Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
I R. AL-HUSSAINY
JUNIOR MEMBER AlME
MEMBERS AlME
I
ABSTRACT to the theory of isothermal flow of gases through porous
The efject o f variations o f pressure-dependent viscosity media. The present state of knowledge is far from being
and gas law deviation factor on the flow of real gases fully developed. The difficulty lies in the non-linearity
t h r o ~ ~ gporous
h media has been considered. A rigorous of partial differential equations which describe both real
and ideal gas flow. Solutions which are available are
gas Pow9 equation was developed which is a secorrd order,
non-linear partial diflerential equation with variahle co- approximate analytical solutions, graphical solutions, ana-
logue solutions and numerical solutions.
eficients. This equation was reduced by a change of vari-
able t o a form similar to the diflusivity equation, but The earliest attempt to solve this problem involved
with potential-dependent di.fusivity. The change o f vari- the method of successions of steady states proposed by
able can be used as a new psezdo-pressure for gus flow Muskat.' Approximate analytical solutions' were obtained
which replaces pressure or prc>ssur.e-squared (4s currently by linearizing the flow equation for ideal gas to yield
rrpplied t o gas flow. a diffusivity-type equation. Such solutions, though widely
Substitution o f the real gas pseudo-pressure has a used and easy to apply to engineering problems, are of
limited value bemuse of idealized assu~mptionsand restric-
nrtmber o f important consequences. First, second degree
pressure gradient terms which have cotnmonly been ne- tions imposed upon the flow equation. The validity of lin-
glected under the assumption that the pressure gradient earized equations and the conditions under which their
solutions apply have not been fully discussed in the literature.
is small everywhere in the .flow system, are rigorously
Approximate solutions are those of Heatherington et a/..'
handled. Omission of second degree terms leads t o rer-
MacRobertsl and Janicek and Katz.' A graphical solution
ious errors in estimated pressure distributions for tight
of the linearized equation was given by Cornell and
formations. Second, flow equations in terrns of the real
Katz.>lso, by using the mean value of the time de-
,sas pseudo-pressure d o not contain viscosity or gas law
rivative in the flow equation, Rowan and Clegg' gave
deviation factors, and thus avoid the need for selection of
a n average pressure to evaluate physical properties. Third, several simple approximate solutions. All the solutions
were obtained assuming small pressure gradients and
the real gas pseudo-pressure can be determined nnmerically
in t e r m of pseudo-reduced pressures and temperatures constant gas properties. Variation of gas properties with
pressure has been neglected because of analytic difficultie5.
f r o m existing physical property correlations t o provide
even in approximate analytic solutions.
generally useful information. The real gas pseudo-pressure
was determined by numerical integration and is presented Green and W i l t s h s e d an electrical network for sirn-
in both tabular and graphical form in this paper. Finally, ulating one-dimensional flow of an ideal gas. Numerical
production of real gas can be correlated in terms o f the methods using finite difference equations and digital com-
real gas pseudo-pressure and shown to be similar to puting techniques have been used extensively for solving
liquid flow as described by di.#usivity equation solutions. both ideal and real gas equations. Aronofsky and Jenkins" I"
and Bruce et al." gave numerical solutions for linear
Applications o f the real gas pseudo-pressnre t o radial
and radial gas flow. Douglas et al." gave a solution for
pow systems under transient, steady-state or approximate
a square drainage area. Aronofsky" included the effect
pseudo-steady-state injection or production have heen con-
of slippage on ideal gas flow. The most important contri-
sidered. Superposition o f the linearized real gas pow
bution to the theory of flow of ideal gases through porous
solutions t o generate variable rate performance was
media was the conclusion reached by Aronofsky and
investigated and found satisfactory. This provides justi-
Jenkins" that solutions for the liquid flow case'" could
fication for pressure build-up testing. It is believed that the
be used to generate approximate solutions for constant
mncept o f the real gas pseudo-pressure will lead to im-
rate production of ideal gases.
proved interpretation o f results of current gas well testing
procedures, both steady and unsteady-state in nature, and A n equation describing the flow of real gases has been
improved forecasting o f gas production. solved for special cases by a number of investigators using
numerical methods. Aronofsky and Ferris'honsidered
linear flow, while Aronofsky and Porter" considered
radial gas flow. Gas properties were permitted to vary as
In recent years a considerable effort has been directed linear functions of pressure. Recently, CartePs proposed
an empirical correlation by which gas well behavior can
be estimated from solutions of the diffusivity equation
Original manuscript received in Society of Petroleum Engineers office using instantaneour values of pressure-dependent gas
.June 28. 1966. Revised m a n u s c r i ~ tof S P E 1243A received Feb. 18, 1966.
Paper was presented a t S P E Annual Fall Meetlng held In Denver,
Colo., Oct. 3-6, 1965.
Presently on the faculty at Stanford U. 'References given a t end of paper.
. . . - (10)
Thus:
T h e velocity vector in Eq. 1 is given by Darcy's law
for laminar flow as:
MAY, 1966
equation has not been presented previously ~n connectron
with gas flow. Equations of this typc have been called
q~rasi-linear flow equati0ns.l "' The real iniportance lie5
If i t is assumed that viscosity and gaz law deviation in the extreme utility of this form of the equation. As
factors change slowly with pressure change. the pressure will be shown, the form of the equation suggehts .t
differential of 1 l n b ~ . ( p ) z ( p )becomes
] negligible. On thc powerful engineering approach to the How o f real gases.
other hand. the assumption that pressurc gradienls arc To solve Eq. 18. i t is necessar-y ro convert the ~ t s ~ l a l
small will permit onlission of lernls of order l o p ' ) ' . I n
initial and boundary conditions into lernis of the neb
either event, Eq. 12, can he si~nplificdto: pseudo-pressure rtl(p). lniportant considerations are a\
follows.
The gas mass flux is:
Eq. 13 is similar in form to the ditfusivity equation.
However, the diffusivity is a function of pressure. even
for a perfect gas. In this form, the close analogy with
liquid flow found by Jenkins and Aronofsky"'. " is empha- In terms of ~ t t ( p )the
, mass flux is:
sized. However, the a.rsurnptiot~ that pre.rsure gr.adients
nre srnall everywhere in the flow s?stern cannot be justified
in many iinporrutlr cases. Thc assumption of small pres-
sure gradients is iniplicit in ~ 1 1 1of the pressure build-up The usual boundary conditions are either specification
and drawdown methods currently in use which are based of pressure or the gas flux across bounding surfaces. When
upon ideal gas flow solutions or liquid flow analogies. We pressure is fixed, m ( p ) can be determined from Eq. 14.
return, then, to the rigorous Eq. 7. If flux is specified, the boundary conditions can be de-
Eq. 7 can be transformed to a form similiar to that of termined from Eq. 20. If the outer boundary is imper-
Eq. 13 ~ v i t k o u t assumir~g srnall pressrire gradiet~rs, by meable, then:
making a scale change in pressure. Define a new pseudo-
pressure nl(p) as follows:
)'(L--
l - , T,,)
-. fcp,, . . . . . . . . . , (25)
The integral can be evaluated generally froni rcduced
,". 1
properties correlations.
Con~pressibilitiesof natural gases have been correlated by
E:V.4I.UATION OF REAL
Trube'" as reduced con~pressibilities, the product of com- (;AS PSEUDO-PRESSURE
pressibility and pseudo-critical pressure. That is:
T o establish the relationship between p,,,and rn(p), the
c,,, = c.,(p). P,,.= f(p,,, T , , ) . . . . . (26) integral must be evaluated numerically for various iso-
therms. The lower limit of the integration (p,,),,,can be
Substitution of Eqs. 23 to 25 in Eq. 14 yields: set arbitrarily. A value of 0.20 was chosen. Selected
isotherms from pseudo-reduced temperatures of 1.05 to
3.0 were used.
Fig. I presents the argument of the integral in Eq. 27
vs pseudo-reduced pressure for various pseudo-reduced
temperatures. T h e dashed line represents the ideal gas case
with both viscosity ratio and gas law deviation factor
equal t o unity. The magnitude of gross variations of gas
properties with pressure and temperature is apparent.
Fig. 2 presents m(p) integrals as functions of pseudo-
reduced pressures and temperatures. The integrals were
evaluated by means of the Trapezoidal rule using an
1BM 709 digital computer. Values of the integrals are also
PSEUDO- REDUCED TEMP.
presented in Table 1. Interpolation between the curves or
between the values presented in the table can be per-
formed easily.
Use of Fig. 2 or Table 1 is limited to gases containing
small amounts of contaminants for which changes in
viscosity and gas law deviation factor can be handled by
02 .1 1 .S .6 .? .8 .9 1.0 2 1 4 5 6 7 8 9 1 0 15 20 25 appropriate changes in the pseudo-critical properties, as
PSEUDO -REDUCED PRESSURE ppr
suggested by Carr el al."' However, useful charts can be
prepared for gases containing large amounts of con-
taminants if complete properties are known. See Robinson
TABLE
P , ~ ( P )
~(PPCI~TP~
.
-
7 prr dpvr
rpr L ( P P ~ zI [PP'I
0.2
Pseudo-
Reduced -- Valuer of Integral For Pseudo-Reduced Temperature Tpr -
of -
Pressure
I .05
-- 1.I 5
- 1 .30
- --1.50 1.75
- -2.00
- -
2.50 3 .OO
-
40
-
$30
P
n. COWDENSATE
s
4
20
m,(b) FOR EILERTS uLO
-Ep
3
10
SOLUTIONS
MAT. GAS: O = 0 . 0 5 ....... 0
= 0.10..........
COWD.GAS: 0- 0.05........0
I IDEAL GAS: ............ --- 1
M A Y . 1966
(Fig. 5). Also shown on J'ig. 5 are [he Aronot'sky-Jenkin\" Senkins finding l'or ideal g a . Suhstit~~tion 01 long-titnc
ideal gas flow resulats. I t i \ cle;~rthat hoth the ide;il ant1 v;~luc.\ fur p,,(r,,) in F4. 3 3 ;dsn l e a d to thia conclu-
real gas cases lead to dimensionless pre\\ure drops which sion. Thus, Eq. 32 hecomes similar in rornm to thc
are lower than the liquid case---and which are flow-rate liquid case pseudo-steady-state equation at times long
the case Q -
dependent. Ano'ther important diH'erencc is illustrated b!.
0.05. The ideal ga3 line terminates at the
point where the well pressure i h zcro. The real gas
enough that thc outer boundary effect is controlling. The
fact that r . eventually becomes constant at 0.472 r , does
not mean the phvsical drainage radius stabilizes ~ ~ O L I I
solutions tern~in'ateat ;I wc'll pressure L>I' 10 pcr cenl half-way out in the reservoir. The entire reservoir volunic
of the initial pressure. Although not shown on Fig. 5 . is being drained. as can I)e seen hy inspection of any o i
the production times for the real gas cases to reach a the Eilerts er (11."' pro'duction figures.
limiting produc'tion pressure are a'bout two and a half times The Eilerts rt (11. solutions have provided an excellent
that required for the ideal gas flow cases. Clearly, pro- set of information to test the linearizaltion of the real
duction forecasts based on the ideal g:19 s~olutions will gas How solutions fo'r production. Eilerts cf a / . specified
be far too conservative. that the effective permeability was a function of pressure
Another important observation can be made from Fig. (assuming pressure drop would result in condensation and
5 by clomparing the real gas solutions for n'atural gas reduction of effective pcrmeabili,ty). Efiective pelrmeability
and condensate for a flow rate Q of 0.05. Although the can thus be taken within the m(p) integral. Correlations
rwtural gas line is close to the liquid case. the condensate in Figs. 4 through 6 do include a pressure-dependent
line is far- below the liquid case line. The terminal pro- permeability. Thus, if an approximation of the efTect of
ducing pressure is reached earlier for the condensate line pressure drop upon liquid condensation and reduction in
than for the natural gas line. This indicates the impmtance permeabi1i.t~ near the wellbore can he made. the per-
of gas property variations upon the results. That is, no formance can be estimated from:
single set of m,(r,,) correlations could be expected to
apply to all real gases at long production times. It is
also cle'ar from Fig. 5 that the real gas results tend to
approach the liquid case results as flow rate decreases. where
and at small production times.
Aronofsky and Jenkins introduced the concept of a
transient drainage radius r,,. This term should not be
confused with the dimensionless radial coordinate r,,.
From the Aronofsky-Jenkins definition of the transient
drainage radius. we write for real gas flow: and k is a known function of pressure.
The usefulness of considering k a function of pressure
to handle condensalte flow might be open to question.
Nevertheless, it is clearly indicated that variation of k as
a function of pressure can be included in the real gas
pseudo-pressure.
The Eilerts et 01."' results can also be correlated as
Correlation of the Eilerts e t al."' data presented pre-
transient drainage radii vs dimensionless time. The results
viously involves calculation of m(p) and determination of
are presented in Fig. 6, and agree with the Aronofsky-
relationships between the Eilelrts et ul. nomenclature and
Jenkins results and the liquid flow results almost exactly.
that used in this paper. (Necessary relationships are in
Actu~ally,the correlation of the real gas flow solutions i n
the Appendix).
terms of the transient drainage radius (Fig. 6) is a much
better correlation than the correlation in termj of ~n,,(t,,) Eilerts et al."' also determined performance with a
(Figs. 4 and 5). The drainage radius correlation is ex- steady-date, non-Darcy flow region near the producing
cellent for all values of production time. Thus, Eq. 32. well. As a result, a steady-state skin effect can also be
provides the mast useful engineering approach to the introduced to yield the fdlowing approximation for the
transient flow of real gases. As recommended by Jenkins radial flow of real gases during production:
and Aronofsky for ideal gas flow, the transient drainage
radius for real gas flow can be found from:"
8 -
*' 6 -
z
-1 LIQUID C A I E . REF. 15 -
N A T U R I L GAS, R E F 20
JO1TRX41. O F P E T R O L E U M T E C H N O L O G Y
where s is the skin effect and D is the non-Darcy flow properties whether physical properties are mon.otonic func-
coefficient. tions or not. The injection problen~has been the subject
of much investigati0.n in the fields of heat transmission
(:ON,qTANT R4'rl;: INJECTION and ground-water movement (Friedmann,"' Storm" and
All of the preceding discussion of r a l and ideal gas Polubarinova-Kochin'a"). As has been sho'wn by these
transienlt flow deals wit.h productton only. Injection results. authors, it cannot always be assumed that evaluation at
as was clearly shown by Aronofsky and Jenkins'" for an average pro,perty will yield good results. Sometimes
radial ideal ga,s flow. cann~otbe linearized in as simple a the answer will vary from one extreme to the other.
fashion. A r ~ ~ n o f s kand
y Jenkins correlated injection well
pressures for radial flow of an ideal gas as functions of SUPERPOSI'TION O F
1,INF:ARIZEI) SOLUTIONS
a dimensionless time based on gas compressibility eval-
uated at the initial formation pressure before injection. Superposition is rigorously correct only for linear
T h e dimen,sionless pressure rise at a given dimension- partial differential equations. Nevertheless. the extremely
less time was generally greater than that for a liquid close check between the linearized real gas solutions cor-
case, and increased with injection rate. Aronofsky and related o n the basis of the m,(?,,), as given by Eq. 3 1,
Jenkins showed that injection case results were very and a t,, given by Eq. 30, and the liquid flow solutions
close t o the liquid case for low injectio'n rztes. Although of van Everdingen and Hurst. indiclates the possibility
injection is of practical importance in itself, the major that superposition might be quite good for matching an
utility of injection case correlations is in applicatiosn of increasing rate production schedule. An increasing rate
the principle of superposition to generate variable rate
production cases, including the important pressure build-up
case.
Superposition, as it has been applied in gas well testing,
requires that dimensionless times for bath injection and
production be based on the same gas physical property
evaluation. Although superposition w u l d be based on
different dimensionless times for injection and production,
the added complexity of such a scheme does not appear
justified. Thus, an obvious question is: will injection
solutions comelate closer to the liquid case if dirnen-
sionless times are based on physical properties evaluated
at a pressure above the initial, low formation pressure?
W e rule out the scheme of using a point evaluation
at the existing injection pressure because this would yield
a result not usable for forecasting. That is, it would be
necessary to know the injwtion pressure-time history
before it could be calculated. An obvious possibility is
to evaluate physical properties at the final, elevated in-
jection pressure, or in the case of s u p e r p i t i o n applied
to reservoir production or build-up, at the initial forma-
tion pressure before produotion was started. This idea is
fundamenmlly the basis for all gas well pressure build-up
applications currently in use.
I n brief, correlations for injection based on a n elevated
pressure are n o better cor worse) than thvrse based o n
1
physical properties evaluated at the initial, low formation
presqure. This is true for both the ideal and real gas
flow cases. Fig. 7 presents the dimensionless real gas
potential rise f o r the Eilerts et al.?' injection case (their
NO WELLBORE STORAGE /
Fig. 8) correlated vs dimensionless times based o n both
the initial, low formation pressure and the final injection
pressure. The dashed line prese.nh the liquid flow solu-
tion. Two fiacts are apparent: the slopes of the correlations
are simil'ar, and correlations based o n final injection pres-
sure are no worse than those based o n initiall, low
formation prossure. From the Jenkins-Aronofsky studies
of ideal gas flow, we oan also conclude that the difference SOLUTION q , MCEID I
between the injeotion case correlations and the liquid
case become smaller as injection rate decreases; in any
case, the differences aren't large.
p,(tJ
Fig. 7 can lead to ano'ther idea. Correlatioa based on
, REF. 15
-1
a dimensionless time evaluated with physical properties
about half-way between the extremes might be quite
good. This idea follows immediately from the theorern
of the mean. That is, if flowing fluid physical properties
vary momtonically with potential, the proper result is
limited by those evaluated at the extreme values of
physical properties. Friedmann" proved that results must
lie between those evaluated at the extremes of physical
M A Y , 1966
s c h e d ~ ~ lwould
e require superposition of positive incre- posilion was 20 psi out of a drawduwn 01 2.150 psi---
mental rates. Howevel.. the real g:~s How solutions do ;I difference of 0.9 per cent. T h e 50-day production
depend slightly upon prtductinn rutc. T h w . the o'nly period was long enough that initial rate changes were
way that the ktpplication 01' the principle 01 supe~positivn influenced hy the outer boundary. Thus, we conclude
!as an acceptable approximation) to real gas flow can be that superposition can he used to reproduce variable-
established is by comparison wi'th finite-difference solu- rate drawdown data witb acceptable accuracy.
tions of variable-rate, re,al gas Row problenls. The previous remarks concerning supe'rpositian of in-
Such a comparison can be made for an increasing cremental rate increases are. of course. directly applicable
pro'duction rate schedule from data for real gas flow to pressure build-up testing. Although insufficient com-
published by Carter.':' Carter studied the eRect of well- parisons between finite-d'ifference build-up so'lutions and
bore stsorage o n gas production. For his solutions, it was superposition solutions [for the real gas flow case have been
assumed that the surface flow rate was held constant. made to completely explore this pro.blem, i,t does appear
but 0.02965 Mcf was withdrawn f r o ' n ~the wellbore per that build-up theory can be used with good accuracy.
psi pressure drop in the wellbore. This resulted in the An interesting test of pressure build-up can be m'ade
sand face flow rate increasing as a functi.on o'f time by co'mparison of Dykstra's" s o l ~ ~ t i o nwith
s superposi-
toward the constant surface flow rate. This case is almost tion solutions. Because Dykstra's cases involved a vari-
exacltly analogous to the wellbore unloading case pre- able-rate production period, permeability was low and
sented by van Everdingen and H u r s t ' y n their Eq. pressure gradients high, it is believed that a faifly
VIII-11. The wellbore storage constant ? for Carter's extreme test results. Fig. I 0 presents the build-up fol-
solutions can be determined from Eq. 6 presented by lowing the drawdown of Fig. 9. As can be seen, the
Ramey."' The value of ? for Carter's solutions does vary -
RATE
slightly with pressu're, but a value of 300 is quiste good. 7000,
Fig. 8 presents a comparison beltween the tn,,(t,,) obtained
from Carter's solutio~ns, both with and witho~lt wellbore
storage, and the van Everdingen-Hurst p,,(/,,) solutions
for the liquid flow case. As can be seen, the comparison
with constanst rate liquid flow w i ~ h o u tstorage is excellent.
This was previously shown for the Eilerts el nl.'" solutions.
Of more interest, the comparison beltween the liquid
flow case witlr wellbore storage and Carter's two solu-
tions with wellbore storage are also excellent. This
establishes that superpositio'n of the linearized real gas
flow solutions for an increasing flow rate should be a
very good approximation-at leasst before outer boundary
effects are controlling.
Although superposition in an increasing production
rate sche,dule appears quite good, ilt is not appare'nt that
a decrensitlg rate schedule is susceptib'le to superposition.
LINES FOR DYKSTRA" SOLUTION FOR
This resul~ts because the dimensionless real gas injection
K=0.25MD,rc/T,, = 1000, h = 179 FT.
pressure increases d o not oorrelate with the liquid case as
COMPUTED BY SUPERPOSITION OF
well as d o productio'n data. Even for transient injection REAL GAS SOLUTIONS
of an ideal gas,"' the resulting dimensionless plressure WITH m,ctd= p j t d
rise appears to depend upon injection rate, but does
approach the liquid case soluti'on as injection rate de- I 10 50
creases. T h e fact t h a injecti'on results do approach the PRODUCING TIME, b I Y S
liquid case as injection rate approaches zero suggests
that supe'rposition of small positive incremental rates FIG.~-COMPAR~SO?~
OF FINITEDIFFERENCE
A N D SUPLKPOSITION
Fr.owl~cPRESSURES
FOR A DECREASING
PRODUCTION RATII.
(injections) would be feasible. Again, the possibility can
only be checked by c'omparison with finite-diflerence
solutions.
F'ortunately, both Carter" and Dykstra" have pre-
sented finite-difference solutions for decreasing flow-rate
production. Dykdr'a's data provide an excel1en.t set for
comparison of finite-difference solutions with superposi-
tion of the linearized solutions. Fig. 9 presents a com-
parison of Dykstra's compufted flowing pressurm with
those obtained by superposition of linearized real gas
*LOPE - SUPERPOSITION OF REAL
GAS FLOW LINEARIZATION
flow solutions. The line is Dykswa's result, while poinlts
represent results of superposition using only four or five
incremental rate changes to represent a rapidly changing
flow rate. The fl'ow rate is shown by the davhed line. K m0 2 5 M D
For the example shown, the prmeability was 0.25 md, h = 179 FT
24
thickness was 179 ft, initial pressure 6,150 psia and r, = I000
rw
flow rate declined quadratically as a function of time
from 6,556 to 2,500 M c f / D by 50 days' producing time.
Superpitioln was accomplished using dimension1e:ss times I 10 100 1000
SHUT-IN TIME, HRS
based o n the initial pressure and the m,(t,) taken equal
to the liquid case pD(t,) values. The maximum difference
between Dykst.ra?s resu1,t and those computed by super-
J O U R N A L O F PETROLEUM TECHNOLOGY
superposition result yields a similar build-up curve of static pressure following a complete pressure build-up.
identical slope, but about 60 psi below Dykstra's It is not a volumetric average pressure. Eq. 33 coupled
finite-difference solutions. Again, the percentage difference with the normal material balance for a bounded drainage
is small; the final static pressure is about 1.1 per cent volume provides a useful means to couple production rate
too low. It appears that s u p e r p i t i o n of the real gas and gas recovery.
fl'ow linearization will always yield a pressure build-up In the case of liquid flow, an equation similar to Eq.
static pressure that is too low, but as good or better 43 can be derived using the concept of pseudo-steady-state
than results of curren't methods. Furthermore, field appli- flaw. That is, a condition is evenltually reached for con-
cation would be t o the field measured data-the real stant rate liquid production when the rate of pressure
solution-which would tend t o co'rrect fo'r this error. decline becomes constant everywhere in the reservoir.
We conclude that pressure-build-up analysis b'ased on This conditioln is expressed mathematically by setting
superpositimon can be do'ne for real gas flow with ac- the Laplacian of the pressure equal to a constant (&her
ceptable accuracy, but that further s~tudy of pressure than zero). Al.th'ough it can be shown that the Laplacian
build-up for re'al gas flow is desirable. of pressure-squared for an ideal gas, or the Laplacian
of the real gas pseudo-pressure cannot be equal t o a
STEADY-STATE A N D PSEUDO-STEADY- constant rigorously, a flow condition similar to pseudo-
STATE FLOW steady-state does appear to exist for both ideal and real
gas flow, for all practical purposes. The existence of
tc.adial gas flow at constant production rate will be such a ~ o n d i t i o nis suggested by Eq. 43. Fig. 1 1 presents
adi side red. A horizontal homogeneous p u s medium an interesting inspection of the pressure behavior during
or constant thickness h with impermeable upper and the period that Eq. 43 'applies f w o n e of the Eilerts et al."'
lower boundary, and a well of radius r , located in the cases. Also shown is the p,(t,) for comparison with the
center of a radial reservoir, constituta the flow system. liquid case. As was seen previously in Fig. 5, the m,(t,)
The outer radius r , represents either the real boundary does no~t change at a constant rate during this period.
or the radius of drainage. T w o cases will be considered: Although it matches the liquid case solutio~n at early
( I ) constant pre5sure at r , , and (2) no flow across r,. times, eventually the m,(t,) drops below the liquid case
solution. The most interwting feature of Fig. I I, how-
(:ONS'TA NT I'RI.:SSURli A?' ever, is that the m,(r,, t,,) for all radial locations are
01I'I'RK BOUNDARY
essentially parallel. Thus, the m b ) profile is essentially
T h e steady-state equation for a real gas in axisymmetrical
independent of time. This condition can be described
coordinates can be written from Eq. 22 as:
approximately by' setting the Laplacian of m(p) equal to
a colnstant. As shmown in Refs. 39 and 42, this leads to
an equation similar to Eq. 32, but in terms of a n
average m ( p ) rather than m ( 2 . Altl-,ough i,t can be shown
The boundary conditions for two concentric cylinders of that these two averages tend to be equivalent for practical
radii r, and r,, are: ranges of conditions, it does not appear worthwhile to
show the de~elopmen~t here. In any event, Eq. 3 2 de-
scribes the long-time flow behavior of closed radial system5
with remarkable accuracy.
Integrating Eq. 38 and using the boundary conditions, the Another consequence of inspecti'on of Fig. I I is that
steady-state pressure distribution in the system is: the m(p) distribution can be obtained readily. For ex-
I I I I 1 / I
Eq. 41 can be evaluated for p = p. at r = r . and re-
arranged t o provide an equation analogous to the normal
radial flow equation:
MAY, 1966
ample, the following equation also describes flow reason- I'usivity of real gas with pressure was similar to that
ably well: OF an ideal gas, it was possible to correlate finite difference
solutions for the ideal radial production of real gab
from a bounded system with the liquid flow solutio'ns
of van Everdingen and JIurst, and the ideal pas solu-
tions of Aronofsky and Jenkins. This correlation avoids
! h e assumption of small pressure gradients in the reser-
voir and offers generally ~lseful solutions fo'r the radial
The purpose of thc pr.eccding was to rlescribe runda- flow of real gas.
mental ~on~siderations which can bc used .successfully to An investigation of the injection of rcal gas into a
analyze the flow of real gases. The concept of the real bounded radial system also gave a reasonable correlation
gas pseudo-pressure promises a consider:~ble sinipfificati~on -but not as good a correlation as production data. The
and improvement in all phases of gas well testing analysis correl,ation was as good as, or better than. the correlation
and gas reservoir calculations. Such applications will be of ideal gas flow results made by Aronofsky and Jenkins.
described in useful engineering form in a companion
paper. An investigation of the possibility of superposition of
the linearized results indicated that superposition can
Several remarks concerning the real gas pscudo-pressure be used as an acceptible engineering approximation to
are in order. N o claim of originality can be made for generate variable rate flow of real gases in a radial
the substitution we have called the real gas pseudo- system. Pressure build-up for real gas flow was thus
pressure. Carslaw and Jaeger3' reviewed application of a jusrtified for the first time. (No justification for pressure
similar 'transformation which was used in solution of build-up for the non-linear problem of ideal gas flow
heat conduction problems as early as 1894 and the has yet been presented.)
early 1930's. Recen'tly, McMordie'" pointed out the u'tility
of this sort of transformation in heat conduction problems. Accurate and simple equati'ons can be written to
There have even been numerous mentions of the use describe unsteady flow of real gases which properly con-
of a transformation similar to the tn(p) function in cnn- sider variation of gas physical properties.
nection with How through porous media. In 194'9,
Muskat" used the same transfo'rmation in a discussion of NOMENCLATURE
the theory of potentiometric models. In 1953, Leibenzon"
used the transformation, and Russian aultho~s refer to o= grad
it as the Leibenzon transformation. In 1951, Fay and V. = divergence
Prats"discussed use of a similar transformation in con- V' = Laplacian operator
nection with transient liquid flew. In 1955, Atkinson A = area, sq cm
and Crawford" evaluated numerically a similar function
h = slope of a straight line in a plot of k(p) vs I / p
but with con,stant viscosity. I n 1962! Carter" used a
gas mobility term M(p), which was defined as: c,(p) = real gas compressibility defined by Eq. 10
h = thickness, c m
k ( p ) = effective permeability, darcies
M = molecular weight
Clearly, the m ( p ) function is proportional to the pres-
m ( p ) = real gas pseudo-pressure defined by Eq. 14
sure integral of Carter's M(p). In 1963, Hurst et. al.'"
used a similar integral, but with constant viscosity. p = pressure, atm
T o our knowledge, however, this paper represents the q = flow rate, cm7/sec
first application of the real gas psttldo-pressure to lineari- r = radius, cm
zation of transient real gas flow. Perhaps the most R = gas constant
surprising fact is that the realization of the utility of t = time, sec
this concept has been so long in coming.
T = temperature, "K
I n the original draft of this paper and the companion
paper,"'," we callcd the m ( p ) function the real gas v = velocity, cm/sec
potential. It was stated in those papers that the m(p) V = pore volume, cm3
transformation was not a true potential. Carslaw and x , y, z = direction notation
JaegerZ8termed a similar substitution in heat conduction z ( p ) = gas deviation factor, a function of pressure
a n effective potential, while Muskat'halled the trans- at constant temperature
formation a potential as a matter of convenience. We p = density, gm/cm3
feel that the m(p) transformation will be a n important
p(p) = real gas visoosity, a function of pressure at
function in gas reservoir engineering, and it is important
constant temperature, c p
that the function be given a suitable name. If we were
to name the transformation as Rus'sian authsors have, we p, = viscosilty ad atmospheric pressure, cp
would call it the Muskat ,tra,nsformation. I n the belief n = normal distance scale
that the name should be reasonably descriptive and 4 = hydrocarbon porosity, fraction
brief, the term real gas pseudo-pressure was finally selected.
This name was originally suggested to us by M. Prats, SUBSCRIPTS
with Shell Development Co. e = external boundary
It appears that the following conclusions are justified. 1 = liquid
The transformation called the real gas pseudo-pressure in pc = pseudo-critical
this paper reduces a rigorous partial differential equation r = radius
for the flow of real gas in an ideal system to a form
similar to the diffusivity equation, but with poltential- sc = standard condit'Ions
t
dependent diffusivity. Because the variation of the dif- w = internal boundary, the well
( A -I )
where
-
P - p z P .K(p)
;
I ~ P p(P),
k(p,)K(P), z(p) 7(/1,)z(p), p(p)
, ) M~(P)
-
= X(P)/ I-C(P)Z(P)
and
M ~ P=) w(p,)W(P) . . . . . . . (A-2)
Dimensi~nlesstime is defined as:
where (A- 1 0)
Let the coefficient on the right side of Eq. A-5 be Thus, in terms of the dimensionless real gas pseudo-
evaluated at the initial conditions, and define: pressure drop:
2
( r )=A m ) . . . . . . . . . (A-I 2)
Q
T h e mfP) for the Eilerts et al. natural gas and con-
densate fluid are shown in Fig. 12. The large difference
Notice that ,p(P) and K(P) are equal to one a t the initial between physical properties of the two fluids is apparent.
P. Hence, Eq. A-5 takes the form: *+