You are on page 1of 11

---

+Q$=rad

Methods for Calculating Well Use of Two-Dimensional Coning Behavior


HENRY J, WELGE A, G, WE.SER MEMBERSAIME

JERSEY PRORUCTION TULSA, OKLA.

RESEARCH

CO,

ABSTRACT A published calculation method for predicting incompressible, mu ltidimension a! /[uid displacenren t has been adapted to the problems of water and gas coning in ail wehs. Since depth and radial distance from the wellbore are the two key dimensions affecting the shape of a gas or water cone, coning calculations are well suited .to the use o~ twodimensionul methods. The a It ernat ing direction imp licit procedure (AD I P ) was used for relaxation ctrlcu lations of two-phase potentials in a twodimensional grid, From the, potentials and the saturation and cap illary pressure re iationsbip, pressure distributions were calculated which trace cone gmwtb with time, Predictions have been made /or we[! pr~ducing histories both before and after core breaktbmugb. To check validity of the method, two-dimensional calculations have matched the con ing behavior and produced water-oil ratio history of a laboratory sand-packed model. They have also matched the coning behavior o! several producing wells, /or which the calm la tio~ were compared w it b produced water or gas cuts and logs showing water or gas cone ,movernents. INTRODUCTION The application of two-phase, two-dimensional calculations using ADIP to various reservoir flow problems has been described in the literature. l-~ When adapted for computer solution, this method has proved o be a powerful tool for simulating well and reservoir behavior. This pgper discusses the method as applied to well coning calculations. Single-well and coning calculations comprise an especially difficult class of two-dimensional problemswhich require special techniques for computer calculation and determination of reservoir characteristics. Refs. 4 through 8 describe, previous approaches to the coning probkrn. Severs 1 examples of water and gas coning .
OrlglnaI manuscript received In Society of Petroleum Engineers office July 27, 1964. Revised manuscript received Oct. 27. Paper presented at SPE Aonual Fall Meeting, held ip Houston, Oct. 1I- 14, 1964. lkeferences given at end of PaP@r. DkCt%MBER, 196,1

ca Iculations, including studies on both laboratory models and producing wells, are presented here. The two-dimensional method accounts realistically for the most critical parameters affecting coning including production behavior, rate, forma tion stratification, horizontal and vertical permeabilities, depth of well penetration, gravity and capillary forces. The method considers the different densities and viscosities of the two phases and the relative permeability and capillary pressure characteristics of the rock and fluids. In adriitiori-to tracing cone growth in the vicinity of the wellbore, the method calculates the overall movement of the fluid interface throughout the wells drainage volume. Incompressible fluid flow is assumed to occur between the producing interval and the wells limit of drainage. Calculations can be made for the producing history both before and after cone breakthrough. A typical two-dimensional grid or array of. blocks used to solve a coning problem contains about 400 blocks. The wells cylindrical drainage volume can be represented by about 20 radial subdivisions suband the formation thickness by 20 vertical divisions, The grid spacing is normally smaller near rhe withdrawal interval to define the cone shape accurately, For this work we used an IBM 7074 digital computer having a core memory of 10,000 tendigit words. A typical study, covering 5 to 10 years of well producing history, required from three to six hours of computing time. MATHEMATICAL SIMULATION CONING BEHAVIOR ,.
BASIC METHOD

OF

In coning calculations, the reservoir volume drained by the producing well is represented by a two-dimensional system of blocks as ahown in Fig. 1 for water coning studies. The horizontal dimensions of the blocks increase with radial distance from the well axis in geometric progression, i.e., the blot k size is--small near the wellbore - - and Iarge near the wells drainage radius (re ). Vertically, the blocks are bounded by horizontal planes located at different depths through the.
54.5

Each block consists of an producing interval. annular ring which is assumed to have radially flow properties, The bounding planes symmetric rire made closer together near the bottom of the producing interval to adequately define the shape of the cone. The vertical grid spacing also depends on the number of permeability layers to be represented, Influx of the displacing fluid is simulated aIong the boundaries of the system at a total rate equal to the fluid withdrawal rate. For a well producing drive, influx of water along the by bottom-water lower boundary is represented. Conversely, if a influx at the flank-water drive is occurring, drainage radius is simulated, Although Fig. 1 the case of water coning, analogous represents concepts apply to gas coniirg calculations. To cnlculare the development and growth of a gas or water cone, wc divide the wells producing histrrry find forecast into time periocls. Normrrlly, the lengrh of a time step varies from 5 to 30 days, although mrrch shorter time step.s must usually be taken shortly before and just after breakthrough of the displacing phast. Changes in pressure and saturation which take plrrce during a time step are computed for each block in the system,
PREDICTION OF PROtXJCEI_l GAS-OIL AND WAIXR-OIL RATIOS AFTER BREAKTHROUGH

For the time periods before cone breakthrough, calculations arc made using AI)IP rLs described in the Appcmdix. For the time periods after breaku+en the displacing phase is being through, must be used to techniques produced, -special obtain rellable results and to avoid excessive computer costs. Difficulties arise because of the small size of

Roclm
*CL.

ii

i
I

i
]~

,.-. . -J !i&!2L
-4 . .

-. 1.

51

1
. .

199

the calculation blocks neat the wellbore. For a normal time step, the fluid throughput in one of these small blocks may be many times the pore volume of the block, The proportions of the two fluids flowing across each fzce of the block are governed by saturations andreiative permcabilitics at the start of the time step. There is no easy way to correct the fluid proportions for saturation changes occttrring during the time step. Thus, the proportions in m,hich the fluids are flowing will always be somewhat out of date, and might cause abnornmlly large saturation changes in the small blocks. In a later time period,; abnormal saturation changes in the opposite direction may be cnmputed. If allowed to continue unchecked, the computed saturations for small blocks near tl~.e wellbore \vi Il fluctuate drastically and yjeld merrningless Produced gas-oil and water-oil ratios results. derived from the calculated saturations also oscillate in an unrehlistic manner. One possible way to overcome this difficulty is to use very smalltime steps. Experience has shown, computing costs wirh ultra-small rime however, that steps (perhaps as small as a few minutes of field time) a~e prohibitive, nnli solution of most coning problems with this approach would bc impractical. -, From the present work, two techniques have been developed which greatly reduce the cost of coning calculations after breakthrough. The first calculati~?nal procedure restricts. how much the saturation in any block can chan:;e during the time step, The maximum permissible change is set at about 0.01 to O, 1, depending on the particular problem, For instance, if the maximum change is set at 0,05, a saturation could change from O.tlX to 0.83 (or from 0.88 to 0.93) in any one time step. The limit in saturation change is rnncle Tligh enough so that most of the blocks in the system will not be affected. On the other hand, the limit is msdc low enough to dampen the saturation oscillations in the small wellbore blocks. In all cases, the saturation of a block is first computed by the iteration procedure described in the Appendix, and then this resulting saturation is changed where necessary to satisfy the above limitations. Calculations on test problems using this device w!ith normal-size time steps give results ful Iy as accurate as more expensive calculations using ultra-small rime steps. The technique has also been employed on several snrdies of well coning behavior. A second procedure for smoothing saturation oscillations consists of combining blocks near the weHbore whenever they appear to have nearly the same amay of saturations. At the start of a coning problem, the displacing phase tends to peak toward the nearest withdrawal perforations. After production of the displacing phase begins, however, this peak. becomes more like a plateau} and the displacing -. phase attains abotm the same saturationr , -. in the three to six blocks nearest the well in a given row. At this time, combining an equal number of blocks in each horizontal row may be
SO CISIY OF PSTROLEIIM F.NGINEElts JO? JRXA1.

very helpful, because the pore volume of the blocks near the wellbore will thereby be increased manyfold. At the same time, less computing is required of the smaller total for each time step, because number of blocks. COMPARISON OF WATER CONING PREDICTIONS WITH A MODEL EXPERIMENT To verify the reliability of coning studies, prediction-type calculations were made to permit comparison with a laboratory model experiment . performed by Soboc~nski and Cornelius, g A pieshaped, Plexiglas model, packed with graded, unconsolidated Ottawa sand, represented a radial Aegment of the drainage area surrounding a producing well. The model was operated to simulate the behavior of a well subject to water coning. Initially, the oil-water interface in the model was horizontal; after the start of production, the movement of the interface was watched through various stages of water cone development. Water and oil were injected across the outer boundary of the model in such a proportion as to keep the oilwater contact ac a nertrly constant level there, Tnc oil-warer transition zone was less than abour I in. thick in this mntfel. Calculations of model behavio~ were made for ac,tual experimental conditions of the run, using known properties of the model and fluids as listed in Table I. ,No attempt was made to scale the mode] data or the com~utations to represent any actual producing well. \!ater-oil interface positions at three different times t-ire shown in Fig, 2. The calculated shapes of the water cone agree closely with observed configurations at corresponding times. Cone occurred after about 7.25 minutes, breakthrough and the production of oil and water from the ntodel was cnntinued for about 80 minutes after the start of the displacement.

TABLE
Radius Depth Radius Depth Porosity kl/k

1 _

PROPERTIES

OF WATER

CONING

MODEL

of Model of Model
Original

48 in.
24 in, I Contact 13 in. 0.5 Interval 0t08 0.37 35 Darcys 1 1.25 Cp l#135cp 0.756 1.15 gin/cc gin/cc in. in.Woter.t3i

Thickness
of

of Wellbore of Producing of Sand of Sand Ratio

Permeability t&l$iEs_:;,ty Oil Density Water Density

.
Fig, 3 shows the computed ancf experimental results obtained for produced fluids, Good agree. ment is shown for both oil and water cumulative production. Since experimental water-oil ratios were obtained from slopes of the cumulative curves, the sma 1I deviations from cnmputcd behavior shnwn for cumulative fluid output becamc magn ified when water-oil ratios were plotted, ffowevcr, the computed water-oil ratio reached the stabilized value of 0.35 as found experimentally, The overall match between cnlcrrlarions and the experiment is considered accepmb[c, CON INK; 131;[[A VIOR OF 13 KO[)[JCINL;
WELL CIi ARAClX~ISTICS

WI;l. 1.S

Several studies of producing wells have been made utilizing the method described in this paper, We selected four of these studies as representative . of typical problems and will discuss them in rhe following sections, Ile main purpose of this \vork was to determine the effectiveness of the calcula- cion method in mrttching the known behavior of actu~l producing wells, The method can obviouslv be used to predict future well behavior, a]thuugh no examples of predictions n re shown in this paper. For later reference, Table 2 summarizes

WELL

COMPUTEO EXPERIMENTAL

____ ORIGINAL

WOC

I
2 I 1.OISTANCE FROM WELL -FEET

1-. .
34?

FIG. 2 CONE BEHAVIOR


DRCEMRER, 196!

IN. LAt30RATORY

MODEL,

L--

.-

TABLE Litholagy

2-

PRODUCING

WELL CHARACTERISTICS

J!!d!A
Clastic stone
Mlddla

~
I Ime. Vuggy and Intergranular I Ime stane Canado Water 120 0.50 70 11 21 1la 236 10.5Q 2,4 0.87 0.812 1,15

well c
Sandy I Imastone

Well D Sandstane

Locat]on Dlsplaclng Fluid Formation Thickness (ft)

East

North Africa Gas 196 15S.198

Southwest Texas Gas 43 33.41 25 23 37 90 90 1.30 0,338 0,0218 0,680 0.172

Woter 170 0-65 130 22 7 182 373 2.50 0,350 (CP) 0.381 O*66 l.li

Depth of Producing Intervals from Tap Of Formotbr(ft) Depth afOrlglnal Fluid Contact from Top of Frwmatlon (ft) Average Paraslty (%) Average Connate Water (%) Average. HarIzantal Permeability frsm Care Analysis (red) Average Horlzorrtal Permeability fram Flow Top?. (red) k~/kv Ratio frarn Caning Calculatlan Oil Vlscasity Dlsplaclng Oil OOrIsltY (cp) Phose Vlscaslty

30 25 11
300 620 2*5O 0.363 0,018 0.635 ~ 0.155

(urn/cc)

Displacing Phase Density (gin/cc)

rock and fluid charact&istics that to the coning calculations for these WATER CONING BEFORE CONE BREAKTHROUGH WELLA ,

are pertinent wells.

The two-dimensional method of prediction was applied to WelI Ar a Middle Eastern producer which had experienced slight water production. The stratification sequence indicated by core analysis in this well is shown in Fig. 4. Because core
analysis reservoir
7000v

data flow ., .

are

often

not

representative the followkrg

of

properties,

we. made

Sooo

COMPUTEO o S%% RIMEN7AL 01 L I d

5000

*000 # soao. r

2000

moo . WATER o >0 F a $ ; s ~ o.10 i & o


TIME-MINUTES

00 /

adjustments of the petmeabilities. While preserving the attatification sequence indicated by core analYsis, the horizontal permeabiliries were alI multiplied by a factor which would give the correct initial productivity index of85 B/D/psi as measured in the field. This Increased the average horizontal permeability from 182 to 373 md. Then, the verticaI perrneabilities were multiplied by a different factor chosen so the time of first water appearance in the well (425 days) was matched by the calculations. The k#ku ratio thus obtained was about 2.5. Once the permeabilities were properly adjusted to match the the first cone breakthrough, additional calculations were made without further adj ustmenr. After cementing off lower perforations ar about 700 days, the well was put back on production at a reduced tate of flow (see Fig. 5). Water production was again detected about 1,370 days after the start of production. The calculations accurately matched this second occurrence of water. The figure alao shows how two typical water saturations in the toner O.I and 0.35, advanced with time. The higher saturations tended to recede during periods of Iow withdrawal rates. In this field, wells are shut in or worked over at the. first to meet market appearance of produced water specific acioiis regarding salt and water content of produced crude. About 1,950 days after the start of production a second workover was performed on well A. ~ring 01 WV8L U mRmIa*lbm.
8801 W::;,:,:::*

0.3 0, WOR a so o

..

FIG. 3 PRODUCTION

BEHAVIOR MODEL.

OFLABORATORY

FIG. 4 PERMEABILITY

DIsTRIBUTION

FROM
JOURNAL

COF$E ANALY81S,WELL A.
sOCrETY OF PETROLEUM ENCINSERS

84S

this operation, the well was drilled out to a depth of 83 ft below the ptevious producing interval and an induction log waa run to determine water saturations. Fig. 6 shows the water saturation distribution obtained from the induction log, together with the computed saturations on the same date for the formation adjacent co the wellbore. The two curves of water saturation vs depth are in good agreement and both show a transition zone about 50 ft high with,in the water cone.
WATER CONING AFTER CONE BREAKTHROUGH , WELL B

Well B offered an opportunity dimensiona~ coning method, on produced a fairly large proportion

to test the twoa well that has of water through-

out most of its life. The producing history of this we]] ia shown in Fig. 7. The same adjustment procedure of horizontal and vertical penneabilities was followed as in the case of Well A above. Thus, the time of first water arrival was again fitted to the field breakthrough time by suitable adjustment of the vertical permeabilities. The agreement between field history and computations is satisfactory, except for the period from about 2,oOO to 2,6oo days after the starr of production when the computed water cut remained constant instead ,of reaching a crest. However, the computed behavior is more typical of other wells in the area than is the peaking of water cut observed in Well B. It is possible that a fissure or channel was opened near the wellbore, possibly by washing out mud cake, which would not be accounted for iri the calculations. The well was worked over after 2,700 days, at which time the lower 10 ft of the 50-ft producing interval was packed off. The water cut declined after the workover, aa shown in Fig. 7; the calculations showed this decline.
GAS CONING BEFORE CONE BREAKTHROUGH - WELL C .

o
/nEco MPLEIIOU

50
t

s..

0.[07

(
a. . 0.ss
I

100

I
I I I

Because gas coning was anticipated at Well C, neutron logs were run at four different dates after the start of production to determine the rate of movement of the gas-oil contact. Fig. 8 shows the measured depths of the gaa-oil contact near the wellbore and the production rate history of this well. The figure also shows the computed movement of the gas-oil contact. The k-hikv ratio was varied until a match was obtained with field history for a value of 2.5. The computed depths of the gas-oil contact are in good agreement with those measured. Computed saturations indicated that the gas cone in this well moved downward with very little transition zone, in contrast to the water coning behavior discussed previously.
GAS CONING AFTER CONE BREAKTHROUGH WELL D

DAYS

Well D produced which was overlain


BEHAVIOR,

from a thin (I6 ft) oil zone, by a Isrge gas cap and. under-

FIG. S cOMPUTED

WATER CONING WELL A.

~: 00 ~:
~ E w

r . COW.JwJ mm, _ *KU

=:

,10 -

1 if

ea

. T

-----

.. E

40

.$f; ~ -~ ;

MB

FIG. 6 WATER SATURATION

VS DEPTH AT 1,950

DAYS, WELL A.
DECEMBER, 1964 ,,

FIG.7 PRODUCING HEYTORY, WELL B.


S49 ., . . .. . . . . .. .

lain by a thin water zone. Because the calculation method handles orLly two phases and gas coning was a more critical problem in this well, the formation was assumed to terminate at the oilwater contact for calculation purposes. Gas cone breakthrough occurred a few days after the start of production and the produced gas-oil ratio increased thereafter aa shown in Fig. 9. In this case, the average pressure drop from the drainage radius to the weHbore was only about 10 psi. Since the average pressure level in the reservoir was about 3,200 psi, there were negligible changes in the density of the flowing gas phase and the incompressible analysis waa adgquate for gas coning calculations on this welI. While computed total withdrawal rates were volumetrically equal to actual well withdrawals, the proportions of gas and oil in the computed production differed slightly from well gas-oil ratios as shown in the figure. Both the compured m-td well gas-oil ratios are quite sensitive to changes in withdrawal rate as would be expecred for gas coning behavior. Computed trends of production rares, cumulative oi,l production and gas-oil ratios agree fairly closely with well history. CONCLUSIONS 1. The ADIP method for calculating twodimensional fluid displacement has been adapted COhandle problems of water and gaa coning. 0. ~ I

2. Special computational techniques must be used after cone breakthrough to achieve reliable results and keep calculation costs within reasonable limits. 3. Calculations have matched the producing histories of a laboratory sandpacked model and of several producing wells experiencing water or free gas production by coning. 4. Since it has matched the producing behavior of several well case ~isrories, the calculation method is considered reliable for predicting future coning behavior, provided that certain critical parameters can be established for wells to be studied. 5. Experience gained from well studies indicates that average horizontal and vertical permeability and the k#ku ratio are critical parameters and should be determined in most cases from production history, logs, or well test information.

?0 }0 10 10 10 0

20 k-

.
60 -- 30

.
COMPUTED
WELL HISTORY

20 I

q ~A-

loo

---+--T--

10

----L----L
OfiS -OIL COMPUTEO COIAT&CT FROM OAS-OIL NE UTROM LOO MOVEMENT CONTACT

so ,

_---

--.... ..-. .,
1

.
40

..
1

30

* .._.

.-.

..

1. ,.

._

-.

.. . . ..

20 .

2 ~~ -

---

10,

~.
0 00

1
lW TIMEOArS

I 200 I ,

1000 DAVS

Im

FIG. 8 COMPUTED
%50 .

AND .OBSERVED PRODUCING


WELL c. FIG. 9 PRODUCING
sOCIETY

HISTORY,

HISTORY, WELL D.
EN CiNttERS 30 EJKNAL

OF PETROLEUM

. .

ACKNOWLEDGMENTS The authors are indebted to Vincent V. Valleroy forhisadvice and assistance inperforming some of the studies described in this paper. We are also grzteful to the engineers from various affiliates of Standard Oil Co, (N. J.) who supplied field and well data used in the calculations. We wish to our appreciation to Jersey Production express to publish the results Research Co. for permission of this investigation. , x area + A length) for ~ = (Permeability flow inthe~or horizontal direction = (k%) (2~i+%AD) . . ~~+.1 - ?i = (Permeability x area + A length) for direction flow in the y or vertical
. (ky) ?7 (raf+~-r+-%)

k = number of iteration)

sweep

(two

sweeps

per

rr = number of time step


REFERENCES

.,

Douglas, H, H.,

J.,
Jr.:

Jr.,
~~A

Peaceman,
Method

D. W.
for

and Rachford, Calculating Multi.

NOMENCLA~RE

(Ax)i+%,

Dimenslcnral Immiscible Displacementr~, ira,ns,, AIME (1959) Vol. 216, 297, 2. Blair, P. M. and Peaceman, D. W.: An Experimental Verification of a Two-Dimensional Technique for Computing Performance of Gas-Drive Reservoirs) ~, Sot. Pet, Eng. ]orm (March, 1963) 19. 3, Sheffield, IM.and Brinkman, l?, H,, Jr,: CcImprovement
of Reservoir Studies Through Use of Two-Dimens ional Mathematical Analysis, SPE 416, presented at

(A ~)i, ~-%

SPE Annual Fall Meeting, Los Angeles (Oct. 7-10, ,, 1962). 4. Muskat, M. and Wyckoff, R. D.: 8An Approximate Theory of Water Coning in Oil Productiont 1, Tram-,, AIME (1935) Vol. 114, 144. of 5, Meyer, H. L and Garder, A. O,: %lechanica

Di

- Dj-1

Two Immiscible Fluids in Porous APf/. PIJys. (1954) Vol. 25, 1400, 6. Piraon,
in z-direction

Media),

jour.

E = error
/=componerrt g = component of error

of error in~direction of error number with respect to

h = componeric iteration

S. J.: Oil Resewoir llngirreerirrg, Third Edition, McGraw-HHl Book Co., Inc., New York (195S) 432. Fingering and Coning of water and 7. Arthur, M. G.: #~ AIME Gas in Homogeneous Oil Sand)$, Trrws,, (1944) VOL 155, 184,
s. Smith,

[-/~.= iteration parameter number) k M .=+ Po k N =JQ.


PO

(a dimensionless

C. R, and Pirson, S. J.: ~Water Coning Control in OH Wells by Fluid Injection, SOC, Pet, Errg. jour. ( 1963) Vol. 22S, 314.

k,w p, k,w
k

p, q = integers

p.

= pressure

in oil phase phase


of terms held

2P

Pw = pressure inwater = x (00 + @w) Q =


stands for temporarily R=%(@oVp = pore @w) volume

groups constant

of a block = rr qi(r2i + ~ -

9. Sobocinski, D. P. and Cornelius, A. J.: A Correlation for Predicting Water Coning Time, SPE S94, presented at SPE Annual Fall Meeting (Oct. 11-14, 1964). 10. Richtmyer, R. D,: Df//ererzce AfetLrods for ItiitialValue Prvblerns, Interscience Publishers, New York (1957) 1010 D. W. and Rachford, H. H., Jr,: The 1 L Peacemsn, Numerical Solution of Psrabolic and Elliptic Differential Equaf ions, J. Sot. fndust. Appt. hl~tb, (1955) vol. 3, 2s. D. W., Rachford, H. H,, 12. Bruce, G. H,, Peacemani of UnsteadyJr. and Rice, J. D.: ~8Calculafions State Gas Flow Through Porous Media*t, Tratis,, 198, 79. AIME (1953) Vol. 13. O*Brien, G, G,, I.fyman, M. A. and Kaplan, S,: A Study of the Numerical Solution of Partial Differential Vol. 29, 223.. Equation, ,/. Math P@s. (1951)

,2

i-ti
(in

) AD
direction perpendicular to

W =

width z

and y) of Te of total by which thickness error is decreaped

APPENDIX
BASIC EQUATIONS

x = fraction y = fraction A = rario

during

an iteration

Ap = pw - pa or po - pg (i.e.
positive) @o = P. - pc.gD

AP is always

To illustrate the coning a problem where water is 12a and 12b from Ref. and may be written in foIlows:

analysis, we will consider the displacing phase. Eqs. I describe srich a system Cartesian coordinates as

4*
SUBSCRIPTS

= PW - PrugD
AND SUPERSCRIPTS

~ = num~er

-of block increasing ~ block (i.e.,

in

direction

of of

j = number of increasing~
DECEMBER,
1913J

in direction downward)

Ssl

.,

.,

.,. :

.- .-

.-

. .. . . .

..

. .

~,

~Ax)?-l/jjR ?-1,31

w(y~)l %}
flo +WAXAY -o % ;(l)

and

+*{[,QQ+*)19
+%{[ kiy+~)1 } % (
k km += W Ax AY

. . . . . . . . . . . . . . . . . . . . (3)
Eq. 3 is written for each block in a horizontal row of blocks such as shown in Fig. L There is one unknown P2k+1 for each block in the row; and, thus, the same number of unknown Ps as there are equations. Each equation, however, (except for the first and last blocks) contains pi and pi-l. three unlmown Ps whibh are Pi+l, Such an array is desig~ated as a tridiagonal system of equations. Details of the method used for solving such a system have been published previously. 10-12 By similar methods, we may solution Eq. 1 in the of develop a corresponding y- direction (see section entitled Iteration , Parameters). The difference equation. corresponding co Eq. 2 for solution in the ~.direction is:

q. .

N
WAx Ay,
-bv~~

d$

Wax

*.

. . . . . . . . . . . . .

...

(2)

We riow multiply Eqs. 1 and 2 by W AX AY, rhe volume of a differential element, and write thein as difference equations for the two-dimensional calculation grid. For solution in the ~-direction, Eq. 1 becomes:

(q?+ ,j [q:; - #-1 .] i-l,J


+ (NAy)~, j+ l-l [F ~,j+l #1-1 i,j 1

.f

,.. .

-,-

___

.. . . .

8s8

SOCIETY

Or

PETROLEUM

ENCINEERS

jOtlKtNAL

,,

. .. .

., -..,..

+ (MAx);~

,j

$U5+l,n+l

- ~2k+l,n+l i,J

[ %+l,j

Certain types of flow conditions at the boundaries of the mathematical model can interfere with relatively simple application of the equations developed above. It is, therefore, convenient to simulate flow across a boundary by fiuid injection or with= drawal at blocks adjacent to the boundary. In our incompressible analysis, we hqve found that total withdrawal need be made equal to total injection only within about 0.01 per cent. ITERATXON PARAMETERS Eq., 3 contains a right-hand term in place of the zero in 13q. I. This right-hand term tends tg vanish when the P~s have been satisfactorily reiaxed by iteration, i.e., when all the P2k+~s nearly equal the P2kso The right-hand term, or one of simiIar form. must be included durine iteration; otherwise, the solution becomes unstable. The convergence analysis outlined below leads to the optimum range of iteration parameters (His) to use in solving Eq. 3 and its counterpart for solution of P in they-direction. Eq. 3 is called the P equation because it is used to give values of P for each block at the new time level. In simplified form this equation becomes:

~2k+l,n+l i,j

- #

i,jl%[(mx)tijtd
(m Y )n 1 i23~
.

Ax

[ 1+1> j

P2k+l

- 2P~l . . 9

+ P2k+l i-l, j

(MAX):+ +
~2k+2,n+l

+(~)?,j+]
. . . . . . (4)

[ i,j

- *2k,n+l i,j

ADIP involves solving Eq. 3 for P values in the x-direction, alternating with solution ,in the ~ ~irection, untiI no further significant changes in the P walues aremoted. After new values of the Ps have been secured by alternate iterations in the ~ and ~directions, these values are heId constant while the new values of R are similarly obtained by, iterative aolucion of Eq. 4 and its counterpart for solution in the ~direction. A complete relaxation of ~e P equations, followed by a compIete relaxation of the R equations, constitutes one time step. To obtain saturations at the end of a time step, we use the relation,

Pc=lo-

*w-

A&D=2R-ApgD

..

(5)

Here we have deleted the factors M which contain relative permeabilities. These factors vary from point to point in the grid in a way not predictable ahead of time. We are solving first for the new in the first bracket in Eq. values of P appearing 6. This ia called sweeping: in the ~-direction. The terms involving the Rs and the qs in Eq. 3 appear as the (temporarily constant) term Q in Ea. 6. A sweep in the ~-direction is alternated with a sweep in the ~direction, and both of these sweeps comprise one iteration. Thus, sweeps occur in pairs; as each complete iteration is completed, the value of k increases by 1. Use of Eq. 6 is alternated with the use of Eq. 7 for solution in the y-direction:

Since p= is uniquely related to saturation, a knowledge of P= gives the oil and water saturations for each ,block in the system. - Aa defined in the Nomenclature, the Axs and .AYs .in .Eq..3 contah the area across which. flow occurs. The change from Cartesian to cyJihdri&l coordinates encompasses appropriate definition of the A%?, Ays and VPS,

A ;2k+2 Y. [ t,j+l

- 2p2k+2 i}j

: ;2k%2 i,$-1

+ Q

--

/ . ..

DECElltBER,

1964

Sss

. . . . . . . . . . . . . . . . . , (7)
In Eq, 7, we solve for the Ps having 2k + 2 superscripts. Inasmuch as the use of Eq. 7 disturbs the solutions obtained by Eq. 6, and vice versa, the equations are used repeatedly and alternately further changes in the Ps are urrtil no significant noted. be valuea of P which There must, o f course, exactly satisfy Eq. 6. CalI these values f*, and set: P= E+ P*,

The quantity the ratio which to be reduced to arrange to possible. Continuing von Neumann

bzk+l /h2k we Will Call ~1. It is defines the residual ,error remaining by further iteration. We should tv ,. minimize this error as rapidly a: application of the usuaI type stability analysisl 1-13 to Eq. of 10

or

-,(.x +Ay)Hk+ si.2(n&) ~y(,


1= ., (Ax+ AY~Hk
!,.

where E represents the error in P, or the amount by wh~ch the correct values of ~~ sought are not P attained. Eq. 6 then becomes:

- Ax(h

si.(~-

..*

(11) be

2k+l Ax [ i+l, j

. 2E2k+l i,j

+ ~2k+l i-l, j

1 +
]

Treatment analogous to applied to Eq. 7, giving

the

foregoing

may

*
Y [

~2k i,j+l

-2E~d

+ E&

,2 .

2(AX +X(4Sin(*) +Y)!k


-2( Ax+ AY)~-Ay(4sin2(~~ \ the decrease will become, 2A 1 2 = sin in error for one cowplete after slight rearra-lgemcmt,

so that iteration

. . . . . . . . . . . . . . . . . ...

(8)

In particular, note that the constant term Q is missing from Eq. 8 because it comprised part of Eq. 6.that is satisfied exactly by the P*. We next assume that the Es may be expressed as a prochrcr of three functions,
.

(2 )-L+Y)ik
Tll ~@)+~x + Ay ) ~

2A;

sin2

2.X

sin.2 &2 x

E~,j where ~ = g = ~, = Subsritutjon rhugh by

=figjhk,

-......(9)

2A

(d-(.x+ &
PAX npAx (T)+ (x + y) %

iunction of & position only function ofy position function of iteracinn number only. of Eq. 9 into 8 gives, after dividing /i g} hzk, L a * i+l 2k+l . 2,++]+ ... A Xh [ fi kk i

. . . . . . . . . . . . . . . . . . . . (12)

The optimum value for H~ in Eq. 12 will make one or the other of the; t<wo numerators zero, or close to zero. The smallest value sin nq,4y/2 can have is when q = 1; we then have far the first numerator, which contains q, to vanish,

Ay .y. [

gj+l
J

gj.1 2+

. .._

=2
354

(AX+AY) H@-,

] .,10)

. . . . . . .*
SOCIETY oF

**.PETROLEUM

-.

(13)

ENGINEERS

JOURNAL

. .
Similarly, for the second vanish, for p = 1, n 2 @)2 2rAy+l numerator in Eq~ 12 CO permeabilities and viscosities were omitted in happens that deriving them, Thus it sometimes these equations give too small an eStimate of the miriimum ff~, so that the P equations do not converge. In this case, we increase the minimum H~ by successive factors of two until satisfactory convergence results. Iti practice, only one range in iteration paramerer is specified. We then define this range as extending from the smallest value given by Eqs. 13 or 14,, to the largest value given by Eq. 15, or 2. The range in Hks may then comprise 6 to 10 values spaced in a geometric progression extending from the minimum to the maximum value, We then assume that one of the values in this range will make the error factor /! 1 A 2 quite small for- all values of p and q. me re~axation calculations are repeated for a number of cycles, using the same range in values of Hk each cycle, until very Iittle further change in any of the Ps results. A convergence analysis for Eq, 4, called the R equation, may be derived in a simi Iar way. The analysis shows that the Iast term in Eq. 4 is not needed for convergence, strictly because the a similar preceding term involving & serves work, it is best to purpose. However, in coning include the term involving the iteration parameters Hk because the .& term may be very sma~l. If it is smaIl, the rate of convergence may be very slow. Note that ECI. 4 uses the new values of P iust obtained by relaxation of the P equation. fiese
vah.res of P are held constant until relaxation of

Hk=

. . . . . ..

(14)

L%

Here Ax and Ay are fractions (dimensionless) of the total radial distance, ~, and of the total vertical thickness, respectively. The smallest Ax and Ay will normally be for blocks near the wellbore. The largest value that sin nq.4y/2 can have is nearly I. For this case, we have for the first numerator, 2A Hk .+= y ~>:1,
J

and for the second

numerator,

=ti
x
As

..

. . (1.5)
H~ will since be

a practical
2.

matter,

the

maximum

essentially

Eqs.

13 and 14 are approximations,

relative

the

R equations

is completed.

***

I
DECEMBEq, 1964 8s6

You might also like