Vous êtes sur la page 1sur 15

1

WHOC12-240

A Review of Methods for Calculating Heat Transfer from a Wellbore to
the Surrounding Ground
P. SKOCZYLAS
C-FER Technologies
This paper has been selected for presentation and/or publication in the proceedings for the 2012 World Heavy Oil Congress
[WHOC12]. The authors of this material have been cleared by all interested companies/employers/clients to authorize dmg events
(Canada) inc., the congress producer, to make this material available to the attendees of WHOC12 and other relevant industry
personnel.

Abstract
Accurately estimating heat transfer from a wellbore to
the formation is important in heavy oil applications. In
primary or cold production, the viscosity of heavy oil
changes substantially as the oil cools, and an error in
estimating the cooling in the wellbore can have a significant
impact on pumping systems. In thermal operations, energy
efficiency is an important design consideration for which
heat transfer rates must be estimated. For 50 years,
engineers calculating the rate of heat transfer from oil wells
to the ground have referred to the classical paper by
Ramey
(1)
. Rameys formulation is simple and effective but
only at times longer than one week. Other authors have
presented improved formulations to work at shorter times.
For shorter times, properly considering the effects of casing
and cement layers becomes significantly more important. At
times less than one day, this may be critical, such as when
examining the stresses in the cement caused by thermal
gradients at the onset of steam injection. Many proposed
methods do not consider the effects of temperature on the
ground properties. This is an issue, particularly in thermal
wells, as the thermal conductivity of rock is reduced at high
temperatures. It is also an issue in wells which pass through
permafrost, as some of the heat is absorbed by the ground as
latent heat. This paper compares results using a
comprehensive model to those based on simple formulations,
and discusses under what conditions the simple formulations
are adequate or not.
Introduction
There are many applications in which the rate of heat
transfer to or from a well needs to be calculated. These
include:
Predicting the temperature of the fluid so that the
viscosity, and hence the flowing pressure losses, may
be estimated in a heavy oil well.
Estimating the amount of heat lost in a steam
injection well and/or the associated production
wells, in order to estimate quantities such as the total
thermal efficiency of a process.
Determining whether flow assurance issues may exist
in a proposed application due to formation/deposition
of paraffin, asphaltenes, hydrates, or precipitates at
lower temperatures.
Estimating the degree to which permafrost
surrounding a well in an Arctic region may thaw as a
result of injection or production through the well, or
due to the use of heat tracing on that well to assess the
effects of thaw on subsidence and wellbore
deformation
(2)
.
For these applications, or other similar ones, a quasi-
steady state result may be adequate. For other applications,
however, a transient solution may be required. An example
could be in an examination of thermal stresses in cement
surrounding casing at the onset of steam injection
(3)
. If these
stresses are too large, a wellbore integrity problem may
result. In order to calculate the thermal stress, the thermal
gradients must be known. The thermal gradients will change
rapidly over the first few minutes or hours of steam injection,
and will in fact be greatest early in the process, so no quasi-
steady state solution will be adequate.
This paper reviews various methods of estimating heat
transfer that have been presented in the literature, and
discusses when these are appropriate for use, and when they
may not be. It also presents a numerical method which can
be used when simpler methods will not give accurate results.
Problem Formulation
For the scenarios presented here and to illustrate the
methods described below, we will focus on the heat transfer
occurring at a single depth in a well. To calculate the overall
heat transfer between the wellbore and the formations that it
penetrates, one must repeat this at every relevant depth in the
well, taking into account the change in wellbore and ground
temperature with depth. This has been covered
before
(1,4,5,6,7,8,9)
, and as such will not be repeated here. The
2
assumption is also made here that heat is conducted evenly in
all radial directions from the wellbore, allowing us to use an
axisymmetric simplification. For this type of problem, axial
conduction of heat is much smaller than radial conduction, so
we will ignore the axial component, yielding a much simpler
problem of calculating a number of one dimensional heat
transfer problems instead of a single two dimensional
problem.
That is, for the purposes of this discussion, we will
calculate heat transfer from a particular depth in a wellbore to
the surrounding formation as an axisymmetric, one-
dimensional, heat conduction problem in cylindrical
coordinates. The differential equation governing this is
(10)
:

2
1

2
+
1

=
1
u
1
t
..................................................................... Equation 1
This can be solved for wellbores using a finite inner
boundary and an infinite outer boundary. It is also
convenient to make the further simplification that the ground
properties are constant in space and time. (There may be
cases where this assumption is not valid; this will be
discussed later.) Some formulations, including in the classic
paper by Ramey
(1)
, make a further simplification that the
inner boundary can be considered infinitesimalthe well is
treated as a line source.
Before the problem can be solved, however, boundary
conditions must be applied. At the outer (far field) boundary,
a constant temperature boundary is generally applied. This
constant temperature boundary condition may not be valid
over the long term when wells are close together, but in most
cases it will be valid for a period of several years. At the
inner boundary, several different boundary conditions may be
considered. The three most commonly addressed boundary
conditions are constant temperature, constant heat flux, and
convection.
While the nomenclature conventions used by the various
sources referred to here were often quite different, for this
paper, all nomenclature has been converted to a consistent
system, described in the Nomenclature section. In addition,
some sources used different definitions for certain
mathematical functions, such as the exponential integral.
Where needed, conversions have been made to one consistent
set of definitions in this paper.
Line Source Solutions
The line source problem has generally been addressed as a
constant flux problem. The solution for this problem, as
presented by Carslaw and Jaeger
(10)
, is:
2n
kA1
q
=
1
2
Ei [
1
4Po
..................................................... Equation 2
This is presented in a non-dimensional form, which will
facilitate certain comparisons later. The 2 is a useful
addition, in part because it makes the term on the left hand
side consistent with a non-dimensionalisation presented by
Hasan & Kabir
(4)
. Carslaw and Jaeger
(10)
noted that the
exponential integral can be simplified for very small or very
large input arguments. However, there is no valid reason in
this case to look at very short times (large input arguments to
the exponential integral). A line source does not suitably
represent a wellbore, which has finite size, at small times. It
is only as the heat front progresses further from the well that
a line source approximation becomes a good representation.
Ramey
(1)
made note of this in his paper, saying the line
source will often provide a useful result if times are greater
than one week.
Carslaw and Jaeger
(10)
presented an approximation of the
exponential integral which works for small input arguments
(large times):
Ei(x) y +ln x x +
x
2
4
............................................... Equation 3
If we take just the first two terms of this approximate
solution and substitute them into Equation 2, we get:
2n
kAT
q
= [
y
2
+ln
1
2Po
..................................................... Equation 4
Rameys
(1)
formulation is:
2n
kA1
q
= [ln
1
2Po
+u.29 .............................................. Equation 5
Since /20.2886, Rameys solution is essentially identical
to Carslaw and Jaegers, which should not be surprising, as
Ramey directly referenced their result. This formulation has
numerical problems at very small times, due to the small
argument (large time) approximation of the exponential
integral. Furthermore, it gives negative results at Fourier
numbers less than approximately 0.45. These clearly cannot
be used, but even for Fourier numbers greater than 0.45 the
accuracy is very poor until the time gets sufficiently large
the error is less than 2% (relative to methods which do
consider the finite wellbore diameter) at Fourier numbers
greater than approximately 30. This translates to roughly a
week for typical wellbore calculations, which is consistent
with Rameys assertion.
Use of a more complete evaluation of the exponential
integral would remove the numerical difficulties at Fourier
numbers below 0.45. These methods, however, do not
significantly improve the accuracy at Fourier numbers much
above 0.45. While Rameys method achieves errors less than
2% at Fourier numbers above 30, a more complete evaluation
of the exponential integral only improves this to 2% accuracy
at Fourier numbers above 27, which is not much
improvement over the approximate solutions.
It is apparent, therefore, that to improve upon the results
of these line source approximationsto estimate heat transfer
at earlier times the finite size of the wellbore must be
considered.
Laplace Transform Methods and
Integral Results
For an infinite outer boundary and finite inner boundary,
the heat conduction differential equation presented earlier is
most easily solved using Laplace transforms. For the
constant flux inner boundary, the result in the transformed
space is:
3
I{I(t, r)] =
q
2n
wb
K
0
__
s
o
]
ks_
s
o
K
1
_
wb
_
s
o
]
........................................ Equation 6
And for the constant temperature inner boundary, it is:
I{q(r, t)] = 2nr
wb
k I
wb
1
s
_
s
u
K
1
__
s
o
]
K
0
_
wb
_
s
o
]
...................... Equation 7
Note that for the constant flux condition, we have solved
for temperature; and for the constant temperature case, we
have solved for flux. Both of these cases were solved as a
function of radius and time. We are generally interested in
the flux or temperature at the outside of the wellbore, so we
can solve for the respective results at that radius. Note that
the solutions presented above are those of Carslaw and
Jaeger
(10)
, but their solutions were presented only as
temperature. The flux result (in the constant temperature
case) was obtained through an extra step done as part of the
present work.
For the case with convection at the inner boundary, Jaeger
and Chamalaun
(11)
presented:
I|I
wb
I
]
| =
1

-1
]
s
_1 +
h
k
K
0
_
wb
_
s
k
]
-_
s
k
K
1
_
wb
_
s
k
]-
h
k
K
0
_
wb
_
s
k
]
_ ....................
.................................................................................................... Equation 8
To calculate flux from this, the following relationship is
used after the inverse Laplace transform is calculated:
q = 2nr
wb
(I
wb
I
]
) ........................................................ Equation 9
The inverse Laplace transforms of these results are not
easily obtainable. Carslaw and Jaeger
(10)
discuss the
mathematics of doing this, but their results were given in the
form of integrals which cannot be evaluated analytically.
Numerical methods can be used to obtain the inverse Laplace
transform in some cases, and these are fast and accurate for
the range over which they can be used. The method that was
used in this work
(12)
did not yield a solution at very small
(<0.001) or very large (>1,000,000) Fourier numbers but
worked well between these values.
The inversions using integrals give the following results
(10)

for the constant temperature case:
q
2nk1
=
4
n
2
]
c
-ou
2
t
u(]
0
2
(
wb
u)+
0
2
(
wb
u)
Ju

0
............................. Equation 10
and for the constant flux case:
2n
k1
q
=
4
n
2

wb
2
]
[1-c
-ou
2
t

u
3
[]
1
2
(u
wb
)+
1
2
(u
wb
)
Ju

0
.................. Equation 11
The formulations in Equations 10 and 11 contain
dimensional values in the integrals, specifically time and the
wellbore radius. This makes comparison to non-dimensional
results difficult, and means that tables of results need to have
an extra dimension. Jaeger
(13)
also provides non-dimensional
versions of these results, which are, for the constant
temperature case:
q
2nkA1
=
4
n
2
]
c
-u
2
Fc
u[]
0
2
(u)+
0
2
(u)
Ju

0
....................................... Equation 12
and for the constant flux case:
2kA1
q
=
4
n
2
]
1-c
-u
2
Fc
u
3
(]
1
2
(u)+
1
2
(u))
Ju

0
...................................... Equation 13
It can be shown that these are equivalent to the
dimensional versions. Jaeger did not state how the
nondimensionalization was done, but simply showed that the
two results were equal. Hasan and Kabir
(4)
, however,
provided some insight into how it may have been achieved
they actually provide a result for the constant flux case,
although to show that their equation is in fact equivalent to
the one above, one needs to use the following relationship:
[
1
(u)
0
(u) [
0
(u)
1
(u) =
2
nu
......................................... Equation 14
The convection solution was presented by Jaeger and
Chamalaun
(11)
in a dimensionless form, as:
q
2nk(1

-1
]
)
=
- 4
n
2
]
c
-u
2
Fc
u_j
u

]
1
(u)+]
0
(u)[
2
+j
u

1
(u)+
0
(u)[
2
_
Ju

0
..................
................................................................................................. Equation 15
where:
[ =

wb
h
k
................................................................................. Equation 16
Note that the temperature differential in Equation 15 is
different from the one in the other versions (in that it refers to
the fluid temperature instead of the wellbore interface
temperature). Jaeger and Chamalaun
(11)
also presented a
revised version of this:
q
2nk(1

-1
]
)
=
-4[
2
n
] c
-u
2
Po
[
Po
u
2
+[
2
+
1
(u
2
+[
2
)
2
(u, [)u Ju

0
........ Equation 17
where:
(u, [) =
n
2
+aig _
u
[
E
1
(1)
(u) + E
0
(1)
(u)_ ................... Equation 18
Note that in order to use this formulation, the arg(z)
function, evaluated over the whole range of the integral,
needs to be continuousthis means that its output is not
always in the to range as might be expected. Rather, it
monotonically increases as the z vector rotates around the
origin of the Argand diagram as u goes from 0 to infinity.
Jaeger and Chamalaun
(11)
also presented simplified results
for very small or very large values of time which do not
4
require integration. These are not presented here, but should
be consulted if needed.
The equations for all three boundary conditions contain
integrals which cannot be evaluated analytically. Rather,
numerical integration is required to evaluate them.
Unfortunately the integrands tend towards infinity at very
small times. A way to deal with this problem is the method
of asymptotic expansions
(10)
. The Bessel functions and
exponentials in the integrals can be replaced with simple
expansions which are valid for either very small times or very
large times. For example, the exponential function in the
integrals (where the argument is negative) tends toward 1 as
the variable of integration (u) gets very small, and towards 0
as it gets very large. Similar simplifications (although not
generally as simple) also exist for the Bessel functions
presented in those integrals. When these are applied, the
integrals can be solved analytically for regions near u=0 and
for very large values of u. Numerical integration is then only
necessary for intermediate values of the variable of
integration. Note that the revised version of the integral
presented by Jaeger and Chamalaun, as shown above in
Equations 17 and 18, is more conducive to being evaluated
numerically, as its integrand does not tend towards infinity as
u approaches 0.
Several authors, including Van Everdingen and Hurst
(14)
,
Willhite
(5)
, Jaeger and Clarke
(15)
, Hasan and Kabir
(4)
, and
Jaeger and Chamalaun
(11)
have given tables of results for one
or more of these integrals, which can be consulted.
While tables can certainly be entered into a computer
program, correlations are much simpler to use. Several
authors have published correlations to these tabulated values.
These include, for constant temperature boundary conditions,
Chiu and Thakur
(6)
and Moini and Edmunds
(7)
, and for
constant flux boundary conditions, Hasan and Kabir
(4,8)
.
These correlations are presented in Appendix A. No
correlations were found in the literature for the convection
boundary condition as part of this work.
Carslaw and Jaeger
(10)
also presented formulations for
both the constant flux and constant temperature cases which
provide solutions at very small or very large values of time.
These are as simple to use as the correlations (i.e. no
integration or numerical inversions are needed), and so may
be quite convenient to use when short or long times are of
interest.
Results Comparisons
In this section, the results for several of the solutions
described above for the cases with constant temperature and
constant flux at the inner boundary are presented. For
comparison, the values are presented as dimensionless
temperatures, as used by Hasan and Kabir
(4)
, and which is
essentially the same as Rameys
(1)
dimensionless time
function:
I

= 2nk
A1
q
........................................................................................ Equation 18
The relative error presented in the plots shown below was
calculated using the following:
e =
|1
D
-1
D0
|
1
D0
............................................................................ Equation 19
where T
D0
is a reference value, which for the plots here was
the value obtained by numerical integration as part of this
work. All of the other values obtained from tabulated data,
formulations, and correlations were compared against this. In
the figures, markers represent tabulated data, while lines
represent continuous sources of values obtained from
correlations, short/long time approximations, numerical
inversions, and numerical integrations.
Figure 1 and Figure 2 show the dimensionless temperature
and relative error for the constant temperature inner boundary
case.

Figure 1 Constant temperature boundary models

Figure 2 Error in constant temperature boundary models
Examination of these figures shows that all the constant
temperature formulations yield similar and suitably accurate
results, with the exceptions of the Carslaw and Jaeger
approximations for small time results at longer times and
long time results at smaller times. These approximations
yield better than 1% accuracy if the Fourier number is less
than 0.32 for the small time version and greater than
approximately 100,000 for the long time version.
Figure 3 and Figure 4 show the dimensionless temperature
and relative error for the constant flux case. In these figures,
the data points labeled Hasan and Kabir Rigorous are
from the table in their 1991 paper
(4)
; they use the phrase
rigorous solution to describe this tabulated data and to
differentiate it from values obtained using their
approximate correlation.
0.001
0.01
0.1
1
10
1.E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06 1.E+08
D
i
m
e
n
s
i
o
n
l
e
s
s

T
e
m
p
e
r
a
t
u
r
e
Fourier Number
Numerical Integration
Inverse Laplace
Carslaw & Jaeger
(Small Time)
Carslaw & Jaeger
(Long Time)
Jaeger & Clarke
Willhite
Chiu & Thakur
Moini & Edmunds
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06 1.E+08
R
e
l
a
t
i
v
e


E
r
r
o
r

i
n
D
i
m
e
n
s
i
o
n
l
e
s
s

T
e
m
p
e
r
a
t
u
r
e
Fourier Number
Inverse Laplace
Carslaw & Jaeger
(Small Time)
Carslaw & Jaeger
(Long Time)
Jaeger & Clarke
Willhite
Chiu & Thakur
Moini & Edmunds
5

Figure 3 Constant flux boundary models

Figure 4 Error in constant flux boundary models
As with the constant temperature case the results
presented in Figures 3 and 4 show that all of the constant flux
formulations yielded suitably accurate results, with the
exception of the Carslaw and Jaeger approximations for small
time results at longer times and long time results at smaller
times. These approximations yield better than 1% accuracy if
the Fourier number is less than 0.05 for the small time
version and greater than approximately 58 for the long time
version.
The correlations which gave the best results, showing the
least disagreement with numerical results over the largest
range of Fourier numbers, were the revised Hasan and
Kabir
(8)
formulation for constant flux cases and the Chiu and
Thakur
(6)
formulation for the constant temperature cases.
Getting into sufficiently small or large times, the Carslaw and
Jaeger
(10)
approximations perform better than the correlations,
but these have the disadvantage of only being accurate in a
narrow range.
Notice that the plots in Figure 1 and Figure 3 look very
much alike. In fact, as Ramey pointed out, these plots
converge toward identical results at long times. However, at
short times, there remain differences. As time approaches
zero, the two results approach a constant ratio, with the
dimensionless temperature for the constant temperature case
being /2 times the value for the constant flux case at the
same Fourier number. At large times, there is less than 1%
difference between the values in Figures 1 and 3 by the time
a Fourier number of one million has been reached. After a
Fourier number of 30, Rameys one week, there is less than
10% difference between themgenerally good enough for
many engineering applications, especially considering the
other sources of error and the probability that the true
wellbore boundary condition is not quite either a constant
temperature or a constant heat flux boundary condition.
Shortcomings of These Methods
While the results discussed above may be very valuable
and useful for many engineering applications in the oilfield,
they do have their shortcomings. These include:
1. The formulations do not consider how the thermal
mass of the wellbore itself affects the results. Several
authors, including Ramey
(1)
, Willhite
(5)
and
Hagoort
(9)
, provided methods to consider the
resistance to heat transfer of the wellbore itself, but
these methods ignored its thermal mass. This is not a
significant issue in long term cases, but if one needs
to observe changes in the short term (less than a few
days), neglecting this thermal mass may be a problem.
2. There is no consideration of boundary conditions
which change over time. Even if one is injecting
constant temperature fluids (e.g. steam), or producing
constant temperature reservoir fluids, over time, the
boundary conditions downstream of the place where
flow enters the wellbore (i.e. the wellhead or the
formation, depending on whether an injection well or
producing well is being considered) will in fact
change over time. This is because the ground
temperature between the point where flow enters the
well and the point under consideration will change
over time. This in turn changes the amount of heat
lost (or gained) by the wellbore fluids over time.
Other instances of changing boundary conditions
include periods during which the well is shut in or
restarted.
3. There is no consideration of changing ground
properties. This is a concern in thermal wells in
particular, as the thermal conductivity of ground
changes substantially between 0C and 300C. This
is illustrated for some soil types in Figure 5, as given
by Clauser and Huenges
(16)
(note their use of for
thermal conductivity). It is also a concern in wells in
frozen ground, as the thermal properties of frozen soil
are different from those of unfrozen soils.

Figure 5 Effect of temperature on ground thermal
conductivity
0.001
0.01
0.1
1
10
1.E-05 1.E-02 1.E+01 1.E+04 1.E+07 1.E+10
D
i
m
e
n
s
i
o
n
l
e
s
s

T
e
m
p
e
r
a
t
u
r
e
Fourier Number
Numerical Integration
Inverse Laplace
Carslaw & Jaeger (Small Time)
Carslaw & Jaeger (Long Time)
Hasan & Kabir "Rigorous"
Van Everdingen & Hurst
Hasan & Kabir
Hasan & Kabir (#2)
1.E-10
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E-04 1.E-01 1.E+02 1.E+05 1.E+08
R
e
l
a
t
i
v
e

E
r
r
o
r

i
n
D
i
m
e
n
s
i
o
n
l
e
s
s

T
e
m
p
e
r
a
t
u
r
e
Fourier Number
Inverse Laplace
Carslaw & Jaeger
(Small Time)
Carslaw & Jaeger
(Long Time)
Hasan & Kabir
"Rigorous"
Van Everdingen & Hurst
Hasan & Kabir
Hasan & Kabir (#2)
6
4. There is no consideration of latent heat effects. This
is most often an issue in permafrost applications,
where the ground around the wellbore is initially
frozen, but thaws over time, consuming significant
thermal energy in the process. It may also be an issue
in many thermal applications, as any ground water
may boil if the temperature reaches or exceeds the
saturation temperature at the local pore pressure
value. (This is more likely to happen closer to surface,
where the pore pressures tend to be lower.)
5. The results do not, in general, tell us anything about
the radial temperature gradients in the ground. This is
relevant for certain problems. For example, one
might be interested in knowing what the thermal
gradients (and therefore the thermal stresses) in the
wellbore cement sheath are following the onset of
steam injection. One might also want to know where
the thaw boundary is around a well in a permafrost
region.
Analytical solutions do not generally exist for these
situations. Other methods, such as numerical solution
methods, must be used to obtain results where they may
occur.
Numerical Methods
Finite element analysis (FEA) can be used to obtain
solutions in such cases. Most commercial FEA packages can
do heat transfer analyses. Finite difference (FD) methods
may also be used. For simple geometries, FD methods are
often easier to develop and to couple with the temperature
calculations in a wellbore simulator.
A detailed, one-dimensional, axisymmetric, finite
difference model was developed for this study. It allowed for
variable spacing of the nodes, varying properties of the
volumes around the nodes, and regions of different
properties. The variable nodal spacing is useful in that
greater accuracy can be obtained where the temperature
gradients are steepest near the wellbore while allowing more
widely spaced nodes further removed from the wellbore
where the temperature gradients are small. If one was forced
to have constant spacing, one would either lose accuracy (due
to not having enough nodes where they are most needed), or
one would have to have a very large number of nodes (which
would greatly increase computation time). The FD model
used in this study can handle all three boundary conditions
referred to above.
The full details of the method are not given here for the
sake of brevity, but an example of the derivation method is
shown in Appendix B. Additional examples were presented
by Skoczylas
(17)
.
Consideration of Latent Heat
Latent heat effects were considered using a method called
apparent heat capacity as referred to, for example, by
Pham
(18)
, Osterkamp
(19)
, and Mottaghy and Rath
(20)
. The
original implementation of this method was used in cases
when the modeled substance would freeze at a single
temperature. As Pham
(18)
says, the latent heat is represented
by a peak of small but finite width in the c(T) curve, where
c(T) represents the specific heat of the material as a function
of temperature. The problem with this method, however, is it
is possible for the calculation to ignore this peak, by
jumping of the latent heat peak
(18)
. This can be prevented
only by making the peak wider or by making the time steps
very short, so that the amount of heat entering a node which
is just below the peak is not enough to take the temperature
above the peak in a single time step.
Ground, particularly when made up of fine grained soils
such as silts and clays, does not freeze at a single
temperature, but rather it freezes over a temperature range.
An example of this phenomenon is shown in Figure 6 below,
as given by Williams
(21)
. Therefore, one does not need to
apply a very narrow, yet very tall, c(T) peak to implement an
apparent heat capacity method. Rather, the peak is spread
out over the full range of temperature over which freezing
occurs. Pham
(18)
actually suggests that smoothing of the peak
is a way to prevent jumping of the peak, but for the problem
of freezing at a single temperature, he noted that this reduces
accuracy. In our case, this should not be an issue, because
the c(T) relationship is naturally smoothed.

Figure 6 Ground percent unfrozen as function of
temperature
To use this concept in a finite difference model, no
changes need to be made to the model itself, but only to the
value of the specific heat at every node at every point in time
based on its temperature at that time.
Note that there is also an analytical solution for a latent
heat problem as presented by ziik and Uzzell
(22)
. This
solution was for a line sink, however, so it is only useful for
longer times. This approach may be useful, for example for
determining the heat loss from a wellbore over its lifetime. It
will not be considered further here, except to note that it was
used to validate the numerical methods developed during this
study.
Comparison of Numerical Methods with
Simple Models
In some cases numerical solution methods (such as FEA
or FD models) may be necessary to get accurate results.
These may include cases when we need to know the thermal
gradients inside one or more layers of cement in a wellbore,
or when the thermal mass of the casing and cement may
affect the results. In other cases, the correlations noted above
may be more than adequate for obtaining useful results. Such
a case may be when estimating the thermal efficiency of a
steam injection well over its life. The question is: in which
cases can we use the simple models, and what, if any,
modifications do we need to make to the inputs to use them?
7
Case 1a: Consideration of the thermal mass of casing and
cement (constant temperature boundary).
For example, let us consider 9.625 casing, concentrically
cemented in a 13 diameter hole. A constant temperature
boundary condition (45C higher than the ground
temperature) was applied at the inner wall of the casing, and
this was compared to the result from the FD model of this
scenario to that obtained using the Chiu and Thakur
(6)

correlation. In using the correlation, the Fourier number was
calculated using the diffusivity of the ground and the
diameter of the wellbore/formation interface. This was also
compared to a Fourier number calculated using the inside
diameter of the casing. The former approach ignores the
thermal mass of the casing and cement, while the latter
assumes that the casing and cement have thermal mass
equivalent to that of the same volume of ground. Results are
presented from four versions of the finite difference model,
labeled as in Figure 7:
1. full consideration of the casing, cement, and ground,
Full Model;
2. consideration of only the ground outside the wellbore
(with the boundary condition applied at the
cement/ground interface), Ground Only;
3. consideration of the casing and cement with the same
thermal properties as the ground, Ground to Casing
ID, and
4. consideration of the casing and cement with no
thermal mass but with their actual resistance to heat
transfer, Ground with Resistance.
The results are shown in Figure 7. Several insights can be
gained from examination of these results:
Ignoring the casing and cement and applying the
boundary condition at the wellbore/formation
interface, as in the Ground Only case, and the
corresponding Chiu & Thakur C&T (Wellbore
Interface) case, is not valid. While these results
eventually approach the full model results, they are
visibly different (error greater than 10%) even after
one year of operation (the longest time shown in the
figure). Note that the Ground Only and C&T
(Wellbore Interface) lines in Figure 7 are practically
identical such that only one line is distinctly visible in
the figure except at very short times (fractions of a
second).
Modeling the ground plus the thermal resistance of
the wellbore (the Ground with Resistance case)
works very well for times longer than approximately
one day, but is very inaccurate at shorter times, in
that it shows much too low a value of heat transfer.
If one is not interested in the results at very short
times, however, the results show that this can be a
very effective method.
The Chiu and Thakur method, considering the casing
and cement to be equivalent to ground (the C&T
(Inside Casing) case), is reasonably good, but with a
small error, for times longer than approximately one
day for the scenario considered here. (The actual
method described by Chiu and Thakur is actually
more like the Ground with Resistance method
plotted in the figure. However, their full method was
not used hereonly their transient calculation for the
ground was used.)
Likewise, a finite difference method that considers
the casing and cement to be equivalent to ground (the
Ground to Casing ID case) yields results that are
reasonably good but with some error as compared to
the full FD model for times longer than one day.
All of these methods diverge substantially from the
results of the full FD model for times shorter than
approximately one day.
In the figure, the three finite difference results
without resistances (the Full Model, Ground
Only and Ground to Casing ID cases) applied
have an upturned result as time approaches zero.
This does not represent a physical phenomenon, but
is a numerical artifact of the finite difference method
which disappears within approximately 10 time steps.
The initial time step in this calculation was 0.1 s, and
the upturned result has disappeared by approximately
1 second. This artifact doesnt appear in the case with
resistance because the resistance serves to damp it
out.

Figure 7 Model comparison: constant temperature boundary
Let us further compare the true result (the full FD
model) to the closest simple result, that of the Chiu and
Thakur correlation, considering the wellbore to have the same
thermal properties as ground (the C&T Inside Casing case).
The following comments relate to the comparison of these
two cases
The full model result gives an order of magnitude
more heat transfer over the first ten seconds. This is
the time when heating of the casing steel is dominant.
The steel has a high thermal conductivity, and easily
absorbs a lot of heat for a short period of time.
From 10-100 seconds, the true heat transfer drops
substantially. This can be regarded as the time when
the full thickness of the casing has essentially
reached the imposed casing ID temperature, and heat
is being transferred to the cement.
From 100 seconds to one day (depending on the
desired accuracy), the true heat transfer is somewhat
less than what is predicted by the simpler solution.
During this time, heat is transferred from the casing
to the cement layer, increasing its temperature. The
cement (as modeled here) has a lower thermal
conductivity than the ground, so the heat transfer rate
is less than it would be if it was replaced with a
material with the same thermal conductivity as the
ground. (If a lower conductivity was used, such as
that of an insulating cement, one would expect the
difference to be greater.)
100
1000
10000
100000
1000000
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (s)
Full Model
Ground Only
C&T (Inside Casing)
Ground to Casing ID
C&T (Wellbore Interface)
Ground with Resistance
8
For times longer than approximately one day (to one
week, depending on the desired accuracy), the results
suggest that the simple solution is adequate for many
engineering purposes. If the problem at hand
requires an analysis at shorter times, however, the
simple method is not adequate. An example of a
calculation which may require the more detailed
solution is a consideration of how the casing and
cement react (with regard to thermal stress) when
steam is initially injected in a wellbore
(3)
.
The results presented here were just for a very simple
system of a single casing and cement in a formation. Clearly,
the existence of other layers will complicate this. One would
expect that the greater the thermal mass and/or thermal
resistance of the combined layers between the inside of the
wellbore and the formation, the longer the time until the
results of the simple calculation approach those of the full FD
model.
Case 1b: Consideration of the thermal mass of casing and
cement (Convection Boundary).
Consider the same wellbore configuration, except that
instead of having a constant temperature boundary, a
convection boundary coefficient was applied in the wellbore.
The problem geometry, thermal properties and far-field and
fluid temperatures were the same as in the previous example,
but a convective heat transfer coefficient of 15 W/mK was
applied at the inside of the casing. This would tend to be
somewhat more realistic than the previous case, in that the
rate of heat transfer at the very early times is not forced to be
extremely high as it is in a constant temperature boundary
case.
Figure 8 shows the results for this scenario. Some
comments about these results are as follows:
The Ground Only FD model, and the
corresponding case using the Jaeger and
Chamalaun
(11)
data, the Jaeger & Chamalaun
(Wellbore Interface case), have greater error in this
case than the corresponding results in the constant
temperature case. (Note that these two results largely
overlie each other in the figure.)
The results for the Ground with Resistance FD
model are somewhat better (closer to the full model
results) in this scenario, but still show a reduced rate
of heat transfer at early times.
Considering the casing and cement to be thermally
equivalent to ground (as in the Ground to Casing
ID FD model and the Jaeger & Chamalaun (Inside
Casing) casesnote that these cases largely overlie
each other in the figure) seems to work reasonably
well at large times, although it slightly overpredicts
the rate of heat transfer.

Figure 8 Model comparison: convection boundary
Case 2: Variable ground properties
For this example, a scenario of a high temperature thermal
recovery well was considered with an injected steam
temperature of 300C. The thermal conductivities for the
samples shown in Figure 5 from 0C to 300C were roughly
linear from 3.2 W/mK at 0C to 1.5 W/mK at 300C. Well
consider the same geometry as the previous problem, as well
as the same casing and cement properties. Well also assume
that the thermal conductivities of the casing and cement
remain constant. The initial ground temperature will start at
0C and steam will be injected at 300C. No phase change of
water in the ground was considered. A constant temperature
boundary condition was used to simulate this case.
Figure 9 (a and b) shows the heat transfer calculated using
a finite difference model considering the thermal conductivity
of the formation varying with temperature. It also shows the
results using the Chiu and Thakur correlation with the
conductivity set to what it is near the wellbore (High
Temp) and what it is far from the wellbore (Low Temp).
The difference is that Figure 9a shows the results with casing
and cement being included along with the formation in the
model, while Figure 9b shows the results with the boundary
condition being applied at the formation interface.

Figure 9a Model comparison: effect of temperature on
thermal conductivity
100
200
300
400
500
600
700
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (s)
Full Model
Ground Only
Ground to Casing ID
Ground with Resistance
Jaeger & Chamalaun
(Inside Casing)
Jaeger & Chamalaun
(Wellbore Interface)
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E+07
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (s)
Full FD Model
Chiu & Thakur (Low Temp)
Chiu & Thakur (High Temp)
9

Figure 9b Model comparison: effect of temperature on
thermal conductivity
The results presented in Figure 9 show that, for the
scenario considered, using both the high and low temperature
thermal conductivity values in the Chiu and Thakur
correlation brackets the heat transfer for times longer than a
few hours. When the casing and cement are included in the
model, the heat transfer is higher at very early times, when
the casing is being heated, and lower just after that, when the
cement (with a lower thermal conductivity than the
formation) is being heated.
It does not make sense to apply the same adaptation to a
simple model in the constant heat flux or convection case, as
the temperatures in the ground near the well will change
substantially over time.
Case 3: Phase change
The figures below show an example for a well flowing
warm fluid in a permafrost interval. In this scenario, the
grounds latent heat is 140 MJ/m. The specific heat and
thermal conductivity were also considered to be functions of
temperature, in that one value was used for frozen ground and
another value was used for unfrozen ground. Partially frozen
values were assigned a weighted average based on the
temperature and the percent of frozen ground. The density
was considered to be constant. The wellbore includes a layer
of casing and cement, with constant properties and no latent
heat. The figures show results with and without phase
changethe only difference between the two cases is that the
no phase change case had the ground latent heat set to zero.
In the example, the ground temperature was set to -15C,
while the wellbore temperature was +25C.
Figure 10 shows the rate of heat transfer for both cases
and Figure 11 shows the ratio of heat transfer in the case
including the phase change to the one that does not.

Figure 10 Model comparison: effect of phase change

Figure 11 Model comparison: effect of phase change (ratio)
For the scenario considered, up to a time of approximately
2000 s, the results for the two cases are nearly identical. This
is the time in which most of the heat is simply heating the
casing and cement (which were modeled the same in the two
cases). After approximately 2000 s, however, the case with
the latent heat has a greater amount of heat transfer. The
difference peaks at just over 20% after approximately
40,000 s, after which it drops gradually to approximately 5%
at 1.2 years. The temperature profile from the casing ID to a
radius of 5 m into the permafrost, after 1 year, is shown in
Figure 12.

Figure 12 Model comparison temperature profiles
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E+07
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (s)
FD Model
Chiu & Thakur (Low Temp)
Chiu & Thakur (High Temp)
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (s)
No Phase Change
Phase Change
1
1.05
1.1
1.15
1.2
1.25
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
R
a
t
i
o

o
f

H
e
a
t

T
r
a
n
s
f
e
r
(
P
h
a
s
e

C
h
a
n
g
e
/

N
o

P
h
a
s
e

C
h
a
n
g
e
)
Time, s
-15
-10
-5
0
5
10
15
20
25
0 1 2 3 4 5
T
e
m
p
e
r
a
t
u
r
e

(

C
)
Radius (m)
No Phase Change
Phase Change
10
Because it takes more heat to increase the ground
temperature in the phase change case (due to the latent heat),
the temperature stays lower longer. Because the temperature
in the ground is lower, there is a higher gradient, which
translates to a greater rate of heat transfer.
To determine how far from the well the ground had fully
thawed (assuming that this is at 0C), Figure 12 shows that a
very different answer would be reached if the phase change is
neglected. The 0C boundary is at 1.12 m when the phase
change is considered, and at 1.28 m when it is not, an error of
14%.
Case 4: Shut-ins
Consider a well which is shut down for three days after
every 90 days of operation over a period of two years. When
in operation, it was modeled as having a constant temperature
boundary condition. The effects of casing and cement were
ignored for this scenario and constant ground properties were
assumed. Figure 13 shows the heat flux from the wellbore
for this case. Results obtained using finite difference
methods with and without consideration of the shut-ins are
compared with results obtained using the Chiu and Thakur
correlation. The shutdowns clearly cause a disruption in the
heat flux profile, but it is important to note that the results
rapidly approach the case without shutdowns during the
operation period after each shutdown.

Figure 13 Model comparison: effect of shut-ins
The results for this case, as presented in Figure 13, show
that the simple correlations may be used to get a reasonably
good engineering assessment of thermal efficiency in cases
like this, provided that having some error during the periods
after restarts does not have a significant impact on the
requirements of the scenario being considered. Note that the
Chiu and Thakur result is not visible in the plot, despite being
on the legend, because it is overlain so closely by the No
Shut Ins casethe difference between the two is 0.6% after
one hour, and declines to 0.3% at the end of the modeled
period.
Case 5: Changing temperature over time
Even when injection or production conditions are constant
with time, at depths further from the top of an injection well
(or conversely, further from the bottom of a production well),
the fluid temperature will change over time. This is because
the amount of heat lost to the formation decreases over time,
as the ground near the well heats up. In Figure 14, the heat
transfer results are shown for a case in which the temperature
in the well was increased linearly from 50C to 70C over a
period of two years, after which it remained constant at 70C
for an additional three years. The cases of constant
temperatures of 50C and 70C, as calculated using the Chiu
and Thakur correlation, are presented for comparison. No
casing or cement is modeled in this case. The properties of
the ground were assumed to be constant over the temperature
range. The initial and far ground temperatures were set to
20C.

Figure 14 Model comparison: effect of changing boundary
condition
As one would expect, the heat transfer profile for the low
temperature case matches that of the full FD model very well
over the first days or weeks. Over time, it deviates, and the
heat flux increases towards the high temperature case. As
one would also expect, when the temperature at the wellbore
after two years is held constant, the heat flux approaches the
high temperature value.
Assumptions
In this study, we examined many methods of calculating
the heat transfer from a short section of wellbore to the
surrounding formation. For each case presented, certain
assumptions were made, depending upon the method. Some
of the key assumptions, pervasive across most or all of the
methods include the following:

There is no axial heat transfer (other than by mass
transfer in the wellbore); all heat transfer is radial.
The section of wellbore being considered is short
enough that there is no significant change in the
temperature of either the wellbore fluid or the far ground
over the length of the section. Sufficiently large changes
in the temperature over a segment length can cause
significant calculation errors and artifacts. In any
wellbore model, one should check that the change in
temperature along any one segment is much smaller than
the temperature difference between the fluid and the far
ground. It is possible, however, to compensate for this;
Ramey
(1)
did so, and Skoczylas
(17)
also did so in the
context of a finite difference model.
The far (undisturbed) ground temperature is equivalent
to the initial temperature, and is constant in time. This
assumption can be overridden in a finite difference (or
1.E+02
1.E+03
1.E+04
1.E+05
0 200 400 600 800
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (d)
With Shut Ins
Chiu & Thakur
No Shut Ins
100
1000
0 500 1000 1500 2000
H
e
a
t

T
r
a
n
s
f
e
r

(
W
/
m
)
Time (d)
Full Model
Chiu & Thakur (Low Temp)
Chiu & Thakur (High Temp)
11
FEA) calculation, where any initial temperature profile
can be considered.
The system is axisymmetric. Some key situations in
which this assumption may be violated are:
o Non-vertical wells; in these, the far
temperature at some distance from the
wellbore, perpendicular to the axis of the well
varies with direction.
o Non-concentric tubularsfor example casing
that is not centralized prior to being cemented
in place.
There is no mass transfer, other than in the wellbore
itself. These models are not designed to handle cases
such as:
o In injection or production intervals; other
methods need to be considered, if necessary,
in intervals open to the reservoir
o Movement of formation fluids adjacent to the
wellbore. This can cause significant increases
in heat transfer, as illustrated by Liu et al
(23)
.
The density of the ground is constant. While changes in
thermal conductivity and specific heat are permitted in
the finite difference models, changes in density imply
movement of the ground or of fluid within the ground,
and the consideration of this was deliberately avoided in
this study.
There is no time-dependent heat transfer other than
conduction. Most of the correlation-based models
ignore transient conduction anywhere other than the
ground, while the finite difference models allow for
transient conduction in casing and cement (and even in
tubing strings within the casing, under certain
conditions). But none of the models described here
consider (for example) the time dependency of
developing natural conduction in a tubing-casing
annulus.
Conclusions
Under these assumptions (and others specific to each
method), several conclusions can be made, including the
following:
Numerical solutions, such as the finite difference
approach are the most flexible and robust of the
methods that can be used to determine wellbore heat
transfer rates. They can handle complicated
problems in an accurate way that the other
approaches simply cannot match. The only real
drawback to numerical methods is the required
computational power and the associated time to reach
a solution. (If overlapping effects from multiple
wells are to be considered, FEA methods would tend
to be required, rather than FD models.)
For many simple cases, correlations fit to the results
of the exact solution methods can work very well.
Such correlations, however, cannot generally be used
at shorter times (under one day in typical wellbores),
regardless of the accuracy of the correlation, because
they do not consider the thermal mass of the casing
and cement (and any other tubulars/annuli in the
well).
Correlation results may not be accurate when the
boundary conditions are changing. This is especially
true early in the life of a well or during other periods
of transition such as during shutdowns and restarts.
Only when conditions have stabilized do the results
obtained from such correlations approach an accurate
result.
Ignoring (such as by the use of correlations) the
effects of phase change in scenarios when it can
occur (e.g. in permafrost) can lead to significant
errors.
Once a changing boundary condition stabilizes, the
prior history seems to have minimal importance, and
the results will, in a reasonable time frame, approach
what they would have been had the boundary
conditions been held at that value from the start.
Correlation methods can therefore be used after some
time from a startup (or a restart), or any other change
in operating conditions.
In cases, where correlations are to be used for
simplicity or computational efficiency, the following
recommendations are made:
o For constant temperature problems, the
Chiu and Thakur
(6)
method is recommended
due to its simplicity and accuracy.
o For constant flux problems, the revised
Hasan and Kabir
(8)
method is
recommended, also due to its simplicity
and accuracy.
o For problems with a convective boundary
condition in the well, the Jaeger and
Chamalaun
(11)
method can work very well.
Unfortunately, no accurate correlation is
available over a full range of times. The
choices for using the Jaeger and
Chamalaun method are currently:
Interpolate from a table of data.
This is a reasonable approach in
many circumstances.
For very small or very large
times, use the approximations
provided by Jaeger and
Chamalaun
(11)
.
Perform a difficult numerical
integration. Unless the
conditions fall outside the range
of conditions shown in the Jaeger
and Chamalaun table, there
should be no real benefit to doing
this, while there is a significant
computational cost.
o With correlation methods, the casing and
cement (and other wellbore tubular/annuli,
as appropriate) should be modeled as
ground, or an equivalent resistance should
be applied. The error from doing this is
significantly less than the error from
applying the wellbore boundary condition
at the formation interface instead. The
results will generally be valid for most
wellbore scenarios after approximately one
day of elapsed time from start-up.
o When the thermal conductivity of the
formation varies with temperature and a
constant wellbore temperature exists, the
thermal conductivity of the formation
12
should be evaluated at the wellbore
temperature.
o There is generally no reason to use line
source methods. They work fine at long
times (so long as you are interested in
constant flux results), but they are no easier
to use and provide no better results than
correlation methods.
A Final Note
While this paper has not addressed the problem of
calculating temperatures throughout the wellbore, a final note
on this topic is warranted. Many authors
(1,4,5,6,7,8,9)
have
presented approaches to predict wellbore temperature
profiles, usually by using steady state or quasi-steady state
solution approaches. But in very short time cases (such as
the cement stress problem mentioned earlier), this is not
adequate.
Consider a simple example of a small 3.5 tubing
cemented inside a 6 diameter vertical hole, 300 m deep. The
ground temperature near surface is 10C, and at 300 m it is
30C. At time 0, the well is filled with water at thermal
equilibrium with its surroundings. Starting at time 0, hot
water at 70C is injected at a rate of 4 litres per second (this
gives a velocity of just over 0.88 m/s). What does the
temperature profile in the well look like after 170 seconds,
when the fluid front has reached 150 m? (Note that this is
assuming there is no mixing between the cold fluid and the
warm fluid; i.e. the fluid front is always at a single depth
across the cross section of the tubing.) This case was
examined using separate finite difference models at several
different depths along the well. Each element of fluid (the
volume contained within the tubing for each segment in the
well) was tracked as it moved from its initial location (or
from surface for injected fluid) down the well, changing
temperature in proportion to the rate of heat transfer.
Figure 15 shows the results of this example for three different
boundary conditions: a constant temperature boundary
(where the casing ID is assumed to be equal to the fluid
temperature), a convection boundary, and a perfectly
insulated wellbore.

Figure 15 Temperature in wellbore soon after initiation of
injection
With examination of the three temperature profiles presented
in Figure 15, a key point is that the fluid in the wellbore
moves downward as fluid is injected. This means that warm
casing/cement that is deeper in the well is first exposed to
colder fluid from higher in the well before it is exposed to the
warmer injected fluid. The thermal stress experienced by the
wellbore casing and cement would therefore be increased
relative to what it would be if this effect was not considered,
such as in a quasi-steady state solution.
Nomenclature
c = specific heat, J/kgK
Fo = Fourier number
h = convection coefficient, W/mK
E
0
(1)
, E
1
(1)
= Hankel functions
i = nodal index
J
0
, J
1
= Bessel functions of the first kind
k = thermal conductivity, W/mK
K
0
, K
1
= modified Bessel functions of the second kind
L{} = Laplace transform operator
q = heat flux per unit length, W/m
r = radius, m
r
wb
= wellbore radius, m
s = complex argument used in the Laplace transformed space
t = time, s
T = temperature, C
T
D
= dimensionless temperature
T
D0
= reference dimensionless temperature
T
f
= fluid temperature, C
T
wb
= temperature at the wellbore/formation interface, C
T

= far ground temperature, C


T = difference in temperature between the
wellbore/formation interface and the far ground, C
u = variable of integration
V = nodal volume per unit length of wellbore, m/m
X,Y = variables used to collect terms
Y
0
, Y
1
= Bessel functions of the second kind
z = complex variable
= thermal diffusivity, m/s
= dimensionless convection coefficient
r = nodal spacing, m
t = time step, s
= 0.5772156649, Eulers constant (also known as the
EulerMascheroni constant)
= density, kg/m
= dimensionless convection function
References
1. RAMEY, H.J. JR., Wellbore Heat Transmission;
Journal of Petroleum Technology, pp. 427-435, April
1962.
2. XIE, J., and MATTHEWS, C.M., Methodology to
Assess Thaw Subsidence Impacts on the Design and
Integrity of Oil and Gas Wells in Arctic Regions;
2011, SPE 149740.
3. XIE, J., and ZAHACY, T.A., Understanding Cement
Mechanical Behavior in SAGD Wells; WHOC11-
557, 2011.
4. HASAN, A.R., and KABIR, C.S., Heat Transfer
during Two-Phase Flow in Wellbores: Part I -
Formation Temperature; 1991, SPE 22866.
10
20
30
40
50
60
70
0 50 100 150 200 250 300
T
e
m
p
e
r
a
t
u
r
e

(

C
)
Depth (m)
Perfectly Insulated
Constant Temperature
Boundary
Convection Boundary
13
5. WILLHITE, G. PAUL., Over-all Heat Transfer
Coefficients in Steam and Hot Water Injection Wells;
Journal of Petroleum Technology, May 1967.
6. CHIU, K. and THAKUR, S.C., Modeling of Wellbore
Heat Losses in Directional Wells Under Changing
Injection Conditions; 1991, SPE 22870.
7. MOINI, B., and EDMUNDS, N., Quantifying Heat
Requirements for SAGD Start-up Phase: Steam
Injection and Electrical Heating; WHOC11-513, 2011.
8. HASAN, A.R., and KABIR, C.S., Fluid Flow and
Heat Transfer in Wellbores; Richardson, TX : SPE,
2002.
9. HAGOORT, J., Ramey's Wellbore Heat Transmission
Revisited; SPE Journal. SPE 87305, 2004.
10. CARSLAW, H.S and JAEGER, J.C., Heat
Conduction in Solids; Oxford University Press, 1959.
11. JAEGER, J.C., and CHAMALAUN, T., Heat Flow in
an Infinite Region Bounded Internally by a Circular
Cylinder with Forced Convection at the Surface;
Australian Journal of Physics, Vol. 19, pp. 475-488,
1966.
12. http://www.cambridge.org/us/engineering/autho
r/nellisandklein/downloads/invlap.m
13. JAEGER, J.C., Heat Flow in the Region Bounded
Internally by a Circular Cylinder; Proceedings, Royal
Society of Edinburgh, pp. 223-228, 1942.
14. VAN EVERDINGEN, A.F., and HURST, W., The
Application of the Laplace Transform to Flow
Problems in Reservoirs; Petroleum Transactions,
AIME, pp. 305-324, December 1949.
15. JAEGER, J.C., and CLARKE, M., A Short Table of I
(0,1;x), Proceedings, Royal Society of Edinburgh, pp.
229-230, 1942.
16. CLAUSER, C., and HUENGES, E., Thermal
Conductivity of Rocks and Minerals. Rock Physics
and Phase Relations - A Handbook of Physical
Constants; AGU Reference Shelf. American
Geophysical Union., Vol. Vol. 3, pp. 105-126.
17. SKOCZYLAS, P., A Method for Calculating
Transient Temperature and Pressure Profiles for
Crude Oil and Water Flowing in a Buried Pipeline;
Univeristy of Alberta, 2001. M.Sc. Thesis.
18. PHAM, Q.T., A Fast, Unconditionally Stable Finite
Difference Scheme for Heat Conduction with Phase
Change; No. 11, 1985, Int. J. Heat Mass Transfer,
Vol. Vol 28.
19. OSTERKAMP, T.E., Freezing and Thawing of Soils
and Permafrost Containing Unfrozen Water or Brine;
No. 12, Water Resources Research, Vol. Vol. 23,
December 1987.
20. MOTTAGHY, D., and RATH, V., Implementation of
Permafrost Development in a Finite Difference Heat
Transport Code; [Online] RWTH-Aachen University.
http://www.eonerc.rwth-aachen.de.
21. WILLIAMS, P.J., Unfrozen Water Content of Frozen
Soils and Soil Moisture Suction; No. 3, Geotechnique,
Vol. 14, September 1964.
22. OZISIK, M.N. and UZZELL, J.C. JR., Exact Solution
for Freezing in Cylindrical Symmetry with Extended
Freezing Temperature Range; Journal of Heat
Transfer, Vol. 101, pp. 331-334, May 1979.
23. LIU, Z., STARK, S., and LUNN, S., Modeling of
Wellbore Heat Loss for Thermal Operations at Cold
Lake A Convection Cell Approach; WHOC11-628,
2011.
Appendix A Correlations
The correlations from the various sources presented in this
paper are listed here.
Ramey
(1)
(line source, constant flux):
2kA1
q
= ln
1
2Po
u.29 ............................................... Equation A-1
Hasan and Kabir
(4)
(constant flux):
2nkA1
q
= _
1.1281Fo(1 u.SFo) Fo 1.S
u.4u6S + u.S ln(Fo) [1 +
0.6
Po
Fo > 1.S
....................
............................................................................................... Equation A-2
Hasan and Kabir
(8)
(constant flux):
2nkA1
q
= ln(c
-0.2Po
+(1.S u.S719c
-Po
)Fo) ..... Equation A-3
Carslaw and Jaeger
(10)
(constant flux, short times):
2nkA1
q
2Fo j
1

u.2SFo[ ..................................... Equation A-4


Carslaw and Jaeger
(10)
(constant flux, long times):
2nkA1
q

1
2
(ln4Fo y) ................................................... Equation A-5
Chiu and Thakur
(6)
(constant temperature):
2nkA1
q
= u.982 ln(1 +1.81Fo) .................................. Equation A-6
Moini and Edmunds
(7)
(constant temperature):
log
10
q
2nkA1
= u.uu24(log
10
Fo)
3
+ u.u446(log
10
Fo)
2

u.Su64(log
10
Fo) u.u126 .......................................... Equation A-7
This correlation should only be used in the range of
Fourier numbers between 0.01 and 1000.

Carslaw and Jaeger
(10)
(constant temperature, short times):
q
2nkA1
(nFo)
-
1
2 +
1
2

1
4
[
Po
n

1
2
+
1
8
Fo ................................ Equation A-8

Carslaw and Jaeger
(10)
(constant temperature, long times):
q
2nkA1
2 j
1
In(4Po)-2y

y
(In(4Po)-2y)
2
[ ................................... Equation A-9
14
Appendix B Derivation of a Finite
Difference Model
This Appendix contains an example derivation of a finite
difference heat transfer model.
Consider three adjacent radial nodes in an axisymmetric
system. These nodes are an inner node (i-1), a middle node
(i), and an outer node (i+1). In the simplest case, which we
will consider here, the properties around the nodes are
assumed to be constant and the nodes are equally spaced.
The heat flux from the inner node to the middle node can
be determined using the following radial heat transfer
equation:
q
1
=
2nk(1
i-1
-1
i
)
In[
r
r-Ar

................................................................. Equation B-1


Similarly, the heat flux from the outer node to the middle
node is:
q
2
=
2nk(1
i+1
-1
i
)
In[
r+Ar
r

.................................................... Equation B-2
Note that if the heat was flowing from the middle node to
one of the other nodes, that flux value would be negative.
Before we can look at what happens during a time step,
we need to know the volume around the centre node; this
volume is considered to be uniformly at the temperature of
the node. The volume per unit length or depth (since q is
measured per unit length) is:
I = 2nrr ......................................................................... Equation B-3
During a time step, the total energy entering (or leaving)
the volume must be the same as the change in energy stored
(or lost) in the volume:
(q
1
+q
2
)t = pIc(I

t+At
I

t
) ................................... Equation B-4
In this notation, I

t
refers to the temperature at node i at
time t.
Combining Equations A1-A4 yields:
2nkAt(1
i-1
-1
i
)
In[
r
r-Ar

+
2nkAt(1
i+1
-1
i
)
In[
r+Ar
r

= p2nrrc(I

t+At
I

t
) ..............
................................................................................................ Equation B-5
Some constants can be cancelled, we can substitute
diffusivity for the combination of density, specific heat and
thermal conductivity, and we can solve the resulting
relationship for the change in temperature of the centre node
during the time step, leaving:
(I

t+At
I

t
) =
uAt(1
i-1
-1
i
)
A In[
r
r-Ar

+
uAt(1
i+1
-1
i
)
A In[
r+Ar
r

................. Equation B-6
Alternatively, a more useful way of writing this might be:
I

t+t
= I

t
+
ut
In[
r
r-Ar

(I
-1
I

) +
ut
In[
r+r
r

(I
+1
I

) ..............
............................................................................................... Equation B-7
or by lumping variables as:
I

t+At
= I

t
+ X(I
-1
I

) + (I
+1
I

) ............... Equation B-8


where:
X =
uAt
A In[
r
r-Ar

.................................................................. Equation B-9


=
uAt
A In[
r+Ar
r

................................................................ Equation B-10
It has not yet been specified whether most of the
temperature terms on the right hand side of the equation are
evaluated at time t or t+t. It might seem obvious to use time
t; this is called an explicit method. If we do this, the
calculation procedures are very simple, but there is a problem
with stability if the time steps are not kept sufficiently small,
which may make our calculation take much longer. On the
other hand, if we evaluate those temperatures at time t+t, we
have to solve a system of simultaneous linear equations, but
we do not have a stability problem with longer time steps.
This is called an implicit method, and was chosen for this
work. Equation A-8 is then rewritten as:
XI
-1
t+At
+ (1 + X +)I

t+At
I
+1
t+At
= I

t
....... Equation B-11
This is now in a form which is easily adapted to matrix
methods for the simultaneous solution of a system of linear
equations, such as by the method of Gaussian elimination.
The methods for setting up the matrix and then solving it are
not discussed further here but were given by Skoczylas
(17)
for
a more complicated two-dimensional heat transfer problem
solved with the FD method.
To derive the equations for different situations, such as
variable nodal spacing, or different properties around each
node, or for the presence of a discontinuity at a boundary
between regions, the same basic process is used. That is,
where energy flowing into the node during a time step is
equated to the energy storage in the volume around the node.
Appendix C Inputs Used in
Scenarios Presented
Unless otherwise specified, the material thermal properties
used in the problems described here are listed in Table C-1.

Material Thermal
Conductivity,
W/mK
Density,
kg/m
Specific
Heat, J/kgK
Casing 45 7850 450
Cement 1.2 2000 1000
Formation 3.0 2200 1200
Table C-1 Thermal Properties
When used, the casing OD is 9.625, the ID is 8.835.
The casing is cemented in a hole with a diameter of 13. In
cases with no casing or cement, the hole diameter is also 13.

15
Case 1a
The far-field and initial temperatures are 5C. The
imposed wall temperature at the inside of the wellbore is
50C.

Case 1b
The far-field and initial temperatures are 5C. The fluid
temperature inside of the wellbore is 50C, and there is a
convection coefficient of 15 W/mK.

Case 2
The ground thermal conductivity varies linearly from
3.2 W/mK at 0C to 1.5 W/mK at 30C.
The far-field and initial temperatures are 0C. The
imposed wall temperature at the inside of the wellbore is
300C. Note that in this case, there is no consideration of
phase change effects.

Case 3
The far-field and initial temperatures are -15C. The
imposed wall temperature at the inside of the wellbore is
25C.
The frozen ground thermal conductivity is 3.6 W/mK and
the unfrozen ground thermal conductivity is 2.5 W/mK. The
specific heat of frozen ground is 1103 J/kgK and the specific
heat of unfrozen ground is 1500 J/kgK. For partially frozen
ground, these properties are a weighted average, based on the
unfrozen content. The unfrozen content is shown as a
function of temperature in Figure B-1. The grounds latent
heat is 140 MJ/m

Figure B-1 Percent Unfrozen with Temperature
Case 4
This case has no casing or cement. The far-field and
initial temperatures are 20C. The imposed wall temperature
at the inside of the wellbore is 75C.

Case 5
This case has no casing or cement. The far-field and
initial temperatures are 20C. The imposed wall temperature
at the inside of the wellbore is 50-70C, as described in the
text.

0
10
20
30
40
50
60
70
80
90
100
-10 -8 -6 -4 -2 0
%

U
n
f
r
o
z
e
n
Temperature, C

Vous aimerez peut-être aussi