studied the effects of water influx on the shape of the plot of p/z vs.

GP and showed that the shape and direction of the deviation from straight line depend on the strength of the aquifer support system and on the aquifer properties and the reservoir/aquifer geometry. If the initial gas in place is known from other sources, such as volumetric estimates, we can calculate We from Eq. 3.32. In practice, however, usually both W, and G are unknown, and calculation of initial gas in place requires an independent estimate of water influx. Therefore, in the next section we discuss three methods for estimating water influx. Methods For Estimating Water Influx. Water influx results from a reduction in reservoir pressure following gas production. Water influx tends to maintain, either partially or wholly, the reservoir pressure. In general, both the effectiveness of the pressure support system and the water influx rates are governed by the aquifer characteristics, which principally include the permeability, thickness, areal extent, and the pressure history along the original reservoir/aquifer boundary. Note that, in practice, estimating water influx is very uncertain, primarily because of the

lack of sufficient data to characterize the aquifer (especially its geometry and areal

continuity) completely.
Because wells are seldom drilled intentionally into an aquifer to gain information, these data must be either assumed or inferred from the geologic and reservoir characteristics. Generally, reservoir/aquifer systems are classified as either edgewater or bottomwater

drive. In edgewater-drive systems, water moves into the reservoir flanks, while
bottomwater drive occurs in reservoirs with large areal extents and gently dipping structures where the aquifer completely underlies the reservoir. van Everdingen and Hurst's and Carter and Tracy's methods are applicable only to edgewater-drive geometries or for combined geometries that can be modeled as edgewater-drive systems, while the Fetkovich method16 is applicable for all geometries

3-19

1. van Everdingen-Hurst's Method. In 1949, van Everdingen and Hurst presented an unsteady-state model for predicting water influx. As Fig. 3.8 shows, the reservoir/aquifer system is modeled as two concentric cylinders or sectors of cylinders. The inside cylindrical surface, defined by radius rr, represents the reservoir/aquifer boundary, while the outer surface is the aquifer boundary defined by ra. Radial flow of water from the aquifer to the reservoir is described

mathematically with the radial diffusivity equation'

a2
PD + 1
c3rD

aPD = D
rD (r D

(3.33)
?tD

where the dimensionless variables are defined in terms of the aquifer properties. The
dimensionless pressure for constant-rate conditions at the reservoir/aquifer boundary is

PD =

0.00708kh(pi - p) q

(3.34)

For constant-terminal-pressure conditions,

PD = Pi-P
Pi -Pr

(3.35)

The dimensionless radius is defined in terms of r1:

rD = r/rn

(3.36)

and for t in days,

3-20

tD =

0.00633kt

(3.37)

4ctrr
van Everdingen and Hurst derived solutions to Eq. 3.33 for two reservoir/aquifer

boundary conditions-constant terminal rate and constant terminal pressure. The water
influx rate for the constant terminal-rate case is assumed constant for a given period, and the pressure drop at the reservoir/aquifer boundary is calculated. For the constant-pressure case, the water influx rate is determined for a constant pressure drop over some finite time period. Reservoir engineers usually are more interested in determining the water influx than the pressure drop at the reservoir/aquifer boundary, so we will focus on water-influx calculations under constant-pressure conditions. van Everdingen and Hurst derived the constant-pressure solutions in terms of a dimensionless water influx rate defined by

na =

qwp 0.00708khAp

(3.38)

Integrating both sides of Eq. 3.38 with respect to time yields

tD fgDdt

D-

0.00633k Ji tr2 0.00708khOp

p

1\

4tc

t q dt j 0 w

90gc

thr2 Op 0

jq dt

(3.39)

In material-balance calculations, we are more interested in the cumulative water influx than in influx rate. Therefore, because cumulative water influx, We, is

3-21

We= f gwdt 0 and dimensionless cumulative water influx is

(3.40)

tD

QpD = JgDdtD

(3.41)

we can combine Eqs. 3.39 through 3.41 to obtain

pD

(3.42)

1.1 19 c t hri Op

Thus, We = 1.1 19 Cthr2 A PQPD

(3.43)

If the total productive reservoir life is divided into a finite number of pressure reductions or increases, we can use superposition of the solution given by Eq. 3.43 to model the water-influx behavior for a given pressure history. This method assumes that the pressure history at the original reservoir/aguifer boundary can be approximated by a series of step-by-step pressure changes. Fig. 18.9 shows the modeling of a pressure history. Referring to Fig. 3.9, we define the average pressure in each period as the arithmetic average of the pressures at the beginning and end of the period. Thus, for an initial aquifer pressure, p;, the average pressure during the first time period Pi = 2 (pi + pl) Similarly, for the second time period, P2 = _(PI + P2)- In general, for then' time period,

3-22

Pn = 2 (Pn-1 +Pn)

We can then calculate the pressure changes between time periods as follows. Between the initial and first time periods,

APo=(pi-P1)=pi- I(pi+p1)= I (pi-P1) 2 2

Similarly, between the first and second time periods,

4Po =P1-P2 = -(Pi +P1)-2(P1 +P2)= 2(Pi P2)

2

In general, for the (n-1) and nt time periods,

1 dPo = Pn_i - Pn = 2 (Pn-2 + Pn-1)

1 1 (Pn-1 + Pn) = 2 (Pn-2 Pn)

During each time increment, the pressure is assumed constant (i.e., constant-pressure solution), and the cumulative water influx for n time periods is

n We(tn) = B YOpiQpD (tn

i=1

-ti-OD

(3.44) (3.45)

where B = 1.1 19 4 ch r2

If the angle subtended by the reservoir is less than 360 (Fig. 3.8). then B is adjusted as follows:

3-23

8) B = 1.119cthr2( r 360

(3.46)

The pressure change during each time increment, as explained above, is calculated with

LPi = 2 (Pi-2 - Pi), i =1,2...n

(3.47)

and pi-2 = po = initial pressure. Each Opi in Eg. 3.44 is multiplied by the dimensionless
cumulative water influx, QPD. evaluated at a dimensionless time corresponding to the time for which Opi has been in effect. For example, ipi will have been in effect for the total productive life of the reservoir, so QPD will be evaluated at (t1 - 0)D. In general, have been in effect for the time period t-t,,_1 so QPD that multiplies L\Pn will be evaluated
at (t-tn_1)D.

6Pn will

To simplify calculations, Tables 3.1 and 3.2 present values for dimensionless cumulative water influx as a function of time for both infinite-acting and finite aquifers. Alternatively, for the special case of infinite-acting aquifers, Edwardson et al. developed polynomial expressions for calculating QPD. These expressions, Eqs. 3.48 through 3.50, depend on dimensionless time: For tD < 0.01,

t7C D

QpD(tD) = 2

(3.48)

For 0.01 < tD < 200,

3-24

( ) _ 1.2838t'12 + 1.19328tD +0.269872t 3 2 + 0.00855294t2

QPD tD

1+ 0.616599t

2 + 0.0413008tD D

(3.49 )

For tD > 200,

_ -429881+ 2.02566t

ht(tD)

D D

(3.50)

Similarly, Klins et al. 2' developed polynomial approximations for both infinite-acting and fuiite aquifers. The van Everdingen and Hurst method also is applicable to linear flow geometries (Fig. 3.10). For linear flow, we define a dimensionless time in terms of the reservoir length, L, as

0.00633kt
tD =

cto

(3.51)

Following a derivation similar to that presented for radial flow, we find that the cumulative water influx for n time periods is

n e(tn) B ApQ i pD(tn -ti-1) D

(352)

where the parameter R is defined in terms of the reservoir length,

B = 0.178 tct hL

(3.53)

Derived from exact solutions to the diffusivity equation, the van Everdingen and Hurst
method models all aquifer flow regimes (i.e., transient and pseudosteady-state) and is

3-25

applicable to both infinite acting and finite aquifers. Example 3.6 illustrates the following

calculation procedure.
1. First, calculate the parameter B for radial flow,

B = 1.1 194cthr2

e
r 360

(3.46)

or for a linear flow geometry,

B = 0.178 4 tct hL

(3.53)

2. Calculate the pressure change, Api. between each time period,

LPi = 2 (Pi-2 Pi), i =1,2...n

3. Calculate the tD that correspond to each time period on the production history. For a radial flow geometry,

tD =

0.00633kt
2
4tct r r

(3.37)

and for a linear geometry,

0.00633kt
tD =

ictL2

(3.51)

3-26

4. For each tD computed in Step 3, calculate a dimensionless water cumulative influx, QPD(tD). For an infinite-acting aquifer we can use Eqs. 3.50 through 3.52, use Klins et al.'s21 equations, or read the values directly from Table 3.1. For finite aquifers, we must use Klins et al.'s equations or Table 3.2. 5. Calculate the water influx:
We(ts)= BEAPiQpD(tn ti-1)D i=1

(3.44)

3. Carter-Trapy Method. van Everdingen and Hurst's method was developed from exact solutions to the radial diffusivity equation and therefore provides a rigorously correct technique for calculating water influx. However, because superposition of solutions is required, their method involves rather tedious calculations. To reduce the complexity of water influx calculations, Carter and Tracy proposed a calculation technique that does not require superposition and allows direct calculation of water influx. If we approximate the water influx process by a series of constant influx intervals, then the cumulative water influx during the jt1' interval is

J-1 We(tDJ)= y gDn(tDf+1 -tDn) n=0

(3.54)

Eq. 3.54 can be rewritten as the sum of the cumulative water influx through the ith interval and between the ith and jth intervals:

J-1 We(tDJ)= y gDn(tDn+1 -tDn) n=0

J-1 + ygDn(tDn+1 _tDn) n=i

J-1
We(tDj W (t Di)+nY eigDn(tDn+l-tDn)

(3.56)

3-27

Using the convolution integral, we also can express the cumulative water up to the j`h
interval as a function of variable pressure:

t D]j

we (t Dj) = B

OP(

d d [Q pD (t a )]da

(3.57)

Combining Eqs.

3.56 and 3.57, we use Laplace transform methods to solve for the

cumulative water influx in terms of the cumulative pressure drop, 0 Pn:

en - wen-1 +(tDn -tDn-1)

n - wen 1P'D (tDfl) tDn)_ PD tDn-1PD(tDn)

BOp

(3.58)

where B and tD are the same variables defined previously for the van Everdingen-Hurst method. The subscripts n and n-1 refer to the current and previous timesteps, respectively, http://www.youtube.com/watch?v=k

and

OOoHWhQNPE
(3.59)

P n= P aq, i- P n

PD is a function of tD and for an infinite-acting aquifer, can be computed from the following curve-fit equation:

370-529(t,,)112 + 137.582(tD) + 5.69549(tD)312

PD(tD) =

(3.60) 328.834+265.488(tD)Ii2 +45.2157(tD)+(tD)3iz

3-28

In addition, the dimensionless pressure derivative, P'D, can be approximated by a

curve-fit equation.

P'D (tD) =

716.441+46.7984(tD) 1/2 +270.038(tD)+71.0098(tD)3"2 (tD)1"2 + 1,204.73(tD) + 618.618(tD) 3/2 + 538.072(tD)2 + 142.41(tD)5i2 1,296.86
(3.61)

Eqs. 3.60 and 3.61 model infinite-acting aquifers; however Klins et al. developed similar polynomial approximations for both infinite and finite aquifers. We should stress that, unlike the van Everdingen-Hurst technique,- the Carter-Tracy method is not an exact solution to the diffusivity equation, but is an approximation. Research conducted by Agarwal, however, suggests that the Carter-Tracy method is an accurate alternative to the more tedious van Everdingen-Hurst calculation technique. The primary advantage of the Carter-Tracy method is the ability to calculate water influx

directly without superposition. The Carter-Tracy method, which also is applicable to infinite acting and finite aquifers, is illustrated with the following calculation procedure and Example 3.7. 1. First, calculate the van Everdingen-Hurst parameter B for radial flow,

B= 11190Cthr2

8 360

(346)

or for a linear flow geometry,

B = 0.178 0 cthL

(3.53)

2. Calculate the pressure change, Opn, for each time period,

3-29

(3.59)

3. Calculate the van Everdingen-Hurst dimensionless times, tD, that correspond to each
time period on the production history. For a radial flow geometry,

tD =

0.00633kt

ctr
2

(3.37)

and for a linear geometry,

tD =

0.00633kt

ctL2

(3.51)

4. For each tD computed in Step 3, calculate a PD and a P'D. For infinite-acting radial aquifer, we can use Eqs. 3.60 and 3.61 to calculate PD and P'D, respectively:

370-529(t,, )112 + 137.582(tD) + 5.69549(tD)312

PD(tD)=

(3.60) 328.834 + 265.488(tD)li2 + 452157(tD) + (tD)3i2

In addition, the dimensionless pressure derivative, P'D, can be approximated by a

curve-fit equation.

716.441+ 46.7984(tD )112 + 270.03 8(tD) + 71.0098(tD )312

P'D (tD) =1,296.86(tD) 1/2

+ 1,204.73(tD) + 618.618(tD)312 + 538.072(tD)2 + 142.41(tD)112 (3.61)

3-30

We also can use Klins et al's equations. For infinite aquifers, we must use Klins et al's equations

5. Calculate the water influx:

BOp W =W en en 1 + (t -t Dn Dn -) 1 n -Wen-1pD (tDfl) pD(tDn) tDn , 1p D (tDn) (3.58)

3. Ferkovich Method. To simplify water influx calculations further, Fetkovich proposed a model that uses a pseudosteady-state aquifer PI and an aquifer material balance to represent the system compressibility. Like the Carter-Tracy method, Fetkovich's model eliminates the use of superposition and therefore is much simpler than van the Everdingen-Hurst method. However, because Fetkovich neglects the early transient time period in these calculations, the calculated water influx will always be less than the values predicted by the previous two models. Similar to fluid flow from a reservoir to a well, Fetkovich used an inflow equation to model water influx from the aquifer to the reservoir. Assuming constant pressure at the original reservoir/aquifer boundary, the rate of water influx is

qw =

dd

= 1(Paq - Pry

(3.62)

where n = exponent for inflow equation (for flow obeying Darcy's law, n = 1; for fully turbulent lfow, n = 0.5).

3-31

Assuming that the aquifer flow behavior obeys Darcy's law and is at pseudosteady-state conditions, n = 1. Based on an aquifer material balance, the cumulative water influx

resulting from aquifer expansion is

W1 (Paq,i

- Paq )

(3.63)

Eq. 3.63 can be rearranged to yield an expression for the average aquifer pressure,

Paq = Paq i

1-

We
CtWiPaq,i

= Paq,i 1- We
Wei

(3.64)

where W

ei = ctPaq,iWi

(3.65)

is defined as the initial amount of encroachable water and represents the maximum possible aquifer expansion. After differentiating Eq. 3.64 with respect to time and rearranging, we have

dW e __ dt

Wei dPag pi dt

(3.66)

Combining Eqs. 3.62 and 3:66 and integrating yields

Paq d P

- ag

=-f

t JP.

Paq -P P r aq, i

(3.67)
p Wei
wit

or Paq-Pr=(Paq-Pyi)e Wei

(3.69)

3-32

Table

3.10

summarizes the equations for calculating the aquifer PI for various

reservoir/aquifer boundary conditions and aquifer geometries. Note that we must use the aquifer properties to calculate J. From Eq. 3.67, we can derive an expression for (P Eq. 3.68 and rearranging, we have -Pr), and following substitution into

JP.ti dWe

dt

= J (Paq,i Pr) e Wei

(3.69)

which is integrated to obtain the cumulative water influx. We: i

We = W ei

JP 1-e

aq,i
ei

.t (3.70)

P (Paq, i Pr) aq, i

Recall that we derived Eq. 3.10 for constant pressure at the reservoir/aquifer boundary. In reality, this boundary pressure changes as gas is produced from the reservoir. Rather than using superposition, Fetkovich assumed that, if the reservoir/aquifer boundary

pressure history is divided into a finite number of time intervals, the incremental water
influx during the nth interval is

JPaq, i Lt n

Wen

Wei (Paq,n-1 - Prn) 1Psi

Wei

(3.71)

3-33

where We,n-1
Paq,n-1 = Paq,i 1Wei Pm =

(3.72)

PM-1+Pm
2

(3.73)

Although it was developed for finite aquifers, Fetkovich's method can be extended to infinite-acting aquifers. For infinite-acting aquifers, the method requires the ratio of water influx rate to pressure drop to be approximately constant throughout the productive life of the reservoir. Under these conditions, we must use the aquifer PI for an infinite-acting aquifer.

The following calculation procedure illustrates this method. I. Calculate the maximum water volume, WCi, from the aquifer that could enter the gas reservoir if the reservoir pressure were reduced to zero.

Wei = CtPaq,iWi

(3.65)

where W1 depends on the reservoir geometry and the PV available to store water. 2. Calculate J. Note that the equations summarized in Table 3.10 depend on the boundary conditions and aquifer geometry. 3. Calculate the incremental water influx, interval.
- aq,i n
W. t1Wen = P e1 (aq,n_-1 -Pm) aq,i

AWen, from the aquifer during the nth time

JP

At (3.71)

W
1-e

3-34

4. Calculate WC11:

Wen = Y AWei
i=1

Estimating Original Gas In Place With Material Balance For A Dry-Gas Reservoir With Water Influx Once the water influx has been calculated, we can estimate the original gas in place with material-balance concepts. The general form of the material-balance equation including

water influx:

GBgi = (G-GP)Bg+We BwWp,

(3.74)

which can be rearranged to yield

GpBg +WpBw =G+ (Bg Bgi)

We (Bg Bgi )

(3.75)

If we define a water influx constant, C, in terms of the cumulative water influx as

We = Cf(p,t),

(3.76)

then Eq. 3.75 becomes

GpBg + WpBw (Bg Bgi)

G+- Cf(p't) (Bg Bgi )

(3.77)

3-35

The form of Eq. 3.77 suggests that, if water influx is the predominant reservoir drive mechanism, then a plot of [(GeBg)+(WpBW)]/(Bg Bg;) vs. f(p,t)/(Bg Bg;) will be a straight line with a slope equal to C and an intercept equal to G. The functional form of f(p,t) varies according to the water influx model used. Any water influx model, such as steady state, unsteady state, or pseudosteady state, can be used with Eq. 3.77. Note that, if the incorrect water influx model is assumed, the data may not exhibit a straight line. Example 3.9 illustrates the application of Eq. 3.77 for an unsteady-state water influx model.

3.3.3 Volumetric Geopressured Gas Reservoirs. We developed the material-balance equation for a volumetric dry-gas reservoir assuming

that gas expansion was the dominant drive mechanism and that expansions of rock and
water are negligible during the productive life of the reservoir. These assumptions are valid for normally pressured reservoirs (i.e., reservoirs with initial pressure gradients between 0.43 and 0.5 psi/ft) at low to moderate pressures when the magnitude of gas compressibilities greatly exceeds the effects of rock and connate water compressibilities. However, for abnormally or geopressured reservoirs, pressure gradients often approach values equal to the overburden pressure gradient (i.e., -1.0 psi/ft). In these and other higher-pressure reservoirs, the changes in rock and water compressibilities may be important and should be considered in material-balance calculations. Following Ramagost et al.'s23 derivation, we begin with Eq. 3.26 for a normally pressured volumetric gas reservoir and include the effects of changing water volume, A VW, and formation (rock) volume, A Vf. As Fig. 3.12 shows, the general form of the material-balance equation for a volumetric geopressured reservoir is

GBg; = (G-Gp)Bg+ A Vw + A Vf ,

(3.78)

3-36