Optimizing product distribution in the heavy oil catalytic cracking (MIP) process

ABSTRACT

The artificial neural network provides an effective way to handle a non-linear and strong coupled reaction system because of its strong nonlinear prediction and self-learning ability. A 19–24-4 type of back propagation (BP) neural network that can predict the product distribution of a fluid catalytic cracking maximizing iso-paraffin unit was established using 19 input variables including properties of feedstock and regenerated catalyst and operating variables. The influences of the operating variables on product distribution were simulated, and the operating variables were optimized to maximize gasoline (GS) yield by a genetic algorithm. The predicting results agreed well with the industrial data, and a significant improvement in the GS yield was gained.

KEYWORDS:

• BP neural network
• FCC
• genetic algorithm
• MIP process
• simulation