(139b) Molecular Reconstruction of Heavy Petroleum Residue Fractions

In petroleum refining, both the accurate design and the optimization of conversion processes require the development of reliable kinetic models. In order to account for the differences in reactivity of the various species, more rigorous models containing molecule-based reaction pathways are needed. Such models expect a molecular description of the feed, however. For petroleum cuts boiling above the naphtha range (gasoil, VGO, ?), a molecular description can no longer be directly obtained, even through cutting-edge analytical techniques. To circumvent this lack of molecular detail, such a detailed description needs to be ?reconstructed? from more ?global? analyses. To this aim, a series of generic algorithms were developed in order to generate a representative mixture of molecules from standardized petroleum analyses. The idea of these molecular reconstruction algorithms is to create a discrete set of molecules that mimics the properties of the petroleum cut to be represented. The application of such an algorithm to generate a representation of heavy petroleum cuts will be described and illustrated for residue and asphaltene fractions.

During previous work (1-5), two different algorithms were developed to generate a complex mixture of molecules from standardized petroleum analyses: a stochastic reconstruction technique and a reconstruction by information entropy maximization. Moreover, both approaches can be advantageously combined to avoid some of the drawbacks of each technique. In the present work, the methods have been improved to be able to treat heavy petroleum fractions that contain large amounts of impurities and heteroatoms.

The stochastic reconstruction technique is based on the idea to create an initial set of molecules, which is then modified until a mixture is obtained with the same properties as the petroleum cut to be represented. First, molecules are constructed from a number of structural blocks (polycyclic cores, rings, chains, ?). The type and number of these blocks are chosen by randomly sampling a set of distributions of structural attributes. The construction procedure is repeated N times in order to obtain an equimolar mixture of N molecules. For each molecule, pure compound properties are calculated from the structure, either directly by inspection (e.g. chemical formula, molecular weight, NMR, mass spectra) or numerically by group contribution methods (e.g. density, boiling point). The average properties of the mixture of N molecules are then obtained through mixing rules and compared to the available data (elemental analysis, density, molecular weight, mass spectrometry, ?) of the refinery cut. The deviation between the experimental and simulated data is minimized by a simulated annealing method that modifies the parameters of the distributions for structural attributes.

The reconstruction by information entropy maximization starts from an a-priori defined set of molecules and adjusts their molar fractions so as to obtain a mixture with the same characteristics as the petroleum cut. Adjusting the molar fractions is done by maximizing the information entropy. This criterion implies that, if no information is available, it is impossible to prefer a specific molecule rather than another. Without constraints (or information), the distribution of the set of compounds is therefore uniform. The introduction of constraints (i.e. analytical data of the petroleum cut) distorts the uniform distribution of the set in order to match this information. Currently, the only restriction of the method is that these constraints must be linear. As opposed to the stochastic reconstruction technique, the entropy maximization method uses classical optimization techniques and the computational effort is much smaller.

The two outlined methods of molecular reconstruction were successfully combined in a two-step method and applied to residue petroleum cuts. The stochastic step allowed to introduce chemical knowledge through the choice of the structural blocks and the molecule construction scheme. In this first step, a set of molecules characteristic of heavy petroleum fractions is generated, whose composition results in properties that are close to properties of the residue fraction. Starting from this representative "mixture", the entropy maximization only needs to slightly modify the molar fractions of the molecules to represent the analytical data in an optimal way.

The reconstruction method was applied to various petroleum residues and is illustrated for an Arabian Light vacuum residue (VR). Starting from the available experimental data, a set of 5000 molecules was generated during the stochastic reconstruction step. After the subsequent entropy maximization step that modifies the mole fractions of these various compounds, the synthetic mixture has the same macroscopic analyticals as the petroleum cut (Table 1). This synthetic mixture is a polydisperse molecular representation of the Arabian Light VR, which can now be used as input to detailed kinetic models.


1. D. Hudebine, C. Vera, F. Wahl & J. Verstraete, AIChE Spring Meeting (New Orleans, LA, March, 10 - 14) Paper 27a (2002).

2. D. Hudebine, Ph.D. thesis, Ecole Nationale Supérieure de Lyon (2003).

3. D. Hudebine & J. Verstraete, Chem. Eng. Sci., 59, 4755?4763 (2004).

4. J. Verstraete, N. Revellin, H. Dulot & D. Hudebine, Preprints - American Chemical Society - Division of Fuel Chemistry, 49(1), 20?21 (2004).

5. K. Van Geem, D. Hudebine, M-F. Reyniers, F. Wahl, J. Verstraete & G. Marin, Comp. Chem. Eng., 31, 1020?1034 (2007).