See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266876326

ANALYSIS OF WELL TESTS IN AFYON ÖMER-GECEK GEOTHERMAL FIELD,

TURKEY

Article

CITATIONS READS

3 676

5 authors, including:

Mustafa Onur Murat Cinar

University of Tulsa Istanbul Technical University
175 PUBLICATIONS 1,112 CITATIONS 33 PUBLICATIONS 380 CITATIONS

SEE PROFILE SEE PROFILE

Niyazi Aksoy Umran Serpen

Dokuz Eylul University Istanbul Technical University
33 PUBLICATIONS 447 CITATIONS 77 PUBLICATIONS 453 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development of 2D (r-z) Fluid & Heat Transport Model for Single-Phase Water Geothermal Reservoirs & History Matching of Pressure and Temperature Transient
Measurements by Gradient-Based & Ensemble Kalman Filter Optimization Methods (TUBITAK Project) View project

Ground Source Heat Pumps View project

All content following this page was uploaded by Umran Serpen on 23 October 2014.

The user has requested enhancement of the downloaded file.

PROCEEDINGS, Thirty-Second Workshop on Geothermal Reservoir Engineering
Stanford University, Stanford, California, January 22-24, 2007
SGP-TR-183

ANALYSIS OF WELL TESTS IN AFYON ÖMER-GECEK GEOTHERMAL FIELD, TURKEY

Mustafa Onur*, Murat Cinar*, Niyazi Aksoy+, Umran Serpen*, and Abdurrahman Satman*
*
Istanbul Technical University, 34469 Maslak, Istanbul, Turkey, +Dokuz Eylul University, 35210 Alsancak, Izmir,
Turkey
E-mails: [email protected], [email protected], [email protected], [email protected], [email protected]

The Ömer-Gecek geothermal system is a convective

ABSTRACT hydrothermal type commonly occur in areas of active
geological faulting and folding, and areas where the
In this work, analyses of various types of pressure regional heat flow is above normal, as in much of the
transient tests (such as multi-rate tests, conventional western Turkey. As for the geology of the system,
drawdown/buildup tests, and interference tests) mica schist and marbles of Paleozoic age forms the
conducted in the Afyon Ömer-Gecek geothermal basement of the field. At the same time, these rocks
field, Turkey, are presented. The pressure transient form the reservoir system. Neogene deposits
tests were conducted at six wells. The pressure data composed of conglomerate, sandstone, clayey
were acquired by downhole quartz gauges, and thus, limestone-sandstone, and volcanic glass-
amenable to the applications of modern well-test trachandesitic tuff unconformably overlie the
analysis techniques such as derivative and Paleozoic basement. A companion paper (Satman et
deconvolution. Deconvolution analysis based on al., 2007) provides further details about the geology,
recently proposed robust algorithms was found useful well depths, well temperatures, geochemical analysis
to extract more information from the variable-rate of the geothermal water.
well tests conducted in the field. In general, the
pressure data analyzed indicate that the wells’
productivities are quite high, but influenced by non-
Darcy flow effects and are producing in a complex
fractured/faulted network system. The estimated
values of permeability-thickness products (kh) from
buildup and interference tests range from 40 to 2000
Darcy-m, whereas porosity-compressibility-thickness
products (φcth) estimated from the interference tests
range from 2.91x10-4 to 1.06x10-2 psi/m.

INTRODUCTION
Located in the central Aegean region of Turkey and
15 km northeast of the city of Afyon (Figure 1), the
Ömer-Gecek geothermal field is one of the important
geothermal fields in Turkey.

The geothermal system can be classified as a low Figure 1. Location map of the Ömer-Gecek
temperature, single-phase liquid-dominated one geothermal field.
containing geothermal water (having salinity of 4 000
to 6 000 ppm and dissolved CO2 content about 0.4% Although the field was explored in 1960s, not much
by weight) with temperatures ranging from 50 to quantitative information on reservoir characteristics
111.6oC. The wells (nearly 30) drilled in the field (permeability, fault/fracture networks, wells’ IPRs
range in depth from 56.8 to 902 m. The total etc.), which is essential for understanding and
production rate from the field is about 236 kg/s and modeling the production performance of the wells
the geothermal water produced has been utilized to and the field, was available. To acquire such
support a district heating system with a capacity of information, pressure transient tests were designed
approximately 4500 residences and some health spa and conducted in the field in 2004. Well tests were
facilities since 1996 (Satman et al., 2007). conducted at six wells; AF-10, AF-11, AF-16, AF-20,
AF-21, and R-260. Further information regarding
these wells (location, depth, temperature of the measurements were taken at the surface using a weir.
geothermal fluid produced, etc) are given by Satman In this example, the measured initial pressure is about
et al. (2007). 279.75 psi. Although not shown here, temperatures
were also recorded at 195 m during the entire test
The objective of this work is to determine wells’ sequence, and the temperature was nearly constant at
productivities, estimation of permeability-thickness 107.8 oC.
and porosity-compressibility-thickness products, as
well as to determine reservoir characteristics (single Figure 3 shows the IPR curve obtained from the
layer, multi-layer, double porosity, etc.) and reservoir multi-rate test conducted at the well AF-21. The IPR
boundaries (faults and their flow characteristics) by curve fitted through measured pressure drop data is
the analyses of pressure transient tests conducted in described best by the steady-state “turbulent” flow
the field. model (Eq. 1). The second term bq2 in the right-hand
side of equation is due to non-Darcy flow. The non-
ANALYSIS OF MULTI-RATE TESTS Darcy effect observed on the IPR curve of the well
AF-21 (as well as on those of other wells tested) is
Here, we summarize the results obtained from the possibly due to a high permeability fracture network
analyses of multi-rate tests conducted at the wells system intersecting the well. Because of this, flow
AF-11, AF-16, AF-20, and AF-21. Multi-rate tests rate near the wellbore is so high that the flow regime
are designed to construct the inflow performance becomes “turbulent” in the vicinity of the wellbore.
relationship (IPR) of those wells as well as to Thus Darcy’s law loses its validity, and hence the bq2
determine reservoir parameters and characteristics term in the right-hand side of Eq. 1 becomes
from the pressure signal recorded by using important on well deliverability.
conventional as well as modern well-test analyses
techniques based on recently proposed deconvolution
algorithms by von Schroeter et al. (2004) and Levitan ∆p = aq + bq 2 , (1)
(2005). Only pressure/rate data for the multi-rate test
of the well AF-21 and its analysis will be presented
0
here because the multi-rate tests conducted at the
other wells give similar behavior to that of AF-21.

The multi-rate test of AF-21 consists of a well-test 4

sequence (4 distinct step rate changes and one shut-in

period) acquired over a 16-hour test. Figure 2
Pressure drop, ∆p (psi)

presents the pressure and rate data for this well-test 8

sequence (note that a single pressure buildup profile
of about 4 hr is acquired at the end of the testing
sequence). 12

281 160

280
140 16
279 Measured
120 Polynomial fit (Eq. 1)
278
100 20
Flow rate, lt/s
Pressure, psi

277
0 20 40 60 80
276 80 Flow rate, q (lt/s)
Pressure
275 Flow rate
60 Figure 3. IPR curve for well AF-21, obtained from
274 multi-rate test data shown in Fig. 2.
40
273
Start of test
20 a and b parameters estimated from the multi-rate test
272
for the well AF-21 and for the wells AF-11, AF-20,
271
0 2 4 6 8 10 12 14 16 18 20 22 24
0 and AF-21 are given in Table 1. IPR curves
Time (h) constructed for these wells are compared in Figure 4.
From Figure 4, it can be seen that IPR curves for all
Figure 2. Pressure and rate data for the multi-rate four wells tested indicate a non-Darcy flow model
test conducted at the well AF-21. represented Eq. 1 and that the well AF-21 is the most
productive amongst four wells.
The pressure measurements shown in Fig. 2 were
taken from a down-hole quartz gauge set a depth of
195 m (total depth of the well is 210 m), and rate
 Table 1. Parameters of IPR curves for Wells AF-11, constant value. Moreover, as the flow rate increases,
AF-16, AF-20, and AF-21, determined the displacement becomes larger. This is possibly due
from multi-rate tests. to rate-dependent skin due to non-Darcy flow.
Although not shown here, we have also looked at the
Well a b pi (psi) Bourdet derivatives (Bourdet et al., 1989) for each
2 flow period, these derivative signals indicate
Name psi/(lt/s) psi/(lt/s) @ depth (m) changing (and/different) wellbore storage effects
AF-11 0.0985 0.00418 147.9 @ 107 m (possibly due to non-isothermal/multiphase flow
inside the wellbore) at early times of each flow
AF-16 0.0408 0.00276 236.4 @ 174 m
period.
AF-20 0.0249 0.00265 131.8 @ 98 m
Production period no. 1 (qn-qn-1 = 8.35 lt/s, qn+qn-1 = 8.35 lt/s)
AF-21 0.0120 0.00321 279.8 @ 195 m Pruduction period no. 2 (qn-qn-1 = 16.65 lt/s, qn+qn-1 = 33.35 lt/s)
Production period no. 3 (qn-qn-1 = 13 lt/s, qn+qn-1 = 63 lt/s)
Production period no. 4 (qn-qn-1 = 3 lt/s, qn+qn-1 = 79 lt/s)
Buildup period (qn-qn-1 = 41 lt/s, qn+qn-1 = 41 lt/s)

AF-11
1
AF-16
AF-20
AF-21
Polynomial Fits (based on Eq. 1)

[p wf(tn )-p wf(tn-1 )]/[q(tn )-q(tn-1 )], psi/(lt/s)

300

0.1
270
Flowing bottomhole pressure, psi

240

210
0.01

180

150

0.001
120 0.001 0.01 0.1 1 10
Time (h)

90
0 20 40 60 80 100 Figure 5. Log-log plot of normalized pressure
Flow rate, q (lt/s) changes vs. elapsed time for each flow
period of the multi-rate test of the well
Figure 4. IPR curves for the wells AF-11, 16, 20 AF-21.
and 21, constructed from multi-rate tests.
To eliminate the multi-rate effects and convert the
Although IPR curves and IPR parameters a and b are multi-rate data into an equivalent unit-rate constant
useful for understanding the deliverability of wells, drawdown response, we apply a robust deconvolution
and for tubular design purposes, unfortunately, the algorithm developed by Cinar et al. (2006) by
reservoir parameters such as permeability-thickness accounting for all flow rate history (Fig. 2). As is
product and skin as well as information about known (Bourdet, 2002), conventional drawdown or
reservoir characteristics and boundaries cannot be buildup analysis based on superposition-time
derived from the “lumped” parameters a and b of IPR transform does not completely remove all effects of
curves. To derive such information, one must analyze previous rate variations and often complicates test
the pressure signal, particularly, recorded during the analysis due to residual superposition effects.
buildup period of multi-rate tests.
It is worth noting that our deconvolution algorithm
Therefore, next, we analyze the pressure signal used here are based on the ideas presented by von
recorded during multi-rate tests. Here, we will Schroeter et al. (2004) and Levitan (2005), and is
present a detailed analysis only for the multi-rate test based on minimization of a weighted least-squares
of the well AF-21. Figure 5 shows log-log plots of (LS) objective function given by
conventional rate normalized multi-rate pressure
change vs. elapsed time for each flow period. This
graph clearly shows that pressure change data for
each flow period are more or less displaced by a
 2
1 p ⎡ p0 − pm,i − ∆pmodel (ti ) ⎤
N we will not treat flow rate data as unknown, and thus
O ( z1 , z2 ,..., z N , q1u , q2u ,..., qNu r , p0 ) = ∑ ⎢
2 i =1 ⎢⎣ σ p ,i
⎥ the second summation term in the objective function
⎥⎦
2 (Eq. 2) is deleted. In Eq. 2, σ c , k represents the
1 Nr ⎡ qm , j − q j ⎤
u

+ ∑ ⎢
2 j =1 ⎢⎣ σ q , j ⎥⎦
⎥ “standard deviation” of the curvature constraint κ c , k .
1 N −1 ⎡ κ c,k ⎤
2 As suggested by von Schroeter et al. (2004) and
+ ∑⎢ ⎥
2 k =1 ⎢⎣ σ c,k ⎥⎦ Levitan (2005), we set σ c , k = σ c for all k, equal to
(2) one constant value. We have often found that
σ c = 0.05 works well. This value has been chosen to
Here, O represents the weighted LS objective provide small degree of regularization and at the
function, and the parameters listed in the left-hand same time not to over constrain the problem and
side of Eq. 2. Note that the model parameters are the create significant bias. In all our applications given
response function z , the rate q , and the initial here, we use σ c = 0.05 .
pressure p0 . Here z is equal to the natural logarithm
of the Bourdet derivative of unit-rate drawdown Given the issues of inconsistent data set with the
response, that is, z = ln[dpu(t)/dlnt] (where pu(t) deconvolution model as discussed by Levitan (2005),
represents drawdown pressure drop if the well were Levitan et al. (2006), and Cinar et al. (2006), we will
produced at constant unit-rate; see von Schroeter et derive deconvolved constant-unit-rate response using
al., 2004). The rate q plays dual role. It can be only the pressures of the buildup period — and we
treated as one of model parameters. It is also the part will assume the rate history is accurate. Our hope is
of the data that must be fitted to the model. that for this multi-rate test, deconvolution will allow
us to better identify the underlying unknown
In Eq. 2, N represents the total number of nodes at reservoir model and extract more information from
which the z-responses to be computed, Np represents the test sequence than that based on the conventional
the total number of measured pressure points to be analysis of buildup data based the multi-rate (radial
history matched, and Nr is the total number of flow) superposition time (Agarwal, 1980). The
measured (or allocated) flow rate steps to be treated deconvolved pressure drop and derivative responses
as unknown in history matching process. It is worth obtained from our deconvolution algorithm using
noting that the objective function considered is quite three different initial reservoir pressure estimates of
general because it allows one to perform 279.75, 279.9 and 280 psi are shown in Fig. 6, and
simultaneous estimation of z responses at each node the deconvolution results are compared to the
and the initial reservoir pressure p0, as well as any conventional pressure buildup derivatives based the
flow rate steps in the rate sequence. In all multi-rate (radial flow) superposition time (Agarwal,
applications given in this paper, we use N = 70. 1980) plotted versus shut-in time.

In Eq. 2, σ p ,i represents the standard deviation of

10
Unit-rate deconvolved and rate-normalized buildup responses, psi/(lt/s)

deconvolved unit-rate response (pi = 280 psi)

error in measured pressure pm,i at time ti. Typically, in deconvolved unit-rate response (pi = 279.9 psi)
applications, we can assume identically distributed 1
deconvolved unit-rate response (pi = 279.75 psi)

normal errors with zero mean and the same specified conventional rate-normalized buildup response

standard deviation for each measured pressure point;

that is, σ p ,i = σ p ; for i = 1, 2,L , N p . Levitan 0.1
(2005) suggests using σ p ,i = σ p = 0.01 psi as
default value. However, as shown later, if pressure
data contain a higher level of noise, then one may 0.01

need to consider values of σ p greater than 0.01 psi to

obtain a smoother deconvolved unit-rate response. It
0.001
should be worth noting that if we have noisy data and
we require to honor each pressure point by assigning
small standard deviation (or equivalently a higher
0.0001
weight) to every pressure point in the objective 0.001 0.01 0.1 1 10 100
function, we may end up with noisy deconvolved Time (h)

unit-rate drawdown response function. Figure 6. Comparison of deconvolved responses

derived from buildup pressures with the
Similarly, σ q , j represents the standard deviation of conventional pressure buildup pressure
change and its Bourdet derivative
error in measured (or allocated) rate step qm,j to be (normalized by the last rate prior to
treated as unknown. In our applications given here, buildup) for the well AF-21.
The deconvolved responses give an indication of a is acceptable due to possible errors in flow rate data
partially penetrating geothermal well (-1/2 slope line for those flow periods. From Table 2, note that
in the interval from 0.07 to 0.2 hr) producing near a permeability-thickness is quite high, and skin factor
highly conductive fault (-1 slope line in the time is highly negative, indicating a highly permeable
interval from 0.3 to 1 h). The late time behavior (after fracture or fault zone intersecting the wellbore.
1 h) indicates, however, completely three different
flow regimes based on the three different values of
initial pressures differing by only 0.25 (max) psi; the 295

one based on 280 psi shows 1/2 slope line indicating 290 measured
a channel or parallel faults, while the one based on model
279.9 shows almost 1/4 slope line, indicating a finite 285

conductivity fault (Abbaszadeh, 1995) and the one 280

based on 279.75 psi shows -1 slope line indicating a

Pressure, psi
constant-pressure (or infinite-conductivity) fault. 275

270
It is also interesting to note that the conventional rate-
normalized buildup pressure change and its 265

derivative data seem to agree better with the 260

deconvolved responses (based on the initial reservoir
pressure estimate of 279.9 psi) until 4 h — which is 255

the total duration of the buildup period. Nevertheless, 250

for sure, we cannot determine the correct boundary 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (h)
model based on the deconvolved late time data for
this example in high confidence due to sensitivity of
Figure 7. Model match of measured pressure data
deconvolved late-time derivative response (after 4 h)
recorded during entire multi-rate test
to the initial reservoir pressure. In addition,
sequence of the well AF-21.
unfortunately, the geological data available also did
not help us to reject one of these boundary models. If
we had designed this multi-rate test sequence to
Table 2. Some input and estimated model parameters
include one or more buildup periods, then we would
for obtaining the model match shown in
have had the chance to check the validity of the
Figure 7.
initial pressure by a trial-and-error procedure as
suggested by Levitan et al. (2006) from
deconvolution, and to determine the most appropriate Model Parameters
boundary model for the well AF-21.
khh (Darcy-m) 308
This example clearly demonstrates how sensitive
kv/kh (anisotropy ratio) 0.11
deconvolution could be to the initial pressure at late
time where we actually desire to extract more pi, psi 279.72
information about the reservoir model and boundaries
than conventional buildup analysis, and emphasize S (skin, dimensionless) -5.2
that when designing tests one should consider at least D, (lt/s) -1
(non-Darcy 6.5x10-3
two buildup periods in the test sequence to determine coefficient)
the correct initial reservoir pressure. hw/h (penetration ratio)length 0.385
of open interval)
Next, we perform parameter estimation for hw, m (length of the open
determining permeability-thickness product (kh), skin 80.8
interval)
(s), non-Darcy coefficient (D), and distance to a fault zw, m (reservoir thickness)
(df) by using a “simple” well/reservoir model based 210
on a partially-penetrating (top of the formation is df, m (distance to the fault) 177
open) well near a constant pressure fault in a
homogeneous anisotropic reservoir (i.e., the model µ , [email protected] C o
0.27
indicated by green curves on Fig. 6). rw, m 0.108
Figure 7 presents a model match of the measured φcth (m/psi) 2.35x10-4
pressure data recorded for the entire multi-rate test
sequence. The estimated model parameters are The permeability-thickness (kh), mechanical skin (S),
summarized in Table 2. Note that the match is not and non-Darcy coefficient (D) values estimated from
perfect, particularly, for the second and third flow
periods. Nevertheless, in general, the match obtained
multi-rate tests conducted at the wells AF-11, AF-16, diameter. As shown in Fig. 9, the first drawdown
AF-20, and AF-21 are summarized in Table 3. period is nearly ten hours at constant production rate
of 33.7 lt/s. The duration of the following pressure
Table 3. kh, S, and D values estimated from multi- buildup (PBU) period is 6 hours. After the PBU
rate tests conducted at the wells AF-11, period, there is another 4-hour flow period with the
AF-16, AF-20, and AF-21. same flow rate of the first flow period. The total test
duration is about 20 hours.
Well kh S D
170 100
Name (Darcy-m) dimensionless (lt/s)-1 Pressure @ 115 m
Flow rate
AF-11 201 -3.3 8.26x10-2 80

Bottomhole pressure, psi

160
-1
AF-16 665 -1.1 1.92x10

Flow rate, lt/s

60
AF-20 1085 -4.3 4.32x10-1 150

AF-21 308 -5.2 6.5x10-3 40

140
20

ANALYSIS OF DRAWDOWN/BUILDUP TESTS

130 0
Here, we summarize the results obtained from the 0 2 4 6 8 10 12 14 16 18 20 22
Time (h)
analyses of conventional drawdown/buildup tests
conducted at the wells R-260, AF-11, AF-16, AF-20, Figure 9. Pressure and rate data for the
and AF-21. A schematic view of the tested well drawdown/buildup test conducted at the
locations together with the other wells in the field, well R-260.
with possible faults given by the geological model, is
shown in Figure 8. The red colors in Fig. 8 show the The deconvolved pressure drop (based on PBU data
highest elevations (above sea level), whereas dark alone) and derivative responses obtained from our
blue colors show the lowest elevations. deconvolution algorithm are shown in Fig. 10, and
the deconvolution results are compared to the
The drawdown/buildup tests are designed to conventional pressure buildup derivatives based the
determine kh, skin factor, as well as reservoir conventional (radial flow) Agarwal’s superposition
characteristics and boundaries if possible. Only the time plotted versus shut-in time.
drawdown/buildup test of the well R-260 and its
analysis will be presented in detail here. 1
Unit-rate deconvolved and rate-normalized buildup responses, psi/(lt/s)

1/4 slope (bilinear flow?)

Fault-I

0.1
Fault-II
Fault-IV
radial 2 (?)
Fault-III
radial 1 (?)

0.01
+ 1 slope
Well where well tests (wellbore storage)
are conducted
-1 slope (infinite-conductivity fault)

0.001
conventional rate-normalized buildup pressure change
conventional rate-normalized buildup pressure-derivative
Figure 8. A schematic view showing the wells where deconvolved unit-rate responses (with noise level of 0.07 psi)
well tests are conducted, and possible deconvolved unit-rate responses (with noise level of 0.01 psi)
faults in the field.
0.0001
0.0001 0.001 0.01 0.1 1 10 100
Elapsed time (h)
Figure 9 presents the pressure/rate recorded during
Figure 10. Comparison of deconvolved responses
the drawdown/buildup test at the well R-260. The
derived from buildup pressures with the
pressure data were measured at a depth of 115 m with
conventional pressure buildup pressure
a down-hole quartz gauge. The temperature of the
change and its Bourdet derivative
fluid recorded at this depth is around 103.6 oC. The
(normalized by the last rate prior to
total depth of the well is 166 m. The open interval is
buildup) for the well R-260.
from 100 m to 166 m, with 8 and 1/2 inches wellbore
In deconvolution, we considered two different values In short, we considered three different plausible
of noise level ( σ p = 0.07 and 0.01 psi in Eq. 2) for models to history match the full pressure history
PBU pressures to investigate the effect of noise level shown in Fig. 9. Model 1 refers to a model with two
on deconvolution (to be discussed below). intersecting faults with a right angle (one is a no-
flow, and the other is a constant-pressure). Model 2
The upward trend observed in conventional rate- refers to a model with a single constant-pressure
normalized buildup-derivative data (blue data points fault, whereas Model 3 refers to a model with a finite
in Fig. 10) near the end of buildup period is due to conductivity fracture intersecting the well located
the right-hand side smoothing effect associated by near a single-constant pressure fault. The “best”
using Bourdet et al. (1989) smoothing method with a match (based on rms values obtained for the matches,
smoothing parameter L = 0.75. Hence, this upward and confidence intervals for parameters) was
trend should not be attributed to the reservoir obtained with Model 1.
boundary effects. Deconvolved unit-rate responses
(red and green curves in Fig. 10) provide a 14-hour Figure 11 presents a model match (based on Model 1)
longer data set than conventional rate-normalized of the measured pressure data recorded for the entire
buildup responses and identify a well-defined -1 test sequence. The estimated model parameters are
slope line for almost one-and-a-half log-cycle near summarized in Table 4. As can be seen from Fig. 11,
the end of the data, indicating an infinite conductivity we have almost a perfect match of measured
(or constant-pressure) fault near the well. pressure. A highly negative skin factor gives an
indication of a highly permeable fracture/fault
The flow regimes indicated by deconvolved unit-rate network intersecting the well.
derivative data in the time interval from 0.0003 to 1 h 170 100
are not very conclusive. It seems that deconvolved Measure pressure
Calculated pressure
unit-rate responses for this time period may indicate (based on Model 2)
Rate 80
different flow regimes depending on the noise level
Bottomhole pressure, psi

160

used for matching pressure data in deconvolution (see

Flow rate, lt/s

60
Eq. 2). The deconvolved unit-rate derivative response
based on using σ p = 0.01 psi in Eq. 2 (green curve in 150

40
Fig. 10) is oscillatory, and indicate changing wellbore
storage effects (perhaps, also including possible non- 140
20
isothermal effects in the wellbore) until 0.1 h, and in
the time interval from 0.1 to 1 h, it indicates a radial
flow period. The oscillatory behavior of deconvolved 130 0
0 2 4 6 8 10 12 14 16 18 20 22
data for σ p = 0.01 psi may indicate that the actual Time (h)

noise level in PBU pressure data could be larger

Figure 11. Model match of entire measured pressure
than σ p = 0.01 psi . To obtain smoother deconvolved data recorded during drawdown and
data, we tried larger values of noise level for the PBU buildup test at the well R-260.
pressure data to be used in deconvolution, and found
that the unit-rate drawdown response generated with
a noise level of σ p = 0.07 psi provides an Table 4. Some input and estimated model parameters
acceptably smooth curve as shown by red curves in for obtaining model match shown in
Fig. 10. Although not shown here, unit-rate responses Figure 11.
generated with noise levels such
that σ p > 0.07 psi were too smooth to be considered, Model Parameters
and root-mean-square (rms) errors for the pressure
match obtained for these noise levels were not kh (Darcy-m) 47.8
acceptable. pi, psi 164.09
The deconvolved unit-rate data based on S (skin, dimensionless) -4.9
σ p = 0.07 psi in Figure 10 indicate changing df1, m (distance to no-flow 63
wellbore storage or double porosity behavior in the fault)
time period from 0.0003 to 0.01 h, and an df2, m (distance to constant- 126
intersecting fault (one with no-flow, and other is pressure fault)
constant-pressure) for times greater than 0.01 h. h, m (formation thickness) 65
Another plausible model is a finite-conductivity fault
intersecting the well nearby a constant-pressure fault. µ , [email protected] C o
0.281
 rw, m 0.108 first 140 hours of production at AF-21, the well R-
260 was shut-in about 6 hours due to some
φcth (m/psi) 2.35x10-4 operational problems occurred at the well AF-21.
This shut-in period provided a 6-hour buildup test
The kh and skin factor values estimated from data in the time interval from 182 to 188 h (in
drawdown/buildup tests conducted at the well R-260, cumulative time) as shown Fig. 12.
AF-11, AF-16, AF-20, and AF-21 are summarized in 190 250
Table 5. Note that the skin factor values given in Measured bottomhole pressure at R-260
Table 5 represents total skin (i.e, st = s + Dq), Flow rate at AF-21

Bottomhole pressure at R-260, psi

180 200
including both mechanical skin (s) and non-Darcy

Flow rate at AF-21, lt/s

skin (Dq). As is known (see, for example, Bourdet,
2002), a constant-rate drawdown/buildup test does 170 150

not allow one to obtain individual values of

mechanical skin and non-Darcy skin (or non-Darcy 160 100

coefficient, D).
150 50
Table 5. Permeability-thickness (kh) and total skin
values estimated from analyses of 140 0
drawdown/buildup tests conducted at 0 20 40 60 80 100 120
Time (h)
140 160 180 200 220

wells R-260, AF-11,and AF-16.

Well kh Total skin Figure 12. Pressure and flow rate history for the AF-
Name (Darcy-m) dimensionless 21/R-260 two-well interference test.
R-260 48 -4.9 In Figure 13, we present the deconvolved responses
AF-11 201 1.96 (pressure drop and derivative functions) derived from
the buildup pressures, and we compare these results
AF-16 665 10.6
with the conventional normalized buildup pressure-
change and its derivative with respect to Agarwal’s
equivalent time plotted versus elapsed time. In
ANALYSIS OF INTERFERENCE TESTS deconvolution, we considered two different values of
initial reservoir pressure; one is 158.5 psi (as the
Here, we summarize the results obtained from measured value), and the other is 157.93 psi. We
analyses of two-well interference tests conducted at assume that the flow rate history prior to buildup is
some of the wells in the field. The well pairs where accurate and can be treated as known in
the interference tests are conducted are AF-21/R-260, deconvolution procedure of Eq. 2. In all deconvolved
AF-21/AF-11, AF-20/AF-10, and AF-20/AF-11. A responses, we set the noise level in PBU data as σp =
well name given before the slash indicates an active 0.01 psi.
well, while a well name given after the slash
indicates an observation well during the two-well Although deconvolution provides about a one-and-a-
interference test. Here, we only present our analysis half cycle longer data than conventional rate-
for the interference test involved between the wells normalized buildup data, the late portions of
AF-21 and R-260 in detail. In this test, AF-21 is the deconvolved responses indicate two different model
active well, while the well R-260 is the observation behaviors due to its sensitivity to the value of initial
well. The distance between the two wells is 78.5 m. pressure. As we do not have another buildup period,
we do not know for sure which value of the initial
Figure 12 presents flow-rate history at the active well pressure is the appropriate value. If we accept that the
(AF-21) and pressure recorded at the observation measured value of pi = 158.5 psi is the appropriate
well (R-260). The bottomhole pressure at the well R- one, then the late-time portion of deconvolved unit-
260 was recorded at a depth of 116 m by a downhole rate derivative response (green curves in Fig. 13)
quartz gauge. gives an indication of a finite-conductivity fault near
the well, whereas if we accept that pi = 157.93 psi is
Although it is not evident from Fig. 12, the the appropriate one, then the late-time portion of
production at AF-21 is felt at R-260 in 100 seconds, deconvolved unit-rate derivative response (red curve
indicating a highly permeable fracture/fault network in Fig. 13) indicates a constant-pressure (or infinite-
existing between the wells. The total test duration conductivity) fault.
after production started at well AF-21 is about 170 h
(or about 7 days). The static pressure measured at the
well R-260 at the depth of 116 m is 158.5 psi. After
 1
conventional rate-normalized pressure change that kh value estimated from this interference test is
conventional rate-normalized pressure-derivative
deconvolved unit-rate responses with pi = 157.93 psi different from kh values estimated from the tests
rate-normalized buildup responses, psi/(lt/s)

deconvolved unit-rate responses with pi = 158.5 psi where AF-21 and R-260 were pulsing wells alone
0.1 (see Tables 2 and 4). These results as well as the
results given in Table 7 indicate that the geothermal
reservoir under consideration is highly heterogeneous
Deconvolved unit-rate and

0.01 and permeable. In addition, interference tests give an

indication that permeability is more developed in the
NS direction than in the EW direction.
0.001

Table 6. Some input and estimated model parameters

for obtaining model match shown in
0.0001
0.001 0.01 0.1 1 10 100 1000 Figure 14.
Time (h)
Figure 13. Comparison of deconvolved responses
Model Parameters
derived from buildup pressures with
conventional pressure buildup pressure
change and its Bourdet derivative kh (Darcy-m) 135
(normalized by the last rate prior to pi, psi
buildup). 157.93
ri, m (distance between the 382
Next, we perform parameter estimation for well R-260 and an imaginary
determining permeability-thickness product (kh) and well)
porosity-compressibility-thickness product (φcth), h, m (formation thickness) 65
and the distance (ri) between the observation well R-
260 and an imaginary well by using a “simple” µ , [email protected] C o
0.281
well/reservoir model considering a fully-penetrating rw, m 0.108
well near a constant-pressure fault in a homogeneous
isotropic reservoir (i.e., the model indicated by red φcth (m/psi) 2.91x10-4
curves on Fig. 13). Figure 14 presents a model match
of the measured pressure data recorded for the entire
interference test sequence, and the match can be The permeability-thickness (kh) and the porosity-
considered as acceptable. compressibility-thickness (φcth) products estimated
from all interference tests conducted in the field are
190 250 summarized in Table 7.

180
Measured pressure at R-260
Computed model pressure 200
Table 7. kh and φcth values estimated from two—well
Flow rate history at AF-21 interference test conducted in the field.
Bottomhole pressure, psi

Flow rate, lt/s

170 150

Well Pairs kh φcth

160 100
Active/observation (Darcy-m) (m/psi)
150 50
AF-21/R-260 135 2.91x10-4
AF-21/AF-11 610 2.98x10-3
140 0
0 20 40 60 80 100 120 140 160 180 200 220
Time (h) AF-20/AF-10 1900 1.06x10-2
AF-20/AF-11 415 2.00x10-3
Figure 14. Model match of entire measured pressure
data recorded at the well R-260 during
AF-21/R-260 interference test.
CONCLUSIONS AND RECOMMENDATIONS
The estimated model parameters are summarized in
Table 6. It is important to note that ri given in Table 6 In this work, we presented analyses of various types
represents the distance between the well R-260 and of pressure transient tests (such as multi-rate tests,
an imaginary well, perpendicular to the fault. As is conventional drawdown and buildup tests, and
known from the work of Vela (1977), one cannot interference tests) conducted in the Afyon Ömer-
determine uniquely the distance to the fault and its Gecek geothermal field, Turkey. In general, the
orientation from a single-interference test. Note also pressure tests analyzed indicate that the wells’
productivities are quite high, but influenced by non-
 Darcy flow effects and are producing in a complex 2006 SPE Annual Technical Conference and
fractured/faulted network system. The estimated Exhibition, San Antonio, Texas, September. 24-27.
values of permeability-thickness products (kh) from
multi-rate, drawdown/buildup and interference tests Levitan, M.M. (2005), “Practical Application of
range from 40 to 2000 Darcy-m. The well test data Pressure/Rate Deconvolution to Analysis of Real
also identify highly conductive (recharging) faults, Well Tests,” SPE Reservoir Evaluation and
where we believe these faults are dominating the Engineering (April) 113.
performance of the geothermal field. Regarding
determining these faults orientations in the field, Levitan, M.M., Crawford, G.E., and Hardwick, A.
additional pressure transient tests and more detailed (2006), “Practical Considerations for Pressure-Rate
geological and geophysical work are recommended. Deconvolution of Well-Test Data,” SPE Journal
(March) 35.
Deconvolution analysis based on recently proposed
robust algorithms by von Schroeter et al. (2004) and Satman, A., Onur, M., Serpen, U, Aksoy, N. (2007),
Levitan (2005) was found useful to extract more “A Study on the Production and Reservoir
information from the well tests conducted in the field. Performance of Omer-Gecek/Afyon Geothermal
However, it should be stated that deconvolved Field,” Proceedings, Thirty-Second Workshop on
responses can be quite sensitive to noise level in Geothermal Reservoir Engineering, Stanford
pressure data and the initial reservoir pressure. University, Stanford, California, Jan. 22-24.
Hence, characterizing and estimating the appropriate
noise level in pressure a priori and designing tests Vela, S. (1977), “Effect of a Linear Boundary on
that include at least two buildup periods to identify Interference and Pulse Tests-The Elliptical Influence
the appropriate initial reservoir pressure are Area,” Journal of Petroleum Technology (August),
recommended to accurately and properly interpret the 947.
results derived from such deconvolution algorithms.
von Schroeter, T., Hollaender, F., and Gringarten,
A.C. (2004), “Deconvolution of Well Test Data as a
ACKNOWLEDGEMENTS Nonlinear Total Least Squares Problem,” SPE
We thank the management of AFJET A.Ş. for giving Journal (December) 375.
us the permission to use the pressure data presented
in this study. The pressure tests were conducted by
Iller Bank of Turkey.

REFERENCES

Abbaszadeh, M. and Cinco-Ley, H. (1995), “Pressure

Transient Behavior in a Reservoir With a Finite-
Conductivity Fault,” SPE Formation Evaluation
(March) 26.

Agarwal, R.G. (1980), “A New Method to Account

for Production Time Effects When Drawdown Type
Curves Are Used to Analyze Buildup and Other Test
Data,” paper SPE 9289 presented at the 1980 SPE
Annual Technical Conference and Exhibition, Dallas,
TX, 21-24 September.

Bourdet, D. (2002), Well Test Analysis: The Use of

Advanced Interpretation Models, Elsevier,
Amsterdam, the Netherlands, 426 pp.

Bourdet, D., Ayoub, J.A., and Pirard, Y.M. (1989),

“Use of Pressure Derivative in Well-Test
Interpretation,” SPE Formation Evaluation (June) 69.

Cinar, M., Ilk, D., Onur, M., Valko, P. P.,

Blasingame, T. A. (2006), “A Comparative Study of
Recent Robust Deconvolution Algorithms for Well-
Test and Production-Data Analysis,” Proceedings, the

View publication stats