You are on page 1of 16

Chemical Engineering Science 90 (2013) 53–68

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Toward reservoir oil viscosity correlation


Abdolhossein Hemmati-Sarapardeh a, Mahdi Khishvand b, Ali Naseri c,
Amir H. Mohammadi d,e,n
a
Department of Chemical and Petroleum Engineering, Sharif University of Technology, Tehran, Iran
b
Chemical and Petroleum Engineering Department, University of Wyoming, Laramie, WY, USA
c
EOR Department, Research Institute of Petroleum Industry (RIPI), Tehran, Iran
d
Institut de Recherche en Génie Chimique et Pétrolier (IRGCP), Paris Cedex, France
e
Thermodynamics Research Unit, School of Chemical Engineering, University of KwaZulu-Natal, Howard College Campus, King George V Avenue,
Durban 4041, South Africa

H I G H L I G H T S

c Reliable viscosity correlations for dead and live oil systems have been developed.
c We report few sets of oil viscosity data measured using a Rolling Ball viscometer.
c A large database consisting of more than 1000 data have been used to develop them.
c Their reliability is successfully examined against independent data.

a r t i c l e i n f o abstract

Article history: Oil viscosity plays a key role in reservoir simulation and production forecasting, as well as planning
Received 18 October 2012 thermal enhanced oil recovery methods and these make its accurate determination necessary. In this
Received in revised form communication, the most frequently used oil viscosity correlations are evaluated using a large
2 December 2012
databank of Iranian oil reservoirs which were measured using a Rolling Ball viscometer (Ruska, series
Accepted 5 December 2012
Available online 12 December 2012
1602). To evaluate the performance and accuracy of these correlations, statistical and graphical error
analyses have been used simultaneously. Three of the most accurate correlations for each region,
Keywords: including dead oil viscosity, viscosity below bubble point, viscosity at bubble point and the under-
Crude oil saturated oil viscosity, are recommended for Iranian oil reservoirs. In the last step, four correlations are
Viscosity
developed for Iranian oil reservoirs which have simplified functional format. Furthermore, the input
Correlation
data of the latter correlations are those ones that are easily measured in oil fields. The ranges of data
Reservoir fluid
PVT used to develop these new correlations cover almost all Iranian oil reservoirs PVT data and
Prediction consequently they could be reliable for prediction of other Iranian oil reservoirs viscosity. These new
correlations are consistent across a wide range of parameters and offer increased accuracy than
previously published correlations for viscosity prediction of Iranian oil reservoirs.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction accurate determination of this property is necessary for petro-


leum industry. This property is traditionally measured through
Viscosity of crude oil is an important parameter for reservoir laboratory analysis on bottom hole samples or by recombination
evaluation in performance calculation, simulating reservoirs, of liquids and gases collected from the separators. In the absence
forecasting production and designing production facilities as well of laboratory PVT data and for saving time and money, fluid
as planning thermal enhanced oil recovery methods (Al-Marhoun, properties are mostly predicted from empirical correlations as
2004; Ali and Thomas, 1996; Elsharkawy and Alikhan, 1999; well as equation of states (EOS).
Ikiensikimama and Ogboja, 2009; Naseri et al., 2005). Therefore, Oil viscosity correlations could be categorized generally in two
types (Naseri et al., 2005). The first type is those that use oil field
n
data that are normally available, for instance oil API gravity,
Corresponding author at: Institut de Recherche en Génie Chimique et Pétrolier
(IRGCP), Paris Cedex, France.
reservoir temperature, saturation pressure and solution gas/oil
E-mail addresses: a.h.m@irgcp.fr, ratio. The second type which refers to those empirical and/or semi
amir_h_mohammadi@yahoo.com (A.H. Mohammadi). empirical models which use some parameters that were not

0009-2509/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ces.2012.12.009
54 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

included in the first one such as reservoir fluid composition, pour The proposed correlations have very simple mathematical
point temperature, molar mass, normal boiling point and acentric format and also use parameters which are always available and
factor as well as critical temperature (Ahrabi et al., 1987; Little easily measured in oil fields. In addition, these correlations are
and Kennedy, 1968; Xu and Khurana, 1996). consistent across a wide range of properties and offer increased
In this study, a comprehensive detailed analysis of previous accuracy over existing methods.
reported correlations performance have been conducted on
Iranian oil viscosity data which indicated the inaccuracies of
these correlations for Iranian oil samples. This is due to high
dependency of oil viscosity on oil nature and source; as it is 2. Literature review
known most of these correlations are based on specific regions/
crude oils (Elsharkawy and Alikhan, 1999; Naseri et al., 2005; Over the last decades, numerous correlations have been
Sutton and Farshad, 1990). Moreover, these correlations were developed for prediction of crude oil viscosity. Most of these
developed base on limited data. Afterward, based on more than correlations are based on specific region and often fail to predict
1000 Iranian oil reservoirs data which are always available, four the oil viscosity in other regions due to variety in crude oil nature
new correlations have been proposed for prediction of viscosity and composition, as mentioned earlier. These correlations are
below bubble point, at bubble point, above bubble point as well as generally developed for three conditions: above bubble point,
dead oil. Performance evaluation of most of previously published below and at bubble point and dead oil. Almost all authors
correlations is based on data which were used to their develop- developed single correlation for below and at bubble point
ment and no independent measurement of correlations perfor- conditions, while Khan et al. (1987) and Labedi (1992) divided
mance has been presented. In this study, the data points including it to two distinct correlations which subsequently led to more
inputs and outputs randomly split in to two sets in order to obtain accurate results for below bubble point oil viscosity. A typical
the most efficient correlation for each individual region: 80% have viscosity curve at reservoir temperature as a function of pressure
been used for developing new correlations and adjusting their has been illustrated in Fig. 1.
parameters and the last 20% have been used for testing and
studying their reliability. The Average Absolute Percent Relative
Error (AAPRE) has been used as the objective function for devel- 2.1. Dead oil viscosity correlations
oping these correlations. In addition, to evaluate their perfor-
mance and accuracy, both statistical and graphical error analysis The most well-known correlations for dead oil viscosity are
have been used simultaneously. those developed by Beal (1946), Beggs and Robinson (1975), Glaso
(1980), Kaye (1985), Al-Khafaji et al. (1987), Petrosky (1990),
Egbogah and Ng (1990), Labedi (1992), Kartoatmodjo and
Schmidt (1994), Bennison (1998), Elsharkawy and Alikhan
(1999), Hossain et al. (2005), Naseri et al. (2005) and Alomair
et al. (2011). Table 1 gives the summery of the ranges and origin
of data used in discussed correlations as well as this study.
All of discussed correlations relate dead oil viscosity to tempera-
ture and oil API gravity while some of authors correlated it to
normally unavailable properties in most common PVT reports such
as molar mass, normal boiling point, critical temperature and
acentric factor (Mehrotra, 1991; Svrcek and Mehrotra, 1988). More-
over, some empirical or semi-empirical correlations have been
proposed from corresponding state equations by Teja and Rice
(1981), Johnson et al. (1987) and Johnson and Svrcek (1991).
Although these corresponding state correlations involve multiple
computations as well as use fluid composition as input variable,
they could not sufficiently estimate dead oil viscosity (Elsharkawy
Fig. 1. Oil viscosity as a function of pressure. and Alikhan, 1999; Naseri et al., 2005).

Table 1
The origin and PVT data ranges used in dead oil viscosity correlations.

Author Source of data T, 1F API mod, cp

Beal (1946) US 98–250 10–52 0.86–1550


Beggs and Robinson (1975) – 70–295 16–58 –
Glaso (1980) North Sea 50–300 20–48 0.60–39
Kaye (1985) Offshore California 143–282 7–41 –
Al-Khafaji et al. (1987) – 60–300 15–51 –
Petrosky (1990) Gulf of Mexico 114–288 25–46 0.72–10.25
Egbogah and Ng (1990) – 59–176 5–58 –
Labedi (1992) Libya 100–306 32–48 0.66–4.79
Kartoatmodjo and Schmidt (1994) Worldwide 80–320 14–59 0.50–586
Bennison (1998) North Sea 39–300 11–20 6.40–8396
Elsharkawy and Alikhan (1999) Middle East 100–300 20–48 0.60–33.7
Hossain et al. (2005) Worldwide 32–215 7–22 12–451
Naseri et al. (2005) Iran 105–298 17–44 0.75–54
Alomair et al. (2011) Kuwait 68–320 10–20 1.78–11,360
This study Iran 50–290 17–44 0.39–70
A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 55

2.2. Saturated oil viscosity correlations involve API or dead oil viscosity at their correlations (Elsharkawy
and Alikhan, 1999; Labedi, 1992). The existing correlations for
To thoroughly analyze the accuracy and validation of existing under-saturated oil viscosity are: Beal (1946), Vazquez and Beggs
saturated oil viscosity correlations, 13 of the most commonly (1980), Khan et al. (1987), Petrosky (1990), Labedi (1992), Orbey
used correlations for saturated oil have been gathered. Only Khan and Sandler 1993), Kartoatmodjo and Schmidt (1994), Elsharkawy
et al. (1987) and Labedi (1992) developed two distinct correla- and Alikhan (1999) and Hossain et al. (2005). The summery of
tions for prediction of oil viscosity below bubble point and at origin and ranges of data used in the aforementioned correlations
bubble point while other authors proposed one correlation for as well as this study are presented in Table 3.
saturated oil. Most of saturated oil viscosity correlations intro-
duce saturated oil viscosity as a function of both dead oil viscosity
and solution gas oil ratio while others express it as a function of 3. Experimental equipment and procedure
dead oil viscosity and saturation pressure. The existing correla-
tions in the literature are: Chew and Connally (1959) [1–3], Beggs In this study, a Rolling Ball viscometer (Ruska, series 1602)
and Robinson (1975), Al-Khafaji et al. (1987), Khan et al. (1987), was utilized to measure reservoir oil viscosity at various pres-
Petrosky (1990), Labedi (1992), Kartoatmodjo and Schmidt sures above and below bubble point. Although this instrument
(1994), Elsharkawy and Alikhan (1999), Hossain et al. (2005), has some limitations for heavy oil it is particularly suitable for
Naseri et al. (2005) and Bergman and Sutton (2007). Table 2 black and volatile oil. Before beginning the measurements, it is
illustrates the origin and ranges of data used in the aforemen- necessary to calibrate the instrument with a known viscosity
tioned correlations. standards liquid similar to the fluid to be measured. The instru-
ment consists of a highly polished stainless steel barrel, which can
2.3. Under-saturated oil viscosity correlations be closed at the top by means of a plunger. A steel ball rolls within
the barrel, its diameter necessarily being slightly smaller than the
The viscosity at under-saturated conditions is a function of bore, be filled completely with the fluid to be studied. The barrel
bubble point oil viscosity, bubble point pressure and pressure. is inclined at a known angle and the ball rolls along it under
Several Authors proposed correlations for oil viscosity at under- gravity for a measured distance. The roll time is determined by a
saturated region (Abdul-Majeed et al., 1990; Almehaideb, 1997; digital timer. When operating correctly, the roll time interval is a
Beal, 1946; Dindoruk and Christman, 2001; Elsharkawy and measure of the viscosity. If the clearance between the bore of the
Alikhan, 1999; Hossain et al., 2005; Kartoatmodjo and Schmidt, barrel and the ball diameter is too small, then the flow of fluid
1994; Khan et al., 1987; Labedi, 1992; Orbey and Sandler, 1993; past the ball will be turbulent. Under these conditions the Rolling
Petrosky, 1990; Sutton and Bergman, 2006; Vazquez and Beggs, Ball instrument does not measure viscosity correctly since
1980). Most of reported correlations in the literature correlate the the theory (such as it is) assumes laminar flow. To control the
oil viscosity as a function of bubble point pressure, pressure and rolling time, measurements can be made at different angles or the
oil viscosity at bubble point while some authors additionally stainless steel ball and/or the barrel may be replaced with a

Table 2
The origin and PVT data ranges used in saturated oil viscosity correlations.

Author Source of data Solution GOR, SCF/STB Saturation pressure, psia mod, cp

Chew and Connally (1959)[1] US 51–3544 132–5,645 0.370–50


Chew and Connally (1959)[2] US 51–3544 132–5,645 0.370–50
Chew and Connally (1959)[3] US 51–3544 132–5,645 0.370–50
Beggs and Robinson (1975) – 20–2070 132–5,265 –
Al-Khafaji et al. (1987) – 0–2100 – –
Khan et al. (1987) Saudi Arabia 24–1901 107–4,315 0.130–77.4
Petrosky (1990) Gulf of Mexico 21–1855 1574–9,552 0.210–7.4
Labedi (1992) Libya 13–3533 60–6,358 0.115–3.72
Kartoatmodjo and Schmidt (1994) Worldwide 2.3–572 15–6,054 0.100–6.3
Elsharkawy and Alikhan (1999) Middle East 10–3600 100–3,700 0.050–21
Hossain et al. (2005) Worldwide 19–493 121–6,272 3.600–360
Naseri et al. (2005) Iran 255–4116 420–5,900 0.110–18.15
Bergman and Sutton (2007) Worldwide 6–6525 66–10,300 0.210–4277
This study (below bubble point) Iran 48–2048 158–4,631 0.580–37.18
This study (at bubble point) Iran 126–3261 365–5,702 0.580–37.18

Table 3
The origin and PVT data ranges used in under–saturated oil viscosity correlations.

Author Source of data P, psia Pb, psia mob, cp mo, cp

Beal (1946) USA – – 0.142–127 0.16–315


Vazquez and Beggs (1980) Worldwide 126–9,500 – – 0.117–148
Khan et al. (1987) Saudi Arabia – 107–4794 0.13–77.4 0.13–71
Petrosky (1990) Gulf of Mexico 1600–10,250 1574–9552 0.211–3.54 0.22–4.1
Labedi (1992) Libya – 60–6358 0.115–3.72 –
Orbey and Sandler (1993) – 740–14,504 – 0.217–3.1 0.225–7.3
Kartoatmodjo and Schmidt (1994) Worldwide 25–6,015 25–4775 0.168–184.86 0.168–517
Elsharkawy and Alikhan (1999) Middle East 1287–10,000 – – 0.2–5.7
Hossain et al. (2005) Worldwide 300–3,400 121–6272 3.6–360 3–517
This study Iran 730–12,500 730–5116 0.177–18.15 0.177–31
56 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

Table 4
PVT data of three Iranian oil reservoirs samples.

Sample no. 1 Sample no. 2 Sample no. 3

API T, 1F Pb, psia API T, 1F Pb, psia API T, 1F Pb, psia

31.73 258 3524 20.37 205 1432 24.35 255 1844


P, psia Rs, SCF/STB mo, cp P, psia Rs, SCF/STB mo, cp P, psia Rs, SCF/STB mo, cp
4986 1241.3 0.331 4986 361.7 4.632 5042 547.1 1.334
4487 1241.3 0.320 3988 361.7 4.255 4043 547.1 1.285
3991 1241.3 0.304 2993 361.7 3.950 3532 547.1 1.237
3791 1241.3 0.299 2493 361.7 3.770 3030 547.1 1.212
3524 1241.3 0.289 1995 361.7 3.627 2523 547.1 1.187
2993 1024.9 0.316 1800 361.7 3.550 2323 547.1 1.163
2393 809.5 0.351 1601 361.7 3.480 2223 547.1 1.153
2000 687.4 0.380 1432 361.7 3.402 2123 547.1 1.148
1495 543.1 0.443 1254 334.2 3.550 2023 547.1 1.143
995 411.5 0.522 1005 293.3 3.718 1925 547.1 1.138
495 273.1 0.647 755 244.4 4.001 1844 547 1.133
14.7 0.0 1.463 505 196.7 4.410 1521 478.1 1.146
255 140.1 4.987 1223 406.7 1.196
14.7 0.0 7.620 921 334.2 1.273
621 259.8 1.370
322 172.7 1.499
14.7 0.0 2.161

different diameter. The governing equation is as follows: models, the primary underlying technique used by LINGO’s
optional nonlinear solver is based upon a Generalized Reduced
mo ¼ Aðtðrball roil ÞÞ þ B ð1Þ
Gradient (GRG) algorithm. However, to help get to a good feasible
where mo represents oil viscosity, t is rolling time in seconds, solution quickly; LINGO also incorporates Successive Linear Pro-
(rball–roil) denotes difference in density between ball and oil and gramming (Naseri et al., 2012). It has several optional strategies
A and B are constants of the system which are determined from to help overcome the problem of stalling at local optimal points.
calibration with fluid of known viscosity. The global solver employs branch-and-bound methods to break a
Oil viscosities are normally measured at reservoir temperature model down into many convex sub-regions (Carvalho et al.,
over a range of pressures both above and below the saturation 2012). LINGO also has a multi-start feature that restarts the
pressure extending down to near atmospheric pressure. Measure- nonlinear solver from a number of intelligently generated points.
ments below saturation pressure are made under differential This allows the solver to find a number of locally optimal points
conditions, i.e., matched as closely as possible to the stage and reports the best one found. Finally, LINGO can automatically
pressures used for the differential vaporization. Rolling Ball linearize a number of nonlinear relationships through the
viscometers are constructed in such a way as to allow a pseudo addition of constraints and integer variables so that the trans-
differential vaporization of gas to be conducted within them, formed linear model is mathematically equivalent to the original
leaving the oil to fill the measuring chamber. In this way, the nonlinear model.
viscosity of the oil in the reservoir can be measured as gas is Proposed correlations include dead oil viscosity, below bubble
depleted from it. The change in viscosity with release of gas is point, bubble point as well as under-saturated oil viscosity.
normally very large. Optimization approach is used to find the optimum formula for
In the present work, a large database consisting of PVT each region. In each individual region based on 80% of experi-
experimental data of more than 1000 series of Iranian oil mental data a new correlation has been developed using non-
reservoirs have been measured to develop a new correlation linear multiple regressions and minimization of AAPRE has been
approach, as pointed out earlier. These data include oil API considered as the objective function of this regressions. The PVT
gravity, reservoir temperature, saturation pressure, solution data ranges used in this study is summarized in Tables 1–3.
gas–oil ratio as well as PVT measurements (oil characterization)
at reservoir temperature. The ranges of these data cover almost all 4.1. Dead oil viscosity correlation
Iranian oil reservoirs PVT data and consequently the developed
correlation based on these data could be reliable for prediction of Similar to discussed correlations, in the present study the
other Iranian oil reservoirs viscosity, as mentioned earlier. Three viscosity of dead oil is considered to be a function of temperature
oil viscosity samples have been reported in Table 4. More detailed and oil API gravity. Based on nonlinear multiple regressions, a
information is available upon request. new simple correlation form was developed and minimization of
AAPRE has been considered as the objective function of these
regressions, as mentioned earlier:
4. Correlation development A C
mod ¼ eD ð2Þ
B
To develop new viscosity correlations (particularly) for Iranian
oil reservoirs, LINGO (Linear Interactive and General Optimizer) These four polynomials are given below:
user-friendly software has been utilized, which takes advantage A ¼ T 2 þ a1 T þ a2 ð3Þ
of an in-house well-proven nonlinear optimization algorithm that
is widely used in science and technology (Carvalho et al., 2012; B ¼ API2 þa3 API þ a4 ð4Þ
Chuang et al., 2012; Dupuy et al., 2005; Fan et al., 2011; Liu et al.,
2012; Vidal et al., 2007; Miao, 2011). For nonlinear programming C ¼ a5 T þa6 ð5Þ
A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 57

D ¼ a7 API þ a8 ð6Þ the best formula was found to be as follow:


     
where mod represents dead oil viscosity in cp, T is temperature in m a2 m a4
mo ¼ mob þ ðPPb Þ a1 ob þ a3 ob þ a5 mob ða6 Pb þ a7 Þ ð9Þ
1F and API is dead oil API gravity. The values of constants a1 to a8 Pb Pb
are given below.
where mo and mob represent oil viscosity at interested pressure
a1 ¼ 160:0514 ; a2 ¼ 12488:07 ; a3 ¼ 3482:605 ; a4 ¼ 43254:99 ; and bubble point pressure, respectively. Also, P and Pb represent
a5 ¼ 0:004525228 ; a6 ¼ 1:329148 ; a7 ¼ 0:004335506; a8 ¼ 0:08006255 ; pressure and bubble point pressure, respectively.
The values of constants a1 to a7 are given below:

4.2. Saturated oil viscosity correlation a1 ¼ 0:0007376096 ; a2 ¼ 0:3895556; a3 ¼ 0:2444663 ; a4 ¼ 1:226398;


a5 ¼ 0:00000560643; a6 ¼ 0:001254195 ; a7 ¼ 0:3026506;

In this study, similar to Khan et al. (1987) and Labedi (1992) The PVT data ranges that used for developing this correlation
two distinct correlations have been proposed. Most of saturated are summarized in Table 3.
oil viscosity correlations introduce saturated oil viscosity as a
function of both dead oil viscosity and solution gas oil ratio while
others express it as a function of dead oil viscosity and saturation 5. Performance evaluation
pressure (Naseri et al., 2005).
In this work, several forms of correlations as a function of dead To evaluate the performance and accuracy of the new pro-
oil viscosity, bubble point viscosity, solution gas oil ratio, pressure posed correlations and previously discussed ones, both statistical
and bubble point pressure have been tested for below bubble point and graphical error analysis have been utilized simultaneously.
oil viscosity, and the results indicated that the best correlation could
be developed based on four parameters including: dead oil viscosity, 5.1. Statistical error analysis
bubble point oil viscosity, pressure and bubble point pressure. In
this correlation, pressure ratio (pressure divided by bubble point To assess the accuracy and performance of the existing
pressure) and pressure differential (pressure minus bubble point correlations and new developed ones some statistical parameters
pressure) have been involved. This correlation is expressed as defined including: average percent relative error, average percent
follow: absolute relative error, standard deviation of error, mean square
 a1 error and coefficient of determination. Definitions and equations
P a4
mo ¼ mob ða2 ða3 ððPPb ÞÞ Þ Þ þ a5 mod a6 mob a7 ðPP b Þ ð7Þ of those parameters are given below:
Pb
where mo represents oil viscosity at interested pressure, mob is oil 1. Average Percent Relative Error (APRE). It measures the relative
viscosity at bubble point, P also shows pressure and Pb is bubble deviation from the experimental data, defined by:
point pressure. The values of constants a1 to a7 are given below:
1Xn
Er ¼ E ð10Þ
a1 ¼ 0:3136116; a2 ¼ 0:06182307; a3 ¼ 0:0002871137; a4 ¼ 0:8435338 ; ni¼1 i
a5 ¼ 0:0001231158; a6 ¼ 0:6575529; a7 ¼ 0:2137226;
where Ei is the relative deviation of an estimated value from an
The same approach was undertaken to find the optimum experimental value and is expresses as Percent Relative Error:
formula for bubble point viscosity. A linear relationship exists " #
between log10(mob) and log10(mod) at any specific solution gas oil ðmÞexp ðmÞest
Ei ¼  100 ) i ¼ 1,2,3,:::,n ð11Þ
ratio (Elsharkawy and Alikhan, 1999). This new proposed correla- ðmÞexp
tion is based on solution gas oil ratio and dead oil viscosity.
Proposed correlation for bubble point in this study is as follow:
2. Average Absolute Percent Relative Error (AAPRE). It measures
   ða3 RsÞ
the relative absolute deviation from the experimental data,
mob ¼ a1 þ 3a1  eða2 RsÞ m0d ðð2a1 þ a1 e Þ
þa4 ð8Þ
defined by:
where mob , mod and Rs are bubble point oil viscosity, dead oil
1Xn
viscosity and solution gas oil ratio, respectively. The values of Ea ¼ 9E 9 ð12Þ
ni¼1 i
constants a1 to a4 are given below.
a1 ¼ 0:25; a2 ¼ 0:002038792; a3 ¼ 0:00004354996 ; This parameter is considered as the main criterion in statistical
a4 ¼ 0:01726128; error analysis throughout this study.
3. Root Mean Square Error (RMSE). It measures the data disper-
The PVT data ranges that used in this study are summarized in
sion around zero deviation, defined by:
Table 2. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u n  2
u1 X
RMSE ¼ t miexp miest ð13Þ
4.3. Under-saturated oil viscosity correlation ni¼1

Most of correlations use pressure differential (pressure minus


4. Standard Deviation (SD). It is a measure of dispersion and a
bubble point pressure) as primary correlating parameter while
lower value of it shows a smaller degree of scatter. However, it
others use pressure ratio (pressure divided by bubble point
is defined as:
pressure). Apart from that, bubble point oil viscosity is the most vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
common correlating parameter. However some methods also use u !2ffi
u 1 X n m  m
SD ¼ t
iexp iest
dead oil viscosity and oil API gravity as correlating parameters ð14Þ
n1 i ¼ 1 miexp
(Elsharkawy and Alikhan, 1999; Labedi, 1992). In this region,
plotting (mo–mob) versus (P–Pb) leads to a series of straight lines
for various oils. The slope of these lines was found to be a function 5. Coefficient of Determination (R2). It is a simple statistical para-
of Pb and mob. However, the results of many functional format, meter shows how good model matches the data and subse-
based on nonlinear multiple regressions, has been checked and quently represent a measure of the utility of the model. In fact,
58 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

the closer the value of R2 to 1, the better the model fits the data. 2. Crossplots: In this technique, all estimated values are plotted
It is defined as: against the experimental values and thus a crossplot is formed.
n  2 A 451 straight line (unit slope line) between the experimental
P
miexp miest values and estimated data points is drawn on the crossplot
R2 ¼ 1i ¼ 1n ð15Þ which shows the perfect correlation line. The closer the
P  2
plotted data to 451 perfect correlation line, the higher is the
miest m
i¼1 reliability of the correlation.
where m is the mean of the experimental data values as in above
formula presented.
6. Results and discussion
5.2. Graphical error analysis
In this part of our study, the accuracy and validation of the
To visualize the accuracy and performance of a correlation, in developed correlations have been investigated. This section is divided
general, two graphical analysis are employed including crossplot in to three parts including dead oil viscosity, saturated oil viscosity
and error distribution. and under-saturated oil viscosity. Afterward, the validation of viscos-
ity is checked based on some well-known trends of viscosity versus
1. Error Distribution: It is a tool to measure error distribution specific properties of oil for each individual section. The accuracy of
around the zero line to indicate if the correlation has an error those correlations is confirmed according to statistical and graphical
trend or not. error analysis which previously discussed. It should be noted that the

Fig. 2. Dead oil viscosity variation with oil API gravity @100 1F using various correlations.

Fig. 3. Dead oil viscosity (this study).


A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 59

application of these new proposed correlations are in ranges that 6.1. Dead oil viscosity
these correlations have been derived from them. In addition, applica-
tion of these new correlations to other regions located around the To investigate the validity of the proposed correlation for dead
world should be checked. oil viscosity three of the most precise correlations including

Table 5
Statistical error analysis of dead oil viscosity correlations.

Author Ea Er R2 RMSE SD

Beal (1946) 891.2  890.1 0.10881 83.14 12.01


Beggs and Robinson (1975 216.7  203.7 0.03762 245.05 6.71
Glaso (1980) 33.4  24.1 0.92706 3.84 0.43
Kaye (1985) 52.0 52.0 0.22684 10.19 0.56
Al-Khafaji et al. (1987) 29.9  2.2 0.72835 6.04 0.39
Petrosky (1990) 41.6  28.6 0.86958 4.18 0.54
Egbogah and Ng (1990) 55.6  48.0 0.92083 3.26 0.72
Labedi (1992) 177.9  163.9 0.39109 14.95 2.47
Kartoatmodjo and Schmidt (1994) 36.8  27.4 0.90653 4.50 0.49
Bennison (1998) 70.9 1.8 0.66894 12.25 0.82
Elsharkawy and Alikhan (1999) 72.9  70.5 0.90656 13.25 0.87
Hossain et al. (2005) 68.9 8.8 0.60003 16.24 0.77
Naseri et al. (2005) 27.5 13.5 0.82334 3.88 0.33
Alomair et al. (2011) 72.4 62.4 0.82753 6.36 0.79
This study (training data) 17.4 3.2 0.95838 2.30 0.23
This study (testing data) 18.5 0.8 0.95732 1.65 0.26

Fig. 4. Percent relative error distribution for dead oil viscosity correlations.
60 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

Naseri et al. (2005), Al-Khafaji et al. (1987), Glaso (1980) correla- gravity increases, the dead oil viscosity decreases so Fig. 2 shows
tions for dead oil viscosity prediction of Iranian oil reservoirs have this trend for the proposed correlation as well as those three
been selected for comparison. As we know, whatever oil API previously mentioned correlations at 100 1F. A general reduction

Fig. 5. Crossplot for dead oil viscosity correlations.

Fig. 6. Experimental and predicted by newly proposed correlation of four below bubble point oil viscosity samples.
A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 61

in dead oil viscosity is expected with increasing temperature and so provide an independent measure of correlations performance.
this effect is more sensible at lower temperature. This trend is As can be seen, most of previously published correlations have
captured by the new proposed correlation for dead oil viscosity huge error and could not precisely predict dead oil viscosity of
(see Fig. 3). Iranian oil reservoirs. Beal (1946) correlation is the worst correla-
Statistical error analysis for the dead oil viscosity correlations tion for dead oil viscosity prediction of Iranian oil reservoirs
are illustrated in Table 5. As previously mentioned in this section, and overestimates all data points while (Naseri et al. (2005)
the last 20% of experimental data have been used for testing correlation has given the best results for these reservoirs
which have no effect on development of the new correlations and among the mentioned correlations. In spite of more accurate

Table 6
Statistical error analysis of below bubble point oil viscosity correlations.

Author Ea Er R2 RMSE SD

Chew and Connally (1959)[1] 19.5  10.1 0.95709 0.70 0.24


Chew and Connally (1959)[2] 28.9 2.1 0.89653 1.26 0.33
Chew and Connally (1959)[3] 15.7  2.9 0.98393 0.38 0.20
Beggs and Robinson (1975) 22.3 20.4 0.86929 0.82 0.27
Al-Khafaji et al. (1987) 15.4  2.7 0.98695 0.33 0.20
Khan et al. (1987) 22.9  22.8 0.98910 0.47 0.29
Petrosky (1990) 15.1 5.2 0.98136 0.36 0.19
Labedi (1992) 9.3 0.7 0.95751 0.51 0.12
Kartoatmodjo and Schmidt (1994) 16.4 4.9 0.98600 0.34 0.11
Elsharkawy and Alikhan (1999) 15.3 7.9 0.94414 0.57 0.19
Hossain et al. (2005) 64.9  54.9 0.91022 0.75 1.80
Naseri et al. (2005) 34.0 32.4 0.87991 0.84 0.38
Bergman and Sutton (2007) 15.0 9.7 0.97847 0.39 0.19
This study (training data) 3.9  0.8 0.99825 0.11 0.05
This study (testing data) 4.3  0.1 0.99963 0.16 0.06

Fig. 7. Percent relative error distribution for below bubble point oil viscosity correlations.
62 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

results of Naseri et al. (2005) correlation among the other closer the value of R2 to 1. In addition, comparison between
correlations, it is not sufficiently precise at all. Table 5 reveals AAPRE and APRE for each individual correlation demonstrates
that the new proposed correlation has the smallest average that some of these correlations such as Beal (1946) correlation,
absolute percent relative error, average percent relative error, Beggs and Robinson (1975) correlation, Glaso (1980) correlation,
root mean square error and standard deviation and also it has the Egbogah and Ng (1990) correlation, Labedi (1992) correlation,

Fig. 8. Crossplot for below bubble point oil viscosity correlations.

Fig. 9. Bubble point oil viscosity vs. solution GOR.


A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 63

Kartoatmodjo and Schmidt (1994) correlation and Elsharkawy have been selected and the experimental data and predicted
and Alikhan (1999) correlation overestimate dead oil viscosity of values by this proposed correlation were plotted on the same
Iranian oil reservoirs while Kaye, 1985 correlation underestimate figure (see Fig. 6). It is evident by increasing pressure, this
this property. correlation correctly predicts lower viscosity and also it is
Furthermore, three of the most accurate correlations including marvelously in acceptable agreement with the experimental data.
Naseri et al. (2005), Al-Khafaji et al. (1987) and Glaso (1980) Statistical error analysis for below bubble point oil viscosity
correlations have been selected to visualize the accuracy and correlations are demonstrated in Table 6. Comparison between
performance of these correlations as well as the new proposed AAPRE and APRE for each individual correlation reveals that some
correlation. For this purpose, two graphical analysis have been of these correlations such as Khan et al. (1987) and Hossain et al.
employed including crossplot and error distribution. Fortunately, (2005) correlations overestimate below bubble point oil viscosity
none of these correlations shows trend error (Figs. 4 and 5). It is of Iranian oil reservoirs while Beggs and Robinson (1975), Naseri
evident that almost all data of the new correlation lay on unit slope et al. (2005) and Bergman and Sutton (2007) correlations under-
line and this shows its high accuracy (Fig. 5). Apart from that, the estimate this property.
new proposed correlation has the smaller error range and least Furthermore, three of the most accurate correlations including
scatter around the zero error line (Fig. 4). Labedi (1992), Bergman and Sutton (2007) and Petrosky (1990)
correlations have been selected to visualize the accuracy and
6.2. Saturated oil viscosity correlation performance of these correlations as well as the new proposed
one. For this purpose, similar to previous section two graphical
As previously mentioned, bubble point pressure has the analysis have been employed including crossplot and error
smallest oil viscosity value and pressure reduction below this distribution. Error distribution of Labedi (1992) correlation shows
point causes gas release and subsequently leads to increase in oil lower relative error at higher pressures gradient while error
viscosity as well as oil density (Fig. 1). Two distinct correlations distribution of Bergman and Sutton (2007) and Petrosky (1990)
for below bubble point and at bubble point have been proposed. correlations is almost uniformly with pressure gradient (Fig. 7).
To check the validity of the new proposed correlation for Moreover, Fig. 7 illustrates that this new proposed correlation has
below bubble point region, four samples of Iranian oil reservoirs the smaller error range and least scatter around the zero error

Fig. 10. Bubble point oil viscosity (this study).

Table 7
Statistical error analysis of bubble point oil viscosity correlations.

Author Ea Er R2 RMSE SD

Chew and Connally (1959)[1] 18.1 11.98 0.86769 0.82 0.23


Chew and Connally (1959)[2] 31.1 17.09 0.77383 1.35 0.36
Chew and Connally (1959)[3] 37.2  4.49 0.89212 0.65 0.94
Beggs and Robinson (1975) 34.0 33.68 0.71387 0.78 0.37
Petrosky (1990) 23.7 22.28 0.89862 0.56 0.27
Labedi (1992) 93.1  88.43 0.47667 3.40 1.15
Kartoatmodjo and Schmidt (1994) 28.9 27.13 0.89815 0.62 0.32
Elsharkawy and Alikhan (1999) 27.1 26.45 0.88960 0.56 0.30
Hossain et al. (2005) 1612.4  1600.52 0.00443 56.11 132.97
Naseri et al. (2005) 21.9 16.60 0.89228 0.47 0.25
Bergman and Sutton (2007) 25.9 25.21 0.88384 0.55 0.29
This study (training data) 13.7 4.49 0.94901 0.29 0.17
This study (testing data) 15.1 5.77 0.98292 0.32 0.19
64 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

line. Crosspolts of these correlations show that all predicted data scatter and larger error around the zero error line and also both of
of the new correlation lay on unit slope line and this confirms its these correlations underestimate bubble point oil viscosity of
perfect accuracy while the three other correlations underestimate Iranian oil reservoirs (see Fig. 11). Apart from that, Fig. 11
viscosity at higher values (Fig. 8). illustrates that the proposed correlation has the smaller
Fig. 9 illustrates the behavior of bubble point oil viscosity error range and least scatter around zero error line. Crosspolts
correlation developed in this study as well as two of the most of these correlations show most of predicted data of the proposed
accurate ones including, Petrosky (1990), and Chew and Connally correlation lay on unit slope line and this proves its perfect
(1959)[1] correlations. This figure correctly displays for these accuracy (Fig. 12).
correlations, that increasing solution gas oil ratio reduces bubble
point oil viscosity and also the greatest oil viscosity decreasing
occurs at low solution gas oil ratio. Fig. 10 indicates that bubble 6.3. Under-saturated oil viscosity correlation
point oil viscosity increases as dead oil viscosity increases. In
addition, this figure demonstrates that as solution gas oil ratio As it is known, at pressure above bubble point, whatever
increases up to a certain point, bubble point oil viscosity pressure gradient increases the viscosity increases so Fig. 13 follows
decreases considerably. such trend for four samples of Iranian oil reservoirs and fortunately
Statistical error analysis for bubble point oil viscosity correla- the proposed correlation for this region is captured this trend and
tions are shown in Table 7. Comparison between AAPRE and APRE also it is in good agreement with experimental data. Statistical error
for each individual correlation illustrates that Hossain et al. analysis of under-saturated oil viscosity correlations are illustrated
(2005) and Labedi (1992) correlations overestimate bubble point in Table 8. From this table, two conclusions could be drawn. The first
oil viscosity of Iranian oil reservoirs while Beggs and Robinson conclusion is that Vazquez and Beggs (1980), Petrosky (1990),
(1975), Kartoatmodjo and Schmidt (1994), Elsharkawy and Labedi (1992) and Hossain et al. (2005) correlations overestimate
Alikhan (1999), Naseri et al. (2005) and Bergman and Sutton oil viscosity at under-saturated region for Iranian oil reservoirs while
(2007) correlations underestimate this property. Kartoatmodijo and Schmidt (1994) correlation underestimate this
Moreover, three of the most accurate correlations including property. The second conclusion is that oil viscosity correlations for
Chew and Connally (1959)[1], Petrosky (1990) and Naseri et al. prediction of under-saturated regions are more accurate than
(2005) have been selected to visualize the accuracy and correlations for dead oil, below bubble point and at bubble point
performance of these correlations as well as the developed regions (see Tables 5–8). This may be due to the fact that oil
correlation. Similar to previous sections two graphical analysis viscosity variation at under-saturated region is governed by pressure
have been utilized including crossplot and error distribution. differential (pressure minus bubble point pressure) and also solution
Petrosky (1990) and Naseri et al. (2005) correlations show wider gas oil ratio is constant at this region.

Fig. 11. Percent relative error distribution for bubble point oil viscosity correlations.
A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 65

Fig. 12. Crossplot for bubble point oil viscosity correlations.

Fig. 13. Experimental and predicted by the newly proposed correlation for four under-saturated oil viscosity samples.

In addition, similar to previous parts three of the most elevated pressure gradients while the proposed correlation in this
accurate correlations including Orbey and Sandler (1993), Beal study has the smallest error range and least scatter around the
(1946) and Elsharkawy and Alikhan (1999) correlations have been zero error line. Crossplots of these correlations displays high
selected to visualize the accuracy and performance of these accuracy of these correlations at lower oil viscosity values while
correlations as well as the proposed one. Although these three at higher values Orbey and Sandler (1993), Elsharkawy and
correlations are more accurate than the others, Orbey and Sandler Alikhan (1999) correlations underestimate oil viscosity. Both of
(1993) and Elsharkawy and Alikhan (1999) correlations under- these Figs. 14 and 15 show exceptional and acceptable perfor-
estimate and Beal (1946) correlation overestimates oil viscosity at mance of the proposed correlation in this study for Iranian oil
66 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

Table 8
Statistical error analysis of under-saturated oil viscosity correlations.

Author Ea Er R2 RMSE SD

Beal (1946) 1.9 0.8 0.99783 0.133 0.031


Vazquez and Beggs (1980) 5.2  4.9 0.97138 0.575 0.094
Khan et al. (1987) 3.2  2.7 0.98812 0.298 0.059
Petrosky (1990) 6.7  5.9 0.83011 0.940 0.107
Labedi (1992) 1078.5  1078.4 0.09697 43.557 16.008
Orbey and Sandler (1993) 1.9 0.5 0.96242 0.495 0.035
Kartoatmodjo and Schmidt (1994) 3.6 3.6 0.99864 0.107 0.060
Elsharkawy and Alikhan (1999) 1.9 1.1 0.93720 0.618 0.036
Hossain et al. (2005) 4.7  4.5 0.9948 0.225 0.072
This study (training data) 1.2 0.2 0.99966 0.039 0.019
This study (testing data) 1.3 0.2 0.99989 0.057 0.022

Fig. 14. Percent relative error distribution for under-saturated oil viscosity correlations.

reservoirs. Fabulously, this correlation matches all experimental oil reservoirs. Unfortunately, few authors have reported the
data through all viscosity values and shows R2 value almost database which they have used to develop their correlations.
equal to 1. Most of the authors only noted about the ranges and the number
of data points which their correlations have been derived from.
6.4. Comparison with the literature database Moreover, some papers reported some data points of their dataset
but their input parameters are different from our correlating
The database used for developing the correlations of this study parameters. At last, we just found few data points from Ahmed
belongs to a particular geographical location (Iranian oil reser- (2010) for the dead oil viscosity and several data points from
voirs). As mentioned earlier, most of the previously published Isehunwa et al. (2006) for the under-saturated oil viscosity which
correlations also belong to a particular region around the world. are reported in Tables 9 and 10, respectively. Tables 9 and 10
However, we conducted a thoroughly search into the literature to show the promising results of this correlation compared to
find a comprehensive and free database for crude oil viscosity experimental data not used in its development. Unfortunately,
around the world to compare the capabilities of the newly no dataset for viscosity at bubble point and below bubble point
developed correlations when estimating the viscosity of other have been found with the same input parameters of our work.
A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68 67

Fig. 15. Crossplot for under-saturated oil viscosity correlations.

Table 9 Table 10
Comparison of the predicted dead oil viscosity and experimental data from Ahmed Comparison of the predicted under-saturated oil viscosity and experimental data
(2010). from Isehunwa et al. (2006).

Temperature API Exp. dead oil Pred. dead oil Absolute relative Pressure, Bubble Bubble Exp. under- Pred. under- Absolute
(1F) viscosity(cp) viscosity(cp) error percent psia point oil point saturated oil saturated oil relative
viscosity, pressure, viscosity, cp viscosity, cp error
250 47.1 0.765 0.570 25.5 cp psia percent
220 40.7 1.286 0.941 26.8
260 48.6 0.686 0.507 26.1 2080 2.600 2122 2.600 2.609 0.3
237 40.5 1.0014 0.843 15.8 1859 4.930 2148 5.110 5.108 0.1
218 44.2 1.009 0.787 22.0 1941 1.070 2207 1.090 1.089 0.1
180 27.3 4.166 4.092 1.8 2058 1.020 2303 1.020 1.036 1.6
2004 1.840 2313 1.860 1.879 1.0
1937 1.880 2327 1.900 1.931 1.6
2284 0.525 2340 0.527 0.527 0.0
7. Conclusions 804 1.270 2403 1.310 1.521 16.1
2029 0.800 2448 0.850 0.823 3.2
Several empirical correlations for oil viscosity have been 2393 1.010 2467 1.020 1.014 0.6
evaluated using a large databank of Iranian oil reservoirs, which 2412 3.270 2568 3.310 3.336 0.8
2485 0.540 2573 0.548 0.543 0.9
were measured using Rolling Ball viscometer (Ruska, series 1602).
2435 0.460 2575 0.467 0.461 1.3
In this study, average absolute percent relative error (AAPRE) has 2557 0.650 2670 0.660 0.651 1.4
been considered as the main screening criterion. It was found for 2649 1.160 2697 1.170 1.215 3.8
dead oil region that all of correlations exhibit high errors, 1971 1.240 2728 1.340 1.301 2.9
however, Naseri et al. (2005), Al-Khafaji et al. (1987) and Glaso
(1980) are the most accurate ones. For below bubble point oil
viscosity Labedi (1992), Bergman and Sutton (2007) and Petrosky however, Orbey and Sandler (1993), Beal (1946) and Elsharkawy
(1990) correlations are recommended as the best ones. For bubble and Alikhan (1999) correlations are the most precise ones for this
point viscosity Chew and Connally (1959)[1], Petrosky (1990) and region.
Naseri et al. (2005) correlations are recommended. All correla- In the last step, four correlations have been developed for
tions tested for under-saturated oil viscosity prediction, with the Iranian oil reservoirs which have simplified mathematical format
exception of Labedi (1992) correlation, exhibit good accuracy, and also input data of these correlations are those ones that are
68 A. Hemmati-Sarapardeh et al. / Chemical Engineering Science 90 (2013) 53–68

easily measured in oil fields. The ranges of data used to develop Carvalho, M., Lozano, M.A., Serra, L.M., Wohlgemuth, V., 2012. Modeling simple
these new correlations cover almost all Iranian oil reservoirs PVT trigeneration systems for the distribution of environmental loads. Environ.
Modell. Softw. 30, 71–80.
data and consequently the developed correlations based on these Chew, J., Connally, C., 1959. A viscosity correlation for gas-saturated crude oils.
data could be reliable for prediction of other Iranian oil reservoirs Trans. Am. Inst. Electr. Eng. 216, 23–25.
viscosity. However, for application of these correlations the Chuang, Y.-F., Lee, H.-T., Lai, Y.-C., 2012. Item-associated cluster assignment model
limitations of parameters which these correlations have been on storage allocation problems. Comput. Ind. Eng. 63, 1171–1177.
Dindoruk, B., Christman, P., 2001. PVT properties and viscosity correlations for
derived from them should be considered. These new correlations Gulf of Mexico oils, SPE ATCE in New Orleans, LA.
can be easily implemented in any reservoir simulation software Dupuy, C., Botta-Genoulaz, V., Guinet, A., 2005. Batch dispersion model to optimise
and also they provide better accuracy and performance for traceability in food industry. J. Food Eng. 70, 333–339.
Egbogah, E.O., Ng, J.T., 1990. An improved temperature-viscosity correlation for
viscosity of Iranian oil reservoirs than previously published
crude oil systems. J. Pet. Sci. Eng. 4, 197–200.
correlations. Elsharkawy, A., Alikhan, A., 1999. Models for predicting the viscosity of Middle
East crude oils. Fuel 78, 891–903.
Fan, X.-h., Yi, J., Bao, Z.-q., 2011. Research on safety input research model for
Nomenclature preventing coal gas explosion. Procedia Eng. 26, 2012–2017.
Glaso, O., 1980. Generalized pressure–volume–temperature correlations. SPE J.
Pet. Technol. 32, 785–795.
API oil API gravity Hossain, M.S., Sarica, C., Zhang, H.Q., Rhyne, L., Greenhill, K., 2005. Assessment and
P pressure, psi Development of Heavy Oil Viscosity Correlations. SPE International Thermal
Pb bubble point pressure, psi Operations and Heavy Oil Symposium, Calgary, Canada.
Ikiensikimama, S.S., Ogboja, O., 2009. Evaluation of empirically derived oil
Rs solution gas oil ratio, SCF/STB viscosity correlations for the Niger Delta crude. J. Pet. Sci. Eng. 69, 214–218.
GOR gas oil ratio, SCF/STB Isehunwa, O.S., Olamigoke, O., Makinde, A.A., 2006. A correlation to predict the
T temperature, 1F viscosity of light crude oils. Nigeria Annu. Int. Conf. Exhibition. Soc. Pet. Eng.,
mo oil viscosity, cp Abuja, Nigeria.
Johnson, S.E., Svrcek, W.Y., Mehrotra, A.K., 1987. Viscosity prediction of athabasca
mob bubble point oil viscosity, cp bitumen using the extended principle of corresponding states. Ind. Eng. Chem.
mod dead oil viscosity, cp Res. 26, 2290–2298.
Ei percent relative error Johnson, S.E., Svrcek, W.Y., 1991. J. Can. Pet. Technol. 26 (5), 60.
Kartoatmodjo, T., Schmidt, Z., 1994. Large data bank improves crude physical
Er average percent relative error
property correlations. Oil Gas J. 92, 51–55.
Ea average absolute percent relative error Kaye, S.E., 1985. Offshore California viscosity correlations, COFRC. SPE/PS-CIM/
RMSE root mean square error CHOA International Thermal Operations and Heavy Oil Symposium,
SD standard deviation TS85000940.
Khan, S., Al-Marhoun, M., Duffuaa, S., Abu-Khamsin, S., 1987. Viscosity Correla-
R2 coefficient of determination
tions for Saudi Arabian Crude Oils. SPE Middle East Oil Show, Manama,
n number of data points Bahrain.
Labedi, R., 1992. Improved correlations for predicting the viscosity of light crudes.
J. Pet. Sci. Eng. 8, 221–234.
Little, J.E., Kennedy, H.T., 1968. Calculating the viscosity of hydrocarbon systems
with pressure temperature and composition. Soc. Pet. Eng. J. 6, 157–162.
Acknowledgement
Liu, T., Bu, R., Huang, J., 2012. A cost-efficiency equilibrium problem of regional
single emergency resource guarantee with multi-objective programming. Syst.
Research Institute of Petroleum Industry (RIPI) is acknowl- Eng. Procedia 5, 29–36.
edged for providing the database and supporting this study. Mehrotra, A.K., 1991. Generalized one-parameter viscosity equation for light and
medium liquid hydrocarbons. Ind. Eng. Chem. Res. 30, 1367–1372.
Miao, Z., 2011. Discussion of optimize method of fire alarm dispatching based on
References operation research principle. Procedia Eng. 11, 689–694.
Naseri, A., Yousefi, S.H., Sanaei, A., Gharesheikhlou, A.A., 2012. A neural network
model and an updated correlation for estimation of dead crude oil viscosity.
Abdul-Majeed, G.H., Clark, K.K., Salman, N.H., 1990. New correlation for estimating
Brezilian J. Pet. Gas 6, 31–41.
the viscosity of undersaturated crude oils. J. Can. Pet. Technol., 29.
Naseri, A., Nikazar, M., Mousavi Dehghani, S.A., 2005. A correlation approach for
Ahmed, T., 2010. Chapter 2—Reservoir-Fluid Properties, Reservoir Engineering
prediction of crude oil viscosities. J. Pet. Sci. Eng. 47, 163–174.
Handbook, fourth ed. Gulf Professional Publishing, Boston 29–135.
Orbey, H., Sandler, S.I., 1993. Prediction of the viscosity of liquid hydrocarbons and
Ahrabi, F., Ashcroft, S.J., Shearn, R.B., 1987. High pressure volumetric, phase
their mixtures as a function of temperature and pressure. Can. J. Chem. Eng.
composition and viscosity data for a north sea crude oil and Ngl. Chem. Eng.
Res. Des. 65, 63–73. 71, 437–446.
Al-Khafaji, A., Abdul-Majeed, G., Hassoon, S., 1987. Viscosity correlation for dead, Petrosky, G.E.J., 1990. PVT Correlations for Gulf of Mexico Crude Oils. MSC Thesis,
live and undersaturated crude oils. J. Pet. Res. 6, 1–16. University of Sowthwestern Louisiana, Lufayette, Louisiana, USA.
Al-Marhoun, M.A., 2004. Evaluation of empirically derived PVT properties for Sutton, R.P., Bergman, D.F., 2006. Undersaturated oil viscosity correlation for
Middle East crude oils. J. Pet. Sci. Eng. 42, 209–221. adverse conditions. SPE Annu. Tech. Conf. Exhibition. Soc. Pet. Eng., San
Ali, S.M.F., Thomas, S., 1996. The promise and problems of enhanced oil recovery Antonio, Texas, USA.
methods. J. Can. Pet. Technol., 35. Sutton, R.P., Farshad, F., 1990. Evaluation of empirically derived PVT properties for
Almehaideb, R.A., 1997. Improved PVT correlations for UAE crude oils. Middle East gulf of Mexico crude oils. SPE Reservoir Eng. (Soc. Pet. Eng.) 5 (79–86), 13172.
Oil Show Conf. Bahrain. Svrcek, W., Mehrotra, A., 1988. One parameter correlation for bitumen viscosity.
Alomair, O.A., Elsharkawy, A.M., Alkandari, H.A., 2011. Viscosity Prediction of Chem. Eng. Res. Des. 66, 323–327.
Kuwaiti Heavy Crudes at Elevated Temperatures, SPE Heavy Oil Conference Teja, A., Rice, P., 1981. Generalized corresponding states method for the viscosities
and Exhibition. Society of Petroleum Engineers, Kuwait City, Kuwait. of liquid mixtures. Ind. Eng. Chem. Fundam. 20, 77–81.
Beal, C., 1946. Viscosity Air, Water, Nat. Gas, Crude Oil Assoc. Gases Oil Field Temp Vazquez, M., Beggs, H.D., 1980. Correlations for fluid physical property prediction.
Pressures. SPE J. Pet. Technol. 32, 968–970.
Beggs, H.D., Robinson, J.R., 1975. Estimating the viscosity of crude oil systems. SPE Vidal, D., Blobel, J., Pérez, Y., Thormann, M., Pons, M., 2007. Structure-based
J. Pet. Technol. 27, 1140–1141. discovery of new small molecule inhibitors of low molecular weight protein
Bennison, T., 1998. Prediction of heavy oil viscosity, IBC Heavy Oil Field Develop- tyrosine phosphatase. Eur. J. Med. Chem. 42, 1102–1108.
ment Conference, London. Xu, D.-H., Khurana, A.K., 1996. A Simple and Efficient Approach for Improving the
Bergman, D.F., Sutton, R.P., 2007. An Update to Viscosity Correlations for Gas- Prediction of Reservoir Fluid Viscosity, SPE Asia Pacific Oil and Gas Conference.
Saturated Crude Oils, SPE Annual Technical Conference and Exhibition. Society 1996 Copyright 1996. Society of Petroleum Engineers, Inc., Adelaide, Australia.
of Petroleum Engineers, Anaheim, California, U.S.A..

You might also like