Journal of Petroleum Science and Engineering 57 (2007) 166 180 www.elsevier.

com/locate/petrol

Classification and characterisation of crude oils based on distillation properties

Peter Behrenbruch , Thivanka Dedigama
Australian School of Petroleum, University of Adelaide, Australia 5005 Accepted 31 October 2005

Abstract True boiling point (TBP) distillation is a widely used batch distillation process for the characterisation of crude oils, traditionally mainly for marketing and refining purposes. The TBP curve is obtained by plotting the cumulative mass or volume distillation fraction with increasing temperature. The shape of these curves is dependent on the volatility of components in a given crude oil. As such, these curves give a footprint of the composition of crude oils. A new method of characterising crude oils based on the shape of TBP distillation curves is proposed. A gamma distribution is used to characterise the TBP distillation curve, and the parameters of the fitted distribution are used as characterisation parameters. The proposed method is found to describe experimental data very well with just two parameters, and as such offers a very practical approach in terms of classifying crude oils. Ranges of values for the characterisation parameters for different types of crude oil are identified for a large set of TBP data. The characterisation parameters can be correlated with a number of crude oil properties. They can also be used to predict the petroleum fractions that can be produced from a given crude oil and hence correlated to the value of that crude. As an alternative, it is shown how crude oil cut fractions may be classified with the aid of a ternary diagram, and the link between this approach and the characterisation parameters introduced above is demonstrated. 2006 Elsevier B.V. All rights reserved.
Keywords: True boiling point distillation; Crude oil; Characterisation; Classification; Gamma distribution; Ternary diagram

1. Introduction Petroleum has an organic origin based on the accumulation of plant and animal matter and the action of heat and pressure over a long period of time on this biological material. It is a complex mixture of hydrocarbons that naturally occurs in reservoir rocks. Petroleum encompasses liquid hydrocarbons referred

Corresponding author. Tel.: +61 8 8303 8020; fax: +61 8 8303 8011. E-mail address: peter.behrenbruch@adelaide.edu.au (P. Behrenbruch). 0920-4105/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.petrol.2005.10.016

to as crude oil, natural gas and solid hydrocarbons such as tars. The chemical constituents of petroleum and techniques available to separate and analyse the various compounds are discussed in Waples (1985). Major groups of compounds found in petroleum are saturated hydrocarbons, including straight chained, branched and cyclic hydrocarb ons, simple aromatic hydrocarbons, small sulphur bearing compounds, resins and very large aromatic asphaltene compounds. The varied composition of petroleum has led to the existence of a number of characterisation systems. Each of these characterisation systems is important for decision making at some stage of exploration, production

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180 Table 1 List of TBP distillation data (source: Chevron, 2005) Country Location Crude oil Date of assay 2000 1996 2001 2001 2002 2002 2002 2000 2000 1995 1987 1998 1997 2001 1994 1997 1999 2000 2002 2000 2000 1999 1994 2003 Whole crude properties Gravity (API) 19.2 32.8 39.3 38.4 20.5 37.7 46.8 35.8 34.4 35.0 38.0 48.1 61.2 41.3 20.8 35.3 45.1 40.1 43.0 43.3 35.9 36.0 35.2 26.0 Specific gravity 0.9390 0.8610 0.8282 0.8329 0.9308 0.8365 0.7936 0.8459 0.8528 0.8497 0.8346 0.7877 0.7344 0.8189 0.9293 0.8482 0.8014 0.8246 0.8108 0.8094 0.8454 0.8447 0.8486 0.8987 Characterization factor (K factor) 11.6 12.3 12.2 12.1 12.0 12.2 12.4 11.9 11.7 11.7 11.7 12.0 12.3 NA 12.0 12.6 12.2 12.4 12.7 12.5 11.6 11.8 12.2 12.1 Sulfur (wt.%) 0.68 0.13 0.19 0.17 0.16 0.09 0.03 0.14 0.15 0.08 0.05 0.04 0.01 0.02 0.20 0.09 0.04 0.06 0.04 0.05 0.34 0.37 0.41 1.55

167

Pour point (C) 32 16 4 10 4 7 21 18 7 10 54 18 51 NA 10 38 1 32 32 16 13 4 1 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Angola Angola Angola Angola Chad Congo Congo Nigeria Nigeria Nigeria Australia Australia Australia Australia Indonesia Indonesia PNG China Thailand Thailand Canada Chile Argentina Venezuela

Africa Africa Africa Africa Africa Africa Africa Africa Africa Africa Australia Australia Australia Australia SE Asia SE Asia SE Asia South China Sea South China Sea South China Sea Canada Sth America Sth America Sth America

Kuito Cabinda Nemba Palanca Doba Blend Kitina N'kossa Bonny Light Escravos Pennington Barrow Island Cossack NWS condensate Thevenard Island Duri Minas Kutubu Light Nanhai Light Benchamas Tantawan Hibernia Rincon Medanito Hamaca

and refining of crude oil. These systems can be broadly grouped as follows. (a) Geochemical parameters (b) Whole crude (live oil) properties (c) Assay properties

Geologists and geochemists use geological parameters to identify and characterise crude oil, for crude oil source rock correlation and to measure the degree of evolution (Tissot and Welte, 1984). A range of geochemical parameters is available for the characterisation of crude oil. Live crude properties are of interest

Fig. 1. TBP distillation curves for a selection of samples listed in Table 1. Numbers correspond to Sample No in Table 1.

168

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

Table 2 Detailed TBP distillation data for Griff in crude oil (source: BHP Billiton, 2005) Cumulative volume percentage 4.0 11.0 14.2 16.5 20.0 26.3 30.3 36.0 40.9 44.2 49.3 52.8 54.6 56.8 60.2 63.6 67.3 70.9 74.5 75.7 76.9 79.6 81.7 83.2 84.8 86.6 88.5 89.5 90.4 91.4 92.6 93.5 94.4 95.3 96.2 96.8 97.7 98.4 99.0 99.3 99.6 100.0 Temperature Temperature Mid Specific API Cumulative (F) (C) volume gravity gravity weight percentage percentage 66.2 140 158 176 194 212 230 248 266 284 302 320 329 338 356 374 392 410 428 446 464 482 500 518 536 554 572 590 608 626 644 662 680 734 752 770 806 842 896 932 986 986.0+ 19 60 70 80 90 100 110 120 130 140 150 160 165 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340 350 360 390 400 410 430 450 480 500 530 530.0+ 2.0 7.5 12.6 15.4 18.3 23.2 28.3 33.2 38.5 42.6 46.8 51.1 53.7 55.7 58.5 61.9 65.4 69.1 72.7 75.1 76.3 78.3 80.6 82.4 84 85.7 87.5 89 89.9 90.9 92 93.1 94 94.9 95.8 96.5 97.3 98.1 98.7 99.2 99.5 99.8 0.5738 0.6351 0.6715 0.6888 0.7002 0.7129 0.7230 0.7271 0.7325 0.7390 0.7434 0.7523 0.7567 0.7590 0.7645 0.7714 0.7866 0.7935 0.7995 0.8050 0.8127 0.8204 0.8272 0.8339 0.8380 0.8419 0.8465 0.8513 0.8556 0.8595 0.8635 0.8706 0.8818 0.8790 0.8794 0.8797 0.8831 0.8928 0.9085 0.9273 0.9536 0.9702 115.1 91.3 79.2 73.9 70.6 67.0 64.2 63.1 61.7 60.0 58.8 56.6 55.5 54.9 53.6 51.9 48.4 46.8 45.5 44.3 42.6 41.0 39.6 38.2 37.4 36.6 35.7 34.7 33.9 33.1 32.4 31.0 29.0 29.5 29.4 29.4 28.7 27.0 24.3 21.1 16.9 14.3 3.1 9.0 11.8 13.9 17.1 23.0 26.9 32.4 37.1 40.3 45.3 48.8 50.6 52.8 56.2 59.7 63.5 67.3 71.1 72.4 73.7 76.6 78.9 80.6 82.4 84.4 86.6 87.7 88.8 89.9 91.3 92.3 93.4 94.5 95.6 96.3 97.3 98.1 98.8 99.1 99.5 100.0 Cumulative weight fraction yact 0.031 0.090 0.118 0.139 0.171 0.230 0.269 0.324 0.371 0.403 0.453 0.488 0.506 0.528 0.562 0.597 0.635 0.673 0.711 0.724 0.737 0.766 0.789 0.806 0.824 0.844 0.866 0.877 0.888 0.899 0.913 0.923 0.934 0.945 0.956 0.963 0.973 0.981 0.988 0.991 0.995 1.000 Temperature difference T (C) 19 41 10 10 10 10 10 10 10 10 10 10 5 5 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 30 10 10 20 20 30 20 30 Weight Weight fraction fraction difference yact density 0.031 0.059 0.028 0.021 0.032 0.059 0.039 0.055 0.047 0.032 0.050 0.035 0.018 0.022 0.034 0.035 0.038 0.038 0.038 0.013 0.013 0.029 0.023 0.017 0.018 0.020 0.022 0.011 0.011 0.011 0.014 0.010 0.011 0.011 0.011 0.007 0.010 0.008 0.007 0.003 0.004 0.005 1.000 0.0016 0.0014 0.0028 0.0021 0.0032 0.0059 0.0039 0.0055 0.0047 0.0032 0.0050 0.0035 0.0036 0.0044 0.0034 0.0035 0.0038 0.0038 0.0038 0.0013 0.0013 0.0029 0.0023 0.0017 0.0018 0.0020 0.0022 0.0011 0.0011 0.0011 0.0014 0.0010 0.0011 0.0004 0.0011 0.0007 0.0005 0.0004 0.0002 0.0002 0.0001

to reservoir engineers, who use these properties to understand fluid flow in the reservoir and to predict changes in reservoir conditions with time, during production. Crude oil assay properties are of most interest to petroleum refiners and crude oil marketers, who need to know the quantities of distillate fractions that a particular crude will produce during distillation. They also need to know the physical and chemical properties of the various fractions. Finally, the composition of crude oil is closely linked to its price.

TBP distillation is one of the most common experimental procedures employed in ascertaining assay properties of a crude oil. Many researchers have investigated the use of TBP distillation data as a basis for characterisation of crude oils. Miquel et al. (1992) recognised that TBP distillation data were the most commonly available information regarding the volatile behaviour of hydrocarbon mixtures. They proposed a characterisation system based on optimally selecting pseudo-fractions from the TBP distillation curve. Haynes and Matthews (1991) proposed a continuous

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

169

Fig. 2. TBP distillation curves for Griffin crude oil.

approach to use TBP distillation data for heavy end characterisation of live crude oils in vapourliquid equilibrium calculations. Specific cut fractions (part of the TBP range), used in the generation of petroleum products directly influence pricing of crude oils. The other major use of these data is in deciding refinery processes needed to refine a given crude oil. TBP distillation can also be used as a method to isolate a specified fraction from a crude oil for testing. The method for performing a TBP distillation experiment is described in ASTM D 2892 (American Society of Testing and Materials, 1999a) and ASTM D 5236 (American Society of Testing and Materials, 1999b) and the procedure can be used to analyse hydrocarbon mixtures, including crude oils, condensates and petroleum fractions. However, the method cannot be used to analyse liquefied petroleum gasses, very light naphtha fractions and fractions with initial boiling points (IBP) greater than 400 C. The TBP experiment is performed by distilling a sample of crude oil or petroleum fraction in a standardised fractionating column that is subject to specified operating conditions. Distillation is carried out at atmospheric pressure from the IBP to about 210 C vapour temperature, and

at reduced pressure beyond this point. Distillation under a partial vacuum avoids cracking of the more complex components at elevated temperatures. Samples of distillate are collected at specified temperature cut points. Mass and density of each fraction are measured, and the distillation yield by mass is calculated. Volumetric yield can be estimated with mass and density data. Vapour temperatures measured at reduced pressure are translated to Atmospheric Equivalent Temperature (AET). Distillation can usually be continued up to an AET of approximately 400 C. Final results of such experiments are TBP curves in mass and/or volume yield versus boiling temperature expressed in AET. The shape of a TPB curve is dependent on the type and quantities of hydrocarbon compounds that make up the mixture being analysed. As such, this curve uniquely describes a given crude oil in terms of its chemical makeup. As an indicator of a crude oil's compositional makeup, a TBP distillation curve has the added advantage that it is extensively performed for marketing purposes on all crude oils. These data are also easily accessed, as crude oil marketing companies publish these data in the public domain. Some TBP distillation data and other assay properties published on the Internet by Chevron (2005) are listed in Table 1. The list covers a range of crude oil types from gas condensates to heavier oils, containing samples from across the world, and constitutes one of two major data sets used in the current study. Fig. 1 shows the TBP curve shapes for a selection of crude oils from this set of data. Three marker crude oils (Brent, West Texas Intermediate (WTI) and Tapis) are also shown for comparison. The following detailed assay reports for Australian crude oils published on the BHP Billiton web site (BHP Billiton, 2005) have also been used in the analysis. Griffin Crude Oil Gippsland Crude Oil Cossack Crude Oil Laminaria Crude Oil North West Shelf Condensate Typical detailed TBP distillation data for Griffin crude oil are shown in Table 2 as an example. The corresponding TBP distillation (cumulative weight fraction distilled versus temperature) curve is shown in Fig. 2(a). This curve can be converted in to a weight fraction density curve, as shown in Fig. 2(b). The weight fraction density can be calculated as follows. Weight fraction density of the ith fraction Dyact =DT where, yact = yact,I yact,i1 T = TI Ti1

170

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

yact,I Ti

Cumulative weight fraction of the ith fraction Boiling temperature of ith fraction

The scatter seen on the experimental differential curve could be attributed to two factors. The first is due to the lumping of compounds with close boiling points. The other is due to fluctuations in conditions during the experiment. The variation is highlighted in the weight fraction density curve, as this is a differential form of the cumulative weight fraction distribution, which is somewhat smoother by comparison. A mathematical characterisation of TBP distillation curves would be useful for a number of applications. These include physical property predictions, heavy end characterisation for EOS modelling and crude oil valuation. In addition, the comparison of shapes of different TBP curves would allow for the identification of groupings of oils based on composition. The following sections detail a method of mathematically characterising a TBP distillation curve using the two-parameter gamma distribution. The fitting methodology used is described and results for a range of crude oil types have been presented, demonstrating the flexibility of the proposed methodology. A correlation between the fitting parameters and API

gravity is also shown. A new classification system based on temperature cut fractions from TBP distillation data, using a ternary diagram is introduced as an alternative to the PNA (Paraffins, Naphthenes and Aromatics) ternary diagram (Tissot and Welte, 1984). 2. TBP curve characterisation 2.1. Gamma distributions The gamma distribution has been widely used as a mathematical function to characterise hydrocarbon mixtures. The main focus of this work has been to describe the molecular weight distribution with the objective of predicting phase behaviour of hydrocarbon mixtures. Whitson (1983) proposed a continuous approach for modelling the molecular weight distribution of a hydrocarbon mixture using the three-parameter gamma distribution for Equation of State (EOS) modelling. Cotterman et al. (1985) also used the three-parameter gamma distribution to describe the molecular weight distribution of hydrocarbon mixtures for process design applications. The probability distribution function (PDF) for a gamma distribution can be described using the

Fig. 3. Probability density functions and cumulative probability distributions for the gamma distribution with varying values of and .

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180 Table 3 Gamma distribution fitting for Griffin crude oil TBP distillation data Experimental Temperature T (C) 19 60 70 80 90 100 110 120 130 140 150 160 165 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340 350 360 390 400 410 430 450 480 500 530 530+ Cumulative weight percentage 3.1 9.0 11.8 13.9 17.1 23.0 26.9 32.4 37.1 40.3 45.3 48.8 50.6 52.8 56.2 59.7 63.5 67.3 71.1 72.4 73.7 76.6 78.9 80.6 82.4 84.4 86.6 87.7 88.8 89.9 91.3 92.3 93.4 94.5 95.6 96.3 97.3 98.1 98.8 99.1 99.5 100.0 Cumulative weight fraction yact 0.031 0.090 0.118 0.139 0.171 0.230 0.269 0.324 0.371 0.403 0.453 0.488 0.506 0.528 0.562 0.597 0.635 0.673 0.711 0.724 0.737 0.766 0.789 0.806 0.824 0.844 0.866 0.877 0.888 0.899 0.913 0.923 0.934 0.945 0.956 0.963 0.973 0.981 0.988 0.991 0.995 1.000 Weight fraction density 0.0016 0.0014 0.0028 0.0021 0.0032 0.0059 0.0039 0.0055 0.0047 0.0032 0.0050 0.0035 0.0036 0.0044 0.0034 0.0035 0.0038 0.0038 0.0038 0.0013 0.0013 0.0029 0.0023 0.0017 0.0018 0.0020 0.0022 0.0011 0.0011 0.0011 0.0014 0.0010 0.0011 0.0004 0.0011 0.0007 0.0005 0.0004 0.0002 0.0002 0.0001 Predicted Cumulative weight percentage ypred 0.005 0.081 0.114 0.150 0.189 0.231 0.274 0.317 0.361 0.404 0.446 0.487 0.507 0.527 0.564 0.600 0.633 0.665 0.695 0.722 0.748 0.771 0.793 0.813 0.831 0.848 0.863 0.877 0.890 0.901 0.912 0.921 0.930 0.950 0.956 0.961 0.969 0.976 0.983 0.987 0.991 Weight fraction density 0.0007 0.0030 0.0035 0.0038 0.0041 0.0042 0.0043 0.0044 0.0043 0.0043 0.0042 0.0040 0.0039 0.0038 0.0037 0.0035 0.0033 0.0031 0.0029 0.0027 0.0025 0.0023 0.0021 0.0019 0.0018 0.0016 0.0015 0.0013 0.0012 0.0011 0.0010 0.0009 0.0008 0.0006 0.0005 0.0005 0.0004 0.0003 0.0002 0.0002 0.0001 (yact ypred)2 0.00070 0.00008 0.00002 0.00012 0.00034 0.00000 0.00002 0.00004 0.00010 0.00000 0.00004 0.00000 0.00000 0.00000 0.00000 0.00001 0.00000 0.00006 0.00027 0.00000 0.00011 0.00003 0.00002 0.00005 0.00005 0.00002 0.00001 0.00000 0.00000 0.00001 0.00000 0.00000 0.00002 0.00003 0.00000 0.00001 0.00002 0.00003 0.00002 0.00002 0.00002 0.00228

171

(yact yact)2 0.39017 0.31994 0.28905 0.26691 0.23487 0.18116 0.14949 0.10998 0.08102 0.06382 0.04106 0.02810 0.02239 0.01629 0.00877 0.00344 0.00043 0.00030 0.00307 0.00467 0.00662 0.01218 0.01779 0.02261 0.02835 0.03548 0.04425 0.04900 0.05399 0.05923 0.06624 0.07148 0.07749 0.08373 0.09022 0.09447 0.10072 0.10586 0.11047 0.11247 0.11517 3.57276

following two-parameter form (Hastings and Peacock, 1975). p x where, x Variable (0 x b ) Shape parameter ( N 0) xa1 ex=b ba Ca 1

()
Z Ca
0

Scale parameter ( N 0) The gamma function with parameter , given by,

l

eu ua1 du
x

The cumulative probability function is given by,

Z PX V x
0

pxdx

172

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

Probability density functions and cumulative probability distributions for different values of parameters and are shown in Fig. 3, indicating the diversity of shapes that can be accommodated. From the definition of the two-parameter gamma distribution, the mean, standard deviation and mode of the distribution are as follows. Mean p Standard deviation a Mode ( 1) 2.2. Methodology The two parameters ( and ) of the gamma distribution described previously are varied to obtain the best fit with experimental data. The data in Table 2 for the detailed assay of Griffin crude oil are used as an example. The predefined function GAMMADIST in MS Excel is used to generate the cumulative weight fraction data and weight fraction density data. The goodness of fit is measured using the coefficient of determination (R2) and the Root Mean Square Error (RMSE). The Solver plug-in

Fig. 5. Gamma distribution fitted TBP distillation curves for Cossack crude oil.

from MS Excel is used to tune and to the optimum values while minimising RMSE. The application of the methodology described above for the Griffin crude oil (Table 2 and Fig. 2(a) and (b)) is shown in Table 3. The fitted values of and are 2.90 and 63.48 respectively, and the cumulative weight fraction and weight fraction density curve are shown in Fig. 4(a) and (b), respectively. The RMSE and R2 are calculated using Eqs. (4) and (5), respectively.
v uX n u u yact ypred 2 t
i1

RMSE

n
n X

yact ypred 2 yact Pact 2 y 5

R2 1 P n
i1

i1

where,
Fig. 4. Gamma distribution fitted TBP distillation curves for Griffin crude oil.

yact ypred

Actual (experimental) value Predicted value from model

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

173

y act n

Mean of actual values Number of points

For the above example, r 0:00228 RMSE 0:0075 41 R2 1 0:00228 0:9994 3:57276

These values, together with a visual comparison of the fitted curve to the actual data, indicate that a good fit has been achieved. RMSE was selected as the parameter for optimisation (i.e., minimisation of error) using the Solver as it showed a better convergence when compared to the results obtained by using R2 (i.e., aiming for an R2 value of 1.0). This fitting is independent of the initial values used for and . The fitted curve shows a good visual agreement with experimental data points and R2 values of 0.99 or better in all cases.

Fig. 7. Gamma distribution fitted TBP distillation curves for Laminaria Crude Oil.

2.3. Results The same methodology described above has been applied to other detailed samples. Fitted distributions obtained for four other detailed TBP distillation curves are shown in Figs. 5 6 7 and 8. As can be seen from these graphs, the methodology described has the ability to fit experimental TPB curves very well for a range of oil types. The partial TBP distillation curves listed in Table 1 consist of weight fraction data collected at eight temperature cut points ranging from 80 C to 570 C. Other basic assay properties such as API gravity, sulphur content and UOP K factor are also available. The UOP K factor (or Watson characterisation factor) is defined as follows (Nelson, 1958) p 3 TB K 6 S where, TB
Fig. 6. Gamma distribution fitted TBP distillation curves for Gippsland crude oil.

Average molal boiling temperature in degrees Rankine Specific gravity at 60 F

174

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180 Table 4 UOP K factor values for paraffin, naphthene and intermediate base crude oils (from Nelson, 1958) UOP K factor 12.912.2 12.211.5 11.510.5 Crude oil base Paraffinic base Intermediate base Naphthene base

The value of the K factor allows the characterisation of the base of a crude oil as paraffin base, naphthene base or intermediate base. The ranges of K factor values corresponding to each base are listed in Table 4. The same methodology described above for detailed assay results was used to fit a gamma distribution to each of these curves. The results are shown in Table 5. As with the detailed assay results, good fitting can be observed in all cases. These results also demonstrate the flexibility of the new methodology. As mentioned previously and shown in Fig. 1, the 24 samples in Table 1 cover a wide range of crude oil types. Represented are oils with API gravity ranging from 19.2 (for Kuito crude oil, a heavy oil from Angola) to 61.2 (for North West Shelf Condensate, a gas condensate). In addition, the list contains crude oils with a wide distribution of total sulphur content (0.0 to 1.55 wt.%), UOP K factor (11.6 to 12.7) and pour point ( 54 C to 38 C). For this diverse group of oils, the methodology is able to fit gamma distributions with the coefficient of

determination (R2 ) being better than 0.99 in each case. A visual comparison of the fitted and actual curves confirms the excellent fit. An attempt was made to correlate the gamma distribution characterisation parameters ( and ) with API gravity. Various functional forms were tested against the available data. The form that gave the best approximation to the experimental data was as follows. APIGravity aba1 bc where, and a, b and c Gamma distribution characterisation parameters Constants 7

Fig. 8. Gamma distribution fitted TBP distillation curves for North West Shelf Condensate.

The functional form of Eq. (7) was selected as the term ( 1) corresponds to the modal temperature of the fitted distribution. As such the expectation is that there will be a relationship between the modal temperature and API gravity. The constants a, b and c were tuned to get the best possible fit to the experimental data by using the Solver plug-in from MS Excel. The results for prediction of API gravity using Eq. (7) are compared to experimental values for all 24 samples in Fig. 9. Values for constants a, b and c, and the coefficient of determination (R2 ) and RMSE for these data are shown on the graph. Although there is reasonable agreement between experimental and predicted values, the degree of scatter on the graph is high, with indication of a secondary trend. This aspect was further investigated in order to determine the reasons for this scatter, in terms of other available crude oil properties. The scatter observed on Fig. 9 appeared to be related to the UOP K factor and outlying points have been annotated on that figure. As evident from Fig. 9, samples with a UOP K factor greater than 12.2 cluster as a separate trend, distinct from the main trend. This group corresponds to crude oils having a paraffin base. Therefore, an

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180 Table 5 Gamma distribution fitting results for the samples listed in Table 1 Crude oil Whole crude properties Gravity (API) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Kuito Cabinda Nemba Palanca Doba Blend Kitina N'kossa Bonny Light Escravos Pennington Barrow Island Cossack NWS condensate Thevenard Island Duri Minas Kutubu Light Nanhai Light Benchamas Tantawan Hibernia Rincon Medanito Hamaca 19.2 32.8 39.3 38.4 20.5 37.7 46.8 35.8 34.4 35.0 38.0 48.1 61.2 41.3 20.8 35.3 45.1 40.1 43.0 43.3 35.9 36.0 35.2 26.0 Characterization factor (K factor) 11.6 12.3 12.2 12.1 12.0 12.2 12.4 11.9 11.7 11.7 11.7 12.0 12.3 NA 12.0 12.6 12.2 12.4 12.7 12.5 11.6 11.8 12.2 12.1 Sulfur (wt.%) 0.68 0.13 0.19 0.17 0.16 0.09 0.03 0.14 0.15 0.08 0.05 0.04 0.01 0.02 0.20 0.09 0.04 0.06 0.04 0.05 0.34 0.37 0.41 1.55 Pour point (C) 32 16 4 10 4 7 21 18 7 10 54 18 51 NA 10 38 1 32 32 16 13 4 1 40 Gamma distribution fitting parameters Alpha 4.21 2.42 2.22 2.54 6.09 2.18 1.95 3.30 3.20 4.56 3.67 2.23 2.42 4.87 4.07 3.32 2.15 3.54 2.43 2.81 2.37 2.77 2.49 3.04 Beta 109.9 182.2 147.2 136.6 88.7 170.1 139.3 93.4 100.5 66.0 69.0 103.7 54.7 50.6 142.0 131.6 112.9 97.9 144.2 111.7 154.4 128.0 146.0 139.2

175

R2 1.000 0.996 0.997 0.996 0.999 0.997 0.997 0.994 0.996 0.998 0.999 0.998 1.000 0.998 0.998 0.998 0.995 0.994 0.994 0.999 0.998 1.000 0.999 0.996

attempt was made to predict the API gravity of crude oils with a paraffin base (K factor greater than 12.2) using the functional form of Eq. (7) and tuning the

constants a, b and c to match the experimental data. Results from this API gravity prediction for paraffin base crude oils are compared to experimental values

Fig. 9. Predicted versus experimental values for API gravity using Eq. (7) for the 24 samples listed in Table 1. The UOP K factor for the outlying samples has been annotated.

Fig. 10. Predicted versus experimental values for API gravity using Eq. (7) for paraffin base crude oils (UOP K factor N 12.2).

176

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

gravity for the non-paraffin base crude oils (K factor less than 12.2) using the same functional form and tuning a, b and c to get the best match with experimental data. Results are shown in Fig. 11, indicating that, with the exception of three scattered point, there is also very good agreement between the predicted and experimental API gravities. 3. Cut fraction characterisation 3.1. Methodology One possible framework for classification of crude oils, based on assay properties is given in Tissot and Welte (1984). They propose a classification based on relative quantities of paraffins, naphthenes and aromatics (PNA) contained in a particular crude oil, displayed on a ternary diagram. They have applied this technique to a large number of samples and show how this ternary diagram can be used for oil oil and oilsource correlations. An example of their classification technique is shown in Fig. 12. One draw back of this PNA framework is that data for plotting the ternary diagram are not readily available.

Fig. 11. Predicted versus experimental values for API gravity using Eq. (7) for non-paraffin base crude oils (UOP K factor b 12.2). The UOP K factor for the outlying samples has been annotated.

in Fig. 10. It can be seen that there is very good agreement between the predicted and experimental values. An attempt was then made to predict API

Fig. 12. Paraffins, naphthenes and aromatics (PNA) composition classification ternary diagram (from Tissot and Welte, 1984, p. 440).

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

177

Fig. 13. Cut fraction temperature definitions used by three companies based on data published on their respective web pages.

An assay report would normally only contain the PNA break down for the lighter fraction (usually boiling at less than 200 C) of the crude oil. As such, it is necessary to perform a separate analysis to populate this type of ternary plot. The latter is perhaps one reason that this terminology appears to be used relatively infrequent. For the abovementioned reasons, a new ternary plot based on assay properties is proposed, utilising three boiling temperature cut fractions from TBP distillation. It should be noted that boiling temperature ranges for petroleum fractions vary regionally and also among companies. Definitions used by three companies are shown in Fig. 13, based on data published on the respective company web sites. This figure indicates that, although companies use slightly different definitions for individual cut fractions, there is some agreement between them. Three cut fractions were selected to correspond to distinct refinery products. The cuts fractions were also picked to have a good distribution of points on the

resulting ternary plot for various types of crude oils available, ranging from gas condensates to heavy oils. Table 6 shows the three cut fractions that were selected for the purpose of drawing the new ternary plot. Weight fractions are read by interpolating cumulative weight fractions distilled at 200 C and 350 C from the distillation curve. The three weight

Table 6 Boiling temperature cut fractions selected for plotting the proposed cut fraction ternary diagram Cut Fraction No. 1 2 3 Temperature range IBP to 200 C 200 C to 350 C 350 C+ Main distillation products Naphtha Kerosene and Gas Oil Atmospheric Residue

Fig. 14. Composition of temperature cut fractions (modified after Hunt, 1996, p. 45).

178

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

Fig. 15. Cut fraction ternary diagram showing the 24 samples listed in Table 1.

fractions, adding up to 1.0, are then calculated as follows. IBP to 200 C fraction 200 C to 350 C fraction 350 C + fraction where, W200 W350 Cumulative weight fraction distilled at 200 C in Cumulative weight fraction distilled at 350 C W200 W350 W200 1 W350

Fig. 14 shows the link between cut fractions selected above and chemical composition for a naphthenic crude oil (Hunt, 1996). It can be seen that there is some degree of correspondence between the selected fractions and composition elements. The IBP to 200 C fraction is richer in normal and iso-paraffin, while the 350 C + fraction is richer in heavy aromatics and nitrogen, sulphur and oxygen (NSO) compounds. 3.2. Results The proposed cut fraction ternary diagram, with the samples from Table 1 plotted, is shown in Fig. 15. It can

Fig. 16. Relationship between gamma distribution characterisation parameters and cut fraction ternary diagram.

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

179

Fig. 17. Constant API gravity lines for paraffin base crude oils (UOP K factor N 12.2).

be seen from the plot that there is a good distribution of points on this ternary diagram, considering the varied fluid types plotted. The location of a crude oil on the proposed ternary diagram is governed by the relative quantities of each fraction produced, being a function of the shape of the TBP distillation curve. As such, there is a unique relationship between the two-parameter gamma distribution characterisation system described previously and this ternary diagram. Fig. 16 shows constant and constant lines superimposed on the ternary plot. It can be seen that a given pair of and lines have only one

intersection. A given TPB curve shape (as defined by its and values) will be uniquely located on the ternary plot. It has also been possible to identify the distribution of API gravity on the new cut fraction ternary diagram. As described previously, a good correlation has been obtained between the gamma distribution characterisation parameters ( and ) and API gravity for paraffin and non-paraffin base crude oils. This relationship, based on the functional form of Eq. (6), was used to calculate constant API gravity lines on the ternary diagram.

Fig. 18. Constant API gravity lines for non-paraffin base crude oils (UOP K factor b 12.2).

180

P. Behrenbruch, T. Dedigama / Journal of Petroleum Science and Engineering 57 (2007) 166180

Eq. (6) can be rearranged as follows. n o b G=a1=c b =a1 where, and G a, b and c gamma distribution characterisation parameters API gravity Constants 8

diagram. It is possible to display a wide range of crude oil types on the cut fraction ternary diagram. This ternary diagram has the advantage that the data required to populate it is easily available from crude oil assay reports. The distribution of crude oils according to their API gravity on the cut fraction ternary diagram has also been described. It has been shown that this distribution is a function of the base of a crude oil as defined by the UOP K factor. The relationship developed suggests that API gravity is a function of the modal boiling temperature of a crude oil. Acknowledgments

This formulation allows the value corresponding to a given value to be calculated for a specified API gravity. As shown previously, a given set of and values results is a unique point on the ternary diagram. As such, a constant API gravity locus on the ternary diagram can be established by varying and calculating , using Eq. (8). Fig. 17 shows constant API gravity lines for the paraffin base crude oils from Table 1 (as defined by having a UOP K factor N 12.2). It can be seen that the data points match well with the predicted constant API gravity curves. Constant API gravity lines for non-paraffin base crude oils (as defined by having a UOP K factor b 12.2) are shown in Fig. 18, indicating good agreement between the data points and predicted curves. A comparison of the two sets of constant API gravity curves is shown in Figs. 17 and 18, indicating that the two sets are significantly different. It can be concluded that the distribution of crude oils according to API gravity on the cut fraction ternary diagram is dependent on their base. 4. Summary and conclusions A system for the characterisation of crude oils, based on TBP distillation curve data using a twoparameter form of the gamma distribution from statistics, has been described. The new methodology has been tested against a broad selection of crude oil types and is able to fit the cumulative TBP distillation curve with a high degree of correlation in all cases. The gamma distribution characterisation parameters ( and ) have been related to API gravity using a single functional form, having different constants depending on the base of the crude oil as defined by the UOP K factor. A cut fraction ternary diagram for crude oils has also been introduced as an alternative to the traditional PNA

The authors wish to gratefully acknowledge the support of the sponsors: BHP Billiton, Chevron, Santos and Woodside Energy, for their financial support, field data and permission to publish the results presented. References
American Society of Testing and Materials, 1999a. ASTM D 2892: Distillation of Crude Petroleum (15-Theoretical Plate Column). Annual Handbook of ASTM Standards, vol. 05.01. American Society of Testing and Materials, 1999b. ASTM D 5236: Distillation of Heavy Hydrocarbon Mixtures (Vacuum Potstill Method). Annual Handbook of ASTM Standards, vol. 05.01. BHP Billiton, 2005. BHP Billiton (online). Available from: http:// globaloil.bhpbilliton.com/ (accessed on 4 April 2005). Chevron, 2005. Chevron (online)Available from: http://www.chevron. com/crudemarketing/ (accessed on 4 April 2005). Cotterman, R.L., Bender, R., Prausnitz, J.M., 1985. Phase equilibria for mixtures containing very many componentsdevelopment and application of continuous thermodynamics for chemical process design. Ind. Eng. Chem. Process Des. Dev. 24, 194203. Hastings, N.A.J., Peacock, J.B., 1975. Statistical Distribution. Butterworth and Co (Publishers) Ltd., London. Haynes, H.W., Matthews, M.A., 1991. Continuous-mixture vapour liquid equilibria computations based on true boiling point distillation. Ind. Eng. Chem. Res. 30, 19111915. Hunt, J.M., 1996. Petroleum Geochemistry and Geology, 2nd edition. W.H. Freeman and Company, USA. Miquel, J., Hernandez, J., Castells, F., 1992. A new method for petroleum fractions and crude oil characterization. SPE Reserv. Eng. 265270 (May). Nelson, W.L., 1958. Petroleum Refinery Engineering, 4th edition. McGraw-Hill Book Company Inc, New York. Tissot, B.P., Welte, D.H., 1984. Petroleum Formation and Occurrence, 2nd edition. Springer-Verlag, Berlin Germany. Waples, D.W., 1985. Geochemistry in Petroleum Exploration. D. Reidel Publishing Company, Dordrecht, Holland. Whitson, C.H., 1983. Characterizing hydrocarbon plus fraction. Soc. Pet. Eng. J. 683694 (August).