logo


Oct-2014

Predicting reactive heavy oil process operation

Characterisation of feed and product yields through component structures for better understanding and prediction of operations

GLEN A HAY, HERBERT LORIA and MARCO A SATYRO, Virtual Materials Group
HIDEKI NAGATA, Fuji Oil Company Ltd

Viewed : 3976


Article Summary

To understand and optimise reactive heavy oil processes encountered in refineries there must be a strong knowledge of the product yields and their physical properties. Sometimes property predictions within models of the reacting material are important due to operational constraints. The coke induction point,1 or point at which solid coke begins to form in heavy hydrocarbon mixtures, is an example of the importance of property predictions since many unit designs need to take into account solid precipitation if it occurs. The simulation modelling application discussed in this article applies to the Eureka process where fluidisation of reactor materials and inhibiting the coke induction point is essential.2

When catalytic or thermal cracking simulation models are developed, enough physics must be encoded into the mathematical development to accommodate for the significant changes in both the normal boiling point of produced material and the associated molecular structures. Many different models for the reactive chemistry mechanisms required to model this class of processes have been proposed. These mechanisms include the use of hydrogen donor components, cyclic ring breaking, dehydrogenation of saturated rings, and cracking versus oligomerised propagation of small to large molecules, to name a few. The required basis for the development of a reliable simulation environment designed to handle the type of detail associated with chemical reaction mechanisms requires 
flexible component chemical structures to represent the products from different chemical reaction pathways. In order to capture this type of behaviour directly within the requirements of industrial process simulation software a new PIONA (n-Paraffin, Iso-paraffin, Olefin, Naphthene, and Aromatic) pure component basis environment was developed to both characterise the feedstocks and the resulting products’ estimated chemical make-up and yields.

The PIONA technique3 consists of using constant groups also known as slates of predefined compounds required to cover the carbon number ranges for feeds and products necessary for the modelling of different refinery reactors, such as the Eureka process thermal cracking vessels. The different combinations of these component slates and the compositions of the components within allows for the matching of the experimental distillation curve of a given feed and the calculation of its chemical characteristics, ranging from simple properties such as molecular weight and standard liquid density all the way to more complex 
physical properties such as heating values, liquid viscosities, and pour points. The key advantage in using this method is its ability to capture the essential chemistry of the feedstock and product mixtures and how the changing compositions upon reaction affect property calculations. The number of components used in the simulation is kept constant and consistency is enforced throughout the simulations.

The PIONA structure group classification was found to introduce an unacceptable property estimation error in studies when modelling feeds with an average carbon number higher than ten, where larger aromatic content was encountered. Further investigation showed that a single aromatic structure group was not enough to differentiate multi-paraffin branched aromatic components against those more reacted compounds that were stripped of straight carbon branches. Therefore, an extra chemical type defined as ‘dehydrated aromatic’ is included in the PIONA technique.3 From a molecular structure configuration, these dehydrated aromatics replace branched contributions to a base aromatic ring with additional dehydrated aromatic rings.

Heavy oil reaction models
Models using detailed specific reaction pathways are commonly found in the technical literature on thermal cracking or pyrolysis of lighter gases such as ethane or even naphtha feedstocks.4,5 This is possible because the overall number of pure component and radical species are still manageable within a simulation environment (typically less than 150 species), assuming good numerical techniques for the integration of the differential component material and energy balances are employed. Even then, many of these models have to be linearised to help achieve faster computational speeds.6 Once heavy residual oil feedstocks are introduced, a change in mathematical solution methods for the component material balances is noticed in solutions using generalised, or lumped reaction pathways7,8 due to the sheer complexity of the feed material. In these models, the heavy residue thermal cracking reactor might use only eight lumped components, a dozen reaction pathways, and resolution of the feed and product material balances would look similar to the representation in Figure 1.   

This type of modelling relies on the availability of a substantial amount of experimental data which limits the quality of results extrapolated from the model. This lack of predictive power is mainly driven by the non-mechanistic approach to the reaction’s kinetic parameter fitting and eventual increase of error outside of the fitted data range. Conversely, this approach also presents advantages such as fast solution speeds and its ability to predict specific properties of 
the generalised yield cuts such as the softening point of the pitch and the contained volatile matter7 through the use of lumped components and effective lumped properties and mixing rules for these properties.

In these lumped models, the resulting temperature profiles calculated for the furnace tubes can also be roughly estimated due to the matched enthalpy of formation for each lumped component, assuming the chemical structure shifts stay consistent and that enthalpies of combustion are available. Significant error would be introduced in these models if hydrogenation versus dehydrogenation occurred since these reaction pathways are exothermic instead of endothermic. In that respect, average boiling point lumping used by these models is completely non-predictive and would need to be rebuilt for specific processes and eventually even specific plants and equipment.

The PIONA slate technique was applied in the commercial process simulator software VMGSim, to represent multiple hydrocarbon feedstocks in lighter and heavier cut ranges.3,9 The focus was on property calculations and characterisation of multiple feeds using mixtures based on the same basis component slate. The ability of this PIONA system to model reactive systems is best illustrated in Figure 2 when compared with the lumped, linearised systems. In Figure 2, the ability to model the transition between carbon number and molecular structure types caused by chemical reactions modelled through reaction pathways is explained as well as the overall heat of reaction effects.
 
Characterising process feedstocks and predicting product yields
The key step for the correct modelling of a thermal cracking process such as Eureka is the definition of a correct mixture of PIONA based components needed to characterise the feedstock. Laboratory analysis of hydrocarbon feed material is used to provide the necessary information to fit the model of material stream values against measured properties. The reacted product yields are similarly characterised. The reaction kinetics of the thermal cracking vessel at the desired operating conditions are then simulated and the resulting material product yield’s properties and flow rates would be compared to known data. At that point, the process is repeated until the adjustable model parameters are properly defined and the errors between model and experiment are minimised.


Add your rating:

Current Rating: 4


Your rate: