Molecular-level composition and reaction modeling for heavy petroleum complex system

Abstract

A new methodology for the molecule-based modeling of heavy petroleum mixtures has been developed. Molecules in the heavy feedstock have been described in terms of three essential structural attributes (cores, side chains, and inter-core linkages) and then statistically juxtaposed into a set of representative molecular compositions that can be constrained by a set of probability density functions (pdfs). In order to obtain the optimal molecular composition, an optimization loop was employed to minimize an objective function in terms of available measurements via adjusting the limited parameters of the pdfs. An example of resid feedstock containing 400,000 components was created using only O(30) parameters. To limit the kinetic model to a practical size, the reaction model was described in terms the reactions of the three constituent types and a set of irreducible molecules. Subsequent product property estimation was a straightforward juxtaposition of attributes. For example, a model for resid coking was built in terms of 2,839 attributes and equations but kept the full compositional details of the 400,000-molecule mixture.