Saturation Distribution and Injection Pressure for

A Radial Gas-Storage Reservoir

E. G. WOODS
STUDENT MEMBER A/ME OKLAHOMA STATE U.
A. G. COMER STILlWATER, OKLA.
MEMBER A/ME

Downloaded from http://onepetro.org/JPT/article-pdf/14/12/1389/2213622/spe-401-pa.pdf/1 by guest on 21 March 2022

ABSTRACT in size depending upon the injection and withdrawal his-
tory of the reservoir. Not only will both of these factors
A mathematical model is presented and solved for deter- influence the required injection pressure, but also they will
mination of the saturation distribution and pressure in a have a definite bearing upon the amount of water produc-
radial gas-storage reservoir. The model consists basically tion occurring upon withdrawal from the storage area.
of two parts: (1) the growing gas-bubble core, and (2) the To solve this problem on an IBM 650 computer, several
surrounding aquifer. Since the total pressure at the injec- simplifying assumptions were made. Only the drainage
tion well is 'a function of the two-phase flow in the gas
portion of the relative-permeability curve was considered,
bubble and the unsteady single-phase .flow in the aquifer, thereby neglecting any hysteresis effects introduced during
the resistance to flow in both zones was taken into consid- an imbibition cycle. The quantity of gas withdrawn during
eration. the period of the field study was only a small percentage
The assumptions involved for both the radial equivalent of the total gas in place; therefore, this assumption is
of the Buckley-Leverett two-phase-flow equation and the justified. For the problem studied in this case, the com-
injection-pressure equation are as follows: (1) the geometry bined effects of capillarity and gravity were assumed to
i.l' radial. (2) the gas bubble is free to expand or contract, be negligible. This assumption becomes less valid as the
(3) compression or expansion of the "as within the hubble formation thickness or the average pore diameter increases
may occur at the beginning of a time step, (4) the fluids markedly. The justification of these assumptions should be
are immiscible, (5) water is incompressible within the ga~­ studied in each case.
storage region whereas it is compressible outside of this
region, (6) a stable gas-water interface exists and (7) gas
injection occurs at a constant rate or a series of constant THEORY
rates. This mathematical model for the prediction of satura-
The nUlthematical model was solved numerically using tion distribution and pressure for a radial gas-storage
an IBM 650 computer. A comparison is presented between aquifer is based upon the equations for radial two-phase
the predicted results of the model, the results assuming fluid flow'" and for radial, unsteady-state, single-phase
steady-state flow and the actual initial injection-pressure fluid flow.' The two-phase flow is considered to take place
history of an operating reservoir. Using the initial field in a "core" whose radius is equal to the maximum radius
pressure for a basis, the average deviation between the that the gas zone will attain, as indicated in Fig. 1. The
predicted pressure and the actual field pressure was less injection-withdrawal history is approximated by a series
than 4.3 per cent. of constant flow rates. Within the "core", gas is assumed
to behave as a semi-compressible fluid; that is, the gas is
assumed to have a constant density based on the average
INTRODUCTION gas-zone pressure for the flow period. Between each con-
stant-rate time increment, the gas density is allowed to
Underground storage of natural gas in abandoned oil change. This involves an iterative solution of the pressure
fields, in abandoned coal mines, in caverns and in aquifers equation. The liquid phase in this zone is considered to
has had varying degrees of success. The lack of these first be incompressible, whereas it behaves as a compressible
three facilities in the vicinity of most major gas marketing liquid outside of this region.
areas is leading more and more to storage in virgin aqui-
fers. For the most part, this storage in aquifers has led The equations of incompressible flow in the "core" may
to the study upon which this paper is based. be written for the gas phase
Several excellent papers have been published discussing
\l·v g
. ~=
oS. 0 , (1)
the problems involved in the underground storage of nat- ...

ural gas; however, to these writers' knowledge, none of

them considered the two-phase flow of fluids in the gas and for the water phase
bubble and the fact that the gas bubble will grow or shrink as.,
\l'V" + -'--=
'f' ot
0 (2)
Original manuscript receivEd in Society of Petroleum Engineers office
Sept. 4, 1962. Revised manuscript received Nov. 2. 1962. Paper presented
at 37th Annual Fall Meeting of SPE, Oct. 7-10, 1962, in Los Angeles. 'R"ferences given at end of paper.

DECEMBER, 1<)62 SPE 401 1389

In radial co-ordinates and upon substitution of two-phase zone. The equation becomes
iul. (S) T,
ig ( drlr
2-rrrh
into Eq. 1, the two-phase flow equation becomes
t.p,
0,07952-rrhk l [k,n(r) + k,,,(r)] , (11)

{tg {t",
iol.' oS" + oS. = 0 . (3)
2-rrrhcp or ot which is readily integrated by numerical methods.
This is a planar partial differential equation which may be The pressure increase due to water flow in the region
separated into two ordinary simultaneous differential equa- exterior to the two-phase zone and inside the semi com-
tions representing a doubly infinite set of curves.' By pressible "core" is approximated by Darcy's law in the
this method, Eq. 3 becomes form of Eq. 12.
!!.G .{t" 1 r,
2:;~/<p dr = ~t = ~ . (4) t.p, = p,-po 0.07952-rrhk ..flt
n-.
r,
(12)

Using the second pair of equations The pressure at the outer radius of the "core" is given
as a function of time by the equation for radial, unsteady-
dS = 0 (5)
dt ' state, single-phase flow.'"

= P. + 2~.~~ .. j § t.t.~t (t.t.~l_,]p" -

Downloaded from http://onepetro.org/JPT/article-pdf/14/12/1389/2213622/spe-401-pa.pdf/1 by guest on 21 March 2022

we obtain the solution
So = Constant . (6) p", 1 [( j

Therefore, the characteristic lines of Eq. 4 represent lines

(13)
of constant saturation. Since 10' is considered to be a
constant for a constant saturation, the first pair of Eq. 4 In this equation t.G, is the incremental volume of
may be written as water moved across the boundary of the "core" as related
.1 I to the incremental change in total gas volume during a
rdr=~dt . (7) time step. With the unsteady-state equation in this form,
2-rrhcp
the injection history of a field may be approximated by a
which, upon applying the boundary condition that r = roo series of constant injection rates.
at t = 0, has the solution

r, 1/. t + r
= ( -;rhcplg 0)'
w- (8)
The well bore pressure is then given by
Pw,. = flp". + flp". + po,. (14)
In the case of injection at a series of constant rates, The solution of Eqs. 9 and 14 for successive time steps
Eq. 8 takes the form yields a saturation distribution and pressure history for a

r, =( :~~ G, + r w') l (9)

radial gas-storage field subjected to a given gas-injection
history. At each time step, the solution is iterated to
determine an average pressure in the gas zone to be used
where r, is the radius of a surface of constant saturation. for a closer approximation of Poe.
The radius of the two-phase zone is determined from
Eq. 9 by using t'(S) at the saturation of the front, Sf' A typical pressure distribution might appear as in Fig.
which may be determined by the tangent construction 2. The pressure gradient in the gas bubble is due to two-
method' where phase flow, and the gradient in the remainder of the
incompressible core results from the flow of water. A
I' - 1.-0 (10) pressure gradient is created in the surrounding aquifer
g" - Sr_o as a result of water movement across the boundary, r,.
The solution of Eq. 9 for the set of saturations S
yields the saturation distribution within the two-phase COMPUTER CALCULATION
zone. These solutions may be combined with Darcy's law
for the two phases and the relative-permeability function The actual computer program was originally pro-
to determine the pressure increase due to flow in the grammed in IBM 650 FORTRAN computer language.
The calculation involves the input of k", and k,g data
from which the fractional flow of the displacing gas phase
was calculated from the equation

BORE
w
c::
=>
<II
(/)
W GAff BUBBLE
c::
Q.

GAS BUBBLE - , R"

/' ,
INCOMPRESSIBLE CORE --,.----'
RADIAL DISTANCE
.
FIG. l-MATHEMATICAL MODEL. FIG. 2-TYPICAL PRESSURE GliADIENTS, INJECTION CASE.

1390 JOURNAL OF PETROLEUM TECHNOLOGY

 5a, 5b and 5c. In the early history of the field, between
I Y =---k-;--- (15) 3 V2 and 7 days after injection was started, hydrate prob-
+ r .. jJ..
lems were encountered and no direct comparison can
kroiJ- ..
be made between the model and the field; however, it
The derivative of this function was detennined by may be noted that the model follows the same pressure
numerical methods,' with 1'•. , being the member of the trend as does the field. A 40-day shut-in period followed;
set 10' that satisfies Eq. 10 with a minimum of error. then injection was resumed. During this period the model
The maximum gas inventory was detennined from input perfonnance compares well with the field history. The
data of flow rates and times, and was used in conjunction maximum deviation during the remaining injection history
with Eq. 9 to determine the maximum radius of the gas- did not exceed 10 per cent of the corresponding field
storage zone r,. This equation was also used to obtain pressure. The average deviation was 30 psi, or 4.3 per
tenns for the numerical integration of Eq. 11. The solution cent of the initial fonnation pressure.
of Eqs. 9 and 14 for each time step began with an as- It also may be noted that the model performance closely
sumed average gas-bubble pressure. Then it was iterated followed the field pressures during a production test be-
to obtain a closer approximation of the actual pressure. ginning at 95 days. However, the effects of relative-per-
meability hysteresis should be included if the perfonn-
For a comparison purpose, it was decided to include ance during large-scale withdrawals is to be matched.
the pressure points assuming steady-state behavior. The Near the end of the test (between 100 and 102 days),

Downloaded from http://onepetro.org/JPT/article-pdf/14/12/1389/2213622/spe-401-pa.pdf/1 by guest on 21 March 2022

following equation was solved. the withdrawal rates decreased and the field pressure also
0.305 h k (p",'-p,')PF continued to decrease. The depletion-type production at
i. = ----=-=.,----:--,--,,---
iJ-g T Z log (relr",)
(16) this point may indicate that the relative-permeability hys-
teresis becomes important at this point since reduced per-
meability to water near the boundary as a result of trap-
FIELD COMPARISON ping of gas would retard the water influx. Deviation be-
tween the predicted and actual field pressure in the region
The perfonnance of the described mathematical model beyond 116 days may be due to a leak in the cap rock
has been compared with the initial injection history of an that was discovered later in the life of the field. This leak
operating gas-storage reservoir. This well was located in the cap rock is believed to lie in an area approximately
near the top of the geologic structure and was the first 112 mile from the injection well.
well into which gas was injected. Therefore, the assump- For comparison purposes, a series of calculations was
tion of radial geometry is valid since the formation ap- made assuming steady-state conditions and using a per-
pears to be reasonably homogeneous. fonnance factor of 0.247 in Eq. 16. The calculated points
A core sample was obtained from the St. Peter forma- are plotted in Figs. 5a, 5b and 5c. Steady-state conditions
tion of the Redfield, Iowa, underground gas-storage field, imply a fixed boundary; during any shut-in time, the
Northern Natural Gas Co., on which the relative-penne- pressure would drop to the initial fonnation pressure.
ability characteristics were determined by using a modified The movement of the gas-zone boundary, along with
Penn State technique. This information then was used to injection rates and total gas injected during the time of
determine the fractional flow of gas, I., and the derivative the field test, is shown in Figs. 6a and 6b. It is noted
of the fractional flow, 1/. The results are shown in Fig. 3 that the gas-zone not only expands during injection, but
plotted as a function of saturation. By using Eq. 9, the also expands during shut-in periods. This is attributed
investigators have determined a radial gas-saturation dis- to the unsteady-state flow in the aquifer.
tribution which is shown in Fig. 4.
The measured pressure history, the calculated perfonn-
ance based on the proposed model and the pressure his-
tory based on a steady-state model are presented in Figs. 100

1.0 i '~--==-'--------~-~I 20.0 80

r- I 1
rI
18 0
.

tJ') 0.8 /--19 : 16.0 z 60

o
<l: i-- --i 14.0 <i !;i
~ J 12.0 ~
~
0.6
rL-.'
I '
i
'-'
10.367<l:
c::
=>
~ 40
(/)

: O'r \:---------- -------=1:°: :

(/)
(3
I-
~ I ~ 15 20
o
e: 0.2 ~ \ ~ 4.0 ~ c::
w
c.. --------------.----~
~
1
L \ 9 2.0 :
o
o
L...J...~--..L:=_ _~_ _ _ _ _ _ _ -,---,I 0
o .20 .40 .60 .80 1.0
o 20 40
PERCENT GAS SATURATION
. 60 80 100
DIMENSIONLESS RADIUS. +-
f
(RSD)

FIG. 3--FRACTIONAL FLOW CURVES. FIG. 4-SATURATION DISTRIBUTION, RADIAL CASE.

DECEMBER, 1962 1391

 I.

Around 90 days there was a slight show of gas in the pressures were predicted with an average deviation of 4.3
nearest offset well, Nelson No.2, which is approximately per cent, and the bubble radius was predicted within 5
1,320 ft from the injection well. At this time, the radius per cent at the time of gas breakthrough into the nearest
of the gas bubble calculated by using the model presented offset well.
in this paper was 1,258 ft, or less than a 5 per cent error. 2. This model, along with reservoir and relative-perme-
If it had been assumed that the gas bubble was at residual ability data from the initial well, can provide valuable
water saturation, the predicted radius would have been engineering information to be used in the initial testing
in error by more than 65 per cent. of the field and in estimating future performance. After
the injection program has begun, and after sufficient field
CONCLUSIONS data have been obtained, the model could be used along
the field pressures to determine better reservoir average
1. The proposed model appears to predict favorably values of porosity, formation thickness and base perme-
the initial history of an operating gas-storage reservoir in ability than are available from a weighted one-well average.
an aquifer that was developed for gas storage. The field
3. The large error in the predicted radius of the gas
bubble when assuming a constant saturation in the bubble
1,000 ; I I I , I TI ,I emphasizes the necessity of considering the saturation dis-
SYMBOLS
tribution resulting from two-phase flow in predictions of
CompuIer DoIo-' this type.

Downloaded from http://onepetro.org/JPT/article-pdf/14/12/1389/2213622/spe-401-pa.pdf/1 by guest on 21 March 2022

Field Dot9--'
900 Steady Stote-.. 4. Extension of the model to the withdrawal case by
inclusion of relative-permeability hysteresis appears to be
possible by using an approach similar to that used by Ribe'
UJ 800 for the nonhysteresis case. This would provide useful
...J
0-
:I: £1
information on the water-production rates to be expected
~Il. '" as weIl as the amount of gas that would not be economic-
0-
t-
t- ,, ally recoverable because of excess water production.
o 700 , ~

III
!- Hydrate 5. Also, the model could be extended to conical geo-
Z
t ~P~
__I._mT-__~S~hu~t~ln~__~ metry which would more closely approximate the domal
"''"
~ !
~-,i-,-,~____
I _________________ J symmetry of many storage fields by applying the method
600
o 10 45 50 55 60 presented by Welge, et aI.' This reference also suggests the
TIME (DAYS) method of solution for the elementary multiwell system of
a ring of uniformly spaced wells.
FIG. 5A-bJECTIO:'> PRESSCRE.

NOMENCLATURE':'
1,000

SYMBOLS
Tc = internal radius of aquifer
Computer OOto-"
Field Oato--'
PF = performance factor
Steady $tote-"
900
'-, /' i
P"n = pressure at radius r, and at time n
[ ~ /- P."n = pressure at radius r" and at time n.
~~ .... , ",,1 "'--,;
"-----~ ~""
,,----- "
I -
800
r[;:.-''', -""'\ .' ;::r' -
ACKNOWLEDGMENT
[-r . ''.-~, --1 --'
'~I \
;'0"
I
='1
. The authors are indebted to Northern Natural Gas Co.
700
[ ~--- L________ l for supporting the research project which led to the devel-
~ opment of the model and for permission to use the actual
!~ Shut In : field data from the St. Peter formation of their Redfield,
I

~
~ I Iowa, storage field. Also, we thank William Granet and
600
60 65 70 75 80 85 90 his staff at the Oklahoma State U. Computing Center for
TIME (DAYS) their helpful suggestions and for the use of the computing
facilities.
FIG. 5B-I!'OJECTlo:,> PRESSURE.
*For other symbol definitions. see AIME Symbols List., Trans, AIME
(1956) 207,363.
~
'= I
1,000
, I
SYMBOLS

OO'~ ~ ~~
Cbmputer Data - ..
field Data - - •
Steady Stat. - •
-I
~ :::1500
~~~--~

80i\I, C
o! 500 , , , ,
! .~--

100 105 '"~c:x 1000

1\ ' ' "'-.:\ TIME (DAYS) \oJ a:: .-.--
il~-O_~ __ ~_
V
rb-(Fectl--·
G,·(M)'~CF)---
!g-(MMGF/DAY)- •
•
I
~ i
-\ .
700 1- L ______ , . §
~ 11 /"
~ \t'f-~ ~-~Q._- ~--.
(f) 500
I -' .10
- Shut In \ I ..,/
11\1 ~ /'
~-- ~~ ----~ ------------------------..f--
6 :' I' ! I ._'O___C!__ _ , ---: .01
95 100 105 110 115 120
90 10 20 30 40 50 60

TIME (DAYS) TIME (DAYS)

FIG. 5C-INJECTION PRESSURE. FIC. 6A-GAS·BuBBLE RADIUS AI"D GAS !I"JECTED.

JOURl'\AL OF PETROLEUM TECHl'iOLOGY

 500.0
;:10.0
100.0
~
U:
~
~~u..---

'"
:: 1500
~ 11 /~ (
.~

-:_~qs_J{ithdrowol . -----j l 10.0

UJ ~ ~~ i / I' / ,.. __ - j ....---------------_ ;J

V if _ ri---.-,.t' -F--
~ ~IOOO
0:--- ,1---- -tt
ZI_.O_ _ ~,,'-I SYMBOLS 1.0
Q ~
, ......
".4'

b aJ
rb (Feet),---
~ g'5
Gj (MMCF ) - - .
~ aJ 500 ig (MMCF/DAY)-o
(f) .50 ~.--- - .10
(3 - - - ( ! ) - - -

Downloaded from http://onepetro.org/JPT/article-pdf/14/12/1389/2213622/spe-401-pa.pdf/1 by guest on 21 March 2022

.10 o ! ! 1 ! ! I I , I ' I I
.010
60 70 80 90 100 110 120

TIME (DAYS)
FIG. 61J-GAS-BuBBLE RADIUS AND GAS bJECTED_

REFERENCES 4. van Everdingen, A. F. and Hurst, W.: "The Application of the

Laplace Transformation to Flow Problems in Reservoirs",
1. Ribe, K. H.: "Production Behavior of a Water-Blocked Oil Trans., AI ME (1949) 186,305.
Well", Trans., AIME (1960) 219, 1. 5. Betz, H., et al: Differential Equations With Applications, Har-
~. Welge, H. J., Johnson, E. F., Hicks, A. L. and Brinkman, per, N. Y. (1954) 230.
F. H,: "An Analysis for Predicting the Performance of Cone- 6. Katz, D. 1., et al: Handbook of Natural Gas Engineering, Mc-
Shaped Reservoirs Receiving Gas or Water Injection", Jour. Graw-HilI Book Co., Inc., N. Y. (1959) 413.
Pet. Tech. (Aug., 1962) 894.
7. Salvadori, M. G. and Baron, B. 1.: Numerical Methods in EII-
3. Collins, R. E.: Flow of Fluids Through Porolls Materials, Rein- gineering, Prentice Hall, Inc., Englewood Cliffs, N. .T. (195~ I
hold Puhlishers, Inc., N. Y. (1961) 149. 69. ***

DECE'IIBER. 196~