Jason K. Perry, First Principles Research

William A. Goddard III, Caltech

Anil Patel, Rawls Frazier, and John Shinn, Chevron

Representation of various structures found in gasoline

Gasoline reforming is the important process of altering the composition of gasoline to achieve a higher octane rating. As shown above, gasoline is a complex mixture of hydrocarbons, generally falling in the range of C6-C10, and different mixtures have different octane ratings. In order for refineries to produce gasoline with a consistent octane rating when the composition of the crude may be highly variable, complex kinetic models need to be developed which can guide the engineer in tuning the conditions of the reactor based on the composition of the feed.

With interests in mind reaching well beyond gasoline reforming, Chevron has challenged us to develop a new kinetic model structured around ab initio derived thermochemical data. The importance of this challenge is to find new ways to develop kinetic models for systems where little detailed experimental data exists. Gasoline reforming is an ideal candidate for the development and validation of such a new kinetic model because a significant amount of detailed experimental data indeed does exists and the process is reasonably well characterized mechanistically.

Of the many goals of this project, perhaps the most important is to develop a model which is strongly linked to the chemistry of the catalyst. Where empirical parameters may be required, these parameters should have physical significance tied to the function of the catalyst. Indeed, a well-structured model should have a minimum of adjustable parameters, depending mostly on well-defined thermochemical and kinetic quantities.

The basic strategy for the development of our gasoline reforming model is as follows:

A) Layout the species involved. These species were enumerated by defining the size range of molecules of interest (C1-C8, in this case) and the specific classes of molecules considered [normal paraffins (nP), branched paraffins (iP), 5-membered ring naphthenes (N5), 6-membered ring naphthenes (N6), and aromatics (A)]. To this list, olefin and cation derivatives of these species were added. The total number of species involved is 540.

B) Determine the types of reactions involved. These include: paraffin isomerization, paraffin dehydrocyclization, naphthene isomerization, naphthene dehydrogenation, cracking, and hydrocracking. As shown below, these reactions are generally achieved over a bifunctional catalyst, which first dehydrogenates the parent alkane to form an olefin, then protonates the olefin to form a cation. Once the cation is formed, it can easily rearrange. The new cation is then deprotonated and rehydrogenated to form a new alkane. The total number of reactions involved is 883.

C) Form a reaction network linking all species. Species considered linked via rapid equilibrium processes were lumped together and treated kinetically as a unit. All parent molecules were lumped together with their olefin and cation derivatives, and 6-membered ring naphthenes were lumped together with their aromatic derivatives. These lumps were linked together via reaction mechanisms that involve the cationic species. While making chemical sense, this lumping procedure significantly reduced the size and complexity of the problem.

D) Calculate energetics of all species. Enthalpies and entropies were calculated using a variety of semi-empirical and ab initio methods for all species up to C7 (over 200 species). Group additivity rules were developed from this set, and C8 energetics were predicted from this.

E) Determine the minimum number of adjustable parameters needed to describe the action of the catalyst. These parameters are related to the acidity of the catalyst and the effect the catalyst has on various reaction barriers. There are seven total.

F) Optimize parameters to fit experimental data for a subset of the data. In this case, detailed experimental data exists for a C6 feed for a variety of temperatures, pressures, and residence times. A set of parameters was optimized for this data, with the results shown below.

G) Prove robustness of the model through comparison with the remainder of the experimental data. Taking the C6 optimized parameter set, the model predicted distributions for C7 and C8 feeds. These results were compared to experimental data, as shown below.

Comparison of theoretical and experimental product distributions shows the model has quantitative accuracy. This achievement is particularly important due to the small number of adjustable parameters involved as well as its strong association with the molecular level chemistry of the system. The model takes only seconds to run.  With an added graphical user interface, the model is now in the hands of Chevron engineers.