Professional Documents
Culture Documents
5 - Spe-2969-Ms
5 - Spe-2969-Ms
BY
D.H. TEHRANI
MASJID–I–SULAIMAN
APRIL, 1969
TABLE OF CONTENTS
PAGE
INTRODUCTION 2
THEORETICAL CONSIDERATIONS 3
FIELD APPLICATION 11 - 12
DIs~~S~I~fi 12 - 17
CONCLUSIONS 17 - 18
NOMENCLATURE 22 - 23
LIST OF REFERENCES 24
ACKNOWLEDGEMENT 25
D.H. TEHRANI
The limitations of the method and the effect of its underlying assumptions have
been discussed.
.
1. INTRODUCTION
Simultaneous solution of oil-in-place and water influx has been attempted many
times by various workers. The volumetric balance equation has been written in
a form similar to:
In the past the function?( x(3), x(4) ) has been independently calculated
using various models and assumptions. Eq. (I), which is linear in terms of
x(1) and x(2), has subsequently been used for least square analysis to obtain
the so-called best values of oil-in-place and x(2).
In this paper two improvements are presented which have proved successful:.
. ../ .. .
-3-
The left hand side is the total reservoir withdrawals which hereafter
will be denoted by WN. NBti is the original reservoir volume of oil-in-
place, which in this paper is unknown and will be denoted by x(l). The
coefficient of NBti, in the brackets, is called unit expansion and is
denoted by UT. The last term, We, is the water influx. The equation
giving water infiux, either in Hurst - Van %=~di!lg~~ form or in an
exponential type, is defined by three constant coefficients. By Hurst -
Van Everdigen method, the three constants are:
These three constants are also unknown and are denoted in this paper, by
x(2), x(3) and x(4).
Assuming that pressure history, production history and all fluid property
data are known, the problem is to find the best set of x(j), j = 1,4 to
satisfy Eq: (1) or its simpler form as under:
WN = x(l). UT +,f( x(2), x(3), x(4) ) (2)
where:
WN= Np(Bt+Bg(R -Rsi))+W
P P
When n>4 the system of Eqs. (3) would have more equations than unknowns
and a type of optimization technique would be required to find the set of
unkn~s which would “best” fit all the n equations. In this paper the
“best” set of unknowns is defined as the set which minimizes the sum of
.
square of deviations (SSD) between the calculate s ~..-.-m~i-m
LUU-. -
=n~ measured
value of the function. For example, the best set of x(l), x(2), x(3)
and ~(~) is ~ha~ ~e~ &ich min~izes the:
i =n i =n
where:
Since aX(j) are small, the function can be approximated by its first
terms on the righthand side of Eq. (9) i.e.:
z
j=~
\F
F( X(j), -j= 1,4) + —s0
Ax(j) %x(j)
j=l
. ..1...
-6-
Newton’s root finding method states that if X(j) are approximate values
of x(j), a set of&X(j) satisfying Eq. (10) would give a better e$timate
of x(j) when added to the values of X(j), (See Ref. 3 and 4). In other
words, our problem is now to find a set of AX(j) which satisfy Eq. (10),
For n data points we have n linear equations of the form (10) with four
j =4
3 F~
Fl(x(j)$ j = 1,4) + AX(j) 3x(j) = O (11)
~
j~~
!
i = l,n
or in expanded form:
>Fl \F~ \ F~
F1(X(l), X(2), X(3), X(4) ) +L)X(l)%~+AX(2)%~ +~X(3)%~+
>
(
F2 > F~
1
> ‘2
(11)’ F2( x(1),.x(2), x(3), x(4) ) +Ax(l).%-+4x(2)%~ +Ax(3)a~+
.
●
%% o
*x(4)jm =
.
. ~ Fn ~ Fn
\ Fn
Fn( X(l), X(2), X(3), X(4) ) +L)X(l)>X(l)
— ‘AX(2) >X(2) ‘*X(3)%X(3) +
\\
System of Eq. (11]* are in iinear form and the method Of lea$t $quares
can be applied as follows:
Let:
j=4
Fi >
Di(&(j) ) =Fi( X(j) ) +
7 AX(j) >x(j) (12)
i = l,n
● ☛☛☛☛✎ /
-7-
i=n
*k.P/.\
The beSt SeK OIQA{J) iS ~~~~ whi~~~i~i~~~eS ‘~D~ Since Di is a linear
function of AX(j), the SSD can be minimized by letting the first Partial
derivatives of SSD in respect to flX(j)equal to zero. (See Ref. 5). A
system of 4 by 4 equations will be obtained as follows:
i =n
> Di
z
>~X(l) ●Di=O
i=l
i =n
\ DioD =0 (14)
>~x(2) i
z
i=l
The system of 4 equations with 4 unknowns (14), are linear and can be
solved by method of elimination. Direct manipulation of these equations,
however, is a cumberscnuejob, and the following matrix calculation can
provide the coefficients of system (14) in a more straight forward manner.
The system of Eq. (11) or (11)’ can be shown in matrix form as follows:
[ddd=-t’l (15)
. . ...*
/
-8-
where:
M= I
(16)
1
(n x 4)
and
(17)
[1
Lx = { AX(l), Ax(2), AX(3), Ax(4)
‘} (4 x 1)
and
‘1(x(l), x(2), X(3), X(4) )
(18)
F=
[1 D
Now if both sides of Eq. (15) are multiplied by transpose of matrix [1M ,
-9-
Now let product [M’] [M] be denoted by [S] and multiply both
[Axl
T
‘U S-J
r 7-1●
[Ml]
●
[n)
- (20)
Evaluation of the right hand side of Eq. (20) will determine the values
of t?JX(j).
As it was mentioned earlier, when the values oflix(j) are added to their
respective X(j), a better estimate of x(j) will be obtained. Now the
values of ( X(j) + Ax(j) ) can be taken as new estimate of x(j) i.e.t
in all previous calculations X(j) will be replaced by ( X(j) +AX(j) )
and the computation is repeated. This iteration process can be continued
until the values of Ax(j) are negligible when compared with a required
tolerance.
It should be noted that in Taylor expansion series, Eq. (9), the third
term onward can be dropped only if the values of AX are small which
means the first guess should be fairly close to the answer; otherwise
convergence may take numerous integrationsor may diverge and fail. In
case of volumetric balance equation, however, the first estimates can
be easily obtained from geological and petrophysical data. No convergenc
problem was experienced even if the first estimate of oil-in-place was
30% or 300% of the final answer.
X(2), X(3) and X(4), is made, the elements of the column matrix (18) can
be directly computed, using volumetric balance equation in its form of
Eq. (6):
* ● ✌ / ✌☛✎
-1o-
)Fi
G
We(i) = (dpj) (X(2) +X(3) (1.-x(4)(i - j)j j (24)
L
j=l
. ● ✎ / ● ☛☛
-11-
j=i
>f. >We(i)
(4
(dpj) (I- x(4)’* = j), (26)
The Haft Kel Field, located in southwest of Iran, was chosen to iiiuscrate
an actual application of the simultaneous solution of oil-in-place and
water influx parameters by Newton Optimization technique. This field was
first placed on production in 1928. It produced a total of 1556 million
stock tank barrels of oil, 537 billion standard cubic feet of gas and
about a milll,ohbarrels of water to end 1967. The original oil COlumn
of 2070 ft. was reduced to 230 feet. The reservoir rock is highly
fractured Asmari limestone with very low matrix permeability.
/
-12-
B. BASIC APPLICATION
IV. DISCUSSION
. ../ . . .
-13-
4.2 Assumption that the water influx model truly represents the
aquifer system.
● ✎ ✎ / ✎ ● ✎
-14-
we = x(2) (Pi - P) dt
J o
or (28)
The simplest of these models Eq. (28) has only one constant parameter.
When this is introduced into volumetric balance equation, the equation
will be linear in terms of oii-in-piace and x(2). 9Q+117
It csr&be e-..~
solved by simple least square method. This, unfortunately, may be
suitable for only a limited number of cases, where the water influx is
steady - state.
Attempts have been made to estimate the values of x(3) and x(4) by direct
methods such as physical rock properties and the size of the aquifer, then
apply the least square technique to determine x(2).
-15-
The most recent and complete of such analysis is the work of Wall and Craven
~alkerlo,
They va~ried both constants of x(3) and x(4) over a range and
applied the least square method to find x(2). Then they plotted the S~
square of deviations against x(3) for various values of x(4). Among the
family of curves so obtained, the point ( x(3), x(4) ) corresponding to the
smallest minimum SSD would be used subsequently to determine the oil-in-place
In almost every study which has been carried out so far, the volumetric
balance equation has been arranged to give the oil-in-place as a separate
term in a form such as:
10
or as Wall and Walker put it:
<+#Xi=Yi (32)
where:
fi ( x(3), x(4) )
~= x(1) x. =
1 UT(i)
= x(z)
f
Thereafter, a least square analysis has been carried out to calculate the
so-called “best” values of~andfi .
Eq. (31) is the same as Eq. (3) with the difference that both sides of
Eq. (3) have been multiplied by (-) to obtain Eq” (31)0 This has ‘he
WA,-.,
same effect on the result as giving a weight factor of (*)2 to each data
point i.
Now let us see what is the effect and significance of weighting the data
points in this manner.
.0./ ...
-16-
In any reservoir analysis some data points are considered less reliable than
others. As a general rule the early data are the weakest. Introduction of
WT(i) as a weight factor for year i in Eq. (4) will result in:
i =n
If we now choose WT(i) = (*)2, substitute this in Eq. (33) and carry out
the optimization, by minimizing SSD, we shall get exactly the same results
that we would get by using Eq. (32) or its equivalent Eq. (31), without
cor.siderin.g
any weight factor.
-I----
It should be emphasized that w,,=,..koosin-gweight factor WT(i), normally~
the least weights are allocated to early data for following two reasons:
2. Normally the production and pressure data of early years are not so
accurate as the data of later years; one simple reason, among many
others, being the advancement in technology of measurement.
On the other hand, the unit expansion UT(i) is very small at early years
and larger later on. For example, in Haft Kel Field UT was 0.0001 after
two years of production by end 1929, and 0.08 after 40 years of production
.
by end 1967. Multiplying both sides of equation by ~ and carrying least
square analysis is equivalent to giving a weight factor of 108 to the data
of 1929 and a weight factor of 156 to the data of 1967.
/
● ☛✎☛☛☛
-17-
The method used in this paper, however, utilizes SSD of reservoir with-
drawals, WN(i), rather than oil-in-place. From Eq. (3) it can be seen that
WN(i) is a function which has both observed values (from production data) and
calculated values (from right hand side of Eq. (3) after x(1) through x(4)
are determined). Analysis based on SSD of this function seems more reasonable
than analysis based on SSD of x(1) which is a coefficient in Eq. (3) with no
observed values, although x(l), i.e. the oil-in-place, is the most wanted
value of the whole analysis.
v. CONCLUSIONS
.*..** /
-18-
APPENDIX 1
----
CALCULATIONS OF WATER INFLUX
1. Basic Notations
~1
mitted in the FORTRAN-II program- ;DP(x)=DELP(a)
; I
+_–_ _;-----
sure at the end of first producing two ---;---
year. Figure 2 presents the treat- ?(s) --,,–l– - --1---- +--– ‘i--%
t’ : ‘~ I t
1 1 1
1 b
ment of the pressure and the defini- I 2 3 4 s
tions of related terms. As is depi- YEARS OF HISTORY
igure 1
cated by this Figure, ‘the actual pref-
sure profile is approximated bjjstep~i~e pressure drops corresponding to the
average pressure during each year.
Where A, B and C are the parameters of water influx which in the text of the
paper and in the computer program are denoted by X(2), X(3) and X(4) respec-
tively.
● ✌ ● / ✎ ✎ ●
-20-
J =1
(1.2)
>= . DP(J) = DELP(I)
3A z
J2=
JI=
(I-J-1) (1.4)
~Z(I) = - ~c (1 - J). DP(J). C
>C z
J2=
3. Recurrence Equaticfis
or
(1.5)
Z(I) = (A+B) . DELP(I) - B. SUM(I)
where:
JI=
(1 - J) (1.6)
SUM(I) = DP(J) . C
z
J2=
. . . I ..*
-21-
or
=
JI-1
(SUM(2) = DP(2)
:SUM(I) = C.SUM(I-1) + DP(I) ForI=2
(Z(I) = (A + B). DELP(I) - B.S~(I) For I 2 (1.11)
>
In the like manner the following sets of recurrence equations were derived
f~~ the partial derivatives:
(SUM(2) = DP(2)
(
~suMl(2) = o.
(SUM1(I) = C. SUM1(I-1) + SUM(I-1) (1.12)
(
(hZ(I)
— = DELP(I) - SUM(I)
(>B
(-
(.
(~z(I) = - B. SUM1(I)
(r
/
● ✎☛✎✎☛✎
-22-
NOMENCLATURE
. . ./ . . .
-23-
We Water influx
WN, W’N(i) Reservoir withdrawals.
w Produced water
W;(i) Weight factor used for data of year i
x(j) Oil-in-place (j = 1), and water influx parameters (j = 2,4)
X(j) Estimated values of x(j).
. . . / .* ●
-24-
LIST OF REFERENCES
5. Williams, A.J., C.E.I.R, Ltd., U.K. and Core Laboratories, Inc., “Engineering
Reference Manual of Mark 6/02 Reservoir Simulation Model” for Iranian Oil
Exploration and Producing Company Appendix IV, pages IV-7 to IV-11.
8. Freeman, H.A., Natanson, S.G., “Proc. 5th world pet. Eng. 11, PaPer 24”*
11. Tehrani, D.H., “Bubble Point Pressure Correlation for Reservoir Crudes of
southwest Iran”, a paper presented to Second Regional Tech. Symposium of
S,P,E, of AIME, in Dhahran, Saudi Ara”bia,March, 1968.
. ., / . . .
-25-
ACKNOWLEDGEMENT
The author wishes to thank the management of Iranian Oil Exploration and
Producing Company for permission to piiblishthis paperj and Messrs.
E. Th. Van der Laan, S.A.E. Sablayrolles,”’;,JC.C.
Pike and E.H. Vlierboom,