Heavy feed characterisation: a molecular approach

Characterising heavy feeds on a molecular basis, together with kinetic studies, accurately predicts the reactivity of a wide range of vacuum residues.

GLEN HAY and LANTE CARBOGNANI, Virtual Materials Group
HIDEKI NAGATA, Fuji Oil Company

Download Complete Article

Viewed : 5122

Article Summary

The characterisation and reactivity of heavy oil fractions (vacuum residue) were studied using a PIONA (paraffins, iso-paraffins, olefins, naphthenes, aromatics) molecular approach. Eleven crude oil assays along with vacuum residue pilot-scale thermal cracking information were provided. These feeds were characterised and their susceptibility to thermal cracking was evaluated all using a PIONA rigorous approach. A reactivity index related to the naphthenes-â€¨aromatics-dehydrogenated aromatics ratio was found for each feed in order to match experimental results. This reactivity index was then correlated to the vacuum residue feed properties. The analysis of these results points to C₇ asphaltenes (or equivalent), carbon residue and density of vacuum residue as key properties to be measured in order to capture the chemical nature of this fraction, and thus proper reactivity in a thermal cracking process.

Identifying and quantifying each component contained in oil fluids has proven to be both impractical and unfeasible. Characterisation of crude oils and their different cuts is commonly performed by measuring different properties. Traditionally, techniques such as gas chromatography (GC) for lighter fractions (light ends and C₁-C₅) and the use of distillation curves are commonly used for the liquid fractions. In this regard, the properties (densities, molecular weight, chemical family, and so on) of the light ends and lighter component ranges are usually well known, however the physical properties of the remainder fractions need to be determined separately and this comes at a price (both time and money). Organisations such as the American Society for Testing and Materials (ASTM) have developed routine tests for determining boiling ranges and properties of crude feedstock, distillation fractions and products, however the bulk of these techniques were originally intended to capture the properties and composition of conventional oils. Heavy oils pose a challenge since they contain a large, non-distillable fraction.^1-3
Commonly measured properties in these non-distillable fractions, such as density and viscosity, can prove to be helpful when determining the physical and transport properties of these fractions. Other properties, for example carbon residue, SARA (saturates, aromatics, resins, asphaltenes) analysis and pour point also prove to be helpful when determining not only the quality of an oil product, but also its chemical nature and reactivity.^2-3
Carbon residue is a particularly important characteristic of crude oil residues, since it not only can indicate the quality of the fraction, but also can be correlated to a number of properties such as hydrogen to carbon ratio (H/C), heteroatomic (S, N) content, asphaltenes content, or viscosity.²

The pour point of an oil fraction indicates the minimum temperature at which this material will flow. This property is affected by the presence of heavy molecules (which increase pour point), thus it can be correlated to molecular weight and density.²
SARA analysis proves to be useful for residue fractions, which usually contain large amounts of aromatics, resins, and asphaltenes. In particular, measurements of asphaltenes are important for heavy oils and residues to determine solid deposition probability, usually an issue in not only the production industry, but also in transportation and refining. Alternatively, the fractions of saturates and aromatics are of less importance due to their redundancy with other lighter range properties typically measured.^2-3

Introduction to heavy oil kinetics

When dealing with reaction kinetics for thermal cracking of heavy oils, a considerable amount of material has been published on the subject in the last few decades. This material focuses more on the molecular structure approach to solving the problem.⁴ This approach is a more rigorous alternative to the lumped kinetic schemes that are being replaced due to lack of predictive accuracy across differing feedstocks. Moreover, this surge in molecular structure modelling comes at a time when personal computers have reached a point at which processors can keep up with the enormous computational demands of approaching the problem. It should be noted that molecular structure based kinetics, as well as physical property calculations, have been around since the 1960s and many influential and detailed papers on the topic were published as early as the 1980s.^5-6
When looking at the thermal cracking of heavy oils, or even approaches to catalytic cracking and processing, there is an emphasis on aromatic groups with or without significant saturated branching.⁷ This distinction is of great interest due to the relationship of the saturated nature of these heavy aromatics to their ability to crack into highly desired liquid product, or alternatively propagating into larger molecules that can lead to solid precipitation or coking. Examples of such molecules can be seen in Figure 1. If these heavier cut molecules are characterised accurately, the approach to applying kinetics to these molecules would then depend upon the ability to break C-C bonds or the more challenging C-H scission at specific operating conditions where cracking occurs. Cracking reactions are usually proposed in three general steps: chain start, growth, and termination.⁸ At the same time, the propagation of such molecules will also occur at a calculated reaction rate and the balance of these reactive cracking and propagation pathways will ultimately lead to resulting product yields.

PIONA modelling approach

Different methods to describe a mixture of hydrocarbons and its chemical or physical properties have been attempted over time. The approach described within this work takes a middle ground within most simulation environments. The first simulation approach, a simplistic single property lumped approach, has dominated the software public domain since the 1900s. In this case, the crude oil feedstocks are divided into pseudo-components based on a range of a single focal property, which is usually the average boiling point or molecular weight. Although little information needs to be known about the feedstock to complete this approach, the drawback is the lack of predictive results due to over-generalisation of the molecular structure groups within the mixture. The properties predicted for mixtures also become direct correlations based on very general reference properties. An example of such a correlation was proposed for calculation of the aniline point of a mixture based on API gravity and mid boiling point alone (see Equation 1): ²

Aniline point = -183.3 + 0.27(API)T_b^1/3+ 0.317T_b (1)

For maximum consideration of molecular structures in hydrocarbon mixtures, the components in a simulation environment can be created with contributions groups.^5-6 In these situations, any molecular type can be created with a combination of cumulative groups. The drawback of this approach is the amount of information required about the feedstock, which is usually only available through very expensive laboratory analysis. For these situations, an actual component list no longer becomes reasonable due to the unlimited number of combinations. Furthermore, visual interpretation of a mixture of components must be shown in a different way to the convention of a list of components and their mixed fractions.

The PIONA based approach was created in order to reach a compromise between these two extremes. In this approach, a limited number of major molecular structure groups consisting of paraffin, iso-paraffin, olefin, naphthene, aromatic, dehydrogenated aromatics, and heteroatomic species (organic molecules including sulphur, nitrogen, iron, nickel, and vanadium atoms) are created. In turn, each of these groups has multiple components relating to different carbon number ranges. In order to bridge the gap between these hard to define structure groups, a fraction of each type of group can be mixed to create the appropriate properties. A representation of this type of approach is shown in Figure 2 with a selection of C₄₀ carbon molecules with differing naphthene and aromatic ring counts as an example.

One main advantage of this PIONA grouped approach is the easier handling of non-ideal interactions between species for the thermodynamics behind the model. Limited property data containing details of all necessary contribution groups in the more detailed molecular approach makes tuning of the increased number of interaction parameters almost impossible when considering potential extrapolation. Poor tuning of these interactions allows for a larger chance of highly irregular results in non-validated ranges which could produce a poorly behaved model. Another advantage of the PIONA approach is that the literature tends to generalise structure groups in a similar resolution and many correlations, such the aniline point correlation in Equation 2, can be applied directly.² It should be noted that this equation, now considering the aromatic fraction of a hydrocarbon mixture and specific gravity, would replace Equation 1 which makes an assumed relationship between density at a given average boiling point to define the fraction of aromatics:

Aniline point = (Aromatic% - 692.4 + 794(SG)) / (12.15(SG) – 10.4) (2)

PIONA kinetic approach

The kinetic pathways created from a PIONA component structure would transverse across the carbon number range through cracking or propagation and across the different structure types with hydrogenation or dehydrogenation. The heat of reaction is easily calculated from the overall balance given the underlying enthalpy of formation of these components as they shift. For each type of structure and carbon number, the kinetic rates for each potential reactive pathway need to be calculated. For thermal cracking systems where hydrogen donors are not readily available, the main available cracking pathway for the larger aromatic components consists of scissor-type reactions of the saturated chain branches. Once these branches have been removed, the aromatic rich core structure left usually will no longer crack and will ultimately become pitch or coke. These compounds are then being tracked as dehydrogenated aromatics in this method’s application.

With the characterisation of crude feeds, the importance of matching the amount of heavy aromatic core components then becomes essential in order to get the kinetic rates that would match what is occurring within thermal cracking reactions in the process. If too many saturated components were estimated in the feed characterisation, the calculated cracking rates would over-predict lighter material yields. On the other hand, if too many ‘aromatic core’ type components were estimated, which would be balanced by very saturated components, this would lead to over-prediction of both lighter and very heavy range yields with a lack of middle range liquid product.

Luckily, with many laboratory analysis methods the carbon residue is something commonly measured for heavier boiling range mixtures and is shown to relate to H/C.⁶ From this measurement, the amount of dehydrogenated aromatic components can also be estimated and the kinetic rates and yields from thermal cracking are properly predicted. In many cases, the yield differences from two crude feeds that have similar distillation curves and densities come from these heavier component structure distributions.

Methodology of characterisation

Oil characterisation information was provided by Fuji Oil Co. Ltd. (FOC) for 11 crude oil feeds. This information, contained in standard assays, is going to be used in a PIONA backed simulation environment to determine the relationship between several physical properties measured (such as density, carbon residue, and so on) and the reactivity of the vacuum residue to thermal cracking.

The first step consists of the characterisation of these feeds. The physical properties of the crude oil (and its vacuum residue cuts) cover a wide range as can be seen in Figure 3. This plot shows the relative densities of the crude oil feed and vacuum residue fraction relative to the lightest sample (see Equation 3):

Relative density = Sample-density/[Lightest Sample density] (3)

This widespread range proved to be important in order to capture the effect of the different physical properties on the reactivity of each vacuum residue.

The oil assays for these crude oil feeds contain yield information for 14 cuts, and physical properties such as density, sulphur and nitrogen content, viscosities, and so on for each cut. Additionally, for all the vacuum cuts (370°C+), information regarding C₇ asphaltenes %, carbon residue, and metal content (V, Ni) is also available.

The PIONA molecular structure approach was used to characterise the different oil feeds. For each cut range, the different information available (S, N, H/C, densities, for instance) is transferred directly to a simulation model. For this work, VMGSim software was used due to the already available reactive PIONA environment. For lighter cut ranges (<200°C), the amount of paraffins/iso-paraffins is calculated automatically in order to match their density. The naphthenes/aromatics content of these cut ranges is predicted directly by internal correlations developed using available literature.² Similarly, for cut ranges from 200-550°C, the percentage of aromatics is controlled to match these cuts’ density.

For heavier fractions, such as vacuum residue, carbon residue was used in order to predict the H/C.² The next step was to set the amount of naphthene molecules in this fraction in order to minimise the experimental error for both H/C and density.

Feedstock case studies and results

The Eureka process is a thermal cracking technology which aims to produce higher liquid/lower gas yields from vacuum residue feeds. One differentiator from conventional coking processes is that it is designed to prevent coke deposition (due to over-cracking). The residual stream of this technology is a pitch, which is intended to flow easily at the reactor’s outlet conditions.^9-11

The reactivity and kinetics of the Eureka process have been studied in the past on a laboratory scale by several authors. Results for different vacuum residue feeds and descriptions of the set-ups used can be found in the literature.^9-10,12

In order to predict properly the reactivity of the vacuum residue fraction, experimental data for thermal cracking experiments for the 11 vacuum residues (corresponding to the 11 crude oil feeds mentioned earlier) was provided by FOC. In order to compare these experimental results, a PIONA simulation model was developed representing a laboratory scale set-up.^9-10,12 This model was then used to run the different feedstock experiments.

Additionally, a detailed simulation model corresponding to FOC’s Eureka industrial site was developed. This model included all the extra complexities typically present in a full plant scale (recycles, distillation towers, cracking heaters, steam stripping, and so on). The vacuum residue feeds were characterised using the PIONA molecular approach, and product yields for different operating conditions were used to fine-tune the kinetics of the thermal cracking unit. An example of the results obtained with tuned kinetics can be seen in Figure 4.

This set of kinetics was then transferred directly to the laboratory scale model. With the proper kinetic parameters, all of the different vacuum residue feeds characterised were evaluated using this model. The steps followed to go from bench to plant scale can be seen in Figure 5.

From the results obtained, the reactivity index of the different vacuum residues was adjusted until the error between the results predicted (pitch yield, pour point, density, for instance) and the experimental data was minimised. A comparison between the experimental data and results predicted can be seen in Figure 6. A ‘soft sensor’ approach was used in order to attempt the prediction of properties that were seen as key indicators such as heptanetoluene-quinoline insolubles and volatile matter. Soft sensor is a common name for software where several property measurements are processed together. The interaction of these measurements can then be used for calculating new properties. Due to the PIONA molecular structure approach taken, properties such as H/C, aromatics/A-dehydrated, D-1160 50% T, are directly available for any stream (or mixture of streams). These values were then used to correlate the pitch properties of interest. Results for these predicted properties can be seen in Figure 7. The reactivity index is related to the distribution between aromatic and dehydrogenated aromatic PIONA groups. However, since the vacuum residue distillation curve estimated by the simulator (usually not available due to the higher boiling ranges) has a direct impact on the yields predicted, a correlation to estimate the shape and tilt of the curve was developed. This correlation was found to depend on the vacuum residue’s initial boiling point, C₇ asphaltenes content and carbon residue. The correlation’s parameters were regressed to minimise the error with the experimental data for all the different vacuum residue cuts studied.

Once the reactivity index was determined for each of the vacuum residue feeds, a relationship for this value was regressed using the different properties (density, H/C, CCR) available for these feeds. The first attempt involved the vacuum residue density, and the results can be seen in Figure 8a. In this figure, it can be seen that more dense vacuum residues (represented in orange) did not seem to follow the trend. However, this deviation can be explained by the increasing asphaltenes content in these heavier vacuum residues, which allows for a correction factor to be applied (see Figure 8b).

Once a correlation was successfully found for the reactivity index of vacuum residue, the different feeds provided by FOC were evaluated using the Eureka plant model (see Figure 5). Since both models use the same PIONA rigorous molecular approach, the transition was seamless and allowed for extrapolation and comparison of results for the different samples provided. Some of these results can be seen in Figure 9.

Conclusions

The PIONA approach offers not only complete mass balance on feed, products and utilities, but also the opportunity to run optimisation studies on the type of feed/process conditions and utilities. In order to model properly a vacuum residue feed, and to be able to capture its thermal cracking reactivity, certain properties proved to be more significant. To obtain the most accurate result, consistent measures of density, carbon residue, and C₇ asphaltenes (or equivalent) should be provided for a vacuum residue. The density property seemed to provide a good starting point when determining the chemical nature of a vacuum residue. However, by itself it could potentially fall short; for example, we could characterise a stream with the same density in many ways using the PIONA approach (non-unique solutions). In this regard, the carbon residue property used to determine the H/C proved to be key. The PIONA composition of a given stream is then shifted in order to minimise the error, not only in the density but in the H/C as well. The C₇ asphaltenes measurement (or equivalent) seemed to be the most advantageous property taken from the SARA analysis provided for the vacuum residue. The samples with higher asphaltenes content would deviate from the reactivity predicted if only the density was used in the correlation. This would point out that this property can shed some light on the chemical family composition of the group, and thus the reactivity of this fraction to thermal cracking. This also means that when considering a SARA analysis for this fraction, the effort should be focused on getting a consistent measurement of the asphaltene fractions, both simplifying the methodology and potentially reducing costs. The PIONA characterisation and kinetic approach used in this work allow for the ability to predict the reactivity of a wide range of vacuum residues. This would mean not only that optimisation studies can be carried out (blend of feeds, different operating conditions, scaling up, and so on), but also implies potential monetary saving by reducing the need to run cracking experiments.

References

Liang X, Yan W, Thomsen K, Kontogeorgis G M, On petroleum fluid characterization with the PC-SAFT equation of state, Fluid Phase Equilibria, 2014, 375, 254-268.
Riazi M, Characterization and properties of petroleum fractions, ASTM international, 2005, Vol 50.
Nji G, Svrcek W, Yarranton H, Satyro M, Characterization of heavy oils and bitumens. 1. Vapor pressure and critical constant prediction method for heavy hydrocarbons, Energy & Fuels, 2007, 22, 1, 455-462.
Kumar H, Froment G F, Mechanistic kinetic modeling of the hydrocracking of complex feedstocks, such as vacuum gas oils, Industrial & Engineering Chemistry Research, 2007, 46, 18, 5881-5897.
Quann R J, Modeling the chemistry of complex petroleum mixtures, Environmental Health Perspectives, 1998, 106, Suppl 6, 1441.
Quann R J, Jaffe S B, Structure-oriented lumping: describing the chemistry of complex hydrocarbon mixtures, Industrial & Engineering Chemistry Research, 1992, 31, 11, 2483-2497.
Wiehe I A, Process Chemistry of Petroleum Macromolecules, CRC Press, 2008.
Simanzhenkov V, Idem R, Crude Oil Chemistry, CRC Press, 2003.
AlHumaidan F, Hauser A, Al-Rabiah H, Lababidi H, Bouresli R, Studies on thermal cracking behavior of vacuum residues in Eureka process, Fuel, 2013, 109, 635-646.
AlHumaidan F, Lababidi H M, Al- Rabiah H, Thermal cracking kinetics of Kuwaiti vacuum residues in Eureka process, Fuel, 2013, 103, 923-931.
Aiba T, Kaji H, Suzuki T, Wakamatsu T, Residue Thermal-Cracking by the Eureka Process, Chemical Engineering Progress, 1981, 77, 2, 37-44.
Takatsuka T, Kajiyama R, Hashimoto H, Matsuo I, Miwa S, A practical model of thermal cracking of residual oil, Journal of Chemical Engineering of Japan, 1989, 22, 3, 304-310.