Express estimation of crude oil TBP curves only from the viscosity of the crude oil

G. Argirov, S. Ivanov, G. Cholakov Bulgaria, 8104 Bourgas, Lukoil Neftochim Bourgas AD, Heavy Residue Upgrading Complex-Project, e-mail: [email protected]

Key words: crude oil, distillation, TBP prediction, viscosity.

Abstract
Crude oil TBP curves are well recognized as an important characteristic, needed to determine the potential of fractions obtained by distillation, to monitor distillation units, to verify the identity of crude oil samples, etc. They can be obtained experimentally or calculated by a variety of methods, which have different precision. The precision of the particular method determines the applicability of the obtained TBP data. Riazis two-parameter distribution model and experimental crude assay data for 117 crudes from all over the world have been used to derive a model for prediction of true boiling point (TBP) curves from the kinematic viscosity of the samples at 37,8 oC. The model was tested by prediction of the cumulative mass fractions of all studied crude samples. It was found that the TBP distribution for fractions boiling up to 340 oC could be predicted with an average relative deviation of the order of 6 %. This is higher than the respective deviation of previously proposed models, e.g. - Riazis model, with an average deviation of 3.3 %. However, these models require experimental data for several properties, while in practical terms the precision of our model is sufficient to detect any significant upsets in the operation of refinery crude distillation units (CDUs), so its computerized procedure may be used as a quick tool for monitoring of the CDU operation.

INTRODUCTION For modern refineries frequent crude oil switching and disturbances in crude distillation unit (CDU) operations often propagate into bigger issues downstream. Managing the change of crudes and diminishing disturbances require regulatory control of the daily performance necessary to respond quickly and adequately to setpoint changes. The quality and value of a crude oil significantly depend on its TBP (true boiling point) distillation curve. Unfortunately, the experimental determination of TBP curves is costly and time consuming, so it is impractical to use them as a tool for daily monitoring of CDUs. This calls for the development of TBP calculation methods, requiring a minimum of experimental data, but having sufficient precision for daily monitoring of CDU operations. Riazi developed a three-parameter distribution model, which is widely used for estimation of the boiling points and other properties of C7+ fractions of petroleum fluids [1, 2]. It predicts TBP distributions with high precision, when at least three experimental points of the TBP curve of the crude are available [3]. Moreover, it can be used for predictions of the rest of the distillation curves, needed in petroleum processing (ASTM D 86, EFV, CD, etc.) If distillation data are not available, the values of at least three crude oil properties (molecular mass, density and refraction index) have to be known. Riazis method can not be used directly for daily monitoring of CDU operations, because of the need of significant amount of tests, but it gives the general lines along which this can be achieved. Stratiev et al. [4, 5] developed further Riazis ideas and showed that the boiling temperature and other property distributions can be calculated when experimental density and distillation data up to 300 oC are an available. The Lukoil Neftochim Bourgas AD (LNB) refinery has developed a data base of potential feedstocks - 117 crude oil assays, characterized by TBP curves, density, sulphur, viscosity, metals, Conradson carbon, and asphalthene content of crude oils and TBP cuts. The crude oils originate from different parts of the world . The aim of this work is to use the LNB database and Riazis ideas, for development of a model for crude oil TBP curve prediction from one routine measurement the crude oils kinematic viscosity at 37.8 oC.

EXPERIMENTAL The TBP distillation of all 117 investigated crude oil samples has been carried out in the AUTODEST 800 Fisher column apparatus, according to STM-D 2892 - for the atmospheric part of the test, and according to STM-D 5236 for the vacuum one. The TBP distillation has been performed in the AUTODEST 800 Fisher column at pressure drop

from 760 to 2 mm Hg and in the AUTODEST 860 Fisher column from 1 to 0. 2 mm Hg. The sulphur content has been analyzed according to STM-D 4294. Density at 20 oC has been determined according to STM-D 1298. The viscosity has been measured according to STM-D 445. Table I characterizies the range of variation of the properties of the crude oils. Table I. Characterization of the properties of the crude oils in the database.
Property or TBP range Total sulfur, % Specific gravity @150C range per group Viscosity, 37.8 oC, mm2/s TBP yields, % IBP 70 oC Number of crude samples in group/range of values in group I II III IV 64/ 0.5 44/0.51 2.0 4/2.01 2.5 5/>2.5 0.7883-0.8886 0.8095-0.8970 0.8685-0.9047 0.8581-0.9024 88/ 10.00 I 9/1 - 2 incl. 62/3 - 5 incl. 16/6 - 8 incl. 1/9 - 10 incl. 5/1 - 2 incl. 73/3 - 5 incl. 10/6 - 8 incl. 23/5 - 7 incl. 50/8 - 10 incl. 15/11-13 incl. 47/5 - 7 incl. 38/8 - 10 incl. 3/10 - 12 incl. 42/6 - 8 incl. 42/9 - 11 incl. 4/12 - 14 incl. 62/7 - 9 incl. 17/10-12 incl. 6/13-15 incl. 3/16-17 incl. 54/10 -12 incl. 20/13 -15 incl. 6/16 -18 incl. 4/19 -21 incl. 2/22 -23 incl. 1/26 1/20 4/21 - 23 incl. 7/25 - 27 incl. 25/28 -30 incl. 33/31 -33 incl. 10/34 -36 incl. 4/37 -39 incl. 1/40 2/41 1/48 37/1 10 incl. 44/11 20 incl. 7/21 24 incl. 17/10.01 20.0 II 9/1 - 2 incl. 8/3 - 5 incl. 8/1 - 2 incl. 9/3 - 5 incl. 7/3 - 5 incl. 10/6 - 7 incl. 1/3 16/5 - 6 incl. 1/4 16/6 - 8 incl. 14/6 - 8 incl. 3/9 14/10 - 12 incl. 2/13 1/15 7/20.01 30.00 III 2/1 - 2 incl. 5/3 - 5 incl. 2/1 - 2 incl. 5/3 - 5 incl. 4/3 - 5 incl. 3/6 7/4 - 6 incl. 7/6 - 7 incl. 1/5 6/6 - 7 incl. 1/8 6/9-11 incl. 1/12 5/30.01 40.00 IV 4/1 - 2 incl. 1/3 5/1 - 2 incl. 5/4 - 5 incl. 5/3 - 4 incl. 5/4 - 6 incl. 1/5 4/6 - 7 incl. 1/7 4/9-11 incl.

70 - 100 oC 100 150 oC 150 190 oC 190 235 oC 235 - 280 oC

280 343 oC

343 565 oC

6/30 33 incl. 10/35 41 incl. 1/50

6/25 35 incl. 1/41

5/33 40 incl.

Above 565 oC

12/13 23 incl. 12/13 23 incl.

6/24 30 incl. 1/35

4/30 31 incl. 1/34

As would be shown later, we needed data also for the volume average boiling points (VABP) and the viscosities at 38 oC of the discrete fractions of the crude oils, presented in Table I. Some of these data were measured experimentally, others had to be calculated or approximated to known values. Table II summarizes the origin of the data. Table II. Estimation of the volume average boiling points (VABP) and data for the viscosity at 38 oC (V38) of the crude oil cuts.
TBP range, o C IBP - 70 70 - 100 100 - 150 150 - 190 190 - 235 235 - 280 280 - 343 343 - 565 565 + VABP, K 326.15* 358.15 398.39 442.57 485.85 529.96 584.29 716.67 949.47** V38, mm2/s 0.32 0.46 0.64 0.96 1.47 2.47 5.11 52.83 1.2x109 Viscosity at 38 oC, mm2/s

estimated as decribed in [6] estimated as decribed in [6] estimated as decribed in [6] estimated as decribed in [6] measured at 38 oC measured at 38 oC measured at 38 oC measured at 38 oC measured at 60 and 99 oC; converted to viscosity at 38 oC as described in [7]

*The initial boiling point, IBP was assumed to be 36 oC (the boiling point of n-pentane). **For calculation of the 565 0C+ range, the mass average boiling points were calculated by Riazis two parameter method [3], and the VABP of the cuts was assumed to be approximately equal to the mass one.

Methodology Riazis three-parameter distribution model for the absolute boiling points can be presented in the following form [3]: 1 B A log To + T 1 x = To B
1

(1),

in which T is the temperature on the distillation curve in kelvin and x is the cumulative volume or mass part of the respective distillate, obtained up to that temperature. In order to distinguish cumulative concentrations of distillates in the crude oil from discrete concentrations, we shall further denote the former as xC, and the latter as xD.

A, B, and To are the three parameters, which have to be determined by regression of available distillation data. For atmospheric and vacuum residues the parameter B can be considered a constant with a value of 1.5. To is the initial boiling point (IBP, T at xC = 0). However, it should be lower than the experimentally determined IBP and can be determined by regression of experimental distillation data for xC > 0 and iterations for its value [3]. Alternatively, as shown by Stratiev et al. [5], To might successfully be replaced by the normal boiling point of isobutane (-11.6 0C), the lowest-boiling component in the crude. For our calculations it is convenient to transform equation (1) into:
A log U = 1 1 x C B
1 B

(2),

in which U is a normalized temperature, defined by:

To + T =U To

(3)

Solving equation (2) for xC (the cumulative mass fraction distilled up to temperature T, CDM) leads to the following relationship:

xc [ A, B , U ] = 1 e

BU B A

(4)

For a complex continuous mixture, such as a crude oil, equation (4) can be represented as:

xC [A, B, U ] = xD [A, B , U ]dU

(5),

where the function xD [A, B, U] represents the mass of a pseudocomponent, distilled within a certain infinitely small temperature range (dU). The function xC [A, B, 0] is equal to zero for To = T, and Riazis two-parameter model can be differentiated in terms of the normalized temperature U:

x D [A _, B _,U ] = U xC [ A, B,U ]

(6)

The output of (6) gives a relation between the concentration of a discrete pseudocomponent, distilled within a narrow temperature range (which for

practical purposes can be represented by its average boiling point) and the parameters of Riazis equation:

xD [ A, B,U ] =

Be

BU B 1+ B U A

(7)

The parameters of (7) can be calculated from experimental data for discrete fractions (pseudocomponents) by an algorithm, described in [3], if the temperatures in U (To and T) are defined for the respective discrete yields of distillate (xD). However, for achieving the aim of this article, namely development of a model for crude oil TBP prediction from crude viscosity, a combined approach including (7) and other suitable correlations needs to be developed.

Development of a correlation between average boiling point and viscosity A basic concept of the proposed model is that the properties of a petroleum mixture are built-up by a linear or non-linear (depending on the particular property) contribution of the pseudocomponents, contained in that mixture. The viscosity contribution of a pseudocomponent can be expressed by its viscosity blending number (VBN), thereby applying the equation of Refutas [6]. According to this equation the crude oil viscosity could be calculated, knowing the viscosity of its constituent hydrocarbon fractions. The calculations are carried out in three steps: 1. Calculation of the VBN of each pseudocomponent (discrete fraction) of the crude oil at the same temperature by:
VBNDi = 14.534 xDi ln [ln (vDi+0.8)]+10.975

(8),

where vDi is the kinematic viscosity of the crude oil pseudocomponent in mm2/s, and xDi its mass %. 2. Additive calculation of the viscosity number of the crude oil, VBNcrude by:
VBNcrude = (xD1 VBND1) + (xD2 VBND2) ++ (xDn VBNDn)

(9),

where xDi is the percentage mass part of each hydrocarbon fraction of the crude oil. 3. Once the viscosity blending number of a crude oil is obtained from equation (9), the viscosity of the crude oil, vcrude, can be determined by using the invert of equation (8):

VBNcrude10.975

vcrude = e

14.534

0.8

(10)

where VBNcrude is the viscosity blending number of the crude oil, determined by (9). Our next task was to develop a correlation for prediction of viscosity blending numbers of the discrete fractions of the crudes from their volume average boiling points. For this purpose the data, presented in Table II were used. For our next steps in the algorithm, it was convenient to present the VABPs of the discrete cuts (Table II) as normalized temperatures (U), calculated by eq. (2) above. Like in [4], the Riazis IBP of the crude - To, was assumed to be equal of the isobutane boiling point (261 K) the lowest-boiling component in the crude. The viscosities at 38 oC, determined as described in Table II, were used to calculate the VBNs of the respective discrete fractions of each of the 117 crude oil samples. Then the desired correlation, giving the dependence of pseudocomponents VBN from its normalized boiling point, was obtained by regression of the VBN and U data, in the following form:
VBNDi [U]=14.8567 (0.919903U+ln(0.919903U))

(11)

Table III summarizes the data for the volume average boiling points (VABPs, their transformation into normalized temperatures (U) and the viscosity blending numbers of the discrete fractions (VBN), calculated by (11). Table III. Volume average boiling points, normalized temperatures (U) and viscosity blending numbers of the discrete fractions.

Discrete fraction boundaries, TBP range, 0C IBP - 70 70 - 100 100 - 150 150 - 190 190 - 235 235 - 280 280 - 343 343 - 565 565 + Discrete fraction boundaries, TBP range, normalized temperature 0.0003 0.314 0.314 0.429 0.429 0.621 0.621 0.774 0.774 0.946 0.946 1.119 1.119 1.360 1.360 2.211 2.211 3.887 VABP, K U, (normalized VABP) VBNcalc

326.15 358.15 398.39 442.57 485.85 529.96 584.29 716.67 949.47

0.25 0.37 0.52 0.70 0.86 1.03 1.24 1.75 2.64

-20.04 -10.25 -3.58 2.76 8.17 13.48 19.41 30.90 49.05

The average absolute deviation of the resulting model was an acceptable 2,69%, and the R2 correlation coefficient was higher than 0.999. The right hand side of (11) can be decomposed as:

VBNDi1[U]=14.8567x0.919903U= 13.6667U VBNDi2[U]=14.8567(ln(0.919903U)) = 14.8567ln(0.919903U)

(12) (13)

In continuation of our study, we used Riazis two-parameter model (with To fixed at 261 K), as transformed above for for discrete pseudocomponents, distilled up within a certain temperature range, eq. (7) and eq. (9) for the linear calculation of the VBN of a blend, with the aim to define a dependence necessary for evaluation of Riazis model parameters AT (A) BT (B). The mathematical notation is as follows:
VBN Di [ A, B, a , b ] = x Di [ A, B ,U ]VBN Di 1[U ]dU + x Di [ A, B,U ]VBN Di 2[U ]dU (14),
a a b b

where a, b are temperature boundaries of the discrete fraction distilled up within temperature range, expressed by normalized temperatures. The definite integral output gave the following expression for VBN of a discrete pseudocomponent, VBNDi in (9) :
Ba Ba Bb Bb 14.8567 B 0.0834867 e A + ln( a ) e A + 0.0834867 e A ln(b ) e A B
B B B B

B B Ei a B + Ei b B A A

1 B B + 13 .6667 A B 1 + 1 , a B 1. 1 + 1 , b B B B A B A

(15),

where is Riazis gamma function and Ei is ExpIntegral gives the exponential integral function Ei (z). However, in the definite integral output (15), both parameters A and B are unknown. Therefore, we used the relation obtained for a VBN of fraction (VBNDi) with AT (A) and BT (B) originating from Riazis model and the pseudocomponents boiling temperature range (expressed as VABP) to form a system of two equations:
VBNDi [A, B, 0.000001, 3.877394]=VBNcrude VBNDi [A, B, 0.000001, 1.360153]=2.905

(16) (17)

The first equation (16) is in fact equation (9) applied to equalize the VBN of the whole crude to the VBN of a pseudocomponent, obtained within the boiling range from IBP to 1000 oC (0 to 3.87, expressed as normalized temperature, U), which covers the crude distillation range up to T98% , as suggested in [4, 5]. The second equation (17) does the same, equalizing to the constant 2.905, the VBN of a fraction distilled up to 343 oC (1.36 expressed as normalized temperature). The above system empirically allows for the determination of Riazis coefficient. The fraction distilled up to 343 oC for equation (17) was chosen because it provided best agreement of the model with the available experimental data. Hence, with the known values of A and B, the prediction of

a crude TBP yields for discrete pseudocomponets expressed with normalized VABPs, as shown in Table III mathematically is as follows:
FractionsTemperatureBoundaries={0.0003,0.314,0.429,0.621,0.774,0.946,1.119,1.360,2.211,3.877}

(18)

FractionMass[A,B,n]=XC[A,B,FractionsTemperatureBoundaries[[n+1]]]XC[A,B, FractionsTemperatureBoundaries[[n]]]

(19),

where an equation (18) defines the discrete fraction temperature boundaries in normalized temperature (U) while an equation (19) gives the mass of each discrete fraction. n is equal of the sum of the boundaries covering each discrete fractions (in this case n equals of 9 and it varies between 1 and 9). Further, the obtained mass concentrations of the discrete fractions (i.e., in o): IBP - 70; 70 - 100; 100 - 150; 150 - 190; 190 - 235; 235 - 280; 280 - 343; 343 565; 565+, can be readily summed up for cumulative fractions: IBP - 70; IBP 100; IBP - 150; IBP - 190; IBP 235, etc.

RESULTS AND DISCUSSION In order to test the validity and accuracy of the new proposed model (15), its predictions for discrete fractions were recalculated into predictions for the TBP yields of cumulative fractions of the 117 crude oil assays in the data base. They were compared with the measured TBP yields and with those predicted by Riazis model. The results are summarized in Table IV. Table IV. Comparison of absolute deviations
TBP cumulative fraction, oC Average absolute deviation of the TBP yields (%) from the 117 crudes, estimated by New model and Riazis model per group crude sample I Riazis model 0.39 0.33 1.03 1.02 0.80 0.96 1.63 1.42 New model 1.09 1.57 2.29 2.23 2.08 1.95 2.74 2.83 Riazis model 0.17 0.16 0.49 0.50 0.56 0.65 0.90 2.06 II New model 1.00 1.59 2.35 2.74 2.96 2.97 2.80 1.81 Riazis model 0.19 0.22 0.46 0.45 0.41 0.63 0.85 1.57 III New model 1.57 2.72 4.18 4.88 5.14 4.83 3.88 3.32 IV Riazis model 0.11 0.10 0.45 0.18 0.57 0.87 0.90 2.08 New model 0.67 1.23 2.27 2.23 3.62 4.10 4.63 3.49

IBP - 70 IBP - 100 IBP - 150 IBP - 190 IBP - 235 IBP - 280 IBP - 343 IBP - 565

Table IV shows that the average absolute deviations of the yields, predicted by Riazis method are several times closer to the experimental values than those estimated by the new method. This not surprising since the parameters of the Riazi model have been calculated from all available TBP data, while the parameters of new model involve estimated data. The calculated average

relative deviation was reasonably well for white oil fractions distilled up to 3430, amounting of 5.95% against 3.30% - calculated through Riazis model. On the other hand, the deviations of the obtained by calculation by any method yields cannot be less than those which can be reproduced experimentally. For instance, the reproducibility of experimental yields, obtained by the ASTM D 2892 method is 1.2 % for atmospheric and 1.4 % for vacuum fractions [9]. So, the deviations of the new method are comparable to the experimental and thus acceptable for its declared practical applications. Moreover, there are possibilities to improve the new method if more experimental data are available and/or other routinely determined properties of the crudes are used. The adequacy of the model was verified, also for predicting the TBP distribution of a test set of three crude oil samples, not included among the 117 samples used to develop the model. The results obtained are given in Table V. Table V. Absolute deviations of the predictions with the new method for the test set
Properties Mellitah Visc. at 38 oC, mm2/s Specific gravity @150C TBP range, oC IBP - 70 IBP - 100 IBP - 150 IBP - 190 IBP - 235 IBP - 280 IBP - 343 IBP - 565 2.07 0.8132 Crude oil Western Desert 2.51 0.8200 REBCO 8.25 0.8680

0.50 3.72 1.42 1.70 0.77 1.68 1.10 5.00

Absolute deviation, % 1.64 3.44 4.08 2.87 1.58 1.12 3.62 3.48

0.58 1.04 1.25 1.07 0.34 0.05 0.51 0.96

The deviations with the test set fall within those for the 117 samples used in the development of the model.

CONCLUSIONS The main aim of this work was to develop a method for express prediction of crude true boiling point (TBP) curves, that would be suitable for daily monitoring of the operarion of crude oil distillation units (CDUs). The main results achieved in the realization of this aim can be summatrized as follows:

- The developed new model requires only data from the routine laboratory analysis of the kinematic viscosity of crude oil at 38 o. In its development assay data for 117 crude oils from different parts of the world have been used. The precision of the predictions of the new model, though somewhat worse than those of the original Riazi model, are comparable to the experimental reproducibility of the standard methods for laboratory distillation. The adequacy of the model was verified, also for predicting the TBP distribution of a test set of three crude oil samples, not used in the development of the model. The results of this test showed the consistency of the model predictions. - In practical terms the reported deviation is sufficient to detect any upsets in the operation of refinery CDUs. The method requires only data for the viscosity of crude oil at 38 o, and uses a computerized calculation algorithm, easily realized in Excel or other popular statistical software. - An integral part of the new model is a sub model describing the dependence of the viscosity blending numbers (VBNs) of crude oil pseudocomponents from their average boiling boiling points. The submodel allows for subsequent calculation of other integral parameters for each narrow cut of crude oil, such as: molecular mass, specific gravity and refractive index, including proxy on TBP-distillation curve at known VBNs. - In terms of mathematics, it can be assumed that Riazis model can be reduced to an essentially one-parameter model, taking into account eq. (17) which allows to estimate its property parameters from only one laboratory measurement. The Riazis model, which is originally in integral form, can be subjected to differentiation in order to identify the properties of pseudocomponent.

REFERENCES [1] Riazi, M., Distribution Model for Properties of Hydrocarbon-Plus Fractions, Ind. Eng. Chem. Res., Vol.28 (1989), pp.1831-35. [2] Riazi, M.,A Continuous Model for C7+ Fraction Characterization of Petroleum Fluids, Ind. Eng. Chem. Res., Vol. 36 (1997), pp 4299-4307 [3] Riazi, M. Characterization and properties of petroleum fractions, ASTM Manual Series: MNL 50, 2005. ASTM [4] Stratiev, D. et. all,Method Calculates Crude cut Properties, Oil and Gas Journal/Jan.7, 2008, pp.48-49.

[5] Stratiev, D. Dinkov, K. Kirilov, Evaluation of Crude oil data, IPC 200743rdInternational Petroleum Conference, September 24-26, 2007, Bratislava. [6] Abbott, M. M., Kaufmann, T. G. and Domash, L., "A Correlation for Predicting Liquid Viscosities of Petroleum Fractions," Canadian Journal of Chemical Engineering, Vol. 49, No. 3, 1971, pp. 379-384. [7] Singh, B., Miadonye, A., and Putagunta, V.R., Heavy oil viscosity range from one test, Hydrocarbon Processing, August 1993, pp. 157-160. [8] Robert E. Maples (2000). Petroleum Refinery Process Economics (2nd ed.). Pennwell Books. [9] ASTM Manual series 37, Fuels and Lubricants Handbook, 2003.