Research Octane Number Prediction Based on Feature Selection and Multi-Model Fusion

The catalytic cracking-based process for lightening heavy oil yields gasoline products with sulfur and olefin contents surpassing 95%, consequently diminishing the Research Octane Number (RON) of gasoline during desulfurization and olefin reduction stages. Hence, investigating methodologies to mitigate RON loss in gasoline while maintaining effective desulfurization is imperative. This study addresses this challenge by initially performing data cleaning and augmentation, employing box plot modeling and Grubbs’ test for outlier detection and removal. Subsequently, through the integration of mutual information and the Lasso method, data dimensionality is reduced, with the top 30 variables selected as primary factors. A predictive model for RON loss is then established based on these 30 variables, utilizing Random forest and Support Vector Regression (SVR) models. Employing this model enables the computation of RON loss for each data sample. Comparing with existing methods, our approach could ensure a balance between effective desulfurization and mitigated RON loss in gasoline products.

@article{Gu2024,
  title     = {Research Octane Number Prediction Based on Feature Selection and Multi-Model Fusion},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {15},
  number    = {3},
  year      = {2024},
  publisher = {The Science and Information Organization},
  author    = {Junlin Gu},
  doi       = {10.14569/IJACSA.2024.01503114},
  url       = {https://doi.org/10.14569/IJACSA.2024.01503114}
}