(434b) Toward Transitive Closure in Refinery-Wide Modeling: Atomically Explicit Kinetics Models

Bennett, C. A. - Presenter, University of Delaware
Hou, Z., University of Delaware
Zhang, L., China University of Petroleum
Klein, M. T., University of Delaware

Toward Transitive Closure in Refinery-Wide Modeling: Atomically Explicit Kinetics Models

Craig A. Bennett1, Zhen Hou1, Linzhou Zhang2, Michael T. Klein1


1 Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, DE

2 State Key Laboratory of Heavy Oil Processing, China University of Petroleum


A refinery is largely made up of reactor units that work together to produce a series of usable end products from an initial stock of crude (or synthetic crude) oil.  Although these reactor units are all tied together through a maze of pipelines, the models used to predict and evaluate the reactors have, for the most part, stood alone.  Lumped models have largely been employed based on the needs of the reactor being modeled.  As the product lumps of one reactor become the feed for a subsequent reactor, the lumping strategy largely has to change.  Every time a lump is created and redistributed, a loss of information occurs.  Therefore, a system of tracking and modeling the specific molecules involved would be preferable.

The in-house software tool known as INGen (the Interactive Network Generator) has the capability of compiling exhaustive kinetic networks for each reactor unit on an atomically explicit molecular-level.  The reaction network is derived from an initial selection of feed molecules, allowed reaction rules, and a series of chemically based bookkeeping approximations.  As was the case of the lumped model, the product stream of one reactor serves as the input stream to the next, but no information is lost in the molecular case as there is no translation from and to reactor specific lumps.  As each successive reactor is likely to produce new chemical species, it is important to insure that every reactor has the ability to interface with the superset of all possible chemical species, the Universal Oil. 

To assure this transitive closure, an initial oil composition consisting of species with five-aromatic rings or fewer (up to the boiling point range of a VGO) was passed through a series of individual reactor simulations: VGO Hydrocracking, Hydrotreating, FCC, Naphtha Hydrotreating, Cat Reforming, and Isom.  The results of the end-of-line Isom unit were then used as the initial feed to the VGO Hydrocracking unit and the series of network generations was continued recursively until no new species were generated.

The resultant Universal Oil defines the common interface for the plant model units.  Linear Programming and other efficiency modeling strategies built upon these models will be based on a level of detail and accuracy formerly unseen.  Specific molecule tracking and monitoring will lead to better understanding of the processes and more efficient flow sheeting and design.