Derivation of Molecular Representations of Middle Distillates

Abstract

The molecular representation of hydrocarbon mixtures is critical to the advanced kinetics modeling of refining conversion processes; however, the achievement of such a representation is considered a significant challenge. The isomeric lump in a homologous series sets the analytical limit in analytical characterization of middle and heavy distillates. This paper proposes a new procedure for de-lumping detailed analytical information generated using a gas chromatography−field-ionization mass spectrometry (GC−FIMS) method into a molecular representation. As a result, concentration distributions of the various molecules in the sample of interest are calculated. This paper presents a deterministic computer-assisted procedure that automatically (i) generates hydrocarbon molecules according to literature-based set of rules, (ii) selects the hydrocarbon molecules that are most thermodynamically stable (likely to exist), (iii) optimizes the three-dimensional geometry of those molecules and calculates their thermodynamic properties, and (iv) calculates the concentration distributions of those molecules. A separate validation calculation involves (i) prediction of the physical properties for pure hydrocarbons using quantitative structure−property relationship (QSPR) correlations; and (ii) prediction of bulk physical properties of this mixture using the calculated concentrations, properties of its pure hydrocarbon components, and suitable mixing rules. This procedure was applied to find molecular representations of five middle-distillate samples and the results were validated through the estimation and comparison of such simulated and measured physical properties as density, refractive index, and simulated distillation curves (the latter of which are used as a consistency check). Good agreement was observed between the predicted and measured properties for all five middle-distillate samples. This agreement validates the molecular characterization algorithm, at least for the purpose of bulk property prediction in middle distillates.

Molecular reconstruction of complex hydrocarbon mixtures: An application of principal component analysis

“…function can be derived from theoretical concepts such as Gibbs free energy 39 or Shannon entropy, 35,41 or can be some sort of cost function. 37,38 Several of these methods start from a predefined set of components, the mole fractions of which are adjusted using a specific optimization algorithm.…”

“…To create such a set of molecules, several possibilities exist, e.g., algorithms using group contribution methods 1,40 or stochastic methods. 30,31,35,36,39 The second type of reconstruction methods uses a rather pragmatic approach, as they are based on a large set of experimental data, the so-called training set. Basically, the composition is reconstructed by interpolation between the samples in the training set using, for example, an artificial neural network 33,34 or other empirical correlations.…”

Kinetic Modeling of Vacuum Gas Oil Hydrotreatment using a Molecular Reconstruction Approach

“…This approach aims at representing a petroleum fraction via a mixture of molecules which are carefully optimized so that the physico-chemical properties of the mixture fit the "global" analyses of the industrial cut. The molecular reconstruction method used in this work consists of a two-step algorithm that couples stochastic reconstruction and entropy maximization approaches [20][21][22][23][24][25][26] (Fig. 5).…”

Data Reconciliation Among PIONA, GC-FIMS, and SimDis Measurements for Petroleum Fractions

“…The use of n-paraffin standards to link the boiling point with RT has been the basis of the ASTM D2887 simulated distillation method, and is assumed to be sufficiently accurate in this work. Applying the n-papraffin RT-BP standard to the GC-FIMS measurements, results in a similar GC-FIMS report-hydrocarbon type by BP distribution (Ha et al, 2005). Summing up all the hydrocarbon types in each BP interval (e.g., 10 ı C) generates an equivalent SimDis curve-"FIMS-SimDis."…”

“…One way to approach the current problem is to better understand the chemical composition of feedstocks, effluents, and intermediate streams in the refinery, generate the corresponding molecular compositions (Ha et al, 2005), and then use it to derive a better understanding of chemical reactions in refinery conversion processes through mechanistic kinetics models (Quann and Jaffe, 1992). Process modelers often encounter incomplete compositional information due to the limitations of individual analytical methods.…”