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

net/publication/266666376

Integration of Pressure-Transient Data in Modeling Tengiz Field, Kazakhstan

—A New Way To Characterize Fractured Reservoirs

Article in SPE Reservoir Evaluation & Engineering · January 2016

DOI: 10.2118/165322-PA

CITATIONS READS

14 1,188

10 authors, including:

Yan Pan Wayne Narr

University of Nottingham, Malaysia Campus Narr NFR Consulting
107 PUBLICATIONS 761 CITATIONS 58 PUBLICATIONS 2,055 CITATIONS

SEE PROFILE SEE PROFILE

Eric A. Flodin Hai X. Vo

Chevron Stanford University
26 PUBLICATIONS 712 CITATIONS 8 PUBLICATIONS 300 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Hai X. Vo on 18 July 2016.

The user has requested enhancement of the downloaded file.

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 5 Total Pages: 13

Integration of Pressure-Transient Data

in Modeling Tengiz Field,
Kazakhstan—A New Way To
Characterize Fractured Reservoirs
Yan Pan, Mun-Hong Hui, Wayne Narr, Gregory King, and Terrell H. Tankersley, Chevron; Steve D. Jenkins,
KPO; and Eric A. Flodin, Philip W. Bateman, Chris Laidlaw, and Hai Xuan Vo, Chevron

Summary reduce the number of simulation iterations. Its application in Ten-

A systematic work-flow to integrate pressure-transient data col- giz is also one of the rare examples in which single-well and mul-
lected from single-well buildup tests in numerical reservoir-simu- tiple-well transient data, production logging, and image-log data
lation models for a fracture/matrix system is presented. The are all available.
results of its application in a sector model in the southeast region
of Tengiz field in Kazakhstan are also discussed.
The procedure starts with a selected numerical-simulation Introduction
model, either a discrete fracture/matrix (DFM) model or a dual- Reservoir characterization is an important component of reservoir
porosity dual-permeability (DPDK) model, and follows with the management. The characterization process integrates information
analysis of a numerically generated buildup test to calculate from two main sources: static data (such as geological interpreta-
the fracture spacing and shape factor of the model. Then, follow- tion that is based on outcrop, seismic, well log, core, drilling-fluid
ing the correlations between pressure-transient-analysis results loss) and dynamic data (such as well test, production logging, and
and the average or representative values of the model-input pa- production history). Although the use of dynamic data in the res-
rameters near the well, which contain the previously obtained ervoir-characterization process was attempted in the past, its
fracture spacing and shape factor, the numerical-model parame- application is still limited and lacking a robust work-flow. In this
ters are adjusted in each iteration to match the pressure-transient study, numerical well-testing techniques (Kamal et al. 2005) were
behavior observed in the buildup test including the interporosity applied to understand and characterize the fracture/matrix pro-
flow between matrix and fracture and the radial flow in the perties of naturally fractured reservoirs (NFRs), which are espe-
total system. cially challenging because of their often-extreme heterogeneity
Before field application, the numerical-simulation results from and, thus, high degree of flow-related uncertainty (Narr et al.
both DFM and DPDK models were validated against analytical 2006). The effective usage of pressure-transient data that contain
pressure-transient solutions for a dual-porosity system. The gridd- rich information about wells and reservoir should narrow the
ing and time-steps were calibrated to reproduce the analytical uncertainty, improve the characterization, and help optimize
transient behavior. Finally, the new work-flow was applied to a field development.
sector model of Tengiz field in the southeast region focusing on An NFR has a network of permeability-enhancing fractures
two wells. Following the developed work-flow, a DFM model that separate matrix. The flow-simulation models used most com-
was constructed, and its fracture and matrix properties were monly for NFRs are single-porosity enhanced-permeability mod-
adjusted to honor buildup-test data at both wells and the transient els, dual-porosity (DP) models, and DPDK models. Other less-
data collected during a pulse test conducted between them. The commonly used methods include discrete-fracture-network (DFN)
study results show that the key factors of a DFM model on models and DFM models. The DP model is derived from the con-
buildup transient response are the fracture permeability, fracture ceptual model proposed by Barenblatt et al. (1960) and Warren
aperture, and matrix permeability in the well-drainage area, and and Root (1963). In this model, the reservoir consists of two inter-
the dominant parameters on pulse-test response are the fracture acting media: a uniformly distributed network of low-volume but
permeability and matrix porosity in the influence area between the highly conductive fractures and isolated matrix blocks that con-
two wells. Using the correlations quantitatively for each simula- tain the bulk of pore volume (PV). In a DFM model, relatively
tion step could reduce the total number of iterations needed to more-realistic geometry, orientation, density, and other properties
converge to the numerical solution. The modified model also gen- of fractures could be considered by using the geological informa-
erated flow distribution along the wellbore, consistent with pro- tion acquired during exploration and development of reservoirs
duction-logging data at one well. The resulting sector map of explicitly (Wei et al. 1998; Casciano et al. 2004; Karimi-Fard
pressure change during buildup test indicates the area with well- et al. 2004; Basquet et al. 2005; Sarma and Aziz 2006; Casabianca
connected fracture network. Dynamic transient data contain rich et al. 2007; Rogers et al. 2007; Hui et al. 2007, 2008; Izadi and
information about reservoirs, and the effective integration of Yildiz 2009).
dynamic and static data would have a big impact on reservoir Conventionally, the analytical models that treat the reservoirs
management by potentially minimizing the number of wells to be as simplistic and idealized abstractions are used for evaluating
drilled, maximizing the production, and optimizing recovery. pressure-transient-test data. These models may not be representa-
The novelty of this study is the quantitative use of the correla- tive for real complex heterogeneous reservoirs. However, the
tions between pressure-transient-analysis results and the represen- obtained analysis results, such as the estimations of formation per-
tative values of the input parameters in a numerical model to meability, the storativity ratio, and interporosity flow coefficients
of a DP model, are valid, effective average properties of the reser-
voir within the tested volume near a well. The analytical solutions
Copyright V
C 2016 Society of Petroleum Engineers
of pressure-transient responses for NFRs were built on DP models
This paper (SPE 165322) was accepted for presentation at the SPE Western Regional & (Warren and Root 1963). Depending on the flow conditions when
AAPG Pacific Section Meeting Joint Technical Conference, Monterey, California, USA, the interporosity flow from the matrix system into the fracture net-
19–25 April 2013, and revised for publication. Original manuscript received for review 14
January 2014. Revised manuscript received for review 11 August 2015. Paper peer work occurs, there are two types of analytical solutions: pseudos-
approved 29 October 2015. teady-state and transient. The analytical pseudosteady-state

February 2016 SPE Reservoir Evaluation & Engineering 5

ID: jaganm Time: 13:34 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 6 Total Pages: 13

75
m 00
00 m
75

1620 m

Well at center of reservoir

(fully penetrating)
Sugar-cube vertical
fractures (250-m spacing)
Matrix blocks (white)

Fig. 1—A reservoir with uniform sugar-cube vertical fractures and a vertical well at the center.

solution for NFRs has a distinct signature on the pressure response realistic DFM sector model for the Tengiz oil field. The ultimate
characterized by a symmetric trough on the pressure-derivative goal of our study is to use pressure-transient data from buildup
curve in the log-log diagnosis plot, whereas the transient solution tests at individual wells and from pulse tests among multiple wells
may not show a well-defined trough. The property contrast in a real field to build and update the numerical reservoir models.
between fracture network and matrix system and their configura- The updated models, which incorporate both realistic geological
tion determine whether the interporosity flow is pseudosteady or static data and actual field dynamic data, are expected to be more
transient. Since 1963, many papers were published on fractured- robust and predictive to assist reservoir management.
reservoir well-test interpretations and studies. Samaniego-V
and Cinco-Ley (2009) give detailed reviews of major works in
the literature. Calibration and Validation
Our work aims at developing a systematic procedure for inte- To fully understand the pressure-transient behavior of different
grating well-test data that contain rich information on effective numerical reservoir models and to calibrate the gridding selection
reservoir properties in a numerical reservoir model that uses a re- and simulation time-steps, a validation step was taken first with
alistic fracture-system geometric description and natural variabili- synthetic cases. Because horizontal fractures are usually not
ty of such fracture-system properties for NFRs. It includes three observed in the reservoir below moderate depth, a fully penetrat-
parts. The first is the validation of numerical solutions from DFM ing vertical well at the center of a single-layer sugar-cube DP res-
and DPDK models generated with a numerical reservoir simulator ervoir with only vertical fractures was considered (Fig. 1). The
against analytical solutions for a DP system. Second, after investi- fracture network is composed of two orthogonal sets of vertical
gating the impact of different fracture/matrix properties of a nu- fractures as long as the reservoir size with a constant spacing of
merical model on pressure-transient responses from single-well 250 m. Matrix blocks fill the space between the fracture planes.
tests with synthetic cases, based on the results, a work-flow for The uniform fracture and matrix properties are shown in Table 1.
integration of transient data in NFR models was developed. This A black-oil formulation was selected to represent the fluids. A
procedure uses numerical models to simulate transient-test drawdown test (3,500 days) at a constant oil-production rate of
responses and to compare with collected pressure-transient data. 5,000 STB/D was simulated numerically with both DFM and
Then, following the correlations between the parameter estima- DP models, and the pressure responses were compared with ana-
tions from transient analysis of the numerical results and the rep- lytical solutions.
resentative or average input parameters near a well in the For our numerical models with uniform fracture and matrix
numerical model, the model-input parameters are adjusted for the properties, the effective properties were calculated with the fol-
next iteration. The steps are repeated until the numerical pressure- lowing equations.
transient responses match the measured data. The analytical • Effective permeability of a DP system,
model used to analyze the numerical responses is the Warren and
ðxm  km þ af  kf Þ
Root (1963) solution. However, our work-flow can use any ana- keff ¼ : . . . . . . . . . . . . . . . . . . . . ð1Þ
lytical or semianalytical solutions as long as they provide valid ðxm þ af Þ
correlations between the transient-analysis results and numerical
• Storativity ratio of fracture to the total system,
model-input parameters to guide quantitatively the transient-data-
matching process. In the third part, this approach was applied in a /f  ctf
x¼ : . . . . . . . . . . . . . . . . . . . . . ð2Þ
/f  ctf þ /m  ctm
• Interporosity flow coefficient,
a km 2 kf  a f KSIGMA  rw2
k¼  r ; kfeff ¼ ; k¼ :
x2m kfeff w xm þ af kfeff
                   ð3Þ
Parameter a is the geometric shape factor that describes the
contact surface area between matrix blocks and fracture network.
For a sugar-cube DP system with 2D fractures, the value of a is
16 according to the Coats (1989) formula and 19.7 from Lim and
Aziz (1995). Our numerical results showed a is closer to 16 for
this synthetic case. For field cases with relatively realistic realiza-
tions in which matrix blocks have irregular shape and fracture dis-
tribution varies, the shape factor can be obtained by analyzing the
numerical transient response and compare with model input,
Table 1—Properties of a uniform sugar-cube DP reservoir. which will be discussed in the next section. The interporosity flow

6 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:34 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 7 Total Pages: 13

Well
One Matrix Block

Fig. 2—The well location and first-level grid refinement in a DFM model for a uniform sugar-cube reservoir with vertical fractures.

coefficient k is a function of matrix permeability km, and the effec- and 2,000 hours is the interporosity flow period, when fluid starts
tive permeability of the fracture network kfeff , which depends on to flow out of matrix blocks into the fracture network. From the
individual fracture permeability kf, aperture af, and fracture spac- depth of the trough in the pressure-derivative curve, the storativity
ing xm. KSIGMA is a parameter used in numerical simulation to ratio of the fractures to the total system x can be obtained. The
describe the transmissibility between matrix and fracture network. interporosity flow coefficient k, which is an indication of how
The numerically simulated drawdown-test pressure responses easy or difficult it is for fluids to flow out of low-permeability ma-
were analyzed with analytical DP models, and the estimated pa- trix and enter the high-conductivity fracture network, can be esti-
rameter values were compared with those calculated directly with mated from the time when the trough ends and the second-radial
Eqs. 1 through 3. flow period starts (2,000 hours). After 40,000 hours, the pressure
derivative shows pseudosteady-state flow behavior (unit-slope
straight line) caused by the finite size of the reservoir. For ideal
DFM Model DP transient behavior (with very low wellbore storage and small
A synthetic one-layer DFM model with 30  30 matrix blocks and or negative skin), there should be a first radial-flow period at early
31  31 fully penetrated vertical fractures as long as the reservoir time governed by flow only inside the fracture network before the
size with a fully penetrated vertical well at the center of the reser- onset of interporosity flow. The simulated response of the DFM
voir was created to simulate the drawdown test in the sugar-cube (green in Fig. 3b) shows linear flow before 10 hours because of
DP system. The well was in the intersection point of two fractures the fractures crossing the wellbore in the DFM, and the first radial
and surrounded by four matrix blocks. An unstructured grid with flow is thus not observed. Nevertheless, the analytical solution
one level of grid-refinement, four gridblocks within each matrix (red in Fig. 3b) assuming no wellbore storage was reasonably
block, and eight fracture segments surrounding the matrix block matched by the numerical-simulation results from DFM (green).
(Fig. 2) was used. The minimum time-step was selected as 1 hour Table 2 shows that the DP-system properties calculated from
for the 3,500-day test, and increased logarithmically based on Eqs. 1 through 3 are reasonably consistent with the analysis
transient behavior. The pressure responses during the drawdown results of the numerical DFM pressure response. The difference in
test simulated by the DFM (green curve) and those of the analyti- permeability is negligible. The difference in interporosity coeffi-
cal solution (red) are shown in Fig. 3a, and the log-log plot of cient is less than 14% and within the same scale of 107. The dif-
pressure change and pressure derivative are shown in Fig. 3b. The ference in storativity is relatively large, 48%. It is challenging to
horizontal portion of the pressure derivative (lower curve) from match this parameter. The input well skin factor in the numerical
2,000 to 40,000 hours indicates the second radial-flow period, model is zero. The skin value of 6.40 obtained by analyzing the
from which the effective permeability of the DP system can be simulated pressure-transient response is actually the equivalent
estimated. The separation between the pressure (upper curve) and skin, because the well is at the crossing point of two fractures in
the pressure derivative (lower curve) during the radial-flow period the DFM model (Fig. 2). The equivalent skin of a 250-m fracture
provides an estimate of well skin factor. The trough between 20 plane crossing this well should be 6.63 according to the
Pressure (psia)

Numerical model Pressure change

Pressure Change and

Pressure drawdown Analytical solution

9,000 s
Derivative (psia)

1,000

7,000
ω k eff
100
Liquid Rate
(STB/D)

2,500 Constant production rate Pressure derivative λ

0 10
0 40,000 80,000 1.2E+5 1 10 100 1000 10000 1E+5
Time (hours) Time (hours)
(a) (b)

Fig. 3—Pressure responses during a drawdown test of a vertical well in a sugar-cube DP reservoir with DFM model.

Table 2—Numerical results of DFM model for a drawdown test of a vertical well in a sugar-cube DP
reservoir.

February 2016 SPE Reservoir Evaluation & Engineering 7

ID: jaganm Time: 13:34 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 8 Total Pages: 13

Pressure spacing near the well. If there are enough data (e.g., cores, image
Numerical model change logs, production log, and analogs) to generate reasonable fracture
Analytical solution
distributions, for any grid size, the representative gridblock prop-
Pressure Change and 10,000
erties, such as fracture porosity and the transmissibility between
Derivative (psi)
s matrix and fracture, can be calculated based on the discrete frac-
1,000 tures within each cell (Hui et al. 2013). Hence, the derived numer-
ical DP model could simulate the interporosity flow behavior as
ω
keff observed during transient tests without relying on a particular grid
100 size.
λ
Pressure derivative
10 Work-Flow To Integrate Single-Well
1 10 100 1,000 10,000 1E+5 Transient-Test Data for NFR
Time (hours) Following the calibration and verification process of numerical-
simulation models, a sensitivity study of fracture/matrix system
Fig. 4—Log-log plot of a drawdown test of a vertical well in a parameters on pressure-transient responses was conducted. Based
sugar-cube DP reservoir with DP model. on the study results, a systematic procedure of incorporating sin-
gle-well transient-test (buildup or drawdown) data in discrete-
fracture simulation models was developed. The goal is for any
correlation sequiv ¼ lnðLf =4=rw Þ. If two fractures cross the well- given numerical discrete-fracture model to adjust the selected-
bore, the skin could be smaller than 6.63. Sensitivity study model input parameters to make the numerically simulated tran-
showed that the fracture conductivity and the grid refinement near sient-test responses match the measured-pressure data at wells.
the wellbore would have impacts on the equivalent skin derived The transient-analysis results with any selected analytical solu-
from the transient analysis of the numerical-simulation results with tions are not used directly in the numerical model. Only the corre-
the DFM model. Further details are beyond the scope of this paper. lations between the transient-analysis results and the numerical-
model input parameters, Eqs. 1 through 3, are used to quantita-
tively guide the selection of input parameters for each iteration
DPDK Model during the process. The work-flow is shown in Fig. 5.
A Cartesian 30  30  1 DPDK numerical model was also created The systematic procedure is as follows:
to simulate the drawdown-test response from the same DP system 1. Select the most likely fracture realization based on geolog-
(Fig. 1). The size of each gridblock was the same as the sugar- ical information, such as well logs, cores, image logs, out-
cube fracture spacing of 250  250 m. The well was in the center crop analogs, and dynamic flow data, such as from
of a gridblock in the middle of the reservoir. Each gridblock was production logging tool (PLT) and lost circulation. In the
assigned two sets of uniform properties, one for matrix and the current study, fracture properties, such as aperture and per-
other for fracture. Fractures were not represented explicitly as in meability, are assumed uniform for each discrete fracture,
the DFM model. The drawdown-simulation results are shown in although properties could vary among different fractures.
Fig. 4. The numerical DPDK model (green curves) captured the At this step, the fracture sizes, orientations, and spatial dis-
interporosity flow trough, the second radial-flow, and the late- tribution are specified, and are combined with the reser-
time pseudosteady-state periods, just as the DFM did, and are con- voir structure, such as bed dipping and pinchout.
sistent with the analytical solution (red curves) starting from 10 2. Generate an unstructured grid that represents the fracture
hours. The early-time (<10 hours) first radial flow was not realization defined in the previous step (Hui et al. 2008).
observed in the numerical response because of the large wellblock Choose the base-case fracture-matrix system properties,
size of 250  250 m. Table 3 shows that the DP properties calcu- such as matrix permeability and porosity, and fracture per-
lated from Eqs. 1 through 3 are reasonably consistent with the meability and aperture, according to information in hand
analysis results of the numerical DP pressure response. (static-data interpretation and well-test-analysis results).
The DFM model provides a skin factor of 6.40 (Table 2), The properties can be heterogeneous, and modifiers can be
which represents the equivalent skin caused by a well crossing used to adjust its properties. Load the fluid PVT properties
two vertical fractures. The DP model, which homogenizes the and the production/injection history into the simulation
fractures in a DP sense, does not exhibit this feature; it gives a model. Select the appropriate grid-refinement level near
skin factor of 0.46, which is close to the input value of 0.0. One the wellbore and time-steps (Kamal et al. 2005). Run this
should use this observation as the guidance when the DFM model base-case numerical simulation of the transient test.
is selected for a numerical well-testing study. The skin effect seen 3. Compare the simulated buildup/drawdown pressure
from DFM response is a combination of well mechanical skin and response at the well with the gauge data collected during
any fractures crossing the wellbore. the transient test. Pressures may need to be adjusted to the
The transient behavior simulated with a DPDK model is sensi- same datum as the gauge depth for comparison. If the nu-
tive to gridblock size. Different from a single-porosity system in merical results match gauge data during the transient test
which finer Cartesian gridlines close to following streamlines sim- in both Cartesian history plot and transient diagnosis log-
ulate more-accurate transient responses, for a DP system, the log plot, the base case is one solution. If not, proceed to
grid-refinement level has to be selected to balance the need to the next step.
reproduce early-time transient radial flow, for which a finer grid 4. Analyze the simulated transient-test pressure response and
may be preferred, and the need to generate the interporosity flow estimate the effective permeability of the fracture-matrix
between the matrix system and the fracture network, which may system keff, fracture storativity ratio x, and interporosity
require the grid size to be in scale similar to that of the fracture flow coefficient k.

Table 3—Numerical results of Cartesian DP model for a drawdown test of a vertical well in a sugar-
cube DP reservoir.

8 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:34 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 9 Total Pages: 13

1. Select 2. Base 4. PTA of num. 5. Use equations

fracture case DFM response => to estimate
realization simulation k eff, λ and ω xm and α

3. Match buildup 10. Are the new

6. n+1 DFM Yes No
dp, dp/dlnt data results moving in the
No simulation expected direction?
in log-log plot?

Yes

Yes 7. Match buildup No 8. PTA of num. 9. Compare with

End dp, dp/dlnt data response => expected, k eff, λ ,
in log-log plot? k eff, λ and ω ω from eqs.

Fig. 5—The work-flow of single-well transient-test-data integration for NFR.

5. Use these estimates and those average or representative input parameters for next iteration. Repeat Steps 6 through
values of the input parameters (km, /m, kf, af) near the well 10 until the simulated responses match test data.
in numerical model in Eqs. 1 through 3 to calculate frac- 12. If “No,” generate a new fracture realization to change
ture spacing xm and geometric shape factor a of the current fracture connectivity (e.g., bigger fracture height/length
fracture realization. Different realizations may produce and adding more fractures) within the bounds of accepted
different fracture spacing and shape-factor values. These geological uncertainty around the subject well; then repeat
two values together with geological information (e.g., res- Steps 1 through 10.
ervoir structure, fracture-size distribution, spatial distribu- This work-flow was developed with DFM models as the
tion, number of fracture sets, and fracture orientation) will illustration, but a similar procedure can also be used for
be used to guide the following numerical-simulation DPDK models.
iterations.
6. Depending on the deviation of the base-case transient
response from the test measurements, and following the Application in Tengiz Southeast-Sector Model
correlations in Eqs. 1 through 3, select the second set of The developed work-flow was applied in a pilot region in Tengiz
model parameters to run the simulation. field. Tengiz is a supergiant oil field with approximately 30 billion
7. Compare the simulated transient-pressure response at the STB of original oil in place in the Pricaspian basin of western
well with the test measurements. If the numerical results Kazakhstan. The field produces from an isolated carbonate plat-
match gauge data well in both the Cartesian history plot form of Devonian-to-Pennsylvanian age, with a flat-lying central
and the transient-diagnosis log-log plot, the current case is platform surrounded by a relatively steep depositional slope
one solution. If not, proceed to the next step. (Fig. 6). Fractures are common, and they significantly enhance
8. Analyze the new simulated transient responses. productivity in the outer platform and slope. The reservoir has
9. Compare the analysis results with the expected values of multiple layers with distinguishable properties. The matrix system
fracture/matrix system properties (keff, x, k) from Eqs. 1 is heterogeneous. Most fractures dip steeply and show a wide
through 3 with the average values of current model-input range of properties (Tankersley et al. 2010; Collins et al. 2013).
parameters and the values of xm and a obtained in Step 4. Pressure-buildup tests were conducted at more than 100 wells
10. Are the new analysis results moving in a direction consist- in Tengiz. In the slope area where fractures are common, 70% of
ent with expected values and in relatively similar scale? the tests detected possible DP behavior with observed radial-flow
11. If “Yes,” keep current fracture realization, from the cur- regime and recognizable trough showing at least the lowest point
rent-analysis results and the expected values; use correla- of the trough and either the beginning or the end of interporosity
tions to quantitatively guide the calculation of the model- flow period. Some tests observed textbook DP behavior com-
pleted with first radial flow, well-defined symmetric trough (pseu-
dosteady-state interporosity flow), and second radial flow. For test
data without recognizable trough, there are several possibilities:
The interporosity flow near the well might happen very quickly
masked by wellbore storage; it might occur at a later time beyond
the test duration; it might be in transient state without recogniz-
Tengiz Field able trough; it might be masked by linear flow if the well was
drilled crossing an existing natural vertical fracture; or the well
was drilled in an area with low fracture density. Because of the
44
00

high degree of heterogeneity of naturally fractured system, the ab-

ar udy

sence of distinguished trough in transient data is not uncommon

00
ea
St

50

in many carbonate reservoirs. If the collected data show any sign

of interporosity flow, pseudosteady-state or transient, the integra-
tion process is the same. Otherwise, a combination of all parame-
52

ters is used to match the measured data as any other history-

00

2 kms matching process. However, the results may have higher uncer-
v.e = 2x tainty because of the difficulty in distinguishing the impacts of
5400 different matrix and fracture properties in this case.
Many wells in Tengiz are equipped with permanent downhole
gauges; together with oil production, long-term pressure and rate
Fig. 6—Tengiz field structure and the pilot region. data were also analyzed to characterize the reservoir. One of the

February 2016 SPE Reservoir Evaluation & Engineering 9

ID: jaganm Time: 13:34 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 10 Total Pages: 13

(m3/D) Pressure (bara)

Well pressure Pressure change
595
570
Buildup 10
545

Oil rate
Oil Rate

1,250 Pressure derivative

1
0 0.1 1 10
10/13/2009 10/16/2009 10/19/2009 10/22/2009 10/25/2009 Time (hours)
(a) (b)

Fig. 7—History plot (a) and buildup log-log plot (b) of well test conducted at T-1 in October 2009.

Table 4—Work-flow application in Tengiz sector DFM model at T-1.

benefits of production-data analysis (PDA) is that it provides the Well Buildup-Test Data
drainage-area estimations of each well, which one can use directly Well T-1 (well names are disguised) was selected as the first well
in the dynamic data-integration process. However, in the case of to test the work-flow. A 100-hour buildup test was conducted at
Tengiz, PDA is not effective in capturing the interporosity flow this well in October 2009 with controlled flow periods before
behavior caused by frequently varying rates. The transient part of shut-in (Fig. 7a). Data quality is good. As the pressure-derivative
the rate-normalized pressure data is likely to appear noisy and dif- curve of the buildup period shows in the diagnosis plot (Fig. 7b),
ficult to pinpoint different flow regimes. Nevertheless, it is a very the test detected the interporosity flow from matrix system into
useful tool to study long-term well performance. fracture network (the trough between 1 hour and 10 hours) fol-
A sector in the southeast slope area was selected as the pilot lowed by the second radial flow (flat derivative between 10 and
region (Fig. 6) and extracted out of the current history-matched 30 hours). The late-time data 30 hours after shut-in were influ-
Tengiz full-field DPDK simulation model (Tankersley et al. 2010 enced by nearby producers that caused the derivative to dive
and King et al. 2012), which contains three major wells with downward. Reliable estimations of reservoir properties (keff, x, k)
available buildup tests, pulse tests in two directions, permanent were obtained (Table 4).
downhole gauges, production-logging data, image logs, and other
well-log data. The buildup tests conducted at these three wells
captured DP transient behavior successfully and provided reliable Sector Model
fracture/matrix-system property estimations. The rich data col- A sector of 7500  7500 m with thickness of 1620 m in the south-
lected in this region make it a good candidate to test the new east pilot region was selected carefully to minimize the influence
work-flow of integrating single-well transient well-test data, and of the artificial sector boundaries on the buildup-test responses at
procedures that will be developed in the future for incorporating the subject wells. The sector includes parts of both platform
pulse tests among multiple wells. (rather unfractured) and slope (fracture-dominated) areas. The
heterogeneous matrix porosity and permeability of the sector
were extracted directly from the history-matched full-field Tengiz
DPDK model (Fig. 8). Water saturation was set to 1% above the
water/oil contact (WOC) and 100% below the WOC. The initial
pressure field of the sector was taken from the history-matched
T-2 full-field model at the beginning of the buildup test at T-1 on 13
October 2009 (Fig. 9). The matrix and facture could have differ-
T-1 ent pressures at the same time. The sector boundaries were
assumed at no-flow condition in current study. Ideally, a material-
balance region for the sector can be defined in the full-field
model, and sector boundary fluxes can be extracted and specified
for sector simulation. Historical rates of other wells in the sector
can be carried with simulation. However, we wanted to focus on
single-well test integration, and the secondary influence from off-
set wells was not considered in current study.

Fracture Distribution
N Permeability (md)
E Fracture density is the main variable describing the distribution of
fractures throughout the model. It is defined as fracture-surface
W S
0 0.1 0.2 0.3 0.4 0.5 area per unit volume, and is determined mainly from borehole
image logs. Fracture density was distributed throughout the sector
Fig. 8—Tengiz sector model matrix-permeability distribution. area as part of a full-field fracture model with the approach

10 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:35 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 11 Total Pages: 13

T-2 T-2
T-1 T-1

N Pressure (psi) N Pressure (psi)

E E

W S W S
9,000 9,200 9,400 9,600 9,800 10,000 9,000 9,200 9,400 9,600 9,800 10,000

(a) (b)

Fig. 9—Tengiz sector model initial pressure in (a) fracture and (b) matrix.

described by Tankersley et al. (2010). Fractures typically strike 2010). We render just the largest 5% of the effective fracture pop-
parallel or normal to the local platform-margin orientation. ulation as discrete features in our flow-simulation models because
Based on these spatially explicit geological model properties, of computer-performance limitations. The largest fractures should
a discrete-fracture network model was created with commercial have the biggest effect on fluid flow in the reservoir. They expose
software, in which length, height, and aperture were specified for the largest surface area of any fractures to the matrix; they con-
each fracture set. Two sets of discrete fractures constitute the frac- nect across greater distances than small fractures; they are more
ture network—one with orientation that locally parallels the con- likely to intersect with other fractures to form percolating net-
tour of the depositional slope of the reservoir, and the other works; and because we define aperture to be proportional to size,
perpendicular to the first set, which is in accord with both the well they have the largest apertures and, hence, permeability.
data and the character seen at outcrop analogs (Fig. 10). To match The fracture realization was then conditioned to image-log
the limitations of our unstructured grid-handling, fractures are data of selected wells (Fig. 10). The impact of 95% of the smaller
defined to dip parallel to the grid z-axis, thereby rendering the fractures that define the remaining component of the total fracture
fractures to dip vertically in the flow grid. Such steep dip reason- density was incorporated by upscaling them into the “matrix” por-
ably approximates the character of fractures in the reservoir. tion of the simulation model as an effective medium-permeability
The size of individual fractures cannot be measured directly in enhancement, with proprietary software (Tankersley et al. 2010).
a subsurface reservoir; hence, their size distribution is based on Based on geological information, static data, and the comparison
studies of fractures in outcrop analogs explicitly to address the of the corresponding transient responses, fracture Realization 5
Tengiz reservoir (Narr and Flodin 2012). The modeled size of (Fig. 11) was chosen as the most-likely fracture distribution. An
individual fractures is determined stochastically from a specified unstructured grid with one level of grid refinement near wells was
size distribution. Only “effective fractures” observed in the reser- generated to reproduce the interporosity flow behavior observed
voir are used as a basis for this geological model. Effective frac- in buildup tests at the wells in the sector (Fig. 12).
tures are the subset of the total fracture population that show
distinct evidence of dynamic response during drilling or logging, Integration of Buildup-Test Data
such as flow into or out of the wellbore based on PLT, tempera-
For the selected fracture distribution, the permeability and aper-
ture, and lost circulation (Narr et al. 2006; Tankersley et al.
ture were assumed uniform in each discrete fracture. The hetero-
geneity of fracture system comes from different fracture density
in different parts of the sector and different length and height of

T–2

T–1

N
E

W S
Fig. 10—Well image-log data used to generate fracture distribu-
tion in Tengiz sector model. Disks on wells show locations of
effective fractures seen on image logs. Fig. 11—Fracture Realization 5 for Tengiz sector model.

February 2016 SPE Reservoir Evaluation & Engineering 11

ID: jaganm Time: 13:35 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 12 Total Pages: 13

T-2
T-2
T-1
T-1

N
E
depth (m)

W S
4,000 4,500 5,000 5,500 6,000

Fig. 12—Unstructured grid for fracture Realization 5.

each fracture. The discrete-fracture model variables selected to be Case 1 also indicated earlier interporosity flow than the gauge
adjusted to match pressure-transient data are the multipliers of the data (red). With the input parameters of Case 1 in Eq. 3, the shape
heterogeneous matrix permeabilities (the original matrix-perme- factor a was estimated at 60 for the current fracture realization
ability distribution remained the same), fracture permeability, and near Well T-1. The values of xm and a were used in the following
fracture aperture. Following the proposed systematic procedure, steps to estimate the model input parameters for next iterations.
the sector-model parameters were adjusted to honor the buildup Because the effective permeability obtained from Case 1 was
data collected at T-1 in October 2009. The input parameters used 0.38 md, approximately 10 times smaller than 3.3 md derived
in each case and numerical well-test analysis results are listed in from gauge data, for Case 2, the fracture permeability kf was
Table 4. The numerical transient responses of each case are shown increased from 1,000 to 10,000 md. Its numerical transient-test
in Fig. 13. responses are shown in orange in the same log-log plot (Fig. 13b).
The base case, Case 1, shown as black curves in the log-log With the input parameters of Case 2 and the xm and a values from
plot (Fig. 13b), indicates much lower effective permeability. With Case 1 in Eq. 1, the expected effective permeability was calcu-
the DFM input parameters of Case 1 (Table 4), average km, kf, and lated as 3.73 md. The transient analysis of the numerical results
af in Eq. 1, the fracture spacing xm was calculated as 150 m, which estimated keff of 2.21 md (Table 4). For realistic heterogeneous
is consistent with the average value used to generate fracture reservoirs, Eqs. 1 through 3 are used as correlations to quantita-
Realization 5 near Well T-1. The numerical transient response of tively guide the selection of the input for next iterations. Because

(a) Fracture Realization #5 (c) One Solution for Realization #5

gauge data
Rate (m3/D) Pressure (bar)

Final DFM (ref)

8,400

buildup

7,900

N E 0
–1,250
0 100 200 300
W S
Time (hours)
(b) NWT with DFM
Pressure Change and Pressure

Pressure Change and Pressure

Case 1
Case 2
Case 3
Case 4
Derivative (bar)

Derivative (bar)
Pressure (psi)

1,000 Data
100

100 Case 4 + WBS

Data
10
0.01 0.1 1 10
10 Time (hours)
0.01 0.1 1 10
Time (hours)

Fig. 13—A systematic procedure of T-1 buildup integration in Tengiz model.

12 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:35 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 13 Total Pages: 13

WTEP_1
110 DEGC 113
K_SM4_COR_2 KH_SM4_1
Final DFM Results Consistent With PLT 0.01 MD 100 D MD-M 5400
QOZI_STBOPD_2
Depth K_SM4_2 KH_SM4_COR_1
(meters) 0 STB/OPD 6500
0.01 MD 100 D MD-M 5400
QOZT_STBOPD_2
K_PLT_NORM_1 PERM BOOST 2 KH_PLT_NORM_1
0.01 MD 100 0.5 FACT 500000 D MD 5400 0 STB/OPD 6500
Oil Rate (STB/D) ×000
800 1600 2400 3200

×100

×200

Most flow (95%) enters here

×300

×400

Fig. 14—T-1 flow profile of one solution for Realization 5.

the new result was moving in the direction as expected and in sim- the sector, indicating better fracture connectivity and interaction
ilar scale as predicted by the correlations, the same fracture real- between matrix and fracture in that area.
ization was kept, and further iterations were continued. Because Starting with the modified sector model that honors T-1
increasing kf 10 times led to the increase of keff of six times, to buildup data, a similar process was applied to Well T-2 in the
adjust the outcome to 3.3 md, kf needed to be increased further at same sector model. A buildup test conducted at T-2 in November
least three times. 2008 shows DP transient behavior. The initial pressure field of the
The transient analysis of Case 2 numerical results provided the sector had to be transferred from the Tengiz full-field model at the
estimation of interporosity flow coefficient k as 3.2  105, beginning of that test. The fracture spacing and shape factor near
whereas the estimation from the gauge data was 2.8  106. To T-2 were also calculated, which were different from those near T-
reduce k 10 times, while increasing kf three times, the average km 1. The challenge in integrating T-2 transient data was that fracture
needed to be decreased with a multiplier of 0.3 according to Eq. Realization 5 could not provide enough fracture connectivity near
3. Then, the input parameters of Case 3 were selected (Table 4). T-2 to generate the expected pressure response. Except for frac-
The numerical results of Case 3 (blue curves in Fig. 13b) tures conditioned to match well-based observations, the specific
brought the estimations of keff and k closer to the values derived locations of individual fractures throughout the reservoir in the
from gauge data (Table 4). The obtained fracture storage ratio x discrete-fracture model are determined stochastically. Therefore,
was 0.12. It needed to be increased to 0.22, the value from col- a new realization with more connected fractures around T-2 was
lected data. To do that, the fracture aperture would be doubled for created to repeat the procedure until a solution matching the
the next case according to Eq. 2. Because keff needed to be buildup data reasonably well was found. This new sector model
increased 1.3 times, from Eq. 1, kf would be decreased 0.8 times. was then checked to validate that it still honored the buildup tran-
The k value from Case 3 was 1.4  105. To match the data with sient response at T-1. The sector properties were adjusted further
k of 2.8  106, a multiplier of 0.3 for matrix permeability km was to ensure consistency.
needed based on Eq. 3. Because the properties in the well-drainage area have the dom-
After four iterations, the final numerical-model results (green inant impacts on the pressure-transient responses at that well, to
curves) matched both the test history in Cartesian plot and the honor transient data collected at multiple wells, adjusting the nu-
buildup pressure and pressure derivative data in log-log plot (Fig. merical model parameters in the drainage area of each well
13c). The early-time response (before 2 hours after shut-in) respectively would provide consistent model solution. The drain-
departed from gauge data because the numerical model neglected age area could be estimated from several methods based on mate-
wellbore-storage effects that were obvious in test data. After rial balance (Kamal 2009). In our study, we calculated the area
the wellbore storage was added, the numerical results could from the relative oil-production rate of each surrounding well and
match the buildup data closely all the way to approximately the PV near each well, and validated with the drainage area
30 hours until offset-well production started to influence the test- derived from the production-data analysis of long-term pressure
well response. and rate data.
This work-flow was successfully applied to integrate T-1
buildup-test data in the Tengiz discrete-fracture sector model. The
flow profile at T-1 of the obtained numerical solution is consistent Integration of Pulse-Test Data
with the flow data from production logging, indicating that most Just before the buildup test at T-2 in November 2008, a pulse test
of the flow enters the wellbore from a short-depth interval that was conducted from T-2 to T-1. The oil-production history at the
corresponds with the location of a natural fracture (Fig. 14). The active Well T-2 during the pulse test is shown at the bottom
pressure results of the sector model are shown in Fig. 15. The big- of Fig. 16, which had five major rate changes to create the
gest pressure change during buildup is in the northeast corner of pulses. The observation Well T-1 was shut in and continuously

February 2016 SPE Reservoir Evaluation & Engineering 13

ID: jaganm Time: 13:35 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 14 Total Pages: 13

T-2 T-2
T-1 T-1

N Pressure change (psi) N

E E Pressure change (psi)

W S S
W
0 50 100 150 200 250 300 0 50 100 150 200 250 300
(a) (b)

Fig. 15—Simulated pressure change in (a) fracture and (b) matrix during 100-hour buildup of one solution for Tengiz sector-model
Realization 5.

collecting pressure responses shown at the top of Fig. 16. The cor- selected interference-test-influenced areas to match the pulse-test
respondence between the pressure at T-1 and the production rate data. Using geological information (e.g., reservoir structure, fa-
at T-2 is not obvious, but the pressure derivative with respect cies, and fracture orientations) and model pressure-change maps
to time (the middle curve in Fig. 16) clearly indicates the direct could help to identify possible fast paths linking wells and the
correspondence between these two wells, implying good influence area of each test.
flow communication. The methods proposed by Deruyck et al. This new sector model was then checked for whether it still
1982 and Araujo et al. 1998 can be used to study the interference honored the buildup transient responses at T-1 and T-2. Further
responses. adjustments of the sector properties might be necessary to ensure
From synthetic case study, we learned that the dominant fac- consistency. For a full-field simulation study, the application of
tors controlling pulse-signal propagation are the conductivity of the proposed work-flow to honor individual well transient-test
the fractures connecting the two wells and the matrix porosity data and between-well pulse tests needs to be combined with
between the wells. This is consistent with previous reporting that global history matching to ensure material balance and the consis-
the highest diffusivity medium controls the response between tency with the long-term performance of the entire reservoir.
wells in multiple well testing (Woods 1970 and Kamal and Brig- The benefits of dynamic transient-data integration include that
ham 1975). Based on the sensitivity-study results, discrete-frac- it decouples the impacts of different formation properties, such as
ture model parameters were adjusted to honor pulse-test data from effective permeability, well skin factor, storativity ratio, and
T-2 to T-1. Higher kf and lower /m would lead earlier responses transmissibility between matrix and fracture; accelerates the data-
to pulses. Starting with the green case of the modified fracture matching process for the near-well local areas; and has the poten-
Realization 5, which matches buildup-test data from T-1 and T-2, tial to reduce the efforts in long-term history matching with full-
the pulse-test responses between these two wells were generated field simulation models.
(Fig. 17), which do not honor the measurements. Then, the frac-
ture permeability was increased 12 times to generate the brown Conclusions and Discussions
case, which has much quicker and bigger responses at T-1 than
1. NFR characterization is a complex problem with a high
the measured data (red curve). Then, a smaller region around T-1
degree of uncertainty. A multidisciplinary team approach
and T-2 in the northeast was selected, and its fracture permeabil-
(spanning geology, field operation, well testing, and numer-
ity was increased to 1,000 darcies and the rest of the sector with
ical simulation) is effective and necessary to solve a prob-
the same fracture permeability of 28 darcies as the green case to
lem as challenging as NFR characterization. Having an
generate the blue case, which matches the data reasonably well.
effective reservoir surveillance plan in place and collecting
In this example, the matrix porosity in the sector model remained
all possible data of all types (well logs, cores, image logs,
the same as the green case. In other situations, it may be necessary
production log, single-well and multiple-well transient
to adjust both fracture permeability and matrix porosity in
tests, and continuous downhole pressure and production
rates) are crucial in obtaining reasonable characterization.
2. A deep understanding of fundamental flow behavior and
dp/dt (bar/hr) p (bara)

692.25 Pressure response at T-1

proper calibration is needed before application to a full-
691
field problem. Our study found that, for a numerical DPDK
Pressure derivative at T-1 model, the grid size needs to be close to the fracture spacing
–0.003
in the reservoir near the well to be able to regenerate the
–0.004 representative interporosity-flow behavior. For a discrete-
fracture model, first level of grid refinement is likely to
reproduce the corresponding interporosity-flow signature.
qo (m3/D)

T-2 oil production

2,000 3. Dynamic data contain valuable information that one should
1,000 use to augment static data to improve reservoir characteri-
zation and to reduce uncertainty. A single-well drawdown/
buildup test provides reservoir properties of the investigated
11/4/2008 11/9/2008 11/14/2008 11/19/2008 11/24/2008 11/29/2008
local area, depending on the test duration. One can use
pulse-test data to derive properties of the fast path or the
Fig. 16—Pressure responses at Well T-1 to oil-production most conductive layers between wells, and to provide infor-
change at Well T-2 during pulse test in 2008. mation about reservoir connectivity.

14 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:35 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 15 Total Pages: 13

(a) Buildup Data

(b) Fracture Realization 5

Pressure and Pressure

Buildup at T-1

Derivative (bar)
T–2
10 T–1

Buildup at T-2
0.1 N
E
0.01 0.1 1 10
Time (hours) W S

(c) Pulse-Test Data

692
Pressure 691
at T-1 (bar)
690

DFM 1
DFM 2
0
dp/dt at T-1 DFM 3
(bar/hr) Gauge data

–0.004

Oil rate at 2,500

T-2 (m3/D)
1,250

11/4/2008 11/10/2008 11/16/2008 11/22/2008 11/28/2008 12/4/2008

Fig. 17—Integration of pulse test between T-2 and T-1 in Tengiz sector model.

4. Theoretical/analytical solutions, when available, should tion. The steps are straightforward to follow, and should be
always be used to guide numerical-simulation investiga- applicable for most conditions of naturally fractured reser-
tions. The representative fracture spacing and shape factor voirs. However, the application of a transient-data-integra-
near a well in any model realizations can be derived by ana- tion work-flow in a large-scale full-field numerical-
lyzing the numerically simulated transient-test responses simulation model, and its impacts on the forecast of field
and by comparing with the model input parameters. performance need further study.
Depending on the fracture/matrix configurations near each
well, the fracture spacing and shape factor may vary for dif-
ferent wells. These two parameters, together with the corre- Nomenclature
lations, Eqs. 1 through 3, between the pressure-transient
a ¼ aperture
analysis results and the average model input parameters
c ¼ compressibility
should be used to select the model input parameters for
C ¼ wellbore-storage coefficient
each iteration during the transient-data-matching process.
h ¼ sector-model thickness
This will reduce the number of iterations in numerical
k ¼ permeability
study.
KSIGMA ¼ product of matrix permeability and geometric shape
5. The work-flow of transient-data integration was applied
factor
successfully in a sector model in the southeast of Tengiz
L ¼ individual-fracture length
field. From this field study, we learned that to honor
p ¼ pressure
buildup/drawdown test data collected at multiple wells,
q ¼ flow rate
adjusting the numerical-model parameters in the drainage
r ¼ wellbore radius
area of each well respectively would provide a consistent
s ¼ well skin factor
model solution. In certain situations, the fracture distribu-
t ¼ time
tion at the first step of the work-flow may need to be
x ¼ fracture spacing
changed to provide the needed connectivity before continu-
a ¼ geometric shape factor in dual-porosity system
ing with the rest of the steps in the work-flow. To match
k ¼ interporosity flow coefficient
pulse-test data between wells, sometimes, it may require
x ¼ fracture storativity ratio
trial and error to identify the influence area and to adjust its
/ ¼ porosity
dominant properties, such as fracture permeability and ma-
trix porosity. The step-by-step integration procedure can be
used to any NFRs with commercial pressure-transient-anal-
ysis software and numerical reservoir simulator. Subscripts
6. A systematic procedure of integrating transient-test data in equiv ¼ equivalent
a numerical model combined with global history matching f ¼ fracture
is expected to lead to an improved reservoir characteriza- m ¼ matrix

February 2016 SPE Reservoir Evaluation & Engineering 15

ID: jaganm Time: 13:36 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 16 Total Pages: 13

t ¼ total Kamal, M. M. and Brigham, W. E. 1975. Pulse-Testing Response for

w ¼ well Unequal Pulse and Shut-in Periods. SPE J. 15 (5): 399–410. SPE-
5053-PA. http://dx.doi.org/10.2118/5053-PA.
Kamal, M. M., Pan, Y., Landa, J. L. et al. 2005. Numerical Well Testing:
A Method To Use Transient Testing Results in Reservoir Simulation.
Superscript Presented at the SPE Annual Technical Conference and Exhibition,
eff ¼ effective Dallas, USA, 9–12 October. SPE-95905-MS. http://dx.doi.org/10.2118/
95905-MS.
Kamal, M. M. 2009. Transient Well Testing, SPE Monograph, Vol. 23.
Acknowledgments Richardson, Texas: SPE.
The authors would like to thank Tengizchevroil and its partners, Karimi-Fard, M., Durlofsky, L. J., and Aziz, K. 2004. An Efficient Dis-
Chevron, ExxonMobil, KazMunayGas, and Lukoil, for the finan- crete Fracture Model Applicable for General-Purpose Reservoir Simu-
cial support and technical input for this work. lators. SPE J. 9 (2): 227–236. SPE-88812-PA. http://dx.doi.org/
10.2118/88812-PA.
King, G. R., Jones, M., Dagistanova, K. et al. 2012. Use of Brownfield Ex-
perimental Design Methods for Post-Processing Conventional History
References Match Results. Presented at the SPE Annual Technical Conference
Araujo, H. N., Andina, S. A., Gilman, J. R. et al. 1998. Analysis of Inter- and Exhibition, San Antonio, Texas, USA, 8–10 October. SPE-
ference and Pulse Tests in Heterogeneous Naturally Fractured Reser- 159341-MS. http://dx.doi.org/10.2118/159341-MS.
voirs. Presented at the SPE Annual Conference and Exhibition, New Lim, K. and Aziz, K. 1995. Matrix-Fracture Transfer Shape Factors for
Orleans, USA, 27–30 September. SPE-49234-MS. http://dx.doi.org/ Dual-Porosity Simulators. Journal of Petroleum Science and Engi-
10.2118/49234-MS. neering 13: 169–178. http://dx.doi.org/10.1016/0920-4105(95)00010-
Barenblatt, G. I., Zheltov, Iu. P., and Kochina, I. N. 1960. Basic Concepts F.
in the Theory of Seepage of Homogeneous Liquids in Fissured Rocks. Narr, W., Schechter, D. S., and Thompson, L. B. 2006. Naturally Frac-
J. Appl. Math. Mech. 24 (5): 1286–1303. http://dx.doi.org/10.1016/ tured Reservoir Characterization. Richardson, Texas: Society of Pe-
0021-8928(60)90107-6. troleum Engineers.
Basquet, R., Cohen, C. E., and Bourbiaux, B. 2005. Fracture Flow Prop- Narr, W. and Flodin, E. 2012. Shallow-Burial Fractures in Steep-rimmed
erty Identification: An Optimized Implementation of Discrete Fracture Carbonate Platforms: Outcrops in Canning Basin, NW Australia, as
Network Models. Presented at the 14th SPE Middle East Oil and Gas Analog for Tengiz Reservoir, Kazakhstan: Extended abs., Fundamen-
Show and Conference, Bahrain, 12–15 March. SPE-93748-MS. http:// tal Controls on Flow in Carbonates. Presented at the AAPG Hedberg
dx.doi.org/10.2118/93748-MS. Conference, Saint-Cyr Sur Mer, Provence, France, 8–13 July.
Casabianca, D., Jolly, R. J. H., and Pollard, R. 2007. The Machar Oil Rogers, S., Enachescu, C., Trice, R. et al. 2007. Integrating Discrete Frac-
Field: Waterflooding a Fractured Chalk Reservoir. Geological Society, ture Network Models and Pressure Transient Data for Testing Concep-
London, Special Publications 270 (1): 171–191. http://dx.doi.org/ tual Fracture Models of the Valhall Chalk Reservoir, Norwegian North
10.1144/GSL.SP.2007.270.01.12. Sea. Geological Society, London, Special Publications 270 (1):
Casciano, C., Ruvo, L., Volpi, B. et al. 2004. Well Test Simulation 193–204. http://dx.doi.org/10.2118/GSL.SP.2007.270.01.13.
Through Discrete Fracture Network Modeling in a Fractured Carbon- Samaniego-V, F. and Cinco-Ley, H. 2009. Chapter 10–Naturally Fractured
ate Reservoir. Petroleum Geoscience 10 (4): 331–342. http:// Reservoirs, 221–280, Monograph Series No. 23. In Transient Well
dx.doi.org/10.1144/1354-079303-590. Testing, ed. M. M. Kamal. Richardson, Texas: SPE.
Coats, K. H. 1989. Implicit Compositional Simulation of Single-Porosity Sarma P. and Aziz, K. 2006. New Transfer Functions for Simulation of
and Dual-Porosity Reservoirs. Presented at the SPE Symposium on Naturally Fractured Reservoir With Dual-Porosity Models. SPE J. 11
Reservoir Simulation, Houston, USA, 6–8 February. SPE-18427-MS. (3): 328–340. SPE-90231-PA. http://dx.doi.org/10.2118/90231-PA.
http://dx.doi.org/10.2118/18427-MS. Tankersley, T., Narr, W., King, G. R. et al. 2010. Reservoir Modeling To
Collins, J., Narr, W., Harris, P. M. et al. 2013. Lithofacies, Depositional Characterize Dual Porosity, Tengiz Field, Republic of Kazakhstan.
Environments, Burial Diagenesis, and Dynamic Field Behavior in a Presented at the SPE Caspian Carbonates Technology Conference,
Carboniferous Slope Reservoir, Tengiz Field (Republic of Kazakh- Atyrau, Kazakhstan, 8–10 November. SPE-139836-MS. http://
stan), and Comparison With Outcrop Analogs. In Deposits, Architec- dx.doi.org/10.2118/139836-MS.
ture, and Controls on Carbonate Margin, Slope, and Basinal Settings, Warren, J. E. and Root, P. J. 1963. The Behavior of Naturally Fractured
ed. K. Verwer, T. E. Playton, and P. M. Harris. SEPM Special Publica- Reservoirs. SPE J. 3 (3): 245–255. SPE-426-PA. http://dx.doi.org/
tion 105, 50–83. 10.2118/426-PA.
Deruyck, B. G., Bourdet, D. P., DaPrat, G. et al. 1982. Interpretation of In- Wei, L., Hadwin, J., Chaput, E. et al. 1998. Discriminating Fracture Pat-
terference Tests in Reservoirs With Double Porosity Behavior–Theory terns in Fractured Reservoirs by Pressure Transient Test. Presented at
and Field Examples. Presented at the 57th SPE Annual Technical Con- the SPE Annual Technical Conference and Exhibition, New Orleans,
ference and Exhibition, New Orleans, USA, 26–29 September. SPE- USA, 27–30 September. SPE-49233-MS. http://dx.doi.org/10.2118/
11025-MS. http://dx.doi.org/10.2118/11025-MS. 49233-MS.
Hui, M. H., Kamath, J., Narr, W. et al. 2007. Realistic Modeling of Frac- Woods, E. G. 1970. Pulse-Test Response for a Two-Zone Reservoir. SPE
ture Network in a Giant Carbonate Reservoir. Presented at the Interna- J. 10 (3): 245–256. SPE-2570-PA. http://dx.doi.org/10.2118/2570-PA.
tional Petroleum Technology Conference, Dubai, 4–6 December. SPE-
11386-MS. http://dx.doi.org/10.2118/11386-MS.
Hui, M. H., Mallison, B., Lim, K. T. et al. 2008. An Innovative Workflow Yan Pan is a senior reservoir engineering adviser with Chevron
To Model Fracture in Tengiz Oil Field. Presented at the International Energy Technology Company. Her interest is focused on well
testing, production-data analysis, and dynamic-data integra-
Petroleum Technology Conference, Kuala Lumpur, 3–5 December.
tion in reservoir models. Pan is the 2010 recipient of SPE West-
SPE-12572-MS. http://dx.doi.org/10.2118/12572-MS. ern North America Regional Formation Evaluation Award and
Hui, M. H., Mallison, B. T., Heidary-Fyrozjaee, M. et al. 2013. The served as a 2011–2012 SPE Distinguished Lecturer. She holds MS
Upscaling of Discrete Fracture Models for Faster, Coarse-Scale Simu- and PhD degrees in petroleum engineering from Stanford
lations of IOR and EOR Processes for Fractured Reservoirs. Presented University.
at the SPE Annual Technical Conference and Exhibition, New Orle-
Mun-Hong Hui is a research scientist with Chevron’s Energy
ans, USA, 30 September–2 October. SPE-166075-MS. http:// Technology Company based in San Ramon, California, and
dx.doi.org/10.2118/166075-MS. has 12 years of experience in the industry. He holds MSc/PhD
Izadi, M. and Yildiz, T. 2009. Transient Flow in Discretely Fractured degrees in petroleum engineering from Stanford University
Porous Media. SPE J. 14 (2): 362–373. SPE-108190-PA. http:// and a BEng degree in chemical engineering from the National
dx.doi.org/10.2118/108190-PA. University of Singapore. Hui’s research interests are in the

16 February 2016 SPE Reservoir Evaluation & Engineering

ID: jaganm Time: 13:36 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058

 REE165322 DOI: 10.2118/165322-PA Date: 27-January-16 Stage: Page: 17 Total Pages: 13

simulation and upscaling of complex recovery physics, espe- Eric Flodin is currently Earth science adviser at Chevron Pacific
cially for NFRs. Indonesia in Minas, Sumatra, Indonesia. Prior roles include de-
velopment geologist at Tengizchevroil, fractured-reservoir
Wayne Narr is senior research consultant for Chevron Energy
consultant at Chevron Energy Technology Company, and
Technology Company. He is a structural geologist, and his inter-
faculty in the Geosciences Department, Indiana University/
est is focused on characterization of NFRs, ranging from analog Purdue University, Fort Wayne. Flodin holds a BS degree in ge-
studies to characterization methods to modeling. Narr has ology from Indiana University/Purdue Univeristy, Indianapolis
taught schools on NFRs and has published research articles on
and a PhD degree in geology from Stanford University.
them as well as a book published by SPE. He holds a PhD degree
from Princeton University, an MSc degree from University of To- Philip W. Bateman is currently the Gorgon Subsurface Man-
ronto, and a BS degree from Pennsylvania State University. ager for Chevron, and his previous position was Asset Develop-
Gregory King is a reservoir engineer for Chevron. He is a coau- ment Manager for Tengizchevroil. He has been with Chevron
thor of the 2001 SPE textbook, Basic Applied Reservoir Simula- for 27 years. Bateman’s interests include carbonate and frac-
tion, and the author or coauthor of more than 40 publications tured reservoir characterization, petrophysics, and reservoir-
and technical papers. King holds BS, MS, and PhD degrees modeling work flows. He holds a PhD degree in geology from
from the Pennsylvania State University, all in petroleum and the Colorado School of Mines. Bateman is a member of SPE.
natural-gas engineering. He has served on numerous SPE com-
Chris Laidlaw is a deputy asset manager for Chevron Thailand.
mittees as a committee member or committee chair. King
He has been with Chevron for 28 years, working in multiple
was awarded the SPE Regional Reservoir Dynamics and international locations. Laidlaw’s professional interests are in
Description Award for the Russia and Caspian Region in 2014. reservoir management and subsurface asset development.
Terrell H. Tankersley is currently senior adviser geologist with He holds an MSc degree in petroleum engineering from Impe-
Chevron Indonesia Company. He has been with Chevron- rial College, London.
related companies for almost 35 years, focusing on develop-
Hai Vo currently works for Chevron Energy Technology Com-
ment geology and reservoir modeling.
pany. He holds MS and PhD degrees, both in petroleum engi-
Steve Jenkins is development geologist with Karachaganak neering, and a PhD degree minor in computational
Petroleum Operating Company, Kazakhstan. He holds mathematics from Stanford University. Vo’s research interests
degrees in geology and geophysics, with an interest in climb- include reservoir characterization and field-production
ing over and in carbonate outcrops. optimization.

February 2016 SPE Reservoir Evaluation & Engineering 17

View publication stats

ID: jaganm Time: 13:36 I Path: S:/REE#/Vol00000/150058/Comp/APPFile/SA-REE#150058