Escolar Documentos
Profissional Documentos
Cultura Documentos
A New Method for History Matching and Forecasting Shale Gas Reservoir
Production Performance with a Dual Porosity Model
Orkhan Samandarli, Hasan Al-Ahmadi, and Robert A. Wattenbarger, SPE, Texas A&M University
This paper was prepared for presentation at the SPE North American Unconventional Gas Conference and Exhibition held in The Woodlands, Texas, USA, 12–16 June 2011.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
This paper presents a new method for history matching production of shale gas wells and future forecasting. The method is
based on linear dual porosity analytical solutions which includes horizontal wells with multistage fractures.
A linear dual-porosity model assumes hydraulic fractures as a secondary porosity system and conduit to flow. Homogeneous
matrix blocks are sources of primary porosity that feed fluids to the hydraulic fractures. A Systematic analysis was done to
determine the main parameters affecting transient flow regimes. Since main flow regimes observed in the field are Bilinear and
Linear, more attention was given to those regimes to match with the field data. Analytical solutions are modified for gas
properties and desorption of gas from matrix surface.
Besides constant bottomhole pressure production, variable bottomhole pressure cases are also included in this paper. The
proposed method was applied to history match the production of shale gas wells from the Barnett, Woodford, and Fayetteville
plays. The main parameters found from history matching are effective matrix and fracture permeabilities and fracture half-
length. In wells with successful matches, future production can be forecast with some confidence.
Introduction
Gas production from shales has been increasing rapidly in United States. Along with the economic success comes the
complexity of understanding and modeling gas production from horizontal wells with multistage hydraulic fractures in shale
gas reservoirs.
Two reservoir engineering practices are often used in the industry – history matching and forecasting. Both are important in
reservoir engineering and management. By history matching production of a shale gas well, we can get an idea about the
quality of the fracturing job, and the quality and size of the reservoir. Forecasting is used to calculate the annual reserves of the
wells and reservoirs.
Different authors have proposed different methods for history matching and forecasting of the shale gas wells. Some of them
use simulation (Cipolla 2009, Freeman, 2010), some use analytical models, (Bello and Wattenbarger 2008, 2010a, 2010b) and
some use empirical models (Ilk 2010, Valko 2009). Lee and Sidle (2010) explained the advantages and disadvantages of
simulation, analytic, and empirical models in forecasting. According to their work, empirical models are easy to apply for
thousands of wells in a short period of time and accurate enough to estimate annual reserves.
Nobakht et al. (2010) described a forecasting procedure for tight/shale gas wells with empirical methods that take the end of
linear flow from analytical solutions. According to them, assumption of the drainage area would be the only unknown in
estimating end of linear flow, which can be used in decline curve forecasting. Since the end of transient linear flow depends on
drainage area (in horizontal well case it is SRV), it directly depends on the fracture half length that is unknown and desired to
be solved for. Ambrose et al. (2011) described a hybrid method for forecasting which is based on the procedure initially
proposed by Nobakht et al. (2010).
2 SPE 144335
A main advantage of empirical methods is their ease of use for forecasting. On the other hand, they are not consistent and need
to be modified for different cases. Especially, the Arp’s b value used in empirical models is not always 0.5 (Nobakht et al.
2010, Ambrose et al. 2011) and usually changes with time as the reservoir is depleting.
The model proposed in this paper combines the rigorousness of analytical solutions in history matching and easiness of
empirical equations in long term forecasting. Application of this method is described by Samandarli et al. (2011) and based on
linear dual porosity model initially proposed by Bello and Wattenbarger (2008). Samandarli et al. (2011) describes how to
apply the model to history match Bilinear and Linear flow regimes with corrected gas properties and adsorbed gas for shale
gas wells. This paper provides a more detailed overview of parameters affecting different transient regions of linear dual-
porosity model. One of the objectives of this paper is to clarify which parameters group can be determined from each region.
Moreover, a more general case, variable rate, variable drawdown is analyzed without including any type of superposition of
time which could result in bias to one or other type of flow regime.
-3 -5 -7
Fig. 1—Illustration of the five flow regions for a slab matrix dual porosity linear reservoir (yeD = 100); λ =10 , 10 , 10 for values of ω =
-3
10 . From (Bello and Wattenbarger, 2010).
Here we will try to show parameters affecting different flow regimes to clarify which parameters can be determined from each
region. Before discussing each region, it would be good to mention some reservoir properties which are assumed to be known:
total porosity, perforated wellbore length, number of fractures, gas saturation, water and formation compressibility, initial
reservoir pressure and reservoir thickness. Below is the summary of results derived by using dimensionless equations showed
by Samandarli et al. (2011) and asymptotic equations proposed by Bello and Wattenbarger (2010b). Appendix shows a
detailed derivation of these results.
Region 1
This region is the transient linear flow from hydraulic fractures to the horizontal wellbore.
qg
m pi m pwf
1
2 1 1 1
Acw Vct F2 k F 2 1 ............………………………………………………………… (1)
0.00633
2844 T t 2
As it is shown in Eq. 1, the only unknowns we need to find to match rates are effective fracture porosity and permeability.
SPE 144335 3
Region 2
This region is Bilinear flow where two simultaneous transient linear flows occur, one from the matrix to hydraulic fractures,
and the other one from hydraulic fractures to the well. Bilinear flow has been seen in shale gas wells from time to time, and
duration might be significantly long, possibly a year or two.
m pi m pwf 12ct
1
2
1 1 1
qg 2
Acw k F 2 k m 4 1 …………………….............………………..…………………….. (2)
14409.126T 0.00633LF t 4
The main unknowns in this equation are effective fracture and matrix permeabilities.
Region 3
This region is equivalent to homogeneous matrix linear flow without hydraulic fractures.
m pi m pwf ct
1
2
12 1
qg 0.00633 k m Acm 1 …………………….........………………………………………….……. (3)
2844 T t 2
If a horizontal well cross section is used, the only unknown in this equation would be matrix permeability. However, this
equation has been used to analyze transient linear flow from the matrix to the hydraulic fractures (Bello and Wattenbarger
2008, Nobakht et al. 2010), which makes Acw equal to Acm, where the latter depends on fracture half-length – one of the
unknown parameters.
Region 4
This region is transient linear flow from the matrix to hydraulic fractures. The half slope seen in field data corresponds to this
region because hydraulic fracture linear flow would occur in a few hours and could not be recorded by gauges.
m pi m pwf 12ct
1
2
1 1
qg 2
Acw k m 2 y e 1 ..........................……………………………………..………….…. (4)
2844 3 T 0.00633LF t 2
To match the rates with field data, effective matrix permeability and fracture half length should be known.
Region 5
This region is related to the boundary dominated flow of SRV or interference effects between hydraulic fractures. No
asymptotic equation is derived for this region, but simulation on synthetic data indicates that if matrix linear flow occurs, the
end of the half slope will depend on the value of fracture half length, that is, the longer the fracture half length, the later the
curve will bend on log-log plot of rate versus time (Fig. 2).
Fig. 2—Beginning of boundary effects for different yeD values in linear dual porosity model. As yeD increases the boundary effects are
later in time.
4 SPE 144335
Table 1 summarizes flow regions and unknown parameters group which govern their behavior.
qg
m pi m p wf
1
2 1 1 1
Acw Vct F2 k F 2 1
1 0.00633
Vct F2 k F 2
1 1
2844 T t 2
qg
m pi m p wf 12ct
1
4
1 1 1
2 2
Acw k F 2 k m 4 1 1 1
14409.126T 0.00633L F t 4 k F 2 k m4
qg
m pi m p wf ct
1
2
12 1
3 0.00633 k m Acm 1 1
2844 T t 2 k m2 Acm
m pi m p wf 12ct
1
2
1 1
4 qg 2
Acw k m 2 y e 1 1
2844 3 T 0.00633L F t 2 k m2 y e
5 ---------------------------------------------------------------------
Superposition of drawdown is not very common in the petroleum industry because most of the time superposition of rate is
done to calculate pressure drops. Assuming that an operator measured bottomhole flowing pressure for each rate variation,
values of drawdown for each time step would be known for history matching. In this problem, constant pressure semi-
analytical solutions are used for history matching. Motivation for this method was driven from the work of Van Everdignen
and Hurst (1949). Their work clearly says that constant terminal pressure solution can be used to calculate “cumulative influx”
at a particular time by adding different pressure drops over time. The same thing is true for the constant terminal rate case,
which has been used in petroleum industry for a long time. Eq. 5 is for superposition of drawdowns by using unit rate function
Fq(t) and Eq. 6 is superposition of rates by using the unit pressure function Fp(t).
n
q t n pi pi 1 Fq t n t i 1 ……………………………………………………………………………………… (5)
i 1
n
p t n qi qi 1 Fp t n ti 1 …..…………………………………………………………………………………... (6)
i 1
Eq. 5 is used to handle variable drawdown production cases of shale gas wells. Fq(t) and Fp(t) are response of reservoir for 1
psi constant pressure drop or 1 Mscf/D constant production rate, respectively.
To check applicability of the method, a simulation model was built with typical parameters of a shale gas well. To mimic the
behavior of shale gas wells, we used production data and reservoir properties of one of the Woodford wells in the model.
Table 2 summarizes important reservoir and horizontal well properties used in the simulation model. After running the
simulation for about 1200 days, rates and pressures were recorded to use in variable drawdown history matching (Fig. 3).
SPE 144335 5
Fig. 3—Rates and bottomhole flowing pressures for synthetic well. Production history was simulated for typical well behavior in
Woodford shale.
Fig. 4 is a log-log plot of the rate versus time and a Cartesian plot of normalized reciprocal rate versus square root of time.
From the log-log plot, it is clear that well was producing under Bilinear flow for more than 50 days, followed by Linear flow.
Boundary effects are also visible from both plots.
Fig. 4—Production history of synthetic well: Bilinear (quarter slope) followed by Linear (half slope) (left). Normalized reciprocal rate
versus square root of time plot for synthetic well: Linear flow period shows straight line (right).
6 SPE 144335
First, the two unknowns problem was solved with history matching program for both constant drawdown and variable
drawdown cases. Fig. 5 is showing history matching results. In both cases, values for fracture permeability and half-length
were found with less than 5% error, but variable drawdown case results are more accurate than constant drawdown case. Then,
three unknowns problem was solved assuming that matrix permeability is also unknown (Fig. 6). Although results of constant
drawdown case are not close to actual values of parameters, this match should be considered as good, because km1/2ye for both
constant and variable drawdown cases are the same. The same conclusion can be drawn for Bilinear part of the flow, too where
kF1/2km1/4 from constant and variable drawdown cases are very similar (Table 3).
Fig. 5—History matching results of synthetic well for constant bottomhole flowing pressure (CP) case with 2 unknowns: kF and ye.
First 400 days were matched (left). History matching results of synthetic well for variable bottomhole flowing pressure (VP) case with
2 unknowns: kF and ye. First 400 days were matched (right).
Fig. 6—History matching results of synthetic well for constant bottomhole flowing pressure (CP) case with 3 unknowns: kF, km and ye.
First 400 days were matched (left). History matching results of synthetic well for variable bottomhole flowing pressure (VP) case with
3 unknowns: kF, km and ye. First 400 days were matched (right).
Finally, the whole production period was used in history matching for both cases. Since boundary effects were seen in the life
of the well, this well should have a more unique solution – 3 unknowns and 3 flow regimes. Fig. 7 is showing the results of
these cases. An excellent match observed in both cases but values obtained for 3 unknowns from variable drawdown case are
closer to actual values.
Fig. 7—History matching results of synthetic well for constant bottomhole flowing pressure (CP) case with 3 unknowns: kF, km and ye.
The hole production period was matched (left). History matching results of synthetic well for variable bottomhole flowing pressure
(VP) case with 3 unknowns: kF, km and ye. The whole production period was matched (right)
SPE 144335 7
Table 3 summarizes the important results of this analysis. One conclusion can be drawn from these observations: History
matching with variable drawdown should give better results than constant drawdown assumption. However, the difference in
accuracy is not significant enough to favor variable drawdown history matching. Since less computation will be done for
constant drawdown matching than for the variable drawdown, for practical purposes the former is recommended. Moreover,
accurate and continuous pressure records may not be available for each well which also forces use of the constant drawdown
matching.
The following sections present analysis of 3 wells from Barnett, Woodford, and Fayetteville shale plays. Table 4 summarizes
completion and reservoir parameters for each well. History matching was done for each well and reservoir parameters
determined based on our proposed model. Given the economic limit, 30 Mscf/D in this case, forecasting was done for each
well and important parameters such as remaining life of the well, expected ultimate recovery (EUR), and recovery factor (RF)
were calculated. Constant bottomhole flowing pressure forecasting was done by using the measured flowing pressure at the
last month. Table 5 summarizes the results of forecasting.
For the first analysis, only 400 days of production period was matched (Fig. 9). Both kF1/2km1/4 and km1/2ye values are calculated
according to found parameters. Then whole production period was matched by excluding points which are below critical rate
not to have liquid loading. New kF1/2km1/4 and km1/2ye values found from history matching are in the same range with previous
ones however, values found for kF, km, and ye are different. Since a good match was obtained with the proposed model,
forecasting was done for Well B-15 with history-marched parameters (Fig. 10).
8 SPE 144335
Fig. 8—Production history of Well B-15: Bilinear (quarter slope) followed by Linear (half slope) (left). Normalized reciprocal rate
versus square root of time plot for Well B-15: Linear flow period shows straight line (right).
Fig. 9—Early part of the production (No shut-ins or rate fluctuation is observed) was matched for Well B-15 (left). The whole
production period of Well B-15 was matched for forecasting (right).
Fig. 10—Forecasting for Well B-15 with parameters found from history matching (left). Normalized reciprocal rate versus time plot for
Well B-15 with parameters found from history matching (right).
Fig. 11—Production history of Well W-2-4: Bilinear (quarter slope) followed by Linear (half slope) (left). Normalized reciprocal rate
versus square root of time plot for Well W-2-4 (right).
Fig. 12—Early part of the production (No shut-ins and less rate fluctuation is observed) was matched for Well W-2-4 (left). The whole
production period of Well W-2-4 was matched for forecasting (right).
Fig. 13—Forecasting for Well W-2-4 with parameters found from history matching (left). Normalized reciprocal rate versus time plot
for Well W-2-4 with parameters found from history matching (right).
Fig. 14—Production history of Well F-3: Bilinear (quarter slope) followed by Linear (half slope) (left). Normalized reciprocal rate
versus square root of time plot for Well F-3 (right).
Fig. 15—Whole production period of Well F-3 was matched for forecasting. The well is still in transient flow and no effects of BDF
were observed
Fig. 16—Forecasting for Well F-3 with parameters found from history matching (left). Normalized reciprocal rate versus square root of
time plot for Well F-3 with parameters found from history matching (right).
SPE 144335 11
Table 4—Summary of completion and reservoir properties of wells used as field examples
Parameters Well B-15 Well W-2-4 Well F-3 Units
Porosity 0.002 0.002 0.002
Hydraulic Fractures Width 0.04 0.04 0.04 ft
Spacing 82 93 76 ft
Porosity 0.06 0.0425 0.092
Matrix
Thickness 300 250 300 ft
Perforated Length 2870 2232 3192 ft
Area of Well Plane 1722000 1116000 1915200 ft2
# of Macro-fractures 35 24 42
Rock Compressibility 0.000004 0.000004 0.000004 1/psi
Water Compressibility 0.0000042 0.0000042 0.0000042 1/psi
Initial Water Saturation 0.3 0.3 0.5
General
Initial Pressure 2950 3280 1759 psi
Bottom-hole Pressure 480 300 200 psi
Temperature 160 200 119 deg. F
Gas Gravity 0.635 0.62 0.58
Langmuir Volume 96 NA NA scf/ton
Langmuir Pressure 650 NA NA psi
Discussion of Results
Three different wells were analyzed in this paper from different shale plays. Most shale gas wells show Bilinear flow followed
by Linear flow. Considering the low gas prices dominating in the market at the moment, operators try to maintain the well at
low costs. Therefore, having good information about the quality of the fracture job and size of the reservoir is crucial. Since
running build-up tests are impractical in low permeability reservoirs, the only way to get information about the formation after
fracturing job is production data analysis (PDA). The method described in this paper handles this problem in an excellent way.
By matching the early undisrupted production data, we can get kF1/2km1/4 and km1/2ye, which gives us information about quality
of the fracturing job and the production capacity of the well.
In the three wells which were analyzed above, the adsorption effect was included only in Well B-15, since reliable information
about Langmuir volume and pressure is available only for Barnett shale at the moment. Although, contribution from desorbed
gas to total production at the early transient life of shale gas wells is not significant, it is expected that adsorption has started
affecting the gas production in Barnett shale because of considerable reduction of reservoir pressure from its initial value. Fig.
17 shows the total and free gas production for Well B-15 with found parameters. It is clear that at late time adsorbed gas has
significant impact on gas production.
Among the analyzed wells, Well W-2-4 has the lowest original gas in place (OGIP) and obviously shorter life than the others.
On the other hand, Well B-15 has the largest OGIP because of adsorbed gas. Adsorbed gas also affects the recovery factor. In
order to get more gas from matrix surface by desorption, average pressure should be lowered as much as possible.
12 SPE 144335
Fig. 17—Comparison of Total and Free Gas production for Well B-15.
Conclusions
A new method is presented in this paper for history matching and forecasting of shale gas production. It is based on analytical
solutions for linear dual porosity, but corrected for real gas properties and adsorbed gas. Following are conclusions drawn
from this study:
The linear dual porosity model matches the production behavior of many shale gas wells.
The method presented in this paper can be used to find important reservoir parameters and is as rigorous as
simulation.
Because of having more unknowns than equations, the estimates of kF, km, and ye would not be unique. Therefore,
parameter groups are evaluated such as kF1/2km1/4 and km1/2ye rather than individual values.
Once a satisfactory match is obtained, forecasting can be done with the matched parameters. Forecasting with this
method can be as fast as fitting different decline curves to production data. Furthermore, forecasted results should be
more reliable than decline curve analysis.
Acknowledgment
The author would like to thank James K. Forrest at Pioneer Natural Resources for helping to edit this paper to its final form.
Nomenclature
Acm = matrix cross sectional area to the flow, ft2
Acw = wellbore cross sectional area to the flow, ft2
ct = total compressibility, 1/psi
kF = effective fracture permeability, md
km = matrix permeability, md
LF = fracture spacing, ft
m p i = pseudo initial pressure, psi2/cp
m pwf = pseudo bottomhole flowing pressure, psi2/cp
OGIP = original gas in place, Bscf
p = pressure, psi
qD = dimensionless rate
qg = gas rate, Mscf/D
SRV = stimulated reservoir volume
Sw = water saturation
T = reservoir temperature, R
tD = dimensionless time
SPE 144335 13
Greek Symbols
g = gas viscosity, cp
= porosity
= interporosity coefficient
References
Ambrose, R.J., Clarkson, C.R., Youngblood, J.E. et al. 2011. Life-Cycle Decline Curve Estimation for Tight/Shale Reservoirs. Paper
presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, Texas, USA. SPE 140519.
Bello, R.O. and Wattenbarger, R.A. 2008. Rate Transient Analysis in Naturally Fractured Shale Gas Reservoirs. Paper presented at the
CIPC/SPE Gas Technology Symposium 2008 Joint Conference, Calgary, Alberta, Canada. SPE 114591.
Bello, R.O. and Wattenbarger, R.A. 2010. Modelling and Analysis of Shale Gas Production with a Skin Effect. 49 (12). DOI:
10.2118/143229-pa
Bello, R.O. and Wattenbarger, R.A. 2010. Multi-Stage Hydraulically Fractured Horizontal Shale Gas Well Rate Transient Analysis. Paper
presented at the North Africa Technical Conference and Exhibition, Cairo, Egypt. SPE 126754.
Cipolla, C.L., Lolon, E., Erdle, J.C. et al. 2009. Reservoir Modeling in Shale-Gas Reservoirs. Paper presented at the SPE Eastern Regional
Meeting, Charleston, West Virginia, USA. SPE 125530-MS.
El-Banbi, A.H. 1998. Analysis of Tight Gas Wells, Phd Dissertation.
Everdingen, A.F.V. and Hurst, W. 1949. The Application of the Laplace Transformation to Flow Problems in Reservoirs.
Freeman, C.M., Moridis, G.J., Ilk, D. et al. 2009. A Numerical Study of Performance for Tight Gas and Shale Gas Reservoir Systems. Paper
presented at the SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana. SPE 124961-MS.
Ilk, D., Rushing, J.A., Perego, A.D. et al. 2008. Exponential Vs. Hyperbolic Decline in Tight Gas Sands: Understanding the Origin and
Implications for Reserve Estimates Using Arps' Decline Curves. Paper presented at the SPE Annual Technical Conference and
Exhibition, Denver, Colorado, USA. SPE 116731.
Lee, W.J. and Sidle, R.E. 2010. Gas Reserves Estimation in Resource Plays. Paper presented at the SPE Unconventional Gas Conference,
Pittsburgh, Pennsylvania, USA. SPE 130102.
Nobakht, M., Morgan, M.D., and Mattar, L. 2010. Case Studies of a Simple yet Rigorous Forecasting Procedure for Tight Gas Wells. Paper
presented at the Canadian Unconventional Resources and International Petroleum Conference, Calgary, Alberta, Canada. SPE
137456.
Samandarli, O., Al-Ahmadi, H., Wattenbarger R.A. 2011. A Semi-Analytic Method for History Matching Fractured Shale Gas Reservoirs,
Western Regional Meeting. Anchorage, Alaska, USA. SPE 144583
14 SPE 144335
Appendix
Dimensionless Variables
0.00633k F t
tD ...………………………………………………………………………………………………………… (A-1)
ct Acw
1
kF
Acw m p i m p wf .....…………………………………………………………………………….…………. (A-2)
qD 1422 q g T
12 k m
Acw .....……………………………………………………………………………………………..…..………. (A-3)
L2F k F
Vct F
....………………………………………………………………………………………………….………….. (A-4)
Vct t
ye
yeD ....……………………………………………………………………………………………………..………. (A-5)
Acw
Region 1
1 2
qD t D ...………………………………………………………………………………………………….……… (A-6)
2
By substituting Eqs. A-1, A-2, and A-4 into Eq. A-6 we get,
qg
m pi m pwf
1
2 1 1 1
Acw Vct F2 k F 2 1 ….............…………………………………………….………. (A-7)
0.00633
2844 T t 2
Region 2
1
4 1
qD tD 4 .....……………………………………………………………………………………….………….……. (A-8)
10.133
By substituting Eqs. A-1, A-2, and A-3 into Eq. A-8 we get,
m pi m pwf 12ct
1
4
1 1 1
qg 2
Acw k F 2 k m 4 1 …............……………………………..…………………….….. (A-9)
14409.126T 0.00633LF t 4
Region 3
1 1
qD t D 2 ...……………………………………………………….…………………………………………..…… (A-10)
2
By substituting Eqs. A-1 and A-2 (For homogeneous case km would be used instead of kF) into Eq. A-10 we get,
m pi m p wf ct
1
2
12 1
qg 0.00633 k m Acm 1 …........………………………………………………………………. (A-11)
2844 T t 2
SPE 144335 15
Region 4
qD
3
1
2
y eD 1
tD 2 ....………………………………………………………………………………..………………… (A-12)
2
By substituting Eqs. A-1, A-2, A-3 and A-5 into Eq. A-12 we get,
m pi m pwf 12ct
1
2
1 1
qg 2
Acw k m 2 y e 1 ….........…...………………………………………………….… (A-13)
2844 3 T 0.00633LF t 2