Prediction of research octane number loss and sulfur content in gasoline refining using machine learning

Fengyu Zhang, Xinchao Su, Aoli Tan, Jingjing Yao and Haipu Li

Energy, 2022, vol. 261, issue PA

Abstract: In this study, the developed machine learning (ML) model elaborated the highly non-linear and coupling relationship using maximal information coefficients, and 35 important variables were filtered out from 353 variables for modeling. The dragonfly algorithm was successfully applied to optimize the back propagation neural network and logistics regression process, and the combined model balanced the local searching and global searching. The evaluation indicators of training and test sets (0.9731 and 0.9622 of the squared correlation coefficient, 0.0241 and 0.0413 of mean square error, and 0.0982 and 0.1505 of mean absolute error, respectively) and cross-validation of gradient boosting decision tree and random forest models demonstrated that the ensemble model was robust with high accuracy and strong generalization ability. After the optimization process, the RON loss of 163 samples was reduced by 70%, and that of 128 samples was reduced by 50%–70%, while the SC of all samples was optimized to less than 5 μg/g. Furthermore, the visualization program dynamically traced the changes of RON and SC in tuning single and multiple variables. This study provided a much-needed ML model in gasoline refining, which was essential for optimizing the main process variables and increasing economic and environmental values.

Keywords: Research octane number (RON); Sulfur content (SC); Machine learning (ML); Maximal information coefficient (MIC); Back propagation neural network (BPNN); Dragonfly algorithm (DA) (search for similar items in EconPapers)
Date: 2022
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (3)

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0360544222017261
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:energy:v:261:y:2022:i:pa:s0360544222017261

DOI: 10.1016/j.energy.2022.124823

Access Statistics for this article

Energy is currently edited by Henrik Lund and Mark J. Kaiser

More articles in Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().