Professional Documents
Culture Documents
SPE 36736
545
2 AHMED H. EL-BANBI, ROBERTA. WATTENBARGER SPE 36736
Results
qg =Jg [m(;)
-m(pwf)]
........... .,, (2) A gas reservoir simulator] 3 was used to generate production
data for hypothetical cases. Two- and three-layer cases were
In addition to these two equations, the relation between simulated. The layers were termed a, b, and c. The basic
the gas flow rate and cumulative gas production is given by reservoir and fluid properties data used in simulation are
Eq. 3. given in Table 1. These layers have the same properties and
differ only in their permeability.
The two-layer case was constructed by adding layers b and
Gp = ~qg (/) di . ...... ...... . .. .. ........ . .. . .... .. ... (3) c performances. The three-layer case was constructed by
o adding the performance of layers a, b, and c. The computer
and, m(p) is the real gas pseudo-pressure defined by Eq. 4. program developed in this work was used to estimate the
parameters of each layer for both the two- and three-layer
cases, Other combinations were tried] 4 but the space does not
rn(p)=2j~dp .. .. . .. . .. . . .. . . .. ... (4) allow to show them all. We assumed that we only knew three
~ozp
years of production history and ran the Layered Stabilized
Flow program to forecast for twenty years. Our model foreeast
Eqs. 1, 2, and 3 can be solved together if the OGIP and and the simulation were in a very good agreement. F@s. 3
the flow coefficient, Jg, are known for the layer. and 4 show the model and the simulation resutts for both the
two- and three-layer cases respectively. Parameter fitting in
Layered Stabilized Flow Model. The performance from the Layered Stabilized Flow Model was done on only
single-layer stabilized flow models can be added for all the production data for production period from month 13 to
layers in the commingled system using Eq. 5. month 36 which is known to be past transient data or
stabilized flow data. The effect of transient data on the
nlayw
analysis is kther investigated in another section. Tables 2
qr(t) = ~qg, (t) .................................................(5)
j=l
and 3 show the comparison between the parameters we
obtained from the Layered Stabilized Flow Model and the
This requires that both OGIP and Jg be known for each simulation for both cases. The flow coefficient, Jg, was
layer in the commingled system. The analysis of the calculated from simulation data using Eq. 8.
commingled system is carried out by assuming values for
OGIP and Jg for each layer and calculating the total J, =
kh
....... (8)
1
performance (rate versus time) for the entire commingled 1 2.458A
system, The error in the model performance can then be 1424 T ~ln +.s+Dqg
[ CAr~
quantified by either Eqs. 6 or 7 for a particular period of the
known production data.
Effect of Transient Flow, The data for layer c was simulated
el=— IN Data - Model
.. ..... (6) to study the effect of having some transient rate-time data in
N x,=, Data the history match data. This effect is worth studying because
546
SPE 36736 ANALYSIS OF COMMINGLED TIGHT GAS RESERVOIRS 3
rate data in tight reservoirs are expected to have long It was also found that matching any commingled reservoir
transients. Several periods of production data were used in the with a fewer number of layers stabilized flow model results in
history matching with the Layered Stabilized Flow Model under-estimation of the total OGIP in the commingled
program. Fig. 5 shows the rate predicted using the Layered system’ 4. This was seen when attempting to match thredayer
Stabilized Flow Model when the first 12 months of data were commingled systems with either one- or two-layer stabilized
used in the matching. It also shows the rate behavior when flow model’ 4,
the first 24 months of data were history-matched. In both
runs, it is evident from the figure that the rate is under-
predicted for the rest of the reservoir life. Field Applications
If these transient data were not included, i,c. we To check the Layered Stabilized Flow Model developed in
selectively analyze the data starting at a later time (month 13 this work, we used a field case described in Ref 11. The field
for this particular case), the model would estimate a much produced for more than twenty years against almost constant
better answer as it did for the cases presented previously. PW}1. The reservoir was described as being a two-layer
commingled gas reservoir with no crossflow’]. A shale barrier
Effect of Non-Darcy Flow. Since the Layered Stabilized averaging 50 ft. in thickness was identified and correlated in
Flow Model assumes that non-darcy flow is neglected, this the field’ 1, Core data indicated a permeability ratio between
section is dedicated to illustrate the effect of non-darcy flow the two layers in the range of 10/1 to 20/1. The initial
on the model answers. The reservoir simulator was used to reservoir pressure and average temperature were reported as
simulate a one-layer case (layer c) taking into account the 428 psia and 80 “F respectively) 1. Gas specific gravity was
non-darcy flow. The non-darcy flow coefilcient is calculated 0.7[ 1. The total field production as well as the individual field
from Eq. 9, analysis] 1 showed that a b value of 0.9 represents the average
decline parameter of hyperbolic equations,
547
4 AHMED H. EL-BANBI, ROBERTA, WATTENBARGER SPE 36736
to the reservoir and fluid data in Table 4 for well H, multi-layer systems. These results suggest that we should
The Layered Stabilized Flow Model was used again to analyze different periods of the rate-time data and check the
calculate the parameters OGIPand J~ for each of two layers model results. If the model results do not signflcantly change
by matching production rate-time data for months 13 to with the production data period we choose, this means that
month 48, The parameters estimated from the model are transient effects are avoided in the data we match. Also, well
shown in Table 5 for the two layers. These parameters were logs and geological studies do aid in understanding the
then used to forecast future rates up to 31 years, The commingled reservoir and qualitatively assess the relative
comparison between the Layered Stabilized Flow Model productivity of the layers. This may give very important
forecast and the actual well data is shown in Fig. 9. information to check the Layered Stabilized Flow Model
Following the same procedure as with well E, a hyperbolic results,
decline curve forecast was also matched to the same data and The main limitations of this model in matching field data
Table 6 shows the hyperbolic decline curve parameters for is that of the limitations of the optimization routine used in
two fits (month 1 through 48 and month 13 through 48). Fig. the analysis, The optimization routine may converge to a
9 shows the production rate forecast of the hyperbolic decline wrong answer (local minimum) especially if the rate-time
curve for the match of months 13 through 48. data used in the analysis are not good enough. It is also
Figs, 8 and 9 both show that the ‘forecast” of the Layered expected that if the commingled system consists of many
Stabilized Flow Model is lower than the decline curve layers, the answers may not be unique especially if the rate is
forecast and is much closer to the well’s actual production not measured accurately, Therefore, the engineer has to
rates, always check the layered model results against his knowledge
of the geology of the reservoir and the log response.
Another limitation may be seen in the analysis of the gas
Hyperbolic Equation Fitting. It was found that using the reservoirs which produce predominantly under transient flow
entire production data for four years or just the data from conditions. As was shown in the paper, the Stabilized Flow
month 13 through 48 for these wells in the non-linear Model prediction tends to under-estimate the performance for
regression tit did not modify the three matching parameters those reservoirs. From the analysis of many simulated cases
(q,, D,, and b) significantly and almost the same predicted in this work, we propose that a permeability range of 0.1 to
rate was obtained. These parameters are given in Table 6 for 10 md. is a suitable range for the application of this method.
the two matching periods. It is understood, of course, that the For those reservoirs which have permeability less than 0,1
hyperbolic tit would improve as more production history is md, the transient flow period may be so long and we may not
available. However, the advantage of using the Layered have enough stabilized flow production to match. On the
Stabilized Flow Model is that a better forecast can be obtained other hand; for those reservoir above 10 md, non-darcy flow
from the early stabilized flow data in addition to an estimate may be important and the use of the flow equation @q. 2)
for the vahres of OGIP in each layer, without the inclusion of non-darcy flow term may give
inaccurate results. The inclusion of non-darcy flow will
change the problem from a two-parameter optimization for
Discussion each layer to a three-parameter optimization for each layer,
Probably, the most famous method attempted to effectively This will of course increase the diffkulty of the problem
analyze commingled gas resemoirs from limited data is especially if we try to match the production of a commingled
decline curves, However, decline curve analysis is strictly resewoir with many layers.
empirical for multi-layer reservoirs. The Layered Stabilized
Flow Model presented here is based on sound fundamental
basis, This model is also considered a very good personal Conclusions
computer (PC) application as far as computer storage and The work presented in this paper is based on matching
speed are concerned. It may be also adapted for spread-sheet production data for stabilized gas flow, It considers wells
applications. More importantly, this model not only presents producing against constant pwf and ignores non-darcy flow.
an easy method for modeling commingled gas reservoirs but The following conclusions can be drawn from this paper:
also effectively analyzes them if it is coupled with an 1. Production data from commingled gas reservoirs ean be
optimization routine., analyzed by Layered Stabilized Flow Models with the use of
As this work shows, the Layered Stabilized Flow Model an optimization routine.
gives very reliable answers for commingled systems if the 2. OGIP is under-estimated if transient data are included.
model assumptions are satisfied. However, if the model is 3. Total OGIP for the commingled system is under-
used to analyze transient data, it gives conservative estimates estimated if a fewer number of layer model is used in the
of OGIP and consequently conservative prediction. The same analysis.
result was obtained when a one-layer model is used to model 4, Effect of non-darcy flow on the analysis using the
548
SPE 36736 ANALYSIS OF COMMINGLED TIGHT GAS RESERVOIRS 5
Layered Stabilized Flow Models for tight gas reservoirs can Acknowledgment
be ignored. Wc thank Mike Fetkovich of Phillips Petroleum Company for
5, History-matching selective intervals of rate-time data vahrable discussions on decline curve analysis. We also thank
rather than the entire production history helps in identifying Phillips Petroleum Company and Frank Verla of Dwights
the end of the transient effects and/or commingled behavior. Data Base for providing field data, This work was done
6. Formation permeability range of 0.1 to 10 md M within the Reservoir Modeling Consortium at Texas A&M
suitable for the application of Layered Stabilized Flow University.
analysis technique.
References
1 Arps, J, J.: “Analysis of Decline Curves,” Tmns., AIME ( 1945)
Nomenclature 160,22847.
A. reservoir drainage area, L2, ftz 2 F’etkovich, M. J.: “Decline Curve Analysis Using Type Curves,”
JPT (June 1980) 1065-77.
b= Arps’ decline exponent
3 lralm, M. L. and Wattenbarger, R. A.: “Gas Reservoir Decline-
CA = shape factor
Curve Analysis Using Type Curves With Real Gas
D= non-darcy flow coetlicicnt, t/L3, (Mscf/D)”]
Pseudopressures and Normalized Time,” .SPEFE (Dec. 1987)
D, = initial decline factor, l/t, day”’ 671-82, Tmns, AIME, 290.
q = normalized error measure given by Eq. 6
4 I’alacio, J, C, and Blasingame, T, A.: “Decline-Curve Analysis
~2 = normalized error measure given by Eq. 7 [Jsing Type Curves--Analysis of Gas Well Production Data,”
G= original gas in place, L3, MMscf paper SPE 25909 presented at the 1993 Joint Rocky Mountain
Gp = cumulative gas produced, L3, MMscf Regional and Low Permeability Reservoirs Symposium,
h= formation thickness, L, ft Denver, CO., April 26-28, 1993.
Jg = real gas flow coefficient, L4t2/m, Mscf.cp/D/psi2 5 Chen, IL Y. and Poston, S. W.: “Application of a Pseudotime
k= permeability, L2, md L’unction To Permit Better Decline-Curve Analysis,” SPEFE
(Sept. 1989) 44148.
mfi) = real gas pseudo-pressure at average reservoir
6 Carter, R D.: “Type Curves for Finite Radial and Linear Gas-
pressure, m/Lt3, psi*/cp
Flow Systems: Constant-Terminal-Pressure Case} SPEJ (Get,
m(pw~ = pseudo-pressure at bottom hole flowing pressure, 1985)71928
rn/Lt3, psi2/cp 7 Ammian, K , Ameri, S., Srark, J. J., and Yost II, A, B.: “Gas-
rdayer = number of layers in a commingled system Well Production Decline in Multiwell Reservoirs,” JPT (Dec.
N= number of rate-time data points 1990) 1573-79.
OGIP = original gas in place, L3, MMscf t? Guardia, M. A. and Hackney, R. M. “A Practical Approach to
~= pressure, m/Lt2, psia original Gas-Jn-Place Estimation: Development of the South
p, = initial reservoir pressure, rr3/Lt2,psia W ilbuton Field,” paper SPE 22925 presented at the 66ti
Annual Fall Technical Conference and Exhibition, Dallas,
p.. - pressure at standard conditions, rn/Lt2, psia
Texas, (oct. 6-9,1 991).
pwf = bottom hole flowing pressure, rn/Lt2, psia
jj= material-balance average reservoir pressure, rn/Lt2, 9 Kcatmg, J. F., Chen, IL Y., Wattenbarger, R. A.: “Griginal Gas
in Place and Decline Curves From Early Stabilized Rate-Time
psia Data,” paper SPE 27666 (March 1994).
qz = gas production rate, L3/t, Msef/D
10 West, S, L , and Cochrane, P. J. R. : “Reserves Determination
qT = total flow rate in a commingled system, LJA, (Jsing Type Curve Matching and Extended Material Balance
Mscf/D Methods in the Medicine Hat Shallow Gas Field,” paper SPE
rw = wellbore radius, L, fl 28609 presented at the 69th. Annual Technical Conference and
~= skin factor Exlubition of the Society of Petroleum Engineers held in New
t= time, t, days odeans, LA, Sep. 25-28 ( 1994).
T. 11 l’ctkovich, M. J., Bradley, M. D,, Works, A. M., and Thrasher,
reservoir temperature, T, “R
‘1. S.: “Depletion Performance of Layered Reservoirs Without
T.c = temperature at standard conditions, T, ‘R Crossflow,” paper SPE 18266(1988).
~= real gas correction factor
12 AI-I Iussainy, R., Rarney, H. J., and Crawford, P. B. : “The
P “ non-darcy flow parameter given by Eq. 10 Flow of Real Gas Through Porous Media,” .TPT (May 1966)
+= porosity 624-36.
~z . gas viscosity, fit, cp 13 Lee, W. J., and Wattenbarger, R. A.: Gas Reservoir
Engtnem-ing, SPE Textbook Series, to be published in 1996.
549
Table 1: Resetvoir and Fluid Data for Simulated Cases
case a b
area, acres 80 80 ;
thickness, ft. 50 50 50
porosity,fraction 01 0.1 0.1
initialpressure, psia 2500 2500 2500
BHFP, psia 500 500 !x41
temperature, “F 150 150 150
gaa gravily, fraction 0.6 0.6 0,6
formationcompressibility,ps[’ 3.E-06 3,E-08 3,E-06
permeabilky, md 10 1.0 0.1
number of gridblock 25 25 25
OGIP, MMecf 2913.77 2913.77 2913.77
I I
Table 3: Matching Parameters for a Three-Layer Case
OGIP
(MMscf)
I OGIP
(MMscf)
I J.
(Mscf .c~/D/psi2)
I J.
(Mscf.c@psi2)
m m
layer a 3096.5 2913.4 7.639E-05 7.567E-05
layer b 2935.8 2913.4 7.477E-06 7.587E-06
layer c 2827.6 2913.4 7.654E-07 7S87E-07
Table 4-input Parameters for Layered Stabilized Flow Model for Wells E and H
Well E Well H
number of Layers 2 I 2
p,, psia - 394.6 428
~, psia 42,5 42.8
gas specific gravity,fraction 0.7 0.7
reservoir temperature “F 80 80
550
1O(IO
low k layer
—.-———___ —.
—-..
(
IQO
.,, ,,
L
,,,
,,
high klayer ~ ~ ~ .’
10
1 10 100
Time (Months)
u Calculate
Gas
Pro erties
1
Estimate
yes
Prediction
I I
Fig. 2-Schematic flow chart of the Layered Stabilized Flow Program.
10000
100
1 10 100 1000
Time (Months)
Fig. 3-Matching total rate for a hypothetical two-layer case (matching period 13-36 months).
100000
I I
1Oouo
1000
I
‘1
100
1 10 100 1000
Time (Months)
Fig. 4-Matching total rate for a hypothetical three-fayer case (matching period 13-36 months).
10GQ ,, ,,<
,,, ,,
,. ,,,
--i
●
Iw ..
,.
!&____
,*,,,,,
●
w“
● ’”
.4ctual
● Matching Per-id: 1-12 Months
● Matching Period: 1-24 Monttw
1 10 102 10YI
Time (Months)
.!
. .——— ——
,,,
,,.
,.
,,
.,,
,,
1 10 lm lCOI
Time (Montha)
Fig. 6-Effect of nondarcy flow on the Stabilized Flow Model results for a tight layer (matching period 13-
24 months).
m....:.
100C4 . .. .
‘Actual
-%-w
100(
10[
,..
,,
.
1[
1 10 100 1000
Time (Months)
Fig. 7-Matching two-layer hypothetical case with a one-layer Stabilized Flow Model (matching period 1340
months).
1000
Matching Period
%
I I I I I
I ●☛
.
+
t
1-
●
I . Actual
Hyperbolic Decline Fit, b=O.t339
— Layered Stabilized Flow Model
I I I I I
1 1 1 1 r
Time (months)
Fig. 8-Well E; comparison between the Layered Stabilized Flow Model prediction and decline cunre
method.
1000
atching Period
100
●
● 4
10
p .-
● Actual ~
Hyperbolic Decline Fit, b=l.012
— Layered Stabilized Flow Model
1
o 50 100 150 200 250 300 350 420
Time (Months)
Fig. 9-Well H; comparison between the Layered Stabilized Flow Model prediction and decline curve
method.