Skip to Main content Skip to Navigation

Compréhension moléculaire et prédiction des propriétés physicochimiques dans les produits pétroliers

Abstract : The rapid decline in light crude oils requires to convert heavy petroleum fractions into more valuable products (naphtha, diesel, lubricants, etc.). In this context, hydrocracking process (HCK) consists on upgrading vaccum gas oil (VGO) into high quality products. The quality of petroleum products is based on some chemical and physical properties that should fulfill prerequisite specifications. The hydrocracking process optimization requires to set up time consuming and costly experiments for developing catalysts and setting operating conditions. High throughput experimentation (HTE) units are then increasingly used at IFPEN. However, these units do not enable to obtain end products. Otherwise, predictive models were developed in order to understand and predict the impact of operating conditions about products quality. However, some complex properties are very difficult to model and require a better understanding. This work is mainly concerned with the understanding of diesel cloud point (CP) and viscosity index (VI) of base oils. Two analytical techniques were used: the two-dimensional gas chromatography (GC×GC) that enables to identify hydrocarbons compounds in petroleum products and the 13C nuclear magnetic resonance (NMR) spectroscopy which provides structural characteristics of these compounds. A sparse multivariate regression (sparse Partial Least Squares) was performed using chromatographic and spectroscopic data. The sparse PLS is derived from classical PLS. It allows to reduce the number of factors by performing a variable selection. The selected factors are the most correlated to the property to model. Globally, this approach enabled to better understand how hydrocarbon compounds (nparaffins, isoparaffins, aromatics,…) and their molecular characteristics (carbon number, degree of branching,…) affect the diesel CP and the VI of base oil. Furthermore, the good performances of developed sparse PLS models show that it is possible to access to the products quality when using HTE units. Kriging models were also developed. Kriging is an interpolation method that predicts the value of a function at a given point by computing a weighted average of the known values of the function in the neighborhood of the point. Kriging models have local aspect which is well adapted to complex data. Its probabilistic approach enables to provide an estimate of predicted value uncertainty. Results show that kriging improves predictive performances for both diesel CP and VI of base oil. This approach is quite innovative in modelling of petroleum products properties. When using HTE units, it allows to estimate the VI of base oil more easily than from chromatographic or spectroscopic data which are not available for the refiners
Document type :
Complete list of metadatas

Cited literature [165 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, March 27, 2018 - 9:46:08 AM
Last modification on : Tuesday, October 20, 2020 - 11:06:13 AM
Long-term archiving on: : Thursday, September 13, 2018 - 7:27:58 AM


Version validated by the jury (STAR)


  • HAL Id : tel-01744088, version 1




Jean-Jérôme da Costa Soares. Compréhension moléculaire et prédiction des propriétés physicochimiques dans les produits pétroliers. Autre. Université de Lyon, 2017. Français. ⟨NNT : 2017LYSE1310⟩. ⟨tel-01744088⟩



Record views


Files downloads