汽油精制过程中的路径优化模型及仿真  

Path Optimization Model and Simulation in Gasoline Refining Process

作　　者：金晶亮[1,2] 温晴岚 张霰月 李晨宇 Jin Jingliang;Wen Qinglan;Zhang Xianyue;Li Chenyu(College of Science,Nantong University,Nantong,226019,China;College of Economics and Management,Nanjing University of Aeronautics and Astronautics,Nanjing,211106,China)

机构地区：[1]南通大学理学院,江苏南通226019 [2]南京航空航天大学经济与管理学院,江苏南京211106

出　　处：《石油化工自动化》2022年第1期6-14,共9页Automation in Petro-chemical Industry

基　　金：国家自然科学基金项目(71603135,71774080)。

摘　　要：根据汽油产品的工业生产实测数据,从数据挖掘和运筹优化等多角度探究汽油清洁化的工艺改进路径。首先,对原始数据和样本数据进行预处理,以保证后续模型的质量;其次,为了降低模型的复杂性,通过梯度下降回归树筛选出预测模型的主要变量;再次,考虑到操作变量之间存在高度非线性和相互强耦联的关系,通过BP神经网络分别建立主要变量对辛烷值损失以及产品硫含量的预测模型;最后,建立主要操作变量优化的两阶段优化模型,并通过可视化输出操作变量的逐步调整效果,供工业生产参考。Based on the measured data of industrial production of gasoline products, the process optimization path for gasoline cleaning production is explored from multiple perspectives such as data mining and planning and management optimization. First, the original data and sample data are pretreated to guarantee the quality of the subsequent model. Secondly, to reduce the complexity of the model, the main variables of the prediction model are screened out through the gradient descent regression tree. Thirdly, considering the highly nonlinear and strongly coupled relationship among operating variables, BP neural network is used to establish the prediction models of main variables on octane number loss and product sulfur content respectively. Finally, a two-stage optimization model of operation optimization of major variables is established, and the effect of gradual adjustment of out-put operation variables is visualized to provide reference for industrial production.

关 键 词：汽油清洁化 梯度下降回归树 BP神经网络 两阶段优化 

分 类 号：N945.1[自然科学总论—系统科学]