关键词: 催化裂化, 焦炭产率, 神经网络, 遗传算法

Abstract: Coke is the main by-product of fluid catalytic cracking (FCC) process. It is of great significance to predict coke yield accurately to enhance stability and economic performance of FCC plant. Artificial neural network (ANN) has a strong self-learning and adaptive ability,and has obvious advantages in nonlinear forecasting. In this paper,a new model combining BP neural network and genetic algorithm (GA) was developed to predict coke yield by choosing 28 key parameters involving feedstock properties,catalyst properties and operating conditions of industrial data of FCC unit,The prediction results obtained from BP neural network and the genetic algorithm optimized BP neural network (GA-BP) were compared. The GA-BP model had a better result in both accuracy and stability. Furthermore,the influence of key parameters,such as reaction temperature,feedstock carbon residue on coke yield was investigated,which further proved the accuracy of BP neural network model optimized by genetic algorithm.

Key words: fluid catalytic cracking (FCC), coke yield, neural networks, genetic algorithm