(551b) Automation of Composition Modeling for Complex Petroleum Hydrocarbons by Homologous Series

This article developed a hybrid statistical approach for the modeling of a complex petroleum feedstock by optimizing a molecular representation based on structural attributes to a set of analytical characterizations.

A general petroleum hydrocarbon conversion was qualitatively described as a set of homologous series. Generally, the structures in those series could be determined by user, literature and automated reaction network generation etc.

Consequentially, the quantitative information of a petroleum feedstock (mole fraction or mass fraction) can be obtained via employing an optimization loop. The molecular compositions of a feedstock were described as a set of structural attribute probability density functions (PDF) first. Through constraining the parameters of those PDFs, an optimization loop was then applied based on the juxtapositions of the mapping between molecules and multiple attributes; the objective function was given in terms of available analytical information. A structure property calculation program had been implemented to evaluate this objective function. Finally, the optimal mole fraction of the feedstock was obtained and could serve as the initial concentration values for the kinetic modeling.

A user-friendly program with an Excel-VBA interface called Composition Model Editor (CME) was developed to fulfill this new approach. In order to verify this approach, selected complex feedstocks were modeled by CME and then validation results were shown in this article.