Vous êtes sur la page 1sur 6

An Improved Pseudo-Time for Gas Reservoirs

With Significant Transient Flow



D.M. anderson, L. Mattar
Fekete Associates, Inc.

methods. It is also used in analytical well/reservoir models, whose


Abstract conventional formulations are only valid for slightly compress-
The use of semi-analytic methods for correcting flow equa- ible fluids with constant properties over a given pressure range.
tions to accommodate changing gas properties with pressure has These models enjoy widespread usage for both history matching
become increasingly common. It is a mainstay of modern pro- and forecasting, and their inclusion of pseudo-time for gas reser-
duction decline analysis, as well as gas deliverability forecasting. voirs is vital.
The use of pseudo-time is one method which enables a time- To illustrate the value of pseudo-time, let us take the simple
based correction of gas properties, honouring the gas material case of a vertical well in the centre of a circular gas reservoir. We
balance within the time-based flow equation. By using pseudo- will assume constant rate production and pseudo-steady-state con-
time, the analytical well/reservoir models, derived for the liquid ditions. Thus, the model that describes the pressure response at
case (slightly compressible fluid), can be modified for gas by re- the well can be reduced to the pseudo-steady-state equation for
evaluating the gas properties as the reservoir pressure depletes. gas(5 3).
These gas correction procedures are well documented in the lit-
2qpi
erature. Also well documented is the iterative nature of the gas p p
pi pwf
=
(cg Z )i Gi
()
f t
properties correction methods, as original gas-in-place is a re-
quired input into the equations. 1.295 Tq re 3
The pseudo-time correction is based on the average reser- +
kh
ln SI units
rwa 4
( )
voir pressure and works very well for boundary dominated flow. ........................................................... (2)
However, when transient flow prevails, the pseudo-time concept The f(t) in Equation (2) is the chosen time function. Figure 1
is not valid and its use can create anomalous responses. This will shows the pressure response plotted against time for two cases:
occur in low permeability systems or in reservoirs with irregular
f(t) = time (t) and f(t) = pseudo-time (ta). Also compared is the
shapes, especially where some of the boundaries are very distant
numerical solution using the same input parameters. Upon com-
from the well.
parison of the solutions, it is clear that pseudo-time has a signifi-
The semi-analytic gas correction has a representative pres-
cant impact on the flow equation for gas. We also note that the
sure at its root, which, in the existing models, is always the
pseudo-time solution appears to be identical to the numerical so-
average reservoir pressure. We propose a straightforward modifi-
lution. Thus, without using pseudo-time, the pseudo-steady-state
cation to the determination of this pressure as follows. The repre-
equation would be incorrect for gas reservoirs.
sentative pressure ought to be based on a radius of investigation
or region of influence (in the case of non-radial systems), rather
than the average reservoir pressure. In the case of a depleting The Trouble With Transient Flow
system, the representative pressure would be the same as the av- In the conventional definition of pseudo-time, the com-
erage reservoir pressure. The following paper outlines the pro- pressibility and viscosity terms in Equation (1) are evaluated at
posed procedure and illustrates its advantages over the existing
method by using synthetic and field data examples.
Flowing Sandface Pressure (psia)

3,000

2,500
Introduction
2,000
Background Pseudo-time
Numerical
Literature on the derivation and usage of pseudo-time is 1,500

prevalent(2 1, 3 2). The definition that will be used in this paper is Time
1,000
shown below:
500
dt
( ) 0 ct
t
ta = ct
i 0
..................................................................................... (1) 0 20 40 60 80 100 120 140

Time (days)
The above is used in the pseudo-steady-state equation for gas,
FIGURE 1: Pressure responses using time and pseudo-time.
which is at the core of most modern production decline analysis
Peer Reviewed Paper (Review and Publication Process can be found on our Web site)

July 2007, Volume 46, No. 7 49


Pwf vs. time- "Three Boundary" Reservoir
3,500

3,000 Bounded (slightly beyond


region of influence)
2,500

pwf (psia)
2,000

1,500
Semi-Infinite
1,000

Initial pressure
500

0
0 50 100 150 200 250

Time (h)

Well flowing
FIGURE 3: Flowing pressure responsebounded and semi-infinite
pressure reservoirs (using conventional) pseudo-time.

Additional Sources of Error


FIGURE 2: Illustration of high localized drawdown due to
boundariessame well rates, same flow duration. It is worth noting that there is a secondary source of error in
the pseudo-time procedure that has nothing to do with transient
average reservoir pressure conditions. Clearly, the average res- or boundary dominated flow. Rather, the error results from the as-
ervoir pressure is a function only of total pore volume and cu- sumption that bulk properties evaluated at a single average pres-
mulative produced fluids. If the well of interest is producing the sure can adequately represent the range of fluid properties observed
reservoir under boundary dominated conditions (depletion mode), throughout the reservoir. This error, usually negligible, can be sig-
the average reservoir pressure is a very reasonable datum at which nificant if there are large pressure gradients. Such would be the
to establish bulk fluid properties. However, if the well production case when the well produces at very high drawdown from a very
is still transient and no reservoir boundaries have been observed, low permeability reservoir. The error would be present during both
average reservoir pressure based on total reservoir volume may be
transient and boundary dominated flow. The solution suggested
a poor datum to use. Consequently, pseudo-time can cause anoma-
lous model responses under certain conditions. below does not address this error.
To illustrate the above, consider two gas reservoirs of very dif-
ferent size but having identical properties (permeability, skin, etc).
If we allow reservoir fluids to flow from identical vertical wells in Solution Method
both reservoirs, we would expect identical pressure responses in
both cases, until the boundaries of the smallest reservoir have been The proposed solution involves altering the definition of con-
observed. Yet, when we use the analytical models with pseudo- ventional pseudo-time such that the gas properties are evaluated
time, we observe a slightly greater pressure drop in the larger res- at the average pressure of the region of influence, rather than the
ervoir. This discrepancy is clearly not based on physical reality, average pressure of the total pore volume. Evaluating this pressure
but rather, is caused by pseudo-time. The error is usually negli- mathematically would involve first defining the region of influ-
gible. However, in cases where the pressure drawdown is severe, ence, and then integrating the pressure response over the domain
not only at the sandface, but over a significant areal extent in the and dividing by the area of the region (assuming no pressure varia-
vicinity of the well, the error can be significant. tion in the vertical direction):
One such case is that of a long and narrow reservoir with the
well located near one bounded end. Figure 2 illustrates the signifi-
cant localized pressure drawdown that occurs in this type of con- pdA
figuration. p = A
A .............................................................................................. (3)
To illustrate the pseudo-time issue, let us plot the pressure re-
sponse at the well for two scenarios:
1. Bounded Narrow Rectangular System where p~ is the average pressure inside the region of influence.
2. Semi-Infinite Narrow Rectangular System For practical purposes, a much more convenient way to eval-
uate the pressure is to use a material balance approach within the
Both reservoirs are the same width. The bounded system has its
end boundaries located beyond what we will call the region of in- region of influence:
fluence. The region of influence is defined as the region outside of
which there is no measurable pressure disturbance. Therefore, we p p G p
= 1
would expect the pressure responses for Cases 1 and 2 to be iden- Z Z Gir
tical. In fact, we can see from Figure 3 that they are quite different. ....................................................................................... (4)
The pressure response for Scenario 2 falls significantly below that
of Scenario 1. This is contrary to physics. where
The pressure responses ought to be identical, but because of
pseudo-time, there is an artificially induced difference between
them. The error arises because the pseudo-time for the semi-in- Gir =
(
Ar h 1 sw )
finite reservoir (Case 2) is based on an average reservoir pressure Bg
................................................................................... (5)
that remains constant at initial conditions through time (infinite
reservoirs do not deplete), while the finite case has a falling av-
erage reservoir pressure. Which one is correct? The answer is that and Ar is the area of the region of influence.
they are both wrong. However, since Scenario 1 allows for some Once in boundary dominated flow, the average reservoir pres-
reduction in the pressure at which the bulk properties are evalu- sure is established using precisely this method, but with the area in
ated, its pressure response will be closer to the correct answer. Equation (3) being defined as the total reservoir area.
50 Journal of Canadian Petroleum Technology
acting circle. Thus, the solution for Ar changes with time and is dif-
Establishing the Region of Influence ferent for different well/reservoir geometries.
Determining the region of influence is not a trivial matter. There To simplify the geometry of the solution significantly, we can
are (at least) two reasonable methods for establishing an affected approximate the region of influence as a rectangle (rather than a
reservoir area. One would be to calculate the pressure distribution circle with subtracted segments) with negligible impact on the re-
through the entire reservoir area and delineate the region, outside sult. The solution for area is divided into four different geometries,
of which there is less than some minimum percentage of pressure any combination of which may be part of a prescribed Ar (t) func-
drop (say 1%). This procedure, although analytically sound, would tion for a given pseudo-time calculation (Figure 4).
be very costly computationally, as a large array of solved pressures For example, well/reservoir geometry such as that shown
would be required at every time-step in the computation of pseudo- in Figure 4(C) would have a region of investigation whose area
time. In addition, some established analytical solutions for pres- would follow the formulas, in succession, for the single boundary,
sure at the wellbore (such as that of a bounded reservoir containing perpendicular boundaries and three boundaries solutions.
an infinite conductivity fracture) do not include the pressure solu- From Figure 4, it appears that the description of time-depen-
tion for all points inside the reservoir domain. dent areas of influence is somewhat cumbersome, as it requires
A more practical method for establishing a region of influence adjustments on the fly that depend on the particular geometry of
is to use the well established concept of the radius of investigation. the model. Since the pseudo-time correction is negligible in all but
The radius of investigation is defined as follows(1, 4). the most severe well/reservoir geometries, we may further approx-
imate the area of the region of influence by identifying the domi-
nant geometry factor (single boundary, perpendicular boundaries,
kt
rinv =
69.4c
(
SI units ) three boundaries or parallel boundaries) and applying the appro-
...................................................................... (6) priate area formula.
For instance, if d1, d2 and d3 are all of similar magnitude in the
The radius of investigation is an accepted measure of the speed three boundaries case, we may apply that area formula from time
of propagation of the pressure transient through diffusive media, zero until the radius of investigation reaches the far-field boundary.
and is a function only of permeability, porosity, fluid properties Likewise, if d1 and d2 are much closer than d3, and d3 and the far-
and time. Thus, the radius of investigation propagates through field boundary are on the same order of magnitude, then we may
any unbounded diffusive media at precisely the same velocity, re- apply the area formula for the perpendicular boundaries case from
gardless of reservoir geometry or changes in pressure drawdown. time zero.
Appendix A presents two synthetic model responses, clearly dem-
onstrating that this is true for an arbitrary reservoir shape and well Average Pressure in the Area of Influence and Corrected
location. Pseudo-Time
It is important to note that the radius of investigation does not The solution for the area developed above is used in conjunction
account for the magnitude of pressure depletion over time, only the with Equations (4) and (5) describing, respectively, average pres-
propagation of the leading edge of the region of influence. sure (p/z) within the region of influence and gas-in-place within the
Given the above, delineation of the region of influence in an iso- region of influence. The corrected pseudo-time is calculated using
tropic reservoir, containing a single vertical well, involves evalu- the following equation:
ating the intersection of the outward progressing circle prescribed
by the radius of investigation with the reservoir area itself:
( )
t
dt
ta = ct
0 ct
i
......................................................................................... (8)
Ar = rinv 2 A
........................................................................................ (7)
In Figure 5, the pseudo-time corrected solution for the semi-in-
where A is the area of the reservoir (defined by the model). finite reservoir case presented previously is compared against the
The formula given in Equation (7) applies to any arbitrary res- numerical solution, showing reasonably good agreement (the re-
ervoir shape and well location, but is restricted to vertical well ge- maining error cannot be reduced within the confines of the pseudo-
ometry and isotropic and homogeneous reservoirs. time approximation). Note, since the three boundaries case causes
More advanced well/reservoir geometries, such as fractured the most severe localized pressure drawdown of any rectangular
wells, anisotropic media, heterogeneities and/or layers, would
follow much the same procedure. However, the definition of the A. Single boundary B. Perpendicular boundaries
velocity of transient propagation would have to be defined sepa-
Ar  2rinv 2 rinvd 1 Ar  rinv 2 rinvd 1 rinvd 2
rately for all directions and possibly as a function of space. The
additional solution complexity and computational expense would rinv

likely preclude the practical application of a pseudo-time correc- d2 rinv

tion in such cases in favour of using a numerical model (which ex- d1


d1
plicitly evaluates fluid properties at each grid block). In order for
analytical models to remain favourable, they must be convenient
and efficient to solve. C. Three boundaries D. Parallel boundaries

Ar  d 1 d 3 (d 2 rinv) Ar  d 1 d 3 ( 2rinv )

Case Studies d2
d3
rinv
d3
rinv

Synthetic Example 1: Vertical Well in a d1 d1

Rectangular Reservoir
The corrected pseudo-time solution for a vertical well arbitrarily E. Three boundaries with hydraulic fracture

located in a rectangular shaped reservoir of arbitrary dimensions is Ar  d 1 d 3 ( d 2 xf rinv )


constructed as follows.
d3
d2 xf
Area of the Region of Influence d1 rinv

The area of the region of influence is defined as A = in an r2


unbounded system. As the region of investigation expands beyond
FIGURE 4: Region of influence areas for common geometries.
system boundaries, area elements are subtracted from the infinite
July 2007, Volume 46, No. 7 51
Pwf vs time- semi-infinite "three boundary" reservoir 7,000
3,500 k = 0.13 md
6,000 xf = 350 ft
3,000 ye = 1,100 ft
5,000

pwf (psia)
2,500 xe
4,000
pwf (psia)

2,000
Numerical 3,000
1,500 Corrected pseudo-time
2,000 xe = 2,400 ft
xe = 5,000 ft
1,000 xe = 10,000 ft
1,000 OGIP = 4.3 bcf
measured data
500
Uncorrected pseudo-time 0
0 50 100 150 200 250 300 350 400 450 500
0
0 50 100 150 200 250 Time (days)
Time (h)
FIGURE 5: Pseudo-time corrected solution for three boundary case FIGURE 7: Comparison of pressure solutions uncorrected pseudo-
compared with numerical simulation. time.

8,000
k = 0.13 md
7,000 xf = 350 ft
ye = 1,100 ft
6,000

pwf (psia)
5,000 xe

4,000

3,000
xe = 10,000 ft
xe = 5,000 ft
2,000 xe = 2,400 ft
measured data
1,000

0
0 50 100 150 200 250 300 350 400 450 500

Time (days)

FIGURE 8: Comparison of pressure solutionscorrected pseudo-


FIGURE 6: Production response analyzed using Agarwal and time.
Gardener Type Curve.
boundaries case is relevant. For the three boundaries system, the
geometry investigated, the approximations are expected to hold modified area formula and geometry is shown in Figure 4(E).
up as well (or better) for any general rectangular reservoir/vertical Upon correcting the pressure response using the new definition
well geometry. of pseudo-time, the previous model match is no longer valid indi-
cating too large an OGIP (see Figure 8). As the far-field boundary
Field Example 2: Fractured Well in a is reduced, the pressure response (properly) remains unchanged
Rectangular Reservoir until its location coincides with the radius of investigation. At that
point, the synthetic pressures decrease.
In this field study, an analytical reservoir model has been used to This case illustrates the value of the corrected pseudo-time func-
simulate the measured pressure response from a hydraulically frac- tion when history matching using analytical models with extreme
tured well. The production response, analyzed using the Agarwal well/boundary geometry. The conventional definition of pseudo-
and Gardener type curves(1, 2), suggests a bounded system with time in this case, generated history matches that did not properly
OGIP of about 28.3 106m3 (1 bcf) (see Figure 6). The objective of illustrate the boundary-dominated response observed from the type
the modelling exercise is to verify the interpreted gas-in-place, and curve analysis. In addition, the directionality of the results did not
improve upon the type curve model by fine-tuning. make sense. Only with the corrected pseudo-time, was the proper
A satisfactory history match is obtained using a narrow, rectan- synthetic pressure response created.
gular reservoir with the well located near a corner (see Figure 7). To complete the exercise, the history match is now refined, using
The simulated OGIP is more than four times higher than that in- the corrected pseudo-time model. The results indicate an OGIP of
terpreted using the type curve analysis. As the far-field boundary about 28.3 106m3 (1 bcf) (see Figure 9).
is moved closer to the well (and thus, OGIP is reduced), the simu-
lated flowing pressures increase! Common sense tells us that the
pressure response ought to decrease upon reduction in total reser-
voir volume in a bounded, depleting system. The anomalous syn- Conclusions
thesized pressure from the model is a result of the pseudo-time
error discussed previously. Figure 7 compares the results from 1. Without pseudo-time (or similar gas property correction rou-
three different boundary locations in the model. It also displays the tines), the analysis and modelling of gas reservoirs would
measured data, illustrating the original history match. Because of not be possible when using analytical solutions designed for
the pseudo-time error, the model does not properly (or uniquely) slightly compressible fluids.
indicate the OGIP or the bounded nature of the reservoir. 2. The conventional definition of pseudo-time is satisfactory
under conditions of boundary dominated flow.
3. During transient/transitional flow, the standard definition
Corrected Pseudo-Time Model of pseudo-time may cause incorrect and misleading results
when there are localized regions of high pressure drawdown.
The area formulas discussed in the previous section can be ex- These localized regions are caused primarily by severe well/
tended to hydraulically fractured wells. In this example, the three reservoir geometry (such as a narrow reservoir with the well
52 Journal of Canadian Petroleum Technology
Z = gas compressibility factor
~
7,000
k = 0.12 md
Z = average gas compressibility factor within the region of
6,000
OGIP = 1 bcf influence
ye = 1,030 ft ~
xf = 350 ft = viscosity at the average reservoir pressure within the
5,000 region of influence, mPas
xe = 2,280 ft
pwf (psia)
4,000 mPas
= viscosity, mPas
3,000
= porosity, %
2,000

1,000 Subscripts
0 i = initial
0 50 100 150 200 250 300 350 400 450 500
r = within the region of influence
Time (days)
References
FIGURE 9: History match using corrected pressures. 1. Agarwal, R.G., Gardner, D.C., Kleinsteiber, S.W. and
Fussell, D.D., Analyzing Well Production Data Using Combined-
located near a corner). Additional errors may result from Type-Curve and Decline-Curve Analysis Concepts; SPE Reservoir
large pressure gradients in very low permeability systems. Evaluation & Engineering, Vol. 2, No. 5, pp. 478-486, October
1999.
4. A new, corrected pseudo-time function is presented that eval-
2. Fraim, M.L. and Wattenbarger R.A., Gas Reservoir Decline-
uates gas properties at the average pressure of the region of
Curve Analysis Using Type Curves with Real Gas Pseudopressure
influence during transient or transitional flow, rather than and Normalized Time; SPE Formation Evaluation, Vol. 2, No. 6, pp.
the average reservoir pressure (based on external no-flow 671-682, December 1987.
boundaries). A general formula [Equation (7)] is given that 3. Blasingame, T.A. and Lee, W.J., Variable-Rate Reservoir Limits
describes the region of influence for an arbitrarily located Testing; paper SPE 15028 presented at the Permian Basin Oil and
well in a reservoir with an arbitrary shape. Specific solutions Gas Recovery Conference, Midland, TX, 13-15 March 1986.
are also derived for some simple (common) well/reservoir 4. ENERGY RESOURCES CONSERVATION BOARD, Theory and
geometries. Practice of the Testing of Gas Wells; Third Edition, ERCB-75-34, En-
5. The new pseudo-time definition solves the problems asso- ergy and Resource Conservation Board, Calgary, AB, 1975.
ciated with transient flow with severe geometry and yields
results that are more consistent with those from a numerical
simulator. Its primary limitation is the complexity involved in Appendix A
calculating the area of the region of influence (which changes
with time). However, with suitable approximations, this can The purpose of the following is to illustrate that the propagation
be simplified greatly. of the radius of investigation along any unobstructed direction is
6. The pseudo-time modifications could be easily incorpo- not affected by the influence of boundaries in other directions.
rated into any software by simply changing the pressure at The radius of investigation is defined as:
which gas properties are evaluated. The new pressure would
be based on the region of influence, rather than the average kt
reservoir pressure. Note that the region of influence pres- rinv =
69.4c
(
SI units )
sure and average reservoir pressure become identical during .................................................................. (A-1)
boundary dominated flow.
As shown, the radius of investigation is a function of the perme-
Nomenclature ability, time, porosity and the fluid properties.
Let us consider two reservoirs with identical properties but of
A = area of the reservoir, m2
different size and shape. Reservoir A is square shaped with the well
Bg = formation volume factor of gas, m3/m3 in the centre, while Reservoir B is long and narrow. Reservoir B
cg = gas compressibility, kPa-1 has a much smaller total volume than Reservoir A. However, the
ct = total compressibility, kPa -1 distance from the well to the x boundary is the same in both res-
c~ t = compressibility evaluated at the average reservoir pressure ervoirs. A schematic is shown in Figure 10. For simplicity we will
within the region of influence, kPa -1 assume that both reservoirs produce single-phase oil.
d = distance to boundary, m
G = gas in place, 106m3 Diagnostic plots for two reservoir models:
cumulative gas produced, 106m3
(Ppi-Ppwf)/q [psi/(bbl/d)], Derivative

Gp = illustrating the radius of investigation


10,000
h = net pay, m Reservoir A
k = permeability, mD 1,000
Reservoir B
YeA = 100,000 ft
q = production rate, 106m3/day
p = pressure, kPa 100 XeA = 10,000 ft
pp = pseudo pressure, kPa2/mPas Reservoir A
Reservoir B
ppwf = pseudo pressure at the sandface, kPa2/mPas 10 YeB = 100 ft
~
p = average pressure within the region of influence, kPa XeA = 10,000 ft Outer Boundary
re = exterior radius, m 1
(Xe) reached at
about 1,700 hours
for both
rinv = radius of investigation, m reservoirs.

rwa = apparent wellbore radius, m 0.1


sw = water saturation 0.0001 0.001 0.01 0.1 1 10 100 1,000 10,000 100,000

t = time, days or hours Time (hrs)


ta = pseudo-time, days or hours
T = temperature, K FIGURE 10: Comparison of radius of investigation between square
elongated reservoir.
xf = fracture half length, m
July 2007, Volume 46, No. 7 53
Data for Reservoirs A and B are as follows:
k = 20 mD Authors Biographies
= 10%
David Anderson (P.Eng) is a technical ad-
= 1.4194 mPas
visor with Fekete Associates, Inc. He has
c = 1.47 e-6 kPa-1 (1.02e-5 psi-1)
eight years of experience in the petroleum
XeA = YeA = 3,050 m (10,000 ft)
industry, including production optimiza-
XeB = 3,050 m (10,000 ft)
tion, gas deliverability modelling and well
YeB = 30.5 m (100 ft)
test analysis. He is currently the Technical
The time to reach the outer boundary X of both reservoirs is
Leader for Feketes RTA (Rate Transient
calculated by applying Equation (A-1). A time of 1,715 hours is
Analysis) Group. He has taught numerous
calculated when the radius of investigation is 1,525 m (5,000 ft)
industry courses on advanced production
(distance to the boundary from a well positioned in the centre of
decline analysis and has co-authored sev-
each reservoir).
eral technical publications on both pressure
The diagnostic plots in Figure 10 illustrate that the propagation
and rate transient analysis.
of the radius of investigation is identical for A and B. This be-
comes clear when comparing the derivative plots for each case.
Louis Mattar is the President of Fekete As-
The influence of the final (Xe) boundary, and subsequent onset of
sociates, Inc. He was the principal author of
pseudo-steady-state (unit slope), begins at exactly the same time.
the world-renowned E.R.C.B. publication,
Furthermore, the time, as shown in Figure 10, is the same as that
Theory & Practice of the Testing of Gas
calculated using Equation (A-1). Thus, we confirm that the radius
Wells, 1975. He specializes in well testing
of investigation along a particular direction is unaffected by the in-
and teaches it all around the world. He has
fluence of boundaries in other directions.
authored 45 technical publications. He is a
distinguished member of the Petroleum So-
ciety. In 1995, he received the Petroleum
ProvenanceOriginal Petroleum Society manuscript, An Improved Society Distinguished Author Award and
Pseudo-Time for Gas Reservoirs With Significant Transient Flow the Outstanding Service Award. In 2003,
(2005-114), first presented at the 6th Canadian International Petroleum Louis was the SPE Distinguished Lecturer
Conference the 56th Annual Technical Meeting of the Petroleum Society), in well testing.
June 7-9, 2005, in Calgary, Alberta. Abstract submitted for review De-
cember 10, 2004; editorial comments sent to the author(s) September 28,
2006; revised manuscript received October 31, 2006; paper approved for
pre-press October 31, 2006; final approval June 18, 2007.

54 Journal of Canadian Petroleum Technology