Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

DOI: 10.

2118/204479-PA Date: 20-August-21 Stage: Page: 529 Total Pages: 13

Horner Analysis for Negative Inflow


Tests of Well Barriers
James Peyton, Joanna Salamaga, Aaron McPhee, and Arthur Jongejan, Woodside Energy, Ltd.

Summary
Negative tests, or inflow tests, are conducted to verify the integrity of well barriers in the direction of potential flow, subjecting a barrier
to a negative pressure differential, while monitoring for signs of a leak. A common practice is to observe the rate of flowback from the
well. Flowback may be a sign of a leak due to an influx of formation fluids into the well. However, even when there is no leak, flowback

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
is commonly observed due to thermal expansion of wellbore fluids. Heat transfer will occur between the wellbore fluids in each annulus
and with the surrounding formation until temperatures reach an equilibrium. This behavior is described by the process of thermal diffu-
sion, with the resulting temperature increase causing expansion of wellbore fluids and flowback from the well.
Industry guidelines state “Horner” analysis may be used when monitoring flowback or pressure buildup during an inflow test. In
doing so, engineers and wellsite supervisors may use a “Horner plot” to determine if flowback or pressure buildup is attributable to ther-
mal effects. Those without a reservoir engineering background may not be aware the method was originally derived from a radial flow
equation for the purpose of monitoring pressure buildup in a well when shut in after a period of production. The apparent similarity of
the radial flow and thermal diffusion equations is what led Horner’s technique to subsequently be applied to the prediction of static for-
mation temperature from well logs. However, although thermal expansion is a function of formation temperature, Horner analysis of
flowback or pressure buildup during an inflow test has remained a black box that is poorly understood.
For the first time, with support from empirical data from offshore wells, we reveal that Horner analysis of thermal expansion is a prac-
tice without theoretical justification. The radial equation on which Horner analysis depends, along with the constraints implied by the
boundary conditions, fails to accurately account for the conditions of an inflow test. As a result, the method should not be used for analyz-
ing flowback or pressure buildup during an inflow test. Instead, a new method is proposed to interpret a trend of flowback when monitoring
well barriers. The findings of this study can help improve understanding Horner analysis and techniques for interpreting inflow tests.

Introduction
Negative tests, or inflow tests, are conducted by subjecting a well barrier to a negative pressure differential, while monitoring pressure
or flowback (see Fig. 1). Observations of pressure change or flow rate are often plotted with respect to Horner time. This semilog graph
is commonly referred to as a Horner plot. This practice is carried out by engineers and wellsite supervisors to help determine whether
observed pressure increase or flowback is an indication of a leak or is attributable to expansion of wellbore fluids due to thermal diffu-
sion (heat transfer from formation to wellbore fluids).
Horner originally derived the equation from Muskat’s radial flow equation, which describes flow of a homogeneous compressible
fluid in an infinite plane reservoir. Thus, Horner’s equation describes pressure buildup in a reservoir, which is far removed from the pro-
cesses involved in an inflow test, and is unrelated to pressure buildup or flowback caused by thermal expansion of fluids. Timko and
Fertl were first to note there is similarity between the radial flow equation and the radial thermal diffusion equation, showing the
method could be applied to the prediction of static formation temperature from well logs. Later, Dowdle and Cobb explored the method
and explained that the boundary conditions implied by the technique were not correct for thermal diffusion. Their contribution was to
describe the conditions under which the technique could produce useful estimates.
If there is any theoretical justification for extending Horner analysis to thermal expansion of wellbore fluids, it is has not been pub-
lished. To determine if the method is useful, we begin with the radial thermal diffusion equation, which is the same form of partial dif-
ferential equation as the radial flow equation used by Horner. This equation is used to derive an equation for temperature buildup,
which is of the same form as Horner’s equation of pressure buildup. We then extend this to describe volumetric expansion caused by
radial thermal diffusion.
The main contribution of this paper is new information revealing why Horner’s equation is not suitable for analysis of inflow tests.
To support this conclusion, we present data from offshore wells to demonstrate the behavior predicted by the equation is not supported
by empirical observations. We then discuss problems commonly encountered when using Horner plots, highlighting the pitfalls inherent
with this practice. Finally, an alternative method for interpreting inflow tests is proposed, supported by results from a field trial.

Radial Flow Equation


The starting point used by Horner (1951), which was originally given by Muskat (1946, Chap. X, Sec. 10.2, Eq. 1, p. 630), is the partial
differential equation for radial flow of a single-phase fluid in a porous medium:
 
1@ @p /p lc @p
r ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð1Þ
r @r @r kp @t
/p lc
with pressure p, time t, radius r, and diffusivity constant . To apply Horner’s technique, we need to obtain an equation for thermal
diffusion, which is of the same form as Eq. 1. kp

Thermal Diffusion
Before turning to the radial equation, we present the equation for thermal diffusion in cylindrical coordinates in Eq. 2 (see Lienhard and
Lienhard 2019, p. 144),

Copyright V
C 2021 Society of Petroleum Engineers

Original SPE manuscript received for review 10 September 2020. Revised manuscript received for review 30 October 2020. Paper (SPE 204479) peer approved 3 November 2020.

September 2021 SPE Drilling & Completion 529

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 530 Total Pages: 13

   
1@ @T 1 @ 2 T @ @T q_ 1 @T
r þ 2 2þ þ ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð2Þ
r @r @r r @/ @z @z k a @t
k
with temperature T, time t, radius r, azimuth /, vertical axis z, and constant of thermal diffusivity a ¼ . We begin with the equation
qcp
in cylindrical coordinates to demonstrate the consequence of using a radial equation. To obtain the radial equation, the second and third
terms of Eq. 2 must be excluded. This requires us to ignore any change in temperature in the vertical axis. The effect of this is to
“collapse” the well into a horizontal plane in the subsurface, ignoring the different rates of heat transfer occurring along the length of
the well.

Topdrive

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Pressure gauge
TIW Cement
head

TIW
Rig floor

Flowline Monitor Cement Reverse


unit out line
Riser

BOP
Pits

Mud pit
Choke line

Wellbore Mud
packer
Lighter fluid
Well barrier
element

Fig. 1—Typical inflow test lineup. BOP 5 blowout preventer; TIW 5 full-opening safety valve.

q_
We are also required to assume there is no heat generation, such that the term becomes zero. Because the process of hydration of
k
cement slurry is exothermic, inflow tests that occur soon after a cement job will not be accurately accounted for by this radial thermal
diffusion equation.
The equation also does not account for heat transfer between various fluids in the well, either due to fluid trains displaced into the
well during an inflow test or the effect of annulus fluids. We will revisit the implication of these constraints later.
The cylindrical thermal diffusion equation reduces the radial equation with no heat generation:
 
1@ @T qcp @T
r ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð3Þ
r @r @r k @t
By applying the constraints above, we obtain Eq. 3, which is of the same form as Eq. 1. Timko and Fertl (1972) were first to note
similarity between these two equations. However, Dowdle and Cobb (1975) state this is only an apparent similarity and explored the
effect of the boundary conditions to determine if there were any limitations in applying Horner’s technique to predicting
formation temperature.
The following are the thermal boundary conditions necessary to apply Horner’s method:
1. T ¼ Ti at t ¼ 0 for all r.
2. T ¼ Ti at r ¼ 1 for all t.
@T Q
3. limr!0 r ¼ .
@r 2pk
Condition 1 means that temperature is equal to Ti at time t ¼ 0 at all distances from the wellbore r (i.e., the starting condition is that
the well is in equilibrium with static formation temperature).
Condition 2 means that temperature approaches Ti as distance from the wellbore increases, for all t (i.e., formation temperature does
not change at some distance sufficiently far from the wellbore).

530 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 531 Total Pages: 13

Condition 3 is the heat transfer rate at the wellbore. This is Fourier’s equation for heat conduction in radial coordinates rearranged.
@T
Q ¼ k2pr (see Lienhard and Lienhard 2019, p. 69). Q is the heat transfer rate. This condition implies there is a constant temperature
@r
gradient at the wellbore. Dowdle and Cobb (1975) showed the gradient does change during mud circulation and that therefore condition
3 is not correct for thermal diffusion. However, as Dowdle and Cobb (1975) demonstrated, providing that mud circulation time is short,
this boundary condition can still lead to a useful estimate of static formation temperature at a given depth.
The full solution to Eq. 3 is given in Appendix A.
 
Q t0 þ Dt
T ¼ Ti  log : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ
4pk n Dt
Eq. 4 is identical in form to the equation presented by Horner (1951, Eq. V, p. 27) but now derived for radial thermal diffusion instead
of radial flow, for a given depth in a well. As Dowdle and Cobb (1975) demonstrated, providing that mud circulation duration is short, this
equation can be used to predict static formation temperature, with the magnitude of underestimation related to mud circulation duration.

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Thermal Expansion
Further steps need to be taken to adapt the formula for use on inflow tests. To the authors’ knowledge, these steps are not described in
literature. First, we would have to assume a single homogeneous fluid in the well with volumetric coefficient of thermal expansion a.
Due to the use of the radial equation, we are required to ignore the temperature gradient of the fluid column and must treat the expan-
sion of the fluid as a bulk process occurring due to change in temperature at a representative point in the well.
Let Vi equal volume at t ¼ 0 and T ¼ Ti , and r equal a small value, rw. The volume of the fluid as it contracts and later expands is
given by V. Because the well is kept full any additional fluid added to the well will be displaced as fluid expands. Adapting Eq. A-24,
we obtain
 
Q cqcp rw2
Vi  V ¼ a logn : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð5Þ
4pk 4kt
If the well is then left static for period Dt, after being cooled for time t0, the volume drop can be obtained by principle of superposi-
tion. Vi  V ¼ (volume drop caused by rate Q for time t0 þ Dt) þ (volume drop caused by rate change Q for time Dt).
   
Q cqcp rw2 Q cqcp rw2
Vi  V ¼ a logn þa logn ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð6Þ
4pk 4kðt0 þ DtÞ 4pk 4kDt
or
 
Q t0 þ Dt
V ¼ Vi  a logn : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð7Þ
4pk Dt

Again, we obtain an equation of the same form as Horner (1951). Eq. 7 gives the volume of fluid as it expands during period Dt after a
contraction of fluid volume during period t0 for a homogeneous change in fluid temperature.

Horner’s Technique
Because the equation has been derived for thermal expansion instead of its intended use (pressure buildup), it is useful to revisit the pur-
pose of this method.
1. First, the equation produces a straight line. This is relevant, considering the lack of electronic calculators when this method was first pub-
lished in 1951. This method provided a practical way to obtain a result (plot data on a graph and measure the straight-line gradient).
2. From this line, it is possible to estimate the permeability of the reservoir (Muskat 1946; Horner 1951; Matthews and Russell
1967, p. 18), which has obvious practical uses for reservoir engineers.
3. In the case of reservoir pressure buildup, the formula allows a prediction to be made of pressure at infinite shut-in time (p ), with
t0 þ Dt
the difference p and pi being an estimate of reservoir depletion (Matthews and Russell 1967, p. 19). The fact that ! 1,
  Dt
t0 þ Dt
and therefore, logn ! 0 as t ! 1 makes it easy to extrapolate the line to infinite shut-in time.
Dt
It is important that observations should follow a straight line because the ability to extrapolate a line to infinite shut-in time to make
a prediction is compromised if data are nonlinear.
 To understand why, consider the Horner plots of the following functions in Fig. 2.
t0 
1. Horner’s equation, limt!1 10  logn 1 þ ¼ 10.
t
2. A logarithmic function, limt!1 logn ðtÞ ¼ 1.
As we can see from the equations, the limit of Horner’s equation is 10, and extrapolating a line to infinite shut-in results in an accu-
rate calculation of the limit. However, the result for logn ðtÞ should be concerning. Although this function approaches 1, extrapolating a
line on a Horner plot results in a y-intercept of 4.77. Furthermore, because the line visually appears to be linear on a Horner plot over
the observed time period, this could give a false impression that it is close enough to Horner’s equation for extrapolation to be accepta-
ble. The result shows that Horner’s technique should be applied with caution.

Horner Analysis for Inflow Tests


As stated in the UK Oil & Gas Well Life Cycle Integrity Guidelines (2019), “Horner plots may be used to differentiate between pressure
build up due to a leak and thermal effects” and to “verify a successful test fluid volumes may be monitored closely.” The guidelines
also state “Horner plot analysis may be used to assist this process.” Similar guidance is provided in API RP 96 (2013), which states “a
Horner plot may be necessary to evaluate the results of the inflow test, which would require monitoring shut-in pressure at the choke
manifold vs. time.” However, there exists no specific guidance on how Horner plot analysis should be carried out, nor are there any
industry standards covering this process.

September 2021 SPE Drilling & Completion 531

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 532 Total Pages: 13

i. Horner equation (limit = 10) ii. Log time (limit = infinity)

10.0 5

4
7.5

Volume (L)
3
5.0
2
2.5 1

0 0
0 1 2 3 4 0 1 2 3 4
Log Horner Time

Fig. 2—Horner plot of volume for Horner’s equation (left) and log(t) (right) (t0 5 60 minutes).

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Although some well control training material provides guidance on this topic, the practice may vary across the industry. However,
Horner analysis necessarily involves some form of interpretation of the trend of pressure, pressure change, total flowback, or flow rate
on a Horner plot. In the case of flow rate, it is common practice to apply a trend line on a Horner plot. If the trend line has a negative
y-intercept, then flow rate is deemed to be reducing, and the barrier is deemed to be “holding.” Conversely, if there is a positive
y-intercept, it is considered an indication that flow is not reducing. In other words, the y-intercept is interpreted as a prediction that flow
rate will reduce to zero at some time before t ! 1. This is a significant deviation from Horner’s method and is pseudomathematical.  t0 
Extrapolating a line on a Horner plot is to make the assumption that future observations will follow a function of logn 1 þ .
Dt
When observed data do not follow this function (i.e., do not follow a straight line), there is no basis on which to make this assertion.
Extrapolation is a technique to calculate the asymptote of Horner’s equation alone and is no predictor of future values of any
other function.
The authors are aware of anecdotal arguments that plotting a function of exponential decay on a Horner plot will yield a straight
line, and that this is why a Horner plot may be used for thermal expansion. There is no basis for this. The fact that Horner’s equation is
t0
asymptotic is due to the principle of superposition applied in Eq. 6, which introduces the term 1 þ . It is incorrect to plot any asymp-
Dt
totic function on a Horner plot and expect the y-intercept to be related to that function’s asymptote.
Although Dowdle and Cobb (1975) showed that this can be done when predicting static formation temperature, even when data are
nonlinear, this only holds for a given depth in a well, and that “the longer the circulating time the greater the error in the estimated
static value.” Such a practice is dangerous when applied to an inflow test, as in the case of a failed barrier, flowback, or pressure buildup
may continue indefinitely. Horner analysis may mislead the assessor as to whether total flowback or pressure buildup is tending toward
a finite value or continuously increasing.
Although Horner analysis of flowback is problematic, API RP 96 (2013) only refers to use of a Horner plot when monitoring shut-in
pressure. As Horner’s equation was derived for pressure buildup analysis, it is conceivable that it has been incorrectly assumed the
method is suitable for analyzing pressure buildup caused by thermal expansion. There is a significant difference between how reservoir
pressure buildup is monitored when compared to monitoring pressure buildup from thermal expansion. Because radial flow is occurring
within the reservoir, there is no need to consider pressure change beyond the reservoir itself. Pressure observed at the surface is simply
pressure at the reservoir less any hydrostatic head in the well (i.e., the assumption of flow in a horizontal plane is acceptable).
This is not the case for thermal diffusion, as formation along the length of the well has been cooled during previous operations and
there often are multiple fluids in the wellbore and annuli. These conditions are not accounted for in the thermal diffusion equation
implied by Horner’s technique. Monitoring actual reservoir pressure buildup is distinct from monitoring well pressure buildup from
thermal expansion. The two should not be confused. Doing so poses a risk because extrapolating what appears to be a straight line on a
Horner plot of pressure to predict a future value of pressure could yield a misleading result (see Fig. 2).
It appears that Horner analysis will yield a useful prediction only if thermal expansion is dominated by temperature change at some
short section of the well. As questions remain regarding the practical utility of Horner’s method, empirical analysis is required.

Empirical Analysis
Three groups of charts are presented to analyze data from offshore wells.
1. Horner plots of flowback volume, Fig. 3
2. Horner plots of flow rate, Fig. 4
3. Linear time plots for flow rate, Fig. 5
Observation of Fig. 3 shows that in no case does a Horner plot of volume follow a straight line throughout the entirety of the test.
This supports what has been concluded from the theory: that the behavior of thermal expansion of fluid in the well is not accounted for
by the radial equation and boundary conditions assumed. This is expected given the assumptions already noted and the limitations
noted by Dowdle and Cobb (1975). Although volume appears to trend toward a straight line as the test continues, failed tests such as
Test B are indistinguishable from successful tests such as Test F. As we have seen in the previous section, unless data lie on a straight
line, there is no way to accurately predict future behavior using Horner analysis.
Although Dowdle and Cobb (1975) showed Horner plots can still predict static formation temperature when data are nonlinear, this
is safely constrained by the fact temperature change is governed solely by the process of thermal diffusion and must tend toward static
formation temperature over time. There is no such guarantee in an inflow test because a leak is not governed by thermal diffusion and
may lead to sustained flow. For inflow tests, the risk is not harmlessly underestimating thermal expansion but neglecting to identify a
failed barrier. Because it is not possible to distinguish a function that is approaching a limit from one that is not, Horner plots of flow-
back volume are of no analytical value. t0
We have shown Horner’s equation is the outcome of superposition, which gives rise to the term 1 þ in the temperature and
Dt
volume equations. There is no theoretical justification to plot rate of change of thermal expansion (flow rate) with respect to the Horner

532 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 533 Total Pages: 13

time. However, because this is one way fluid volume may be monitored during an inflow test, we will review the results (see Fig. 4).
Note, the trend lines shown in Fig. 4 are applied to the full set of data for each test and the lines do not represent the prevailing trend in
all cases. However, we can see in several tests there appears to be a clear decreasing trend: Test D, Test F, Test G, Test I, Test M, Test
N, and Test O. Conversely, several tests seem to show sustained flow, such as Test B, Test C, and Test J.

Test A Test B Test B (truncated) Test C


25
60 30
20
20
40 20 15

10 10
20 10
5
0 0 0 0
0 1 2 3 4 5 0 2 4 6 0 2 4 6 0 2 4 6

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Test D Test E Test F Test G
5 30 25
4 15 20
20
3 15
10
2 10
10
5
Flowback Volume (L)

1 5
0 0 0 0
0 2 4 6 0 2 4 6 0 2 4 6 0 2 4 6

Test H Test I Test J Test K


400
16
6
900 300
12
4
8 600 200

2 300 100
4

0 0 0 0
0 1 2 3 4 5 0 2 4 6 0 1 2 3 4 0 1 2 3

Test L Test M Test N Test O


50
400 75 30
40
300
30 50 20
200 20
25 10
100 10
0 0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 0 2 4 6
Log Horner Time

Fig. 3—Flowback volume vs. log Horner time (t0 5 200 minutes).

Test A Test B Test B (truncated) Test C


1.25
y = 0.351 + 0.0518 x y = −0.173 + 0.262 x y = −0.415 + 0.364 x 0.5 y = 0.156 + 0.016 x
1.00 0.75 0.75
0.4
0.75 0.3
0.50 0.50
0.50 0.2
0.25 0.25
0.25 0.1
0 0 0 0
0 0.5 1.0 1.5 2.0 2.5 0 1 2 3 0 1 2 3 0 1 2 3

Test D Test E Test F Test G


0.25
y = −0.0653 + 0.0581x 0.4 y = 0.0849 + 0.0231 x y = −0.023 + 0.12 x y = −0.0662 + 0.132 x
0.20 0.75 0.75
0.3
0.15 0.50 0.50
0.10 0.2
Flow Rate (L/min)

0.25 0.25
0.05 0.1

0 0 0 0
0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3

Test H Test I Test J Test K


y = 0.0325 + 0.0235 x y = −0.0537 + 0.0639 x y = 1.73 + 0.387 x 8 y = 1.34 + 0.836 x
3
0.2 0.2 6
2
4
0.1 0.1
1 2

0 0 0 0
0 0.5 1.0 1.5 2.0 2.5 0 1 2 0.0 0.5 1.0 0 1 2 3

Test L Test M Test N Test O


0.8
5 y = 0.453 + 0.796 x 2.0 y = −0.142 + 0.389 x y = −1.06 + 1 x y = −0.015 + 0.203 x
4 3 0.6
1.5
3 2 0.4
1.0
2
0.5 1 0.2
1
0 0 0 0
0 0.5 1.0 1.5 2.0 2.5 0 1 2 3 4 5 0 1 2 3 4 0 1 2 3
Log Horner Time

Fig. 4—Flow rate vs. log Horner time (t0 5 200 minutes)—instantaneous flow rate (gray) with moving average (black) and trend
line (blue).

September 2021 SPE Drilling & Completion 533

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 534 Total Pages: 13

Test A Test B Test B (truncated) Test C


1.25
0.5
1.00 0.75 0.75
0.4
0.75 0.3
0.50 0.50
0.50 0.2
0.25 0.25
0.25 0.1
0 0 0 0
0 50 100 150 0 20 40 60 0 10 20 30 40 50 0 25 50 75 100

Test D Test E Test F Test G


0.25
0.4 0.75
0.20 0.75
0.3
0.15 0.50 0.50
0.10 0.2
Flow Rate (L/min)

0.25 0.25
0.05 0.1

0 0 0 0
0 10 20 30 40 50 0 50 100 150 200 250 0 20 40 60 0 20 40 60

Test H Test I Test J Test K

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
8
3
0.2 0.2 6
2
4
0.1 0.1
1 2

0 0 0 0
0 50 100 150 200 0 25 50 75 100 0 100 200 300 400 0 50 100 150

Test L Test M Test N Test O


0.8
5
2.0
4 3 0.6
1.5
3 2 0.4
1.0
2
0.5 1 0.2
1
0 0 0 0
0 100 200 300 400 0 20 40 60 80 0 25 50 75 100 0 20 40 60 80
Time

Fig. 5—Inflow tests—flow rate vs. linear time axis with trend lines fitted to first and second half of each test—instantaneous flow
rate (gray) with moving average (black) and trend lines (blue).

In the above cases, use of a Horner plot may be somewhat superfluous and harmless. However, for the following tests Horner analy-
sis is ambiguous: Test E, Test H, Test K, and Test L. A discussion of some of the pitfalls of this practice is given in the next section.
Fig. 4 shows that in these tests the trend lines appear to have a decreasing slope, which would suggest flow rate is decreasing, yet
there is a positive y-intercept. On a Horner plot a positive y-intercept is typically interpreted as an indication flow is not decreasing to
zero. In these cases, Horner plots can lead to confusion because it seems to suggest that flow is decreasing but not to zero. This analysis
is problematic for assessors. It is also an interpretation without theoretical justification. The use of Horner plots in such situations is
therefore not harmless and may lead to invisible lost time as tests are extended or repeated.
For all these tests, it is worth reviewing Fig. 5 and comparing this to the Horner plot for each test. In all these examples, any trend
that is apparent on a Horner plot is also easily identified on the equivalent linear time plot. In many cases, a trend is more obvious to the
eye on the linear time plot, such as Test B (truncated). The Horner plot does not appear to provide any further clarity as to whether flow
is decreasing. Instead, it potentially obfuscates any trend because data density increases with time due to the use of a log scale.

Pitfalls of Horner Analysis for Inflow Tests


Predictive Value of Horner Plots. Fig. 6 shows linearly decreasing data plotted against Horner’s equation and a linear time axis. It is
clear that extrapolating a line on a Horner plot does not predict the future values of this function. The trend line changes over time,
despite there being a constant trend. The problem it presents for well construction operations is that, despite a clear deceasing trend, the
trend line has a positive y-intercept, which would be interpreted that flow rate is not decreasing to zero. Although it is easy to see with
artificial data, in real-world inflow tests, the interpretation is more difficult. A Horner plot does not assist in determining whether flow is
decreasing to zero or over what period of time.

Horner plot (t0 = 200 minutes) Linear time plot


Flow Rate (L/min)

0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 0 100 200 300 400
Log Horner Time Time

Fig. 6—Trend lines on Horner plot with linearly decreasing data (t0 5 200 minutes).

Subjectivity. From Fig. 7 shown earlier it can be seen many of the Horner plots do not meet the criteria of a negative y-intercept
required for a successful test. In these cases, the parameter t0 has been arbitrarily set to 200 minutes. As shown in the theory, t0 should

534 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 535 Total Pages: 13

be the time in which formation temperature has been disturbed. It is important to note the y-intercept is itself a function of t0. Eq. 8 may
be solved iteratively for any value of t0ðnewÞ , which gives a negative value of the y-intercept ct0 ¼t0ðnewÞ.
8     9
>
> t0ðnewÞ t0 >
>
>
> >
>
< logn 1 þ x mlogn 1 þ
x2 =
2
ct0 ¼t0ðnewÞ ¼ c  ðy2  y1 Þ  (  )þ     : . . . . . . . . . . . . . . . . . . . . . ð8Þ
>
> x1 x2 þ t0ðnewÞ t0 t0 > >
>
> mlogn 1 þ  mlogn 1 þ >
:logn 
x2 x1 þ t x2 x1 > ;
0ðnewÞ

Test G Test G (altered)


8
y = −0.834 + 0.543 x y = 0.0732 + 0.0719 x
Flow Rate ( L/min)

0.75
6

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
4 0.50

2 0.25

0 0.00
0 1 2 3 4 5 01 2 3 4 5
Log Horner Time

Fig. 7—Horner plot with first data point altered—instantaneous flow rate (black) and trend line (blue) (t0 5 200 minutes).

It is frequently not clear what value of t0 should be used. All operations in the well since drilling commenced will affect formation
temperature to a varying degree. Using a value of t0 that is too large will result in a Horner plot with a negative y-intercept and a suc-
cessful test. Using a value of t0 that is too small will result in a positive y-intercept and a failed test. This makes it difficult for an asses-
sor to objectively select a value of t0. The very fact that the interpretation of the test is dependent on a somewhat arbitrary parameter,
rather than on data alone, explains why Horner analysis is often inconsistent and ambiguous.

High Leverage Observations. There are situations in which altering a single data point can lead to a contradictory interpretation of
the Horner plot. This is due to high leverage observations (James et al. 2018, p. 97). Test G will be used to demonstrate this problem.
The first observation is changed from 7.5 to 1.0 L/min. This results in the trend line gradient changing significantly, and Horner analysis
suggests a pass instead of a fail. This demonstrates how Horner analysis can result in contradictory interpretations with very small
changes in data and is therefore not a reliable form of analysis for inflow tests.
One way of addressing this problem is to apply linear quantile regression instead of least squares. This method is discussed later;
however, it should be noted that tools in commonly available spreadsheet applications that are often used for Horner plots make use of
least squares regression and are therefore vulnerable to high leverage observations.
Real-world situations are often difficult due to the presence of noise, in the form of rig motion, ambient temperature changes, and
error introduced in measurement. When relying on a Horner plot, assessors may often have to deal with contradictory results. A trend
on a linear time plot may be confounded by an ambiguous Horner plot, leaving decision makers in a difficult position.

Alternative Method of Interpreting Inflow Tests


As there is no theoretical justification to use Horner’s technique to evaluate a trend in an inflow test, a trend must instead be obtained
by plotting flow rate or pressure with respect to time (see Fig. 5). By applying linear regression, it is possible to obtain upper and lower
bound estimates of the gradient of a linear line of best fit (i.e., a linear model) (see Table 1). The method of linear regression applied
here is L1 linear quantile regression.

Linear Model Evaluation. Tests such as the Test B (truncated) present a problem because data are nonlinear. There are two distinct
periods during this test. The first 30 minutes involves a reducing trend, which then develops into constant flow. Fitting a linear line to
data without considering the goodness of fit could result in incorrect interpretation of the trend. To understand if data are linear, we can
plot the difference between the linear model and flow rate to obtain residuals. James et al. (2018) explained that if data are linear there
should be no discernible pattern in the residuals. Residuals from perfectly linear data would appear as noise.
Fig. 8 shows that in many instances this is not the case, and there is a discernible pattern (e.g., Test B, Test A, and Test D). This is
behavior is known as “tracking” of residuals (James et al. 2018, p. 94). In these cases, this is because the trend that is apparent from the
beginning of the test changes at some point later on in the test.
1. Test A is on an increasing trend until  80 minutes, after which it decreases.
2. Test B (truncated) is decreasing until  30 minutes, when flow rate becomes constant.
3. Test D has a similar sharp decrease before settling out to a more gradual decline around  30 minutes.
An option is to fit a model to a subset of data, such that data will be better approximated by a linear model. The new residuals are
shown in blue in Fig. 8. In several cases, fitting a model to the last 50% of the test makes a significant difference to the pattern of resid-
uals. It is always important when fitting a model to review if the fit is appropriate. When used on inflow tests fitting a model to a rele-
vant subset of data, such as the most recent 50% of a test, can be a practical way to do so. Linear time plots of flow rate with trend lines
are shown in Fig. 5 and it can be seen this method is effective in discriminating the initial trend from the prevailing trend.
However, Test B presents a problem. Even after applying a trend line to the last 50% of the test, the upper bound gradient is nega-
tive, which suggests a pass (see Table 1), yet we know this soon develops in a failed test. This suggests that in this case 60 minutes may
be an insufficient duration in which to confidently make an assessment of the trend. However, rather than considering just a snapshot in
time, we may also observe how the trend lines evolve over time.

September 2021 SPE Drilling & Completion 535

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 536 Total Pages: 13

Duration Estimated Gradient Upper Bound Estimated Gradient Upper Bound


Well (minutes) (1st half) (1st half) (2nd half) (2nd half)
Test A 186 5.10 5.29 –2.56 –2.12
Test B (truncated) 60 –28.49 –25.12 –1.85 –1.42
Test B 83 –15.98 –11.81 0.19 0.24
Test C 110 –0.67 –0.58 –0.14 0.03
Test D 60 –5.24 –4.42 –0.59 –0.43
Test E 247 –0.32 –0.19 –0.11 –0.08
Test F 79 –5.27 –4.50 –1.28 –1.08
Test G 72 –1.32 –0.68 –2.19 –0.75
Test H 248 –0.35 –0.30 –0.07 0.02

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Test I 120 –0.97 –0.73 –0.27 –0.15
Test J 527 –1.15 –1.09 –0.08 0.06
Test K 181 –39.26 0.97 –9.72 17.97
Test L 400 –4.32 –3.50 –1.73 –1.68
Test M 74 –19.12 –13.93 –17.92 –16.69
Test N 95 –53.33 –26.33 –4.57 –2.82
Test O 91 –6.50 –5.51 –3.38 –3.08
2
Table 1—Estimated gradients of flow rate for trend lines shown in Fig. 8, double derivative of thermal expansion (mL  min ).

Test A Test B Test B (truncated) Test C


100

50

−50

−100
0 50 100 150 0 20 40 60 0 10 20 30 40 50 0 25 50 75 100

Test D Test E Test F Test G


100

50

−50
Residuals (%)

−100
0 10 20 30 40 50 0 50 100 150 200 250 0 20 40 60 0 20 40 60

Test H Test I Test J Test K


100

50

−50

−100
0 50 100 150 200 0 25 50 75 100 0 100 200 300 400 0 50 100 150

Test L Test M Test N Test O


100

50

−50

−100
0 100 200 300 400 0 20 40 60 80 0 25 50 75 100 0 20 40 60 80
Time

Fig. 8—Residuals plot with residuals for single model (black) and for first half (cross) and second half (triangle) models.

Interpretation of Trend Line Gradients. In Fig. 9, we can observe how the estimated gradient of a trend line for a given test changes
as the test progresses. The trend line is applied to the most recent 50% of the test data. This approach provides a more complete picture
of what the inflow test is telling us. If the gradient is relatively consistent throughout the test, we can be confident in interpreting
the trend. However, if the gradient is still changing it is an indication that any trend is still developing. Reviewing results in Fig. 9
demonstrates that, although gradients can be expected to change over time, in many cases it will tend toward a stable value after some
period of time.
However, Test E provides an interesting example. In this case, the gradient fluctuates from negative to positive values and seems to
tend toward zero. If we consider only a snapshot in time, we may incorrectly conclude there is a negative trend, when in fact the gradi-
ent of the trend line may be zero (constant flow). The problem of such an approach is that it is necessary to distinguish between a gradi-
ent that is approaching zero, and one that is close to zero but remaining negative.

536 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 537 Total Pages: 13

Test A Test B Test B (truncated) Test C


0 0
0.010
−0.005 −0.005 0×100
0.005 −0.010
−0.010
−5×10–4
0 −0.015 −0.015
−0.020 −0.020 −1×10–3
−0.005
40 80 120 160 20 40 60 10 20 30 40 50 25 50 75 100

Test D Test E Test F Test G


0 0.001 0 0

−0.001 0
−0.002
−0.001 −0.002
−0.002
−0.002 −0.004
−0.003 −0.003 −0.004
Gradient (mL.min−2)

−0.004 −0.004 −0.006

10 20 30 40 50 0 50 100 150 200 20 40 60 20 40 60

Test H Test I Test J Test K


0.002 0.2

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
0
0×100 0
0
−0.0005
–5×10–4 −0.2
−0.002
−0.0010
–4×10–3 −0.4
−0.004
−0.0015
50 100 150 200 25 50 75 100 100 200 300 400 0 50 100 150

Test L Test M Test N Test O


0 0 0
0.01
0 −0.0025
−0.01 −0.05
−0.01 −0.0050
−0.02 −0.10
−0.02
−0.0075
−0.03
−0.15
100 200 300 0 20 40 60 0 25 50 75 20 40 60 80
Test Duration (minutes)

Fig. 9—Inflow tests—evolution of trend line gradients over time (band showing upper and lower bounds).

This raises an important question, and one which is not generally considered when using Horner plots—at what rate does flow rate
need to be decreasing for an inflow test to be deemed successful? In the case of Test B, a sharp decrease, which if continued, would
have led to zero flow in a short time, in the course of an hour had changed to constant flow. Any answer to this question must be quali-
fied by the caveat that any test must be carried out for a sufficient time for any trend to fully develop.

Distinguishing Constant Flow from Decreasing Flow. One approach to addressing this issue is to make an estimation of the time
until flow will approach zero. Because this calculation is based on a linear approximation, it will only be indicative. However, it will
provide a relative measure of whether any decreasing trend that is observed during a test is reasonable. A gradient that approaches zero
will result in this estimated time tending toward 1. This provides an advantage over the plot of gradient itself because the asymptotic
behavior is visually easier to interpret on a plot. Fig. 10 demonstrates this concept.

Test A Test B Test B (truncated) Test C


1,000

500

−500

−1,000
40 80 120 160 20 40 60 10 20 30 40 50 25 50 75 100

Test D Test E Test F Test G


1,000
Forecast Time to Zero Flow (minutes)

500

−500

−1,000
10 20 30 40 50 0 50 100 150 200 20 40 60 20 40 60

Test H Test I Test J Test K


1,000

500

−500

−1,000
50 100 150 200 25 50 75 100 100 200 300 400 0 50 100 150

Test L Test M Test N Test O


1,000

500

−500

−1,000
100 200 300 0 20 40 60 0 25 50 75 20 40 60 80
Test Duration (minutes)

Fig. 10—Inflow tests—linear estimate of time reach zero flow (band showing upper and lower bounds).

September 2021 SPE Drilling & Completion 537

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 538 Total Pages: 13

By reviewing Fig. 10, we can make a judgment of whether a test appears to be stable and whether a trend is decreasing at a reasona-
ble rate. Tests such as Test O are stable and also suggest flow will reduce toward zero in a reasonable period of time. By reasonable, we
suggest a duration of a similar order of magnitude as the test duration. A test such as Test L may be a basis on which to claim a trend is
decreasing at a reasonable rate.
Conversely, tests such as Test E demonstrate a gradient that appears to be oscillating between slightly negative and slight positive
values. The estimated time to approach zero shows diverging behavior, fluctuating between increasingly positive and negative values.
Negative values in this context simply mean a trend line that is actually increasing. Other tests such as Test C and Test H also exhibit
similar behavior.
In the case of Test B (truncated), we can see the estimated time to approach zero flow is continuously increasing from the beginning
of the test. This is a useful indication that the trend has not yet stabilized, and testing should continue until a consistent trend
becomes clear.

Application
A trial of the proposed methodology was conducted with a web application developed for this purpose. In this case, the new methodol-
ogy proved useful as the Horner plot produced an ambiguous result. With t0 ¼ 660 minutes, the trend line had a positive y-intercept

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
(see Fig. 11), while with t0 ¼ 1; 500 minutes, the y-intercept was negative.
The output of the application is shown in Fig. 12:

y = 0.101 + 0.0617 x

0.9
Flow Rate (L/min)

0.6

0.3

0
0 2 4 6
Log Horner Time

Fig. 11—Horner plot with instantaneous flow rate (gray), moving average (black), and trend line (blue) (t0 5 660 minutes).

Test duration (minutes) Rate of change of flow rate (mL.min–2), Flow rate vs. time
95% upper confidence interval

328 0.9
0 500 –0.21
Flow Rate (L/min)

–20 20

0.6

Time to approach zero flow (upper Volume (bbl)


bound)
0.3

0.45 0.0
855 0 2 0 100 200 300
0 3.3k
Time (minutes)

Line approximation of time to approach zero flow Evolution of trend line gradient over time
Time to Approach Zero Flow (minutes)

1,000 0.001
Gradient (L.min–2)

500 0

0 –0.001

–500 –0.002

–1,000 –0.003
0 100 200 300 400 500 0 100 200 300 400 500
Test Duration (minutes) Test Duration (minutes)

Fig. 12—Inflow test analysis.

538 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 539 Total Pages: 13

• Top left: A summary of key information such as current test duration, the gradient of the trend line, time to approach zero flow,
and total volume returned.
• Top right: A plot of observed flow rate, showing moving average (yellow) and trend lines.
• Bottom left: A plot of estimated time to approach zero flow. This shows a green area indicating that increasing test duration
increases the acceptable time to approach zero flow.
• Bottom right: A plot displaying development of the trend line gradient during the course of the test. This clearly shows the trend
had stabilized after approximately 180 minutes at a negative (decreasing) value.
The additional information provided by the application provided the confidence to make the assessment that flow rate was decreas-
ing. More importantly, without first developing an understanding of the theory behind Horner plots, there would likely have been reluc-
tance to accept the result, leading to a longer test that would never have yielded an acceptable Horner plot.
The authors provide this as an example of their attempts to address the problems identified in this paper. The advantage of this study
is to present a set of real data to help readers relate in a practical way to the theoretical conclusions that we reached. We recognize that
others will have different experiences of Horner analysis, and the application of new data and alternative approaches may yield valuable
insights and improvements on this solution.

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Conclusions
Although industry guidelines support the use of Horner analysis when conducting inflow tests, the lack of literature on the application
of this technique means there is a gap between practical application and theory. We believe the main contribution of this study is to lift
the lid on the black box that is Horner analysis to reveal why the method is not appropriate for inflow tests.
• The mathematical trick in Horner’s technique to extrapolate t ! 1 is due to the principle of superposition, introducing the term t0.
There is no clear way to determine an appropriate value of t0 in the context of an inflow test, and this may often be set to an arbitrary
value. Because the y-intercept of a line on a Horner plot is dependent on t0, Horner analysis is subjective. Although in some cases,
inflow test Horner analysis may have been harmless, in other situations the result may have been invisible lost time due to a Horner
plot that produced a positive y-intercept despite the decreasing trend.
• We have shown that data must be linear to be accurately predicted by Horner’s equations. As predicted by theory and supported by
empirical evidence, the radial equation and boundary conditions, from which Horner’s equation is derived, do not accurately account
for volumetric expansion caused by thermal diffusion in a well, leading to nonlinearity. Extrapolating a line on a Horner plot is to
t0 
assert that future observations will follow a function of logn 1 þ . When data are nonlinear, there is no sound theoretical justifica-
t
tion to make this assertion.
• In an inflow test, the phenomenon of flowback has more than one potential cause—thermal expansion and a leak. The consequence
of non-linearity in inflow test Horner analysis is not harmlessly underestimating flowback from thermal expansion. Instead, flowback
may be incorrectly attributed to thermal expansion rather than a leak from a failed barrier. Because the technique is without theoreti-
cal justification and does not produce satisfactory results, Horner analysis should not be used to determine if flowback is attributable
to thermal expansion or a leak.
• A solution is proposed to apply a linear model to observations of flow rate and check if the estimated gradient of a trend line, includ-
ing the upper bound of the estimate, is consistently negative (decreasing) for a reasonable period of time. In addition to checking for
a negative gradient it is also suggested that making a prediction of the time to reach zero flow based on the observed trend can assist
decision makers. The solution was tested with data from offshore wells and provides decision makers with a more transparent, com-
prehensible, and robust method by which to make an assessment.

Nomenclature
c ¼ isothermal compressiblity, Pa1
cp ¼ specific heat capacity, J/kgK
k ¼ thermal conductivity, W/mK
kp ¼ permeability, darcies
p ¼ pressure, Pa
q_ ¼ heat generation, W/m3
r ¼ radius, m
t ¼ time, seconds
t0 ¼ Horner time parameter
T ¼ temperature,  K
z ¼ vertical axis, m k
a ¼ thermal diffusivity a ¼ , m2/s
qcp
c ¼ Euler-Mascheroni  0.5772
Dt ¼ duration of inflow test after time t0
l ¼ viscosity, Pas
q ¼ density, kg/m3
/ ¼ azimuth, degrees
/p ¼ porosity, fraction

Acknowledgments
The authors would like to acknowledge Stefan Revets, who provided valuable contributions to this paper.

References
API RP 96, Deepwater Well Design and Construction. 2013. Washington, DC, USA: API.
Dake, L. P. 1998. Fundamentals of Reservoir Engineering, Developments in Petroleum Science, 17th edition, Vol. 8. Amsterdam, The Netherlands:
Elsevier Science.

September 2021 SPE Drilling & Completion 539

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 540 Total Pages: 13

Dowdle, W. L. and Cobb, W. M. 1975. Static Formation Temperature from Well Logs—An Empirical Method. J Pet Technol 27 (11): 1326–1330. SPE-
5036-PA. https://doi.org/10.2118/5036-PA.
Horner, D. R. 1951. Pressure Build-Up in Wells, Vol. II, 503. Leiden, The Netherlands: E. J. Brill.
James, G., Witten, D., Hastie, T. et al. 2018. An Introduction to Statistical Learning, with Applications in R, eighth edition. New York, New York, USA:
Springer ScienceþBusiness Media.
Lienhard, J. H. IV and Lienhard, J. H. V. 2019. A Heat Transfer Textbook, fifth edition. Cambridge, Massachusetts, USA: Phlogiston Press. http://
ahtt.mit.edu.
Matthews, C. S. and Russell, D. G. 1967. Pressure Buildup and Flow Tests in Wells, Vol. 1. Richardson, Texas, USA: Society of Petroleum Engineers.
Muskat, M. 1946. The Flow of Homogeneous Fluids through Porous Media, first edition. New York, New York, USA: McGraw-Hill.
Timko, D. J. and Fertl, W. H. 1972. How Downhole Temperatures, Pressures Affect Drilling. World Oil 175 (1): 45–50.
UK Oil & Gas Well Life Cycle Integrity Guidelines, Issue 4. 2019. London, UK: Oil & Gas UK.

Appendix A—Derivation
To solve an equation in the form of Eq. 3, Dake (1998) showed a Boltzmann’s transformation can be applied.

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
qcp r 2
s¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-1Þ
4kt
@s qcp r 2 s
¼ ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-2Þ
@t 4kt2 t
@s qcp r 2s
¼ ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-3Þ
@r 2kt r

Substituting Eqs. A-1, A-2, and A-3 into Eq. 3 gives


 
1@ @T @s @s qcp @T @s
r ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-4Þ
r @s @s @r @r k @s @t
   
1@ @T qcp r qcp r qcp @T qcp r 2
r ¼  : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-5Þ
r @s @s 2kt 2kt k @s 4kt2
 
1 qcp r @ qcp r 2 @T q2 c2p r 4 4 @T
¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-6Þ
r 2kt @s 2kt @s 16k2 t2 r 2 @s
 
1 2s @ @T q2 c2p r 4 4 @T
2s ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-7Þ
r r @s @s 16k2 t2 r 2 @s
 
2s @ @T 4 @T
2s ¼ s2 2 : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-8Þ
r 2 @s @s r @s
 
@ @T @T
2s ¼ 2s : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-9Þ
@s @s @s
@s @T @2T @T
þ s 2 ¼ s : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-10Þ
@s @s @s @s
Finally,
@T @2T @T
þ s 2 ¼ s : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-11Þ
@s @s @s
As expected, Eq. A-11 is identical in form to the pressure buildup formula (see Dake 1998, p. 151).
@T
Let T 0 ¼ ,
@s
@T 0
T0 þ s ¼ sT 0 : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-12Þ
@s
@T 0 ðs þ 1ÞT 0
¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-13Þ
@s s
Integrating Eq. A-13:

es
T0 ¼ C : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-14Þ
s
To solve for the constant C, boundary condition 3 is applied. Integrating Eq. A-13:
@T Q @T @s @T
lim r ¼ ¼r ¼ 2s : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-15Þ
r!0 @r 2pk @s @r @s
@T es
2s ¼ 2sT 0 ¼ 2sC : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-16Þ
@s s

540 September 2021 SPE Drilling & Completion

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047


DOI: 10.2118/204479-PA Date: 20-August-21 Stage: Page: 541 Total Pages: 13

As r ! 0 so does s ! 0.
Q
C¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-17Þ
4pk
Therefore,
@T Q es
¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-18Þ
@s 4pk s
qcp r 2
Eq. A-18 can now be integrated between the limits t ¼ 0 (s ! 1) and the current value of t, for which s ¼ , Ti (initial tempera-
ture) and the current temperature t , 4kt
0
2
ðT ð qcp r
Q 4kt es
dT ¼ ds: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ðA-19Þ
Ti 4pk 1 s
ð1

Downloaded from http://onepetro.org/DC/article-pdf/36/03/529/2484050/spe-204479-pa.pdf/1 by Norwegian University of Science & Technology user on 21 February 2022
Q es
T ¼ Ti  qcp r 2
ds: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-20Þ
4pk s
4kt

Eq. A-20 is the line source solution of the diffusivity equation giving temperature as a function of radius from well center and time for
a plane in the subsurface. The equation is of the same form as given by Dake (1998, p. 152) for pressure buildup.
The integral is a special integral called the exponential integral.
Q
T ¼ Ti  EiðsÞ: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-21Þ
4pk
Horner (1951) makes use of the fact Eið xÞ  logn x  0:5772 ¼ logn cx, when x < 0.01.
The value 0.5772 is the Euler–Mascheroni constant (c). For the approximation to be accurate, x < 0.01, which implies
0:01qcp r 2
t> . As we are inspecting the well, the value of r is small (r ¼ rw ), and values of t will typically be large enough for this
4k
approximation to be valid.
Q
T ¼ Ti þ log ðcsÞ: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-22Þ
4pk n
 
Q cqcp rw2
T ¼ Ti þ log : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-23Þ
4pk n 4kt
0:01qcp rw2
Eq. A-23 is now in a simple form that does not require integration to solve, providing t > .
4k
Rearranging the equation, we obtain the temperature drop Ti  T, which occurs over a period t due to heat transfer Q.
 
Q cqcp rw2
Ti  T ¼  log : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-24Þ
4pk n 4kt
If the well is then left static for period Dt, after being cooled for time t0, the temperature drop can be obtained by principle of superposi-
tion (adapted from Matthews and Russell 1967, p. 18). Ti  T ¼ (temperature drop caused by rate Q for time t0 þ Dt) þ (temperature
drop caused by rate change -Q for time Dt).
Refer to Fig. 7.7 in Dake (1998, p. 171), which is illustrative of how superposition works in this context.
In the context of inflow tests, the first term in the superposition is the period during which the well is disturbed (cooled) during oper-
ations, whereas the second term is the inflow test, in which temperature tends toward equilibrium.
   
Q cqcp rw2 Q cqcp rw2
Ti  T ¼  logn þ logn ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-25Þ
4pk 4kðt0 þ DtÞ 4pk 4kDt
 
Q Dt
Ti þ T ¼ logn ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-26Þ
4pk t0 þ Dt
 
Q Dt
T ¼ Ti þ logn ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-27Þ
4pk t0 þ Dt
or
 
Q t0 þ Dt
T ¼ Ti  log : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ðA-28Þ
4pk n Dt

September 2021 SPE Drilling & Completion 541

ID: jaganm Time: 11:55 I Path: S:/DC##/Vol03603/200047/Comp/APPFile/SA-SPE-DC##200047

You might also like