催化裂化气体烃收率神经网络模型

Establishing Yields Model of Gas Hydrocarbons for Catalytic Cracking by Using Artificial Neural Networks

在小型固定流化率反应装置上进行蜡油与渣油的催化裂化反应实验。考虑人工神经网络在处理复杂系统的建模问题上具有优越性，经过网络结构和学习样本的确定、网络的训练、模型预测能力的考察，利用神经网络建立蜡油与渣油催化裂化气体烃产率模型，该模型直接将气体烃产率与原料油的结构族组成和物性，气体＋ 焦炭产率，汽油产率进行了关联。研究结果表明该模型具有较好的计算精度和满意的预测能力。

Based on the advantage of artificial neural networks for establishing the mathematical model for complex system, this paper is to develop yield models of gas hydrocarbons for catalytic cracking by using artificial neural networks (ANN) through the determination of networks structure and study samples, networks training and model prediction. The cracking reactions of gas oils and residue oils were carried out on commercial equilibrium zeolite catalysts in a fixed fluid bed reactor. The model correlated directly gas hydrocarbon yield with the structure group composition of gas oil and residue oil, gasoline yield, gas plus coke yield. The results point out that the yield models of gas hydrocarbon for catalytic cracking have high calculation precision and perfect prediction ability.