融合深度学习与过程机理的FCC装置关键参数软测量模型  

Soft Sensing Model Integrating Deep Learning and Process Mechanism for Key Parameters of FCC Unit

作　　者：魏彬 谭硕 周华[1] WEI Bin;TAN Shuo;ZHOU Hua(Department of Chemical and Biochemical Engineering,College of Chemistry and Chemical Engineering,Xiamen University,Xiamen 361005,China;SINOCHEM Quanzhou Petrochemical Co.,Ltd.,Quanzhou 362100,China)

机构地区：[1]厦门大学化学化工学院化学工程与生物工程系,福建厦门361005 [2]中化泉州石化有限公司,福建泉州362100

出　　处：《石油学报（石油加工）》2024年第6期1624-1634,共11页Acta Petrolei Sinica(Petroleum Processing Section)

基　　金：国家自然科学基金项目(21576228)资助。

摘　　要：产品产率作为催化裂化(FCC)装置的关键参数,构造其软测量模型对提升装置效益具有重要的现实意义,而原料与催化剂性质的缺失往往使得产率软测量模型性能迅速恶化。为此,以基于半监督学习的深度置信网络-极限学习机(DBN-ELM)算法为基础,将工艺过程机理模型与数据驱动模型集成,提出了可用于预测商业催化裂化装置产品产率的软测量混合建模方法。此外,还提出基于流程模拟的灵敏度分析-相关系数矩阵(SA-CCM)策略用于软测量模型主要输入变量的选择。结果表明,混合模型相比于数据驱动模型具有更优的模型性能,即预测精度提升43.9%、数据相关性(皮尔森系数)提升29.3%。这说明所提出的产率软测量混合建模方法使得模型的预测性能提高,能较好地适应原料与催化剂性质的变化。Constructing a soft-sensing model for product yield,a key parameter of fluid catalytic cracking(FCC),has important practical significance for improving unit efficiency.However,the lack of feedstocks and catalyst properties often leads to a rapidly deteriorating performance of the yield soft-sensing model.To this end,the hybrid soft-sensing model for predicting key product yield in commercial FCC units is built upon a deep belief network-extreme learning machine(DBN-ELM)algorithm based on semi-supervised learning,as well as the integration of process mechanism models with data-driven models.Moreover,the sensitivity analysis-correlation coefficient matrix(SA-CCM)strategy based on process simulation is utilized for key input variable selection of the hybrid model.The results show that compared with the data-driven model,the hybrid model has better model performance,with a prediction accuracy improvement of up to 43.9%and a data correlation(Pearson coefficient)improvement of up to 29.3%,demonstrating that the hybrid model can nicely capture the variation of feedstock and catalyst properties.

关 键 词：软测量 催化裂化 深度学习 产率预测 混合模型 先进控制 

分 类 号：TE624[石油与天然气工程—油气加工工程]