Thermodynamics Prediction of Wax Precipitation in Black Oil Using Regular Solution Model and Plus Fraction Characterization

Abstract

The precipitation of wax/solid paraffin during production, transportation, and processing of crude oil is a serious problem. It is essential to have a reliable model to predict the wax appearance temperature and the amount of solid precipitated at different conditions. This paper presents a work to predict the solid precipitation based on solid-liquid equilibrium with regular solution-molecular thermodynamic theory and characterization of the crude oil plus fraction. Due to the differences of solubility characteristics between solid and liquid phase, the solubility parameters of liquid and solid phase are calculated by a modified model. The heat capacity change between solid and liquid phase is considered and estimated in the thermodynamic model. An activity coefficient based thermodynamic method combined with two characteristic methods to calculate wax precipitation in crude oil, especially heavy oil, has been tested with experimental data. The results show that the wax appearance temperature and the amount of weight precipitated can be predicted well with the experimental data.

1. Introduction

Wax precipitation is a serious problem in the petroleum industry because it may result in the plugging of well bores, production facilities, and transportation pipelines during oil/gas production. The change of temperatures, pressures, and oil/gas compositions may cause wax precipitation. To design oil/gas production process efficiently, it is of great importance to predict wax appearance temperatures (WAT) and the amount of precipitated wax using a thermodynamic wax model.

Several thermodynamic models for wax precipitation have been published in the literature. Wax precipitation models can be divided into two categories: an equation of state plus activity coefficient (EOS + GE) approach and an EOS approach. In the former approach, a cubic EOS is applied for vapor-liquid equilibrium (VLE) and an activity coefficient model (based on either regular solution theory or polymer solution theory) for solid-liquid equilibrium (SLE). At equilibrium, it is necessary that, for every component i

f_{i}^{v} = f_{i}^{l} = f_{i}^{s},

(1)

where f is the fugacity and i is the number of component.

Researchers use a cubic EOS for vapor-liquid equilibrium and an activity coefficient model for solid-liquid equilibrium [1 –6]. These models are based on solid solution theory which assumes that all the components in the solid phase are miscible in all proportions.

Proper characterization of the heavy end hydrocarbons has an important effect on thermodynamic modelling of wax precipitation. The thermophysical parameters of multiphase mixture and the wax deposition rate are relative to phase equilibrium. Due to its important role in predicting the phase behaviour of hydrocarbon fluids, methods of characterization have been investigated using different thermodynamic models. The constituents of a hydrocarbon system are classified in two categories: the well-defined components and the undefined petroleum fraction which are those heavy compounds lumped together and identified as the plus fraction. Characterization of plus fraction usually consists of three steps:

Estimation of the molar distribution by splitting the plus fraction into single carbon number (SCN), with known molar amount and molecular weight.

Estimation of critical properties, the melting enthalpy and the heat capacities of the resulting carbon number fractions.

Lumping of the generated SCN.

A couple of works have been done to estimate the mole distribution fraction. Katz and Firoozabadi [7] proposed a simple method for splitting the plus fraction which requires only the mole fraction of C₇⁺. This equation is as follows:

Z_{n} = 1.38205 * Z_{7}^{+} * e^{- 0.25903 * n} .

(2)

Whitson [8] proposed a three-parameter gamma probability (TPG) distribution function to describe the relation between mole fraction and molecular weight of SCN components of the plus fraction. Pedersen et al. [2, 3] proposed a logarithmic relationship between the mole fraction and the carbon number for carbon numbers (C_n) above 6, as follows:

\ln Z_{i} = A + B * C_{i} .

(3)

In the present work, a thermodynamic model based on regular solution theory is combined with improved methods for plus fraction characterization. Predictions are further compared with published precipitation data of petroleum mixtures, where a reliable prediction of solid prediction is found with the current method.

2. Thermodynamic of Wax Precipitation

A molecular thermodynamic approach is used to describe the solid-liquid phase behaviour of crude oil. The method takes into account the effects of temperature, composition, and nonideality of each phase. For each component i at thermodynamic equilibrium between solid and liquid must satisfy the following equation:

f_{i}^{s} = f_{i}^{l},

(4)

where f is the fugacity; the superscript represents the phase of the component. The solid-liquid equilibrium coefficient K_i^sl is given as follows:

\frac{x_{i}^{s}}{x_{i}^{l}} = K_{i}^{s l} = \frac{γ_{i}^{l}}{γ_{i}^{s}} \exp (\frac{Δ H_{i}^{f}}{R T} (1 - \frac{T}{T_{i}^{f}}) - \frac{1}{R T} \int_{T}^{T_{i}^{f}} {(Δ C p)}_{i} d T + \frac{1}{R} \int_{T}^{T_{i}^{f}} \frac{{(Δ C p)}_{i}}{T} d T),

(5)

where γ_i^l, T_i^f, ΔH_i^f, and ΔCp are activity coefficient, fusion temperature, heat of fusion of component i, and heat capacity change of fusion.

2.1. Theory

Won [1] used regular solution theory to get activity coefficients of the individual components in the liquid and solid phase. The activity coefficients are determined by the solubility parameters of each phase as follows:

\begin{matrix} \ln γ_{i}^{l} = \frac{V_{i} {({\bar{δ}}^{l} - δ_{i}^{l})}^{2}}{R T}, \\ \ln γ_{i}^{s} = \frac{V_{i} {({\bar{δ}}^{s} - δ_{i}^{s})}^{2}}{R T}, \end{matrix}

(6)

where γ_i, $\bar{δ}$ , and δ_i represent activity coefficient, average solubility parameter, and the solubility parameter of component i separately. The superscripts s and l represent the state of phase. V_i is the molar volume of component i.

Combining equations above, the activity coefficient ratio is

\frac{γ_{i}^{l}}{γ_{i}^{s}} = \exp {\frac{V_{i}}{R T} [{(\bar{δ} - δ_{i})}_{l}^{2} - {(\bar{δ} - δ_{i})}_{s}^{2}]} .

(7)

While the average solubility parameters are

\begin{matrix} {\bar{δ}}^{s} = \sum ϕ_{i}^{s} δ_{i}^{s} \\ {\bar{δ}}^{l} = \sum ϕ_{i}^{l} δ_{i}^{l} . \end{matrix}

(8)

It can be involved from (5) and (7) that accurate description of the above parameters as γ_i^l, T_i^f, ΔH_i^f, and ΔCp is crucial to the prediction method.

Thermodynamic model is established based on a modified regular solution theory. A systematic work has been carried out on the existing models, where a new method combined with existing models and correlations is found and helps to predict the precipitation of solid in crude oils.

2.2. Models of Properties

2.2.1. Solubility Parameters

The solubility parameter of component i in liquid solution of paraffin is calculated with the correlation proposed by Riazi and AL-Sahhaf [9]:

δ_{i}^{l} = 8.6 - \exp (2.219195 - 0.54907 {M W}_{i}^{0.3}) .

(9)

As to naphthenic, aromatic, and asphalt, the correlation proposed by Leelavanichkul [10] will be used:

\begin{matrix} δ_{i . nap}^{l} = 8.7 - \exp (2.219195 - 0.54907 {M W}_{i}^{0.3}) \\ δ_{i . arom}^{l} = 8.8 - \exp (2.219195 - 0.54907 {M W}_{i}^{0.3}) \\ δ_{i . asphal}^{l} = 9.3 - \exp (2.219195 - 0.54907 {M W}_{i}^{0.3}) . \end{matrix}

(10)

The solubility parameter in solid solution can be obtained as follows:

δ_{i}^{s} = {(\frac{{Δ H}_{i}^{f}}{V} + {δ_{i}^{l}}^{2})}^{0.5} .

(11)

2.2.2. Heat Capacity Change of Fusion

The relationship between the heat capacity change of fusion, temperature, and molecular weight in the process of phase transform is expressed by Pedersen [11] as follows:

(C_{p}^{l} - C_{p}^{s}) = {Δ C}_{p i} = a {M W}_{i} + b {M W}_{i},

(12)

where a and b are constants, approximately valued 0.3033 and – 4.635e⁻⁴.

Molecular weight is calculated by the SCN and the classes as:

paraffinic:

{M W}_{i} = 12.01 C_{i} + 1.008 (2 C_{i} + 2),

(13)

naphthenic:

{M W}_{i} = 12.01 C_{i} + 1.008 (1.8 C_{i}),

(14)

aromatic:

{M W}_{i} = 12.01 C_{i} + 1.008 C_{i} .

(15)

2.2.3. Fusion Temperature and Heat of Fusion

Fusion temperature can be expressed as a function of molecular weight. Won [1] proposed the following correlation for paraffin:

T_{i}^{f} = 374.5 + 0.02617 {M W}_{i} - \frac{20172}{{M W}_{i}} .

(16)

When the molecular weight is higher than 450, the correlation above is converted to

T_{i}^{f} = 411.4 - \frac{32326}{{M W}_{i}} .

(17)

Correlation proposed by Lira-Galeana et al. [12] for isoparaffin is considered

T_{i}^{f} = 333.46 - 419.01 \exp (- 0.00855 {M W}_{i}) .

(18)

A set of correlations [13] is considered for heat of fusion with differences between the odd carbon number and even carbon number:

odd (7 < C_n < 21)

{Δ H}_{i}^{f} = 0.5754 {M W}_{i} T_{i}^{f},

(19)

even (8 < C_n < 22)

{Δ H}_{i}^{f} = 0.8064 {M W}_{i} T_{i}^{f},

(20)

22 < C_n < 38

{Δ H}_{i}^{f} = 0.4998 {M W}_{i} T_{i}^{f},

(21)

C_n > 38

{Δ H}_{i}^{f} = 0.6741 {M W}_{i} T_{i}^{f} .

(22)

Correlation for naphthenic and aroma is expressed separately as follows [14]:

\begin{matrix} {Δ H}_{i}^{f} = 0.2216 {M W}_{i} T_{i}^{f} \\ {Δ H}_{i}^{f} = 46.89216 {M W}_{i} . \end{matrix}

(23)

Thermodynamic model demands an accurate characterization of the plus fraction. Insufficient description of the plus fraction will affect the prediction accuracy of wax precipitation. So, a proper description of the plus fraction is analyzed below.

3. Plus Fraction Characterization

The plus fraction, that is C₇₊, is a combination of the infinite number of hydrocarbon components heavier than SCN 7. It is normal that certain data would be provided by a laboratory or field for the plus fraction, such as mole percent, molecular weight, density, and the specific gravity. Though the information may be more or less different from the accurate number, the properties of plus fraction are essential for the characterization. Through two improved methods [2, 3, 8, 15], key properties of the thermodynamic model can be calculated.

3.1. Method 1

It is commonly considered that distribution of the mole fraction of components heavier than 7 is exponentially. Pedersen et al. [2, 3] proposed that reservoir fluids for carbon numbers (C_n) above 6 can be expressed with a logarithmic relationship between the mole fraction and the carbon number:

\ln Z_{i} = A + B * C_{i},

(24)

where Z_i is mole fraction of the carbon number i in the SCN; C_i is value of carbon number; A, B are constants.

The molecular weight can be gained from a chart of molecular weights of SCN groups proposed by Katz and Firoozabadi [7]. While the carbon number is higher than 45, molecular weight can be calculated as

M n = 14 C_{i} - 6 .

(25)

Al-Meshan and McCain [16] proposed that the exponential distribution starts from C₁₁ rather than C₇ recommend by Pedersen. In the present work, neither C₇ nor C₁₁ was used. A procedure is characterized to determine the initial carbon number.

Find the carbon number, C_max, with the highest mole fraction in the defined fraction.

If C_max > 11, then C_max is the initial number.

If 7 < C_max < 11, then 11 is the initial number. If C_max < 7, then 7 is the initial number.

3.2. Method 2

Whitson [8] proposed a three parameter gamma (TPG) probability distribution function to describe the relation between mole fraction and molecular weight of SCN components of the C₇₊ plus fraction. The Gauss-Laguerre quadrature method was used to generate the mole fraction of SCN. The probability density function p(M) is as follow:

p (M) = \frac{{(M - η)}^{α - 1} \exp [- (M - η) / β]}{β^{α} Τ (α)},

(26)

where α, η, and β are characteristic parameters defining the distribution. η is the minimum molecular weight in C₇₊ fraction. The plus fraction such as C₇₊ is not restricted, where other plus fractions can be used. The limit η is the minimum molecular weight found in the C₇₊ fraction. The minimum molecular weight is not a constant. It varies with different reservoir fluids. Usually the range is from 86 to 100. α is the fitting parameter, which defines the shape of the mole fraction distribution. Reasonable limits for α are 0.5 to 3.0. β is defined as

β = \frac{(M_{C_{7 +}} - η)}{α} .

(27)

The frequency of occurrence (normalized mole fraction) in the interval η to x is a cumulative probability function, p(X ≤ x), which is the integral of p(X) from η to x:

p (X \leq x) = \int_{η}^{x} p (x) d x .

(28)

The mole fraction of SCN is then obtained by calculating the cumulative frequency of occurrence between the limits M_{i – 1} and M_i multiplied by the mole fraction of the C_n+ fraction:

\begin{matrix} Z_{i} = Z_{C_{n +}} \int_{M_{i - 1}}^{M_{i}} p (M) d M = Z_{C_{n +}^{}} [P (M \leq M_{i}) - P (M \leq M_{i - 1})], \end{matrix}

(29)

where i is the SCN and M_i is the molecular weight of the SCN usually defined as the molecular weight of the normal component.

The developed method here is based on the fact that SCN may represent hundreds of different compounds with the same number of carbon atoms. Uncertainty of the types and distribution of compounds present in each SCN indicate that the average value of the molecular weight of an SCN is different from one sample to another. The modification of Whitson's method includes a simple procedure to determine the value of the parameter α and the new definition of the limits used to obtain the frequency of occurrence for each SCN. According to the statistical characterization of TPG, the mean and variance can be obtained by (27) and (30).

Consider the following:

α β^{2} = \frac{\sum_{n}^{N} z_{i} {(M_{i} - M_{C_{7 +}})}^{2}}{\sum_{n}^{N} z_{i}} .

(30)

The new limits used to calculate the cumulative frequency of occurrence are defined by the best fit between TPG and the experimental mole fraction (Z_iexp). To establish the function of limiting molecular weight (LMW) to SCN, at least two previous carbon numbers and their mole fraction must be known to use this approach. Once an initial value of the parameters is calculated, the second step is to iterate with a controlled error E, expressed as follows:

E (α, η) = \sum_{n}^{N} | Z_{i \exp} - Z_{i cal} |,

(31)

where Z_iexp and Z_{i cal} are experimental and calculated mole fractions of component i, respectively, and N is the final number of components extended. n is the initial number of component that needs to be extended.

The length between molecular weight boundaries is assumed as constant as 14. It could be achieved by minimization of E with the following iterative process.

Taking η = 86 for C₇₊ plus fraction. α is calculated from (27) and (30), then Z_{i cal} and E could be calculated.

Adjusting parameter as α_lower = α – 0.01, and α_higher = α + 0.01; E_lower and E_higher are calculated accordingly. If E_lower < E, Δ = – 0.01; If E_higher < E, Δ = + 0.01.

Let α_test = α + Δ and compare E_test with E; the process is repeated until E_test ≥ E.

α_final = α_test; E_final = E_test; then η = η + 1; repeat processes 1 to 3; if E_final > E_test, then α_final = α_test E_final = E_test. This process is repeated until η > 100.

The η limits for SCN (i) is between 13i and 14i + 2. Once the parameters α, β, and η have been generated, dividing the TPG into SCN with the new definition of the limits by the LMW, (29) is then transformed to

Z_{i cal} = Z_{C_{n +}} \int_{M w \min}^{L M w} p (M) d M,

(32)

where M_wmin = η is taken as an initial value. LMW is calculated by the best fit between Z_iexp and Z_{i cal} with the following iterative.

Z_{i cal} is calculated with the estimation of the limiting molecular weight: LMw = Mw_min + dM.

If Z_{i cal} < Z_iexp, then adjust the LMw for each iteration using secant method. The process is repeated until Z_{i cal} ≥ Z_iexp.

If i ≤ N, then Mw_min = LMw, i = i + 1; the processes 1 to 2 will be iterated with a new i and the LMw for each i will be stored.

A function of LMw versus SCN can be achieved. Experimental mole fractions are compared with the calculated results, with the average relative error (ARE) and the absolute average relative error (AARE):

\begin{matrix} ARE = \frac{\sum_{n}^{N} ((Z_{i cal} - Z_{i \exp}) / Z_{i \exp})}{N} \\ AARE = \frac{\sum_{n}^{N} (| Z_{i cal} - Z_{i \exp} | / Z_{i \exp})}{N} . \end{matrix}

(33)

4. Results and Discussion

For the validation of the proposed methods, 7 different experimental extended fluid compositions have been used. These fluid compositions have different plus fraction such as C₂₀₊ and C₃₀₊. The main characterization for these fluids is presented in Table 1. As shown from Table 1 this set of data covers a wide range of C₇₊ mole fraction of 78.32%–98.46%, that is, all heavy oil and black oil. (oil 1#–4# from [2, 3], 5#–7# from [10]).

Table 1:

Properties of the oil samples.

Fluid	C_n+	Z _n+	M _n+	γ_n+

1#	20	11.3	388	0.872
2#	20	38.8	473	0.963
3#	20	17	544	0.934
4#	30	13.23	624	0.954
5#	30	5.613	728	—
6#	30	3.425	745.1	—
7#	30	5.585	775.7	—

4.1. Characterization of Plus Fraction

The above two methods are used to calculate the cumulative frequency of occurrence. Figure 1 shows the comparison between experimental and predicted SCN mole fraction for selected oil samples. While Table 2 shows the AREs and AAREs that are calculated using (33) for all fluids where the experimental extended composition is grouped to C₇₊ and then extended to the original plus fraction.

Table 2:

Comparison between predicted and measured mole fractions.

Fluid	Average relative error		Absolute average relative error
Fluid	Method 1	Method 2	Method 1	Method 2

1#	0.0363	0.0059	0.1010	0.0761
2#	0.0099	0.0002	0.0321	0.0253
3#	0.0535	0.0036	0.0930	0.0704
4#	−0.0197	0.0384	0.1311	0.1132
5#	0.0107	0.0340	0.0250	0.0488
6#	0.2622	−0.0387	0.2622	0.0728
7#	−0.2035	0.0332	0.2038	0.0792

Figure 1:

Comparison between measured and predicted mole fraction using the method 1 and 2 (oil samples 1#, 3#, and 7#).

Figure 1 shows the comparison between the experimental and calculated mole fraction for fluids using methods 1 and 2. The complex and diversity of the heavy oil make the characterization of plus fraction more difficult than that of light oil or gas condenser oil. As shown in Table 2, both methods result in reasonable mole fraction for SCN, proved by the low ARE (0.01 for method 2 and 0.002 for method 1) and AARE (0.03 for method 2 and 0.02 for method 1). The LMW for each carbon number is determined to fit the mole fraction that is corresponding to the carbon number. So the calculation results of the mole fraction are well with those of experiments.

It is observed from the combination of results presented in Figure 1 and Table 2 that the prediction for all oil samples is acceptable; it is sufficient to use both methods to yield an excellent match of the complete molar distribution.

4.2. Wax Precipitation

Predicted wax appearance temperature (WAT) with proposed characterization methods are compared with experimental data for all oil samples using defined C₅₀₊ plus fraction, as shown in Table 3. The absolute percentage error is calculated as (34), where a good agreement could be found.

Table 3:

Comparison between the predicted and measured WAT for oil 1#–7#.

Fluid	Experiment	Leelavanichkul (2003)'prediction [10]	Method 1 prediction	Deviation %	Method 2 prediction	Deviation %

1#	305.1	—	297.0	2.67	296.0	3.00
2#	311.2	—	309.0	0.69	306.0	1.66
3#	295.1	—	301.0	1.98	300.0	1.64
4#	304.2	—	304.0	0.05	302.0	0.71
5#	298.2	296.6	294.0	1.41	298.0	0.07
6#	295.2	293.4	290.0	1.76	294.0	0.41
7#	294.2	296.0	293.0	0.41	295.0	0.27

Consider the following:

AD % = \frac{| T_{\exp} - T_{cal} |}{T_{\exp}} \times 100 % .

(34)

Further, an accurate prediction method proposed by Leelavanichkul [10] is considered and compared with our modified models, where a total of 148 components including pseudo-components are characterized. Techniques such as refraction index (RI) measurement, high performance liquid chromatography (HPLC), and Fourier transform-infrared (FT-IR) were used to divide the defined fraction up to C₅₀ into paraffinic-naphthenic-aromatic (PNA). Molecular weight and mole fraction of C₅₀₊ plus fraction are divided into defined asphaltene and resin subfractions. Combined with regular solution method, a more accurate prediction of wax precipitation has been established and compared with predictions from the current work for oil samples 5#–7#, as shown in Table 3. The WAT and weight percentage of precipitated wax are further compared in Figure 2 with the two methods and Leelavanichkul [10] method.

Figure 2:

Comparison between predicted and measured weight of precipitated wax (oil samples 5#–7#).

As can be seen from Figure 2 and Table 3, predictions obtained with the modified methods in the current work show an acceptable trend of fitting and agree with Leelavanichkul's accurate prediction method [10], which could be attributed to proper characterization of plus fraction proposed and a reasonable combination of correlations from the literatures.

5. Conclusions

An activity coefficient based regular solution model combined with improved methods for plus fraction characterization has been established in the present work. Predictions of the wax appearance temperature and the amount of precipitated weight are further compared with published data of petroleum mixtures. Data covers a wide range of C₇₊ mole fraction of 78.32%–98.46%, that is, all heavy oil and black oil, where a reliable prediction of solid prediction is found with the current method.

Future work is highly demanded to apply the proposed method in the prediction of wax precipitation at gas-liquid-solid multiphase systems. It could provide an important fundament for the precipitation prediction in multiphase system with the presence of gas phase, where the proposed liquid-solid thermodynamic equilibrium method is demanded to be further coupled with the equation of state for gas-liquid equilibrium.

Footnotes

Acknowledgments

The authors wish to thank the Program for New Century Excellent Talents in University (NCET-12-0969), Foundation for the Author of National Excellent Doctoral Dissertation of PR China (201254), National Natural Science Foundation of China (51104167), National Science & Technology Specific Program of China (2011ZX05026-004-03), and Science Foundation of China University of Petroleum, Beijing (BJ-2011-02) for providing support for this work.

References

Won

K. W.

, “Thermodynamics for solid solution-liquid-vapor equilibria: wax phase formation from heavy hydrocarbon mixtures,” Fluid Phase Equilibria, vol. 30, no. C, pp. 265–279, 1986.

Pedersen

K. S.

Skovborg

, and Rønningsen

H. P.

, “Wax precipitation from North Sea crude oils. 4. Thermodynamic modeling,” Energy and Fuels, vol. 5, no. 6, pp. 924–932, 1991.

Pedersen

W. B.

Hansen

A. B.

Larsen

Nielsen

A. B.

, and Rønningsen

H. P.

, “Wax precipitation from North Sea crude oils. 2. Solid-phase content as function of temperature determined by pulsed NMR,” Energy and Fuels, vol. 5, no. 6, pp. 908–913, 1991.

Coutinho

J. A. P.

Andersen

S. I.

, and Stenby

E. H.

, “Evaluation of activity coefficient models in prediction of alkane solid-liquid equilibria,” Fluid Phase Equilibria, vol. 103, no. 1, pp. 23–39, 1995.

Hansen

J. H.

Fredenslund

, and Pedersen

K. S.

, “A thermodynamic model for predicting wax formation in crude oils,” AIChE Journal, vol. 34, pp. 1937–1942, 2004.

Pedersen

and Christenen

, Phase Analysis of Petroleum Reservoir Fluids, Taylor & Francis, Boca Raton, Fla, USA, 2006.

Katz

D. L.

and Firoozabadi

, “Prediction phase behavior of condensate or crude oil system using methane interaction coefficient,” Journal of Petroleum Technology, vol. 30, no. 11, pp. 1649–1655, 1978.

Whitson

C. H.

, “Characterizing hydrocarbon-plus fraction,” Society of Petroleum Engineers Journal, vol. 23, no. 4, pp. 683–694, 1983.

Riazi

M. R.

and Al-Sahhaf

T. A.

, “Physical properties of heavy petroleum fractions and crude oils,” Fluid Phase Equilibria, vol. 117, no. 1, pp. 217–224, 1996.

10.

Leelavanichkul

, Crude oil characterization and thermodynamic modelling of solids precipitation in crude oils [Ph.D. dissertation], The University of Utah, 2003.

11.

Pedersen

K. S.

, “Prediction of cloud point temperatures and amount of wax precipitation,” SPE Production and Facilities, vol. 10, no 1, pp 46–49, 1995.

12.

Lira-Galeana

Firoozabadi

, and Prausnitz

J. M.

, “Thermodynamics of wax precipitation in petroleum mixtures,” AIChE Journal, vol. 42, no. 1, pp. 239–248, 1996.

13.

Chen

Zhao

Zhang

, and Wang

, “Thermodynamic phase equilibria of wax precipitation in crude oils,” Fluid Phase Equilibria, vol. 255, no. 1, pp. 31–36, 2007.

14.

Pan

and Firoozabadi

, “Thermodynamic micellization model for asphaltene aggregation and precipitation in petroleum fluids,” SPE Production and Facilities, vol. 13, no. 2, pp. 118–125, 1998.

15.

Hassan

S. N.

, “Characterizing pure and undefined petroleum components,” International Journal of Engineering & Technology, vol. 10, pp. 39–68, 2010.

16.

Al-Meshan

A. A.

and McCain

W. D.

, “Validation of splitting the hydrocarbon plus fraction: first step in tuning equation of state,” in 15th Middle East Oil and Gas Show and Conference (MEOS '07), vol. 104631 of Proceedings of SPIE, pp. 171–189, March 2007.