You are on page 1of 19

SPE-199933-MS

Introduction of Steam-Assisted Gravity-Drainage Oil Rate Prediction Using


the 5-LINE Model

Mazda Irani and Sahar Ghannadi, Ashaw Energy Ltd.

Copyright 2020, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Canada Heavy Oil Techncial Conference originally scheduled to be held in Calgary, Alberta, Canada, 18 – 19
March 2020. Due to COVID-19 the physical event was postponed until 29 September – 2 October 2020 and was changed to a virtual event. The official proceedings
were published online on 24 September 2020.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Summary
Steam-assisted gravity drainage (SAGD) is the preferred thermal recovery method used to produce bitumen
from Athabasca deposits in Alberta, Canada. SAGD operation is experiencing five stages: ramp-up,
initial plateau, full-length plateau, wind-down and coalescence. The physics controlling the production
mechanisms in each stage is different. Ramp-up is controlled by sweeping and injection pressure and water
mobility are the most important factors. In initial plateau and full-length plateau, the chamber growth
and bitumen viscosity-temperature dependency are the main controllers. Wind-down initiates as heat-loss
overcomes the input enthalpy, and production controls by reduction of heated front close to steam front.
Finally, coalescence is a result of reduction of oil availability at the edge of steam chamber. Such reduction
is modeled with linearly reducing oil pathway as chambers are coalescing.
Butler's model commonly used for history matching and prediction of SAGD oil rate is mainly meant to
model the full-length plateau stage and that is why it is over-predicting the ramp-up stage and not estimating
the oil rate trend in wind-down and coalescence. This work is a continuation of a previous part discussing
the predicve model for SAGD process (Irani, 2019). The purpose of this work is to create an end-life stage:
winddown and coalescence; and then use the decision tree to optimize the solution. The model that includes
ramp-up, early plateau, plateau, wind-down and coalescence is called 5-LINE model is a mechanistic model
that controls main physic on each stage. Although the 5-LINE model is mainly derived based on physics
controlling each stage, it has enough flexibility to match different geological characteristics. The 5-LINE
model is structured in regression tree to minimize the error and then a decision-tree (DT) learning that
branches from it to honour dynamics that cannot be honoured by 5-LINE model. This model is tested vs.
oil production results of Suncor/MacKay River and Devon/Jackfish, as a result the final predictive model
can predict oil rate reasonably good enough that can compete with results of dynamic reservoir numerical
simulation.

Introduction
Steam-assisted gravity drainage (SAGD) is one successful thermal recovery technique applied in the
Athabasca and Peace River reservoirs in Central and Northern Alberta, Canada. In this technique, two
2 SPE-199933-MS

horizontal wells are drilled parallel with vrertical distance of roughly 5m. The steam is injected from the
upper well at high pressures, and lower well acts as a production well. The injected steam loses its heat to the
cold bitumen at the edge of steam chamber. The viscosity of the bitumen falls several orders of magnitude
as its heated and its mobility rises several orders of magnitude. The immobile bitumen now flows under
gravity toward a production well.
SAGD was first proposed by Butler (1977), then Alberta Oil Sands Technology and Research Authority
(AOSTRA) tested this process in Underground Test Facility (UTF) at Fort McMurray, Alberta, Canada in
1985. In UTF a shaft and tunnel access facility in the Athabasca deposit was built, and horizontal injector and
producer are drilled from tunnels. Following the success of the UTF pilot the SAGD operated in elsewhere.
And after development of horizontal drilling the SAGD wellpairs are nowadays are drilled from surface.
SAGD is Is essentially a conduction heat transfer ablation process were by steam transfers its heat by
conduction to cold bitumen at the edge of chamber and in turn is condensed to water which together with
heated bitumen or oil fall under the influence of gravity, exposing a new element of oil sands to steam
interface.
Due to long run-time associated with numerical simulations for history matching and forecasting of
steam-assisted gravity drainage (SAGD) process, analytical solutions such as Butler's mathematical model
are used for oil production prediction and forecasting. The major pitfall in using the Butler type analytical
models to predict the oil production in rising phase, that his original theory (Butler and Stephens, 1981) that
presents a "S shape" chamber. As shown in Figure 3 the chamber in Butler's original theory rise rapidly at
infinite small time and then become asymptotic to the top of the pay zone. In other words, Butler's chamber
starts from top of the reservoir or as he stated starts as a vertical hot plane. In lab test by Butler the Hele-Shaw
cell have the injector at the top of the cell and the steam chamber starts from top of the cell, which is mostly
shown in pictures from those tests. But the chamber starts to grow from injector at the bottom of the reservoir.
So, using similar shape to calculate the height of the chamber provides height of the chamber which starts
from the top of the reservoir and even do not contain injector and producer at early times (Figure 3). The
other pitfall is embedded in that the physics of ramp-up (or rising phase) and lateral spreading are totally
different. The oil rate in ramp-up is highly dependent to the upward growth of the steam chamber which
is controlled by frontal instability of steam condensate front into heated bitumen. In this study, evolution
of disturbance (spatial growth rate) was explored by solving the initial value problem governing the linear
stability of the pressure and water-phase velocity normal to the edge of the steam chamber that would be
found in the SAGD process.

Figure 1—Cross section of SAGD process: circulation phase (a), early phase
(b), and steam-injection phase (c) (modified from Irani and Ghannadi 2013).
SPE-199933-MS 3

Figure 2—Steam chamber shape and a flat front at the edge of steam chamber in well-pair C6 in MacKay
River operation using seismic cross section. Seismic cross section (Section a); identification lines
shown over seismic cross section (Section b); and components and phases in well-pair C6 (Section c).
4 SPE-199933-MS

Figure 3—Calculated interface positions for an infinite reservoir using the original Butler theory.

Figure 4—The minimum steam chamber interface velocity for an instability at different steam injection temperatures.
SPE-199933-MS 5

Currently, data mining approaches developed for conventional reservoirs applications are commonly
employed for SAGD application but with not much of success. There two primary reasons the application
of data mining approaches failed on SAGD process: firstly, SAGD is time dependant, and secondly, there is
a large delay in response after the operator changes the operational conditions. For example, AI techniques
such neural network cannot model SAGD history since its production rate is time dependant and its time-
dependance a function of reservoir properties, and also for a AI technique we need many data points to create
a decent model but unfortunately there are limited number of SAGD wells, and their rate is highly geolgy
dependant that one cannot use the data set of one asset to another asset. And techniques such as timeseries
forecasting that conducted on systems that models the periodic behavior considering lagged variables cannot
work for SAGD since the spatial change of the chamber and the liquid pool, that effects on heat-loss and
recovery is not incorporated in such models.

Mathematical Model of Ramp-up: Equilibrium at Minimum Instability Velocity


The major pitfall in using the Butler type analytical models to predict the oil production in ramp-up or rising
phase, that these type of models (Butler and Stephens, 1981) rise rapidly at infinite small time and then
become asymptotic to the top of the pay zone (see Figure 3). Therefore, the chamber starts as a vertical hot
plane, and will not grow from the bottom of the reservoir. Even in his lab-test by the Hele-Shaw cell, the is
located at the top corner, and the producer at the bottom corner. In the Hele-Shaw cell the porous medium
permeability is modeled by gap width between two flat plates, and to minimize the Reynolds numbers the
gap is infinitesimally small. Therefore, the injector and producer communicate at the early time and the idea
of a vertical hot plane is achieved from day-one.
Chung and Butler (1988) were one of the first studies that shows the Butler's theory (Butler et al., 1981;
Butler and Stephens, 1981; Butler, 1985) are not applicable for early rising period which is happening in
the early stage of the SAGD process when a horizontal injector is used. They also showed that TANDRAIN
theory is applicable from early stages when steam is injected at the top of the reservoir through the vertical
well similar to configuration that is used in IPC/Orion Lake.
Irani (2019) studied the evolution of disturbance (spatial growth rate) by solving the initial value problem
governing the linear stability of the pressure and water-phase velocity normal to the edge of the steam
chamber that would be found in the SAGD process. He calculated the instability velocity or as he called
"critical velocity" as:

As shown in Figure 5a, the vertical/horizontal permeability ratio (Ωk) of the reservoir changes the shape
of the chamber growth, and as shown in Figure 5b for isotropic permeability reservoirs (Ωk = 1) the injection
pressure improves the growth of the chamber in all direction. He suggests that in ramp-up the interface will
grow at the equilibrium velocity of critical velocity. His ramp-up model the oil production rate increases
linearly with time ( ) as:

2
6 SPE-199933-MS

Figure 5—The chamber growth variation in vertical and horizontal direction for different
horizontal/vertical permeability ratio (Section a); and for different injection pressures (Section b).

, that is in contradiction with Butler's ramp-up formula that the rate is correlated to third root of time
(i.e., ) as:

Irani (2019) calulate the ramp-up time as:

In his steam-chamber-rise model, the maximum oil-production rate is:

Recently, the rising- or ramp-up-phase has been studied by many researchers: Zargar and Farouq Ali
(2018) and Keshavarz et al. (2019), that are modification of initial Butler theory (1991). Both Zargar and
Farouq Ali (2018) and Keshavarz et al. (2019) suggested that in the early stage the steam chamber is an
inverted triangle fixed at the producer with a fixed inclination angle of the side interfaces. Zargar and Farouq
Ali (2018) ramp-up model follows the linear correlation with time ( ) similar to Irani (201), but its
governing parameters are completely different. Both Zargar and Farouq Ali (2018) and Keshavarz et al.
(2019) suggested different temperature distribution along the interface from what Butler-plateau model
suggested. The temperature distribution in Zargar and Farouq Ali (2018) model is a function of the geometry
of the steam chamber. Since in Keshavarz et al. (2019) model at the rise-up phase the side interfaces of the
chamber are stationary the temperature distribution uses error function instead of the exponential form for
the Butler (1991) model quasisteady state at the plateau phase.
SPE-199933-MS 7

In this study the modification is applied in Irani (2019), to consider the length of high-quality section
of the producer. The common belief is that in ramp-up phase only the good-quality section is controlling
the production rate, and the average permeability of the high-quality section should be used for ramp-up
evaluation:

And the ramp-up duration is:

Mathematical Model of Plateau: Rotational Effect


Plateau or sideways-expansion phase starts as ramp-up finished. Butler original model and his succeeding
models (TANDRAIN and LINDRAIN) are all based on solution for conduction heat transfer ablation
process in a moving boundary coordinate. The assumption of quasi-steady state for the moving boundary
problem of heat transfer in SAGD has been a basis for the the development of the temperature profile ahead
of steam-interface. In an elegant approach, Butler and Stephens (1981) incorporated an interface-advance
velocity into the material balance, in the manner the final oil rate solution is independent of the interface-
velocity. The general solution for such conduction heat transfer ablation process is given by:

where, Cmodel is a coefficient that is different for models: the values 2, 1.5, 1.3, 1.25, and 1, respectively,
for the models by Butler et al. (1981), Butler and Stephens (1981), Butler (1994), Reis (1992), and Zargar
and Farouq Ali (2018). In this study the TANDRAIN model is used as the only significant modification.
As shown in Figure 7 the original Butler model creates the horizontal section that there is no hydraulic
potential along it. The TANDRAIN model eliminates the horizontal section and the area of the sweeped
are is decreasing.
Another important parameter in Eq. 8, is m- exponent (mμ) that is the Butler (1991) suggested bitumen-
viscosity exponent, that is given by:

Butler and Stephens (1981) proposed the bitumen-viscosity exponent to corroelate viscosity to
temperature in the manner that simply integrate for oil-rate calculation:

10

where μst is viscosity of the bitumen at steam temperature. The m-exponent (mμ) from Eq. 9 is temperature
dependant and is typically in the range of 3 to 5 (see Figure 6). In Appendix B a consistent method that
honours both Eq. 9 and log-log curve fitting is suggested, such approach provides a constant value for m-
exponent (mμ) that not varies with temperature.
8 SPE-199933-MS

Figure 6—Variation of m-exponent (mμ) with operational injection temperature. To


create the graph the viscosity data has to be fitted to Mehrotra and Svrcek (1986, 1987)
viscosity-model other and knowing the viscosity the integration in Eq. 9 is implemented.

Figure 7—Illustration of TANDRAIN model and hydraulic pressure effect.


SPE-199933-MS 9

The oil rate calculated form Eq. 8 is constant and not varies with time, such constant rates are also
suggested by most researchers that follows the Butler analytical approach: Butler et al. (1981), Butler and
Stephens (1981), Butler (1994), Reis (1992); Zargar and Farouq Ali (2018); Keshavarz et al. (2019); Dixon
and Nguyen (2019); but some reserachers such as Akin (2005) and Azad and Chalaturnyk (2010) uses the
Butler analytical approach and their plateu model varies with time. Akin (2005) considered the asphaltene
content change into the oil rate, and Azad and Chalaturnyk (2010) includes the permeability enhancement
into their model.
Although many suggested that oil rate is constant in most SAGD operations there are minimum two
changes along the plateau that are physically should be evaluated. The first change is due to change in
initial conformance to well general conformance. Initial conformance is a section of the well that is fully
conformed in the ramp-up and that is correlataed to high-perm section of the well that impact on most of
the oil rate. And the general conformance is a section of the well that its conformance achieved at the end
of the plateau and the well conformance would not achieve higher than this value even at the end life. This
section is the sandy part of the well and the impermeable sections should be eliminated from this region.
So the oil rate of the first point is calculated as

11

And the duration that this rate is achieved is calculated as:

12

where, λc is the section of the high perm-zone.


To calculate the plateu rotation effect is considered. To calculate we suggest the upper part of the chamber
are not producing as the chamber is spreading more circularly:

13

Where the angle is calculated knowing the interface velocity:


14
To calculate the second point we should initial calculate the duration:

15

And then knowing the time we can calculate the oil rate:

16

The rotational factor is multiplied to the end of plateu. Therefore, the other points of the plateu is
calculated accordingly. Plateu ends at the start of decline phase, this phase is a topic of next section.
10 SPE-199933-MS

Mathematical Model of Wind-down: Heat-loss Effect and Coalescing Effect


The heat-loss to the caprock increases as the chamber grows. As illustrated in Figure 8, the heat-loss to the
seal/caprock caretes a condition that provided enthalpy by injection is not enough to keep the plateau. The
decline condition is satisfied as:

17

Figure 8—Illustration of heat-loss and its effect on decline onset.

Eq. 17 is compared with historical data in Figure 9.


SPE-199933-MS 11

Figure 9—Decline phase validation vs. a well-pair oil production.

SAGD well pads consist of a repeated pattern of the 6 to 10 well-pairs. The symmetric line between
two adjacent well-pairs acts as a confined boundary. The developing steam chamber during the sideways
expansion, mostly in declining phase, as it approaches the confined boundary the rates are declining due to
limited oil availability. Such reduction is exposed in reduction of the edge of steam chamber (Figure 10):

18
12 SPE-199933-MS

Figure 10—Illustration of coalescence effect on oil rate.

Conclusions
This study explains the 5-LINE forecasting model, that includes 5-different phases: rise-up, initial plateau,
spreading, declining, coalescence. This model consists of the physical tuning constants (not empirical). A
new approach was presented to overcome the challenges in previous models. The model can be structured
in regression decision tree that history matched the previous models at low computational costs, that can
be used for future forecasting. Results indicate that the 5-LINE model can provide a good forecast with
minimum geological and operational data.

References
Akin, S. 2005. Mathematical Modeling of Steam-Assisted Gravity Drainage. SPE Res Eval & Eng. 8(5): 372–376. http://
dx.doi.org/10.2118/86963-PA
Arthur, J.E., Gittins, S.D., and Chhina, H.S. 2009. Recovery Process. Patent No. US7556099B2, CA2591498A1,
CA2591498C, US20070295499, WO2007143845A1, Filing date: June 14, 2006, Original Assignee: Encana
Corporation.
Azad, A. and Chalaturnyk, R.J., 2010. A Mathematical Improvement to SAGD Using Geomechanical Modelling. JCPT
49(10): 53–64. http://dx.doi.org/10.2118/141303-PA
Butler, R.M., 1985. A New Approach to the Modelling of Steam-Assisted Gravity Drainage, JCPT 24(3): 42–51. http://
dx.doi.org/10.2118/85-03-01
Butler, R.M., 1991. Thermal Recovery of Oil and Bitumen, Englewood Cliffs, New Jersey: Prentice Hall.
Butler, R.M., 1994. Horizontal Wells for the Recovery of Oil, Gas and Bitumen; Petroleum Society Monograph No. 2,
Canadian Institute of Mining, Metallurgy & Petroleum, Calgary, AB.
Butler, R.M., Bombardieri, C.C., Slevinsky, B.A., 1978. Recovery of hydrocarbons by in situ thermal extraction, Patent
No. US4116275A. Filing date: March 14, 1977, Original Assignee: ExxonMobil Upstream Research Company.
Butler, R.M., McNab, G.S. and Lo, H.Y., 1981. Theoritical Studies on the Gravity Drainag of Heavy Oil During Steam
Heating, Canadian Journal of Chemical Engineering, Vol. 59, No. 2, pp. 455–460.
Butler, R.M., Stephens, D.J., 1981. The Gravity Drainage of Steam-Heated Heavy Oil to Parallel Horizontal Wells, Journal
of Canadian Petroleum Technology, Vol. 20, No. 2, pp. 90–96.
Chung, K.H., & Butler, R.M., 1988. Geometrical Effect of Steam Injection on the Formation of Emulsions in the Steam-
Assisted Gravity Drainage Process. JCPT 27(1): 36–42. http://dx.doi.org/10.2118/88-01-02
SPE-199933-MS 13

Dixon, D., & Nguyen, A., 2019. An Empirical Oil, Steam, and Produced-Water Forecasting Model for Steam-
Assisted Gravity Drainage with Linear Steam-Chamber Geometry. SPE Res Eval & Eng, 22(4): 1615–1629. http://
dx.doi.org/10.2118/195675-PA
Edmunds, N.R., Kovalsky, J.A., Gittins, S.D., & Pennacchioli, E.D., 1994. Review of Phase A Steam-Assisted Gravity-
Drainage Test. SPE Res. Eng., 9(02), pp. 119–124. doi:10.2118/21529-PA
Irani, M. and Ghannadi, S. 2013. Understanding the Heat-Transfer Mechanism in the Steam-Assisted Gravity-Drainage
(SAGD) Process and Comparing the Conduction and Convection Flux in Bitumen Reservoirs. SPE J. 18(1): 134–145.
http://dx.doi.org/10.2118/163079-PA
Irani, M., & Ghannadi, S., 2020. Modeling the Conformance Improvement Using Flow Control Devices in Infill Wells
Adjacent to SAGD Well Pairs: No Flashing. SPEJ, Preprint. http://dx.doi.org/10.2118/199359-PA
Irani, M., 2019. Oil-Rate Prediction Model for the Ramp-up Phase of a Steam-Assisted-Gravity-Drainage Process:
Stability Approach. SPEJ, 24(03): 1016-1036. http://dx.doi.org/10.2118/195674-PA
Irani, M., and Gates, I., 2013a. On the Stability of the Edge of a Steam-Assisted-Gravity-Drainage Steam Chamber, SPE
J. 19(2): 280–288. http://dx.doi.org/10.2118/167260-PA
Irani, M., and Gates, I., 2013b. Understanding the Convection Heat Transfer Mechanism in Steam-Assisted Gravity
Drainage (SAGD) Process, SPE J. 18(6): 1202–1215. http://dx.doi.org/10.2118/167258-PA
Ito, Y., Ichikawa, M., & Hirata, T., 2001. The Effect of Gas Injection on Oil Recovery During SAGD Projects. JCPT,
40(01): 38–43. doi:10.2118/01-01-03
Keshavarz, M., Harding, T. G., & Chen, Z. J., 2016. Modification of Butler's Unsteady-State SAGD Theory to Include the
Vertical Growth of Steam Chamber. Presented at the SPE Canada Heavy Oil Technical Conference, Calgary, Alberta,
Canada. June 7-9, http://dx.doi.org/10.2118/180733-PA
Keshavarz, M., Harding, T.G., & Chen, Z., 2019. A New Approach to the Analytical Treatment of Steam-Assisted Gravity
Drainage: A Prescribed Interface Model. SPEJ, 24(02): 492–510. http://dx.doi.org/10.2118/194203-PA
Nukhaev, M., Pimenov, V., Shandrygin, A., et al, 2006. A New Analytical Model for the SAGD Production Phase.
Presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 24-27 Septemper, SPE
102084-MS.
Onaisi, A., 2007. Geo-Mechanical Insights into the May 18th 2006 Joslyn Steam Release, Summary of Investigation into
the Joslyn May 18th 2006 Steam Release
Peery, L.J. and Todd, F.C., 1965. Linear Heat Conduction with Temperature Dependent. Physical Properties, Proceeding
of the Oklahoma Academy of Science, Vol. 45, pp. 150–158.
Poling, B.E., Prausnitz, J.M. and O'Connell, J.P., 2001. The Properties of Gases and Liquids, 5th Edition. McGraw-Hill
Companies.
Powell, R.W., Ho, C.Y. and Liley, P.E., 1966. Thermal Conductvity of Selected Materials, National Standard Refernce Data
Series−National Bureau of Standards−8 (Category 5−Thermodynamic and Transport Properties), November 25.
Rajeshwar, K., Jones, D.B.; DuBow, J. B., 1982. Thermophysical Characterization of Oil Sands. 1. Specific Heats, Fuel,
Vol. 61, No. 3, pp. 237–239.
Reis, J.C. 1992. A Steam-Assisted Gravity Drainage Model for Tar Sands: Linear Geometry. JCPT. 31(10): 14-20.
Shalev-Shwartz, S., & Ben-David, S., 2014. Decision Trees. In Understanding Machine Learning: From Theory to
Algorithms (pp. 212–218). Cambridge: Cambridge University Press. doi:10.1017/CBO9781107298019.019
Yee, C.T., & Stroich, A., 2004. Flue Gas Injection into a Mature SAGD Steam Chamber at the Dover Project (Formerly
UTF). JCPT, 43(01): 54–61. doi:10.2118/04-01-06
Zargar, Z., & Farouq Ali, S.M., 2018. Analytical Treatment of Steam-Assisted Gravity Drainage: Old and New. SPEJ,
23(01): 117–127. http://dx.doi.org/10.2118/180748-PA
Zargar, Z., & Farouq Ali, S.M., 2019. Effect of Confinement and Well Interference on SAGD Performance: An Analytical
Assessment. SPEJ, 24(04): 1595–1612. http://dx.doi.org/10.2118/189715-PA
Zhao, L., Law, D. H.-S., Nasr, T. N., Coates, R., Golbeck, H., Beaulieu, G., & Heck, G., 2003. SAGD Wind-Down: Lab
Test and Simulation. JCPT, 44(01): 49–53. doi:10.2118/05-01-04
14 SPE-199933-MS

Appendix A
Viscosity variation using Mehrotra and Svrcek (1986, 1987) correlation
This appendix focuses solely on the matching the viscosity data into Mehrotra and Svrcek (1986, 1987). the
Mehrotra and Svrcek (1986, 1987) correlation is given as following:
A-1
Eq. A-1 is similar to the correlation suggested by ASTM D341 for high viscosity oils (ASTM, 1982). To
calculate parameters for Mehrotra and Svrcek (1986) correlation, if the dead bitumen viscosity for Surmont
bitumen across a wide range of pressure and temperature is plotted in environment suggested by Mehrotra
and Svrcek (1986, 1987) that is vs. ln(T + 273.15); the slope and the intercept will give the
and , respectively (see Fig. A-1). As it can be seen in Fig. A-1, the pressure parameter ( ) is ignored
in Fig. A-1. Neglecting pressure parameter creates a consistent method to calculate model's constants, and
anyone applied this method would calculate 3.824 and 24.3983, for and , respectively.

Figure A-1—The viscosity-temperature measurements of ConocoPhilips/Surmont gas-free bitumen in


environment suggested by Mehrotra and Svrcek (1986) and their matched viscosity-temperature relationship.

Including pressure parameter ( ) slightly improves the prediction but worsen the consistency.
A comparison of experimental and predicted viscosities for best-fit values is given in Figure A-2b.
For example, for ConocoPhilips/Surmont the average absolute relative deviation (AARD) considering
pressure effects is 8.93% compared to 11.35% neglecting pressure effects. Fig. A-2a provides a graphical
representation of the pressure effect into bitumen viscosity and prediction improvement including .
As noted, the pressure dependency is small and even smaller at higher temperatures. Knowing error in
viscosity measurements is exacerbating with temperature reduction, the importance of including the pressure
parameter is questionable.
SPE-199933-MS 15

Figure A-2—The viscosity-temperature measurements and prediction of Suncor/MacKay


River and ConocoPhilips/Surmont gas-free bitumen at two pressure measurements (Section
A and B); and comparison of calculated and measured for both bitumens (Section C and D).
16 SPE-199933-MS

Appendix B
Application of different Methods to Calculate the m─exponent
This appendix focuses on suitable methods to calculate the m-exponent (mμ). As discussed, the m-exponent
based on Eq. 9 varies with temperature, and the viscosity-temperature data should incorporated into the
viscosity model, such as Mehrotra and Svrcek (1986, 1987) correlation, and then apply the integration. So,
we cannot directly calculate a constant value for m-exponent from viscosity-temperature experimental data.
One practical approach is "slope method" that is proposed by Keshavarz et al. (2016) to calculate the m-
exponent from viscosity-temperature experimental data. He tried to honour the Butler and Stephens (1981)
viscosity-temperature model as:

B-1

And to honour the temperature range of the experimental data:

B-2

In this approach to plot (μmax/μo) vs. , and appropriate value of m-exponent should result in a line with
a slope of 1 that passes through the origin. In Fig. B-1, the data set used in m-exponent calculation in Figure
6 (experimental data from a core in UWID: 114/12-09-093-12W4 from Suncor/Firebag). As shown in Fig.
B-1a the slope for m-exponent of 5 is 1.26. Therefore, we should change the m-exponent till the slope of 1
is achieved. To simplify the slope of the trend in plot (μmax/μo) vs. for each m-exponent is calculated and
plotted vs. m-exponent. Then the m-exponent at 1 is readed from the plot (see Fig. B-1b).
We can apply log on both sides of the Eq. B-2:

B-3

If one plot viscosity vs. temperature (μo vs. T) in log-log scale based on Eq. B-3, the appropriate value
of m-exponent can be calculated from slope of the linear end of the curve. This method provides consistent
m-exponent value, and different users calculate similar value. But the m-exponent fitted in log-log model
is not similar to what calculated from Eq. 9. Other concern with m-exponent calculation using Eq. 9 is that
if the experimental data is fitted to different viscosity models then the variation of the m-exponent with
temperature is changing. In Fig. B-3a the comparison of the m-expoenet variation for the same data set but
fitted to Mehrotra and Svrcek (1986, 1987) and log-log model are compared, as shown the concave behavior
in Mehrotra and Svrcek (1986, 1987) is not presented in the case that is fitted to log-log model.
If the error of the calculation of m-exponent from fitted log-log model and the calculation using Eq. 9
of the same log-log model were compared (see Fig. B-3b); then it can be seen for the practical range of
operational temperature, the error is about 0.35. Therefore, if the log-log model is used, simply one can
calculate the m-exponent as:
B-4
SPE-199933-MS 17

Figure B-1—Evaluation of m-exponent (mμ) by plotting (μmax/μo) vs. (Section a); and variation
of line-slope vs. m- exponent (Section b). The m-exponent is calculated at the line-slope of 1.
18 SPE-199933-MS

Figure B-2—Viscosity-temperature data presented in the log-log model for different depth measurements.
SPE-199933-MS 19

Figure B-3—Comparison of the m-exponent dependency with temperature for different viscosity models 9 (Section
A); Variation of erro of the m-exponent fitting the data to log-log model and the integration in Eq. 9 (Section B).

You might also like