Optimal Control Of Fluidized Catalytic Cracking Unit Using Artificial Neural Networks

摘要

The Fluidized Catalytic Cracking Unit (FCCU) is one of the complex process unit in a petroleum industry, with large catalytic reaction, catalyst regeneration, and fractionation and separation equipments. The FCCU cracks heavier hydrocarbons into lighter and more valuable components. Most of the economic gains from FCCU control development have come from optimization level. In this work Actor-Critic Reinforcement Learning based expert optimization control is used to maximize the final yield. As the change in controlled variables affects the final conversion rate, the Actor-Critic (Neural network blocks) are developed to alter the set point of those variables in order to move the process in a profitable region even in the presence of disturbances like feed density and reactor pressure. The Actor-Critic Neural network blocks have been developed for individual as well as integrated loops. In the integrated form set points of the controlled variables are varied simultaneously to the desired level.