Adaptive optimal output regulation for industrial hydrocracking process

Hydrocracking is an important petrochemical process, in which the reactor temperature determines the final product distribution and quality. In this paper, a new data-driven optimal reactor temperature control method is proposed through the integration of reinforcement learning, adaptive dynamic programming (ADP) and output regulation theory. Different from the existing literature, the reactor temperature control problem is formulated as an output regulation problem, and a policy iteration (PI) based ADP algorithm is employed to find the adaptive optimal controller. The simulation results show that the actual nitrogen content in the stream of R101 outlet can be regulated to the desired value and keep the reaction temperature of each bed in R101 to a minimum while disturbance rejection is achieved.