| 摘 要: 针对汽油催化裂化过程中减少辛烷值损失值的问题，基于某企业催化裂化汽油精制脱硫设备的样本数据库数据，通过数据挖掘技术建立汽油精制过程中的辛烷值(RON)模型。首先，对初始数据进行规范化，然后运用随机森林法对数据变量进行降维，提取出因变量贡献程度较大的30 个主要变量；其次，利用BP神经网络，建立辛烷值损失模型；最后，在建立的模型中确定初始样本，结合遗传算法对操作变量进行优化。结果表明：优化后的辛烷值损失值下降的幅度为42.14%，降幅大于30%，有助于在实际生产中减少辛烷值损失，降低企业经济损失。 |
| 关键词: 随机森林法；汽油辛烷值；BP神经网络模型；遗传算法 |
|
中图分类号: TP183
文献标识码: A
|
|
| Establishment and Analysis of Gasoline Research Octane Number Loss Model |
|
YE Yihao, ZHONG Liangwei
|
(University of Shanghai for Science and Technology, Shanghai 200093, China)
yeyihao999@163.com; zlvcad@126.com
|
| Abstract: Aiming at the problem of reducing octane number loss in the process of gasoline catalytic cracking, this paper proposes to establish a research octane number (RON) model in the gasoline refining process through data mining technology, based on the sample database data of a company's catalytic cracking gasoline refinement and desulfurization equipment. Firstly, initial data is normalized. Then, random forest method is used to reduce the dimensionality of the data variables, and the 30 main variables that contribute to the dependent variable are extracted. Secondly, BP neural network is used to establish the RON loss model. Finally, the initial sample is determined in the model, and the operating variables are optimized in combination with genetic algorithm. Results show that the optimized RON loss value decreases by 42.14%, which is more than 30%. The proposed model helps to reduce the octane loss in actual production, so to reduce the economic losses of enterprises. |
| Keywords: random forest method; gasoline octane number; BP neural network model; genetic algorithm |
