Application to Fluid Catalytic Cracker Unit

Model predictive control(MPC) poses an significant role in process industry. However, due to the characteristics of difficult to solve and poor control performance of nonlinear MPC with disturbances, this paper proposes an iterative unit-based adaptive dynamic programming (IU-ADP) aiming to solve the nonlinear MPC problem with disturbances. In the IU-ADP structure, an actor-critic scheme is adopted together with the iterative method, in which the critic network intends to approximate the performance index, the actor network with an iterative method aims to approximate the optimal control. Moreover, IU-ADP with data parallelization can achieve fast time efficiency. Finally, it is tested on the fluid catalytic cracker unit (FCCU) plant compared with the traditional MPC method. The results illustrate that IU-ADP obtains a little better control performance.