Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Gas condensate is a challenging fluid in reservoir due to difficulties arising from condensate dropout
Received 15 January 2014 around the well. Condensate blockage complicates simulating the behavior of this fluid especially in well
Received in revised form testing. The common approach in gas condensate well testing is using gas single-phase pseudo-pressure.
17 March 2014
In this study, statistical approach is applied to well test analysis in order to investigate the effect of fluid
Accepted 18 March 2014
Available online
composition on reservoir behavior. Mixture design controls the number of otherwise numerous well
testing of fluids to a limited number. The output data of well testings are then used for multivariable
regression to create proxy models of reservoir and fluid properties including initial pressure of reservoir,
Keywords:
Gas condensate
gas permeability in condensate zone and maximum liquid dropout of fluids. Results show that proposed
Well testing model efficiently simulates mentioned properties with high R-squared values. Results also show that the
Design of experiment intensity of condensate formation in the reservoir is a strong function of rich components composition in
Mixture design the reservoir.
Ó 2014 Published by Elsevier B.V.
http://dx.doi.org/10.1016/j.jngse.2014.03.015
1875-5100/Ó 2014 Published by Elsevier B.V.
A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376 369
mc4 vmðPÞ
V2 mðpÞ ¼ (1)
k vt
Where,
Zp
pdp
mðpÞ ¼ 2 (2)
mZ
p0
Zpwf !
rg krg ro kro
mðpÞ ¼ þ dp (3)
mg mo
pwfs
Two-phase pseudo pressure has a higher accuracy than single- Fig. 3. Procedure of this study.
phase pseudo pressure because in its definition both gas and
condensate properties have been included. However, in two-phase
compositions. Equation (4) shows that molar composition should
pseudo pressure, we need relative permeability curves, GOR and
add up to one:
PVT properties to be accurate. Roussennac (2001) showed that er-
ror in these parameters has significant effect on calculation of two-
zC14 þ zC10 þ zC7 þ znC4 þ zC1 ¼ 1 (4)
phase pseudo pressure and consequently property calculations will
not be accurate. Hence using two-phase pseudo pressure could Compositions of a very lean and a very rich fluid are added as
have high uncertainty and therefore if one cannot trust the input boundary constraints for design of this study which are constraints
data, single-phase gas pseudo pressure might be more suitable to in addition to the essential constraint of mixture designs. Therefore
be used. For more details, see Yousefi et al. (2014). DOE software will always select the design points between these
boundary constraints. Table 1 shows composition of these fluids. D-
optimal algorithm was employed to cover all feasible mixture
4. Methodology
compositions.
In this section produced condensate near the wellbore and the
4.2. Constructing fluids by PVT software
effect of variation of components composition on the interpretation
of well testing in gas condensate reservoirs are investigated. To do
In this part, by employing a commercial PVT module, and using
this, twenty gas condensate fluids are constructed using appro-
three variable PengeRobinson equation of state maximum liquid
priate module of a PVT software. Compositions of twenty fluids are
dropout is determined. PVT properties of the two boundary fluids
designed by a statistical software. Subsequently, using the
are shown in Table 2.
mentioned synthetic fluids, reservoir simulation is carried out with
a compositional simulator to produce pressure versus time data.
4.3. Reservoir model simulation
This data is imported to a well test analysis software to obtain
reservoir and fluid properties which is a back calculation procedure.
To simulate gas condensate reservoir, a compositional simulator
Obtained reservoir and fluid properties are finally returned to the
was used.
statistical software as response to develop polynomial models by
Following assumptions are considered in reservoir simulation:
multivariable regression. This algorithm is illustrated in Fig. 3.
The reservoir has no geological complexities. The model is
single-well and geometry is one-dimensional radial. Flow regime is
4.1. Designing synthetic fluids by statistical software assumed infinite acting by definition of large grid blocks at outer
boundary to dissipate no flow boundary effect. A fine-grid block
In order to study effect of fluid richness on different properties, a system is defined around the wellbore to better simulate conden-
D-optimal mixture design was chosen to determine component sate formation effects. The statistical petrophysical properties of
A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376 371
Table 1 Table 3
Molar compositions of twenty two synthetic fluids. Well and reservoir data in the simulation model.
Lean fluid 0.77 0.111 0.102 0.009 0.008 Initial reservoir pressure, psi Dew point pressure þ 250
Rich fluid 0.89 0.083 0.022 0.003 0.002 Reservoir temperature, F 200
1 0.816 0.083 0.086 0.006 0.008 Reservoir thickness, ft 50
2 0.849 0.100 0.045 0.003 0.002 Top of reservoir height from ground surface, ft 9000
3 0.801 0.099 0.083 0.009 0.008 Rock and fluid compressibility, psi1 6.02E-6
4 0.871 0.095 0.023 0.009 0.002 Absolute permeability, mD 5
5 0.881 0.083 0.022 0.006 0.008 Porosity, % 20
6 0.804 0.083 0.099 0.009 0.004 Well radius, ft 0.354
7 0.860 0.111 0.023 0.004 0.002 Reservoir drainage radius, ft 7000
8 0.797 0.093 0.102 0.006 0.002 Produced flow rate, MMscf/day 5
9 0.805 0.083 0.101 0.003 0.008
10 0.782 0.111 0.102 0.003 0.002
11 0.841 0.084 0.070 0.003 0.002
12 0.775 0.111 0.102 0.009 0.003 skin are negligible and therefore at pressures above the dew point
13 0.818 0.111 0.059 0.007 0.005 (single phase) model is acceptable. Consequently, any variation in
14 0.822 0.096 0.077 0.003 0.002 well and reservoir properties below the dew point is because of
15 0.887 0.084 0.022 0.003 0.004
condensate blockage around the wellbore. Fluid composition below
16 0.867 0.083 0.042 0.006 0.002
17 0.849 0.083 0.051 0.009 0.008 the dew point pressure controls condensate blockage. Hence, well
18 0.847 0.111 0.025 0.009 0.008 test analysis was carried out separately for each of the twenty
19 0.828 0.111 0.049 0.004 0.008 synthetic fluids.
20 0.773 0.111 0.102 0.006 0.008
In all reservoir models, the initial pressure of the reservoir has
been taken to be 250 psi above the dew point pressure, which
makes equal condition for all twenty fluids.
model are assumed to be constant in the reservoir and the me- Single-phase pseudo pressure method was used for well test
chanical skin factor is zero. analysis. As it was mentioned, when reliable PVT data is unavai-
Data of synthetic reservoir and well are shown in Table 3. In lable, single phase gas pseudo pressure is utilized. Radial composite
addition, the relative permeability data for gas-oil and water-oil are model is the most common model for well test interpretation. In
imported to model. radial composite analysis, there are two concentric zones around
Also, in all simulation models, effects of non-Darcy flow and the wellbore with different permeabilities. In this study, the inner
capillary number were considered in order to make the results zone represents gas effective permeability in condensate and the
more reliable. Increasing of capillary number near the wellbore has outer zone represents single phase gas permeability.
two important effects namely decrease of residual oil saturation For the twenty synthetic fluids, production rate is assumed to be
and changing relative permeability curves from immiscible to fixed at 5 million cubic feet per day and production and shut-in (for
miscible situation. buildup test) times were assumed equal and to be 10.11 days. As
Forchheimer added non-Darcy term to Darcy flow equation to example, for fluids 3 and 11, pressure versus time response in
explain the relationship between velocity and drawdown in porous logarithmelogarithm scale is illustrated in Figs. 6 and 7 respec-
media in high gas rates (Forchheimer, 1914). tively. With reference to these figures, it is apparent that perme-
ability estimated for richer fluid (11) would be higher than the
leaner one (3).
4.4. Well test analysis
Outputs of the software (pressure versus time data) and physical 4.5. Reservoir and fluid property modeling by statistical software
properties of the model were used as input to the well test
software. Since well testing in gas condensate reservoirs could be a
At first, to check the model validity, well test analysis is carried problematic task, a polynomial proxy model is proposed to simu-
out above dew point where there is no condensate in reservoir late the results. To do this, gas permeability of condensate zone,
and only single-phase gas is present. Initial reservoir pressure is initial reservoir pressure and maximum liquid dropout are taken as
set at 5000 psi. Other features of the model are listed in Table 3. response for the proposed statistical software. The polynomials are
Fluid No. 19 is selected and drawdown test for 19 days was then derived from multivariable regression in terms of component
considered.
Fig. 4 illustrates bottomhole pressure drop versus time for this
model. Logarithmelogarithm diagram of this model is shown in
Fig. 5. This figure shows a good agreement between input charac-
teristic data and the calculated ones. Skin factor is predicted to be
0.008 and permeability to be 4.89 mD. Errors for permeability and
Table 2
General properties of lean and rich fluids in the model.
Fig. 5. Logarithmelogarithm plot for fluid No. 19 in drawdown test above dew point pressure.
Fig. 6. Logarithmelogarithm plot for fluid 3 in buildup test below dew point pressure.
compositions. Backward elimination reduces the number of co- insignificant ones in polynomial. In this study 0.05 was taken as the
efficients needed for polynomial model. Because it is unknown uncertainty value.
which terms are important in polynomial model, backward elimi-
nation is a more suitable approach for choosing terms in poly- 5. Results and discussion
nomial instead of forward or stepwise elimination approach.
Backward approach requires an uncertainty value, between zero The procedure illustrated in Fig. 3 is applied to develop a robust
and one, in order to distinct between significant terms and model to the three mentioned reservoir and fluid properties.
Fig. 7. Logarithmelogarithm plot for fluid 11 in buildup test below dew point pressure.
A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376 373
Table 4
Reservoir and fluid properties in the simulation model.
Fluid sample Gas absolute permeability, Gas permeability in condensate Maximum liquid drop out, Initial pressure, psi
mD zone, mD percent
Fig. 8. Crossplots for gas permeability of damaged zone, initial pressure and maximum liquid drop out.
374 A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376
Fig. 10. 3D plot for gas permeability in condensate zone. (a): high level of heavier components, (b): low level of heavier components.
A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376 375
Fig. 11. 3D plot for initial pressure. (a): high level of heavier components, (b): low level of heavier components.
Fig. 12. 3D plot for maximum liquid dropout. (a): high level of heavier components, (b): low level of heavier components.
For maximum liquid dropout a logarithmic transformation is Scattered trend in these plots show desired behavior of polynomial
used to model the data more accurately as shown by equation (7): models in different design points. Scattered trend and sets of plus
and minus residuals indicate that mathematical model predictions
lnðMLÞ ¼ ð1:1833ÞzC1 ð17:2948ÞzC4 ð144:537ÞzC7 are higher than experimental values in some design points and
ð425:512ÞzC10 þ ð77:1334ÞzC14 þ ð219:229ÞzC1 zC7 lower in other ones.
þ ð5174:3ÞzC4 zC10
5.3. Model interpretation
(7)
In Figs. 10e12, 3D plots for the three properties are illustrated.
5.2. Model validation Each reservoir and fluid property is investigated in two cases: a-
‘high level’, measuring high concentration of a particular compo-
R-squared values of model are shown in Table 5. High R-squared nent and b-‘low level’, measuring low concentration of the same
values indicate efficient modeling of properties with polynomials. component.
Adjusted R-squared value determines accuracy of model with re- Gas permeability 3D plots (Fig. 10) show that in high levels of C10
gard to the number of model terms and is an indicator of increase in and C14 maximum permeability happens at highest value of C1. It
model accuracy by adding a new term. It should be noticed that by means that simultaneous high composition of C1, C10 and C14
adding new terms Adjusted R-squared value goes up at initial components leads to increasing permeability and consequently
stages but after reaching a certain number of terms in the poly- reducing condensate blockage. In low levels of C10 and C14, model
nomial, Adjusted R-squared value starts to decrease by adding new shows maximum permeability where C4 is at its maximum value.
terms. Small difference between R-squared value and Adjusted R- Therefore, in case of low level C10 and C14 any increase in C4 will
squared values indicate polynomials have no inappropriate co- result in permeability increase which means decrease in conden-
efficients and corresponding terms. sate blockage. Consequently, in situations where C1 and C4 have
Crossplots for these properties are shown in Fig. 8 which con- equal chances of injection into a reservoir, C1 could be a good choice
firms efficiency of models. In order to investigate existence of pure for injection in rich condensate reservoirs and C4 a better candidate
error for models, residual plot vs. run number is plotted in Fig. 9. for lean condensate reservoirs.
376 A. Azamifard et al. / Journal of Natural Gas Science and Engineering 18 (2014) 368e376
Initial reservoir pressure 3D plots (Fig. 11) show that for high Dejean, J.-P., Blance, G., 1999. Managing Uncertainties on Productions Using Inte-
grated Statistical Methods. paper SPE 56696, presented at the SPE Annual
levels of C10 and C14, increasing C1 leads to increasing initial pres-
Technical Conference and Exhibition held in Houston, Texas, 3e6 October.
sure. For low levels of C10 and C14, the higher C7 composition, the Eide, A.L., Holden, L., Reiso, E., Aanaonsen, S.I., 1994. Automatic History Matching by
higher initial pressure. In both cases, increasing C4 composition, Use of Response Surfaces and Experimental Design. presented at 4th European
decreases initial pressure. Conference on Mathematics of Oil Recovery, Røros, Norway, 7e10 June.
Fevang, O., Whitson, C.H., 1995. Modeling Gas Condensate Deliverability. SPE 30714.
3D plots for maximum liquid dropout is illustrated in Fig. 12. For, Forchheimer, P., 1914. Hydraulik. Teubner, Leipzig, pp. 116e118. Chap. 15.
both cases liquid dropout increases by increasing intermediate Friedmann, F., Chawathé, A., Larue, D.K., August 2003. Assessing uncertainty in
component (C7). While for high levels of C10 and C14 the liquid drop channelized reservoir using experimental designs. SPE Reserv. Eval. Eng., 264e
274.
out decrease by decreasing C4, for low levels of C10 and C14, the Fussel, D.D., 1973. single well performance for gas condensate reservoir. J. Pet.
liquid dropout increase by decreasing C4. Technol. 255, 860e870.
Gary, W. Oehlert, 2010. A First Course Design and Analysis of Experiments. Uni-
versity of Minnesota,.
6. Conclusions Gringarten, A.C., Al-Lamki, A., Daungkaew, S., Mott, R., Whittle, T.M., 2000. Well Test
Analysis in Gas Condensate Reservoirs. SPE 62920.
In this study, damaged permeability in condensate zone, initial Gringarten, A.C., Ramey, H.J., Raghavan, R., 1975. Applied Pressure Analysis of
Fractured Wells.
reservoir pressure and maximum liquid dropout in terms of com- Jones, J.R., Raghavan, R., 1988. Interpretation of Flowing Well Response in Gas
ponents compositions were studied by means of statistical design. Condensate Wells. SPE paper 14204.
Condensate fluids obtained by mixture design and well test simu- Landa, J.L., Güyagüler, B., 2003. A Methodology for History Matching and the
Assessment of Uncertainties Associated with Flow Prediction. paper SPE 84465,
lation data are used to create polynomial models for reservoir and
presented at the SPE Annual Technical Conference and Exhibition held in
fluid properties. On the basis of introduced procedure, the Denver, Colorado, USA, 5e8 October.
following conclusions are made: Manceau, E., Meaghani, M., Zabalza-Mezghani, I., Roggero, F., 2001. Combination of
Experimental Design and Joint Modeling Methods for Quantifying the Risk
Associated With Deterministic and Stochastic Uncertainties e an Integrated
1. Gas permeability in condensate zone, initial reservoir pressure Test Study. paper SPE 71620 presented at the 2001 SPE Annual Technical
and maximum liquid dropout were successfully modeled with Conference and Exhibition held in New Orleans, Louisiana, 30 Septembere3
good accuracy employing multivariable regression and back- October.
Miller, C.C., Dyes, A.B., Hutchinson, C.A., 1950. Estimation of Permeability and
ward elimination. Reservoir Pressure from Bottom-hole Characteristics. AIME.
2. The simple polynomial models developed in this study can Moses, P.L., Donohoe, C.W., 1962. Gas Condensate Reservoirs, Petroleum Engineer-
replace complex well test analysis of gas condensate reservoirs. ing Handbook, vol. 39. SPE, pp. 1e28,.
Mott, R., 2002. Engineering Calculation of Gas Condensate Well Productivity. SPE
3. Variations of heavy components compositions in fluid samples 77551.
lead to different behaviors in damaged zone permeability, initial Natrella, M., 2014. NIST/SEMATECH e-Handbook of Statistical Methods. http://
reservoir pressure and maximum liquid drop out. www.itl.nist.gov/div898/handbook/.
O’Dell, H.G., Miller, R.N., 1967. Successfully cycling a low-permeability high yield gas
4. The 3D plots obtained in this study suggest that; in situations condensate reservoir. J. Pet. Technol., 41e47.
where C1 and C4 have equal chances of injection into a reservoir, Raghavan, R., Chu, W.C., Jones, J.R., 1995. Practical Considerations in the Analysis of
C1 could be a good choice for injection in rich condensate res- Gas Condensate Well Tests. SPE paper 30576.
Roussennac, B., 2001. Gas Condensate Well Test Analysis (MSc thesis). Stanford
ervoirs and C4 a better candidate for lean condensate reservoirs.
University, USA.
Schroeter, T., Hollaender, F., Gringarten, A.C., 2004. Deconvolution of well test data
References as a nonlinear total least squares problem. SPE J. 9 (4), 375e390.
Theis, C.V., 1935. The relationship between the lowering of the piezometric surface
Boe, A., Skjaeveland, S.M., Whitson, C.H., 1981. Two-phase Pressure Test Analysis. and the rate and duration of discharge using ground-water storage.
SPE paper 10224. Wall, C.G., 1982. Characteristics of Gas Condensate Reservoirs and Traditional Pro-
Bourdet, D., Ayoub, J.A., Pirard, Y.M., 1983. Use of the Pressure Derivative in Well duction Methods. Oyze Technical Service, pp. 1e12.
Test Interpretation. Yadavalli, S., Jones, J., 1996. Interpretation of Pressure Transient Data from Hy-
Chopra, A., Carter, R.D., 1986. Proof of the Two-phase Steady State Theory for Flow draulically Fractured Gas Condensate Wells. SPE 36556.
Through Porous Media. SPE 14472. Yousefi, S.H., Eslamian, A., Rashidi, F., 2014. Investigation of well test behavior in gas
Damsleth, E., Hage, A., Volden, R., 1992. Maximum information at minimum cost: a condensate reservoir using single-phase pseudo-pressure function. Korean J.
North Sea field development study with an experimental design. J. Pet. Tech- Chem. Eng. 31 (1), 20e28.
nol., 1350e1356.