石油学报(石油加工) ›› 2019, Vol. 35 ›› Issue (2): 283-288.doi: 10.3969/j.issn.1001-8719.2019.02.009

• 研究报告 • 上一篇    下一篇

虚拟光谱识别法快速测定LTAG原料与产物烃组成

李敬岩,褚小立   

  1. 中国石化 石油化工科学研究院,北京 100083
  • 收稿日期:2018-01-25 修回日期:2018-06-14 出版日期:2019-03-25 发布日期:2019-05-22
  • 通讯作者: 李敬岩,男,高级工程师,博士,从事原油及油品的快速分析工作;E-mail:lijy.ripp@sinopec.com E-mail:lijy.ripp@sinopec.com
  • 基金资助:
    中国石油化工股份有限公司项目(115056)资助

Rapid Analysis of Feedstocks and Products in LTAG Process by Near-Infrared Spectroscopy Combined with Virtual Spectral Identification Algorithm

LI Jingyan, CHU Xiaoli   

  1. Research Institute of Petroleum Processing, SINOPEC, Beijing 100083, China
  • Received:2018-01-25 Revised:2018-06-14 Online:2019-03-25 Published:2019-05-22

摘要: 为减轻近红外光谱化学计量学模型维护工作负担,基于468个LTAG(劣质LCO(轻循环油)转化为催化裂化汽油或轻质芳烃技术)原料与产物样本建立了测定其烃族组成的近红外光谱数据库。从数据库中选出一组与预测样品相似的临近光谱,采用蒙特卡洛虚拟光谱方法对数据库局部进行密化处理,采用移动窗口相似系数法对预测样品进行识别,根据与之吻合的虚拟光谱预测LTAG原料与产物样本链烷烃、环烷烃、烷基苯、茚满或四氢萘、单环芳烃、双环芳烃、三环芳烃,其预测标准偏差分别为1.5%、1.4%、0.9%、0.8%、1.3%、0.8%和0.5%。该方法准确性高于PLS模型法,无需建模,成本低,且数据库维护工作量相对较少。

关键词: 近红外光谱;劣质轻循环油转化 为催化裂化汽油或轻质芳烃技术(LTAG), 识别;数据库;化学计量学

Abstract: Over four hundred different LTAG (Light cycle oil to aromatics and gasoline) feedstocks and product samples were collected to establish the near infrared spectroscopic database for their hydrocarbon group determination. In the proposed method, a group of samples which are similar to the unknown samples, were selected from database in the first step. Then, Monte Carlo virtual spectrum method was used for local densification treatment of the database. Furthermore, the unknown samples were identified by applying Moving Window Coefficient Method. Finally, the composition of LTAG feedstock and product samples is predicted according to the virtual spectrum that coincides with them. Not only is the prediction accuracy of the Monte Carlo method higher than that of PLS method, but also the proposed method does not need modeling and model maintenance. Our results show that, by using the recognition algorithm, the corresponding standard prediction errors for alkane, cycloparaffin, alkylbenzene, indane, monocyclic aromatics, bicyclic aromatics and tricyclic aromatics are 1.5%,1.4%,0.9%,0.8%,1.3%,0.8% and 0.5%, which can satisfy the fast assessment requirement. The proposed method can save the modeling cost and also reduce database maintenance workload.

Key words: near infrared spectroscopy, light cycle oil to aromatics and gasoline(LTAG), identification, database, chemometric