Maximizing margins and optimizing operational conditions for residue fluid catalytic cracking with an artificial intelligence hybrid reaction model

Abstract Because of the recent declining demand for gasoline, the key to making refineries competitive is to maximize the yields of propylene and aromatics by converting heavier feedstock into basic petrochemicals through the residue fluid catalytic cracking (RFCC) process. This study presents an artificial intelligence (AI) hybrid reaction model to optimize the catalyst make‐up rate and maximize the product yield in a real‐time operation by (1) developing a catalyst activity evaluation method, (2) integrating the catalyst to oil (Cat/Oil) ratio to evaluate the reaction performance, and (3) incorporating the yield prediction model into the latest digital technologies. To this end, the catalyst deactivation function, which uses a deep neural network of the basic machine learning method, was added to the past RFCC reaction model. Under actual operational conditions, this study shows that the AI hybrid reaction model using the catalyst deactivation function can minimize catalyst loss and produce an accurate yield prediction as a production planning support tool.