Professional Documents
Culture Documents
SPE 38653 A Procedure To Integrate Well Test Data, Reservoir Performance History and 4-D Seismic Information Into A Reservoir Description
SPE 38653 A Procedure To Integrate Well Test Data, Reservoir Performance History and 4-D Seismic Information Into A Reservoir Description
A Procedure to Integrate Well Test Data, Reservoir Performance History and 4-D
Seismic Information into a Reservoir Description
Jorge L. Landa, SPE, and Roland N. Horne, SPE, Stanford University
Abstract
This paper deals with the problem of estimating the
distributions of permeability and porosity in heterogeneous
and multiphase petroleum reservoirs by matching the dynamic
behavior. The dynamic data is in the form of field
measurements from well testing, production history,
interpreted 4-D seismic information, and other data such as
correlations between permeability and porosity, geostatistics in
the form of a variogram model and the inference of large scale
geological structure.
The issue was posed as an inverse problem and solved by
using nonlinear parameter estimation. The procedure
developed here is capable of processing all the information
simultaneously and this results in a fast and efficient method.
The procedure is also able to determine the uncertainty
associated with the estimated permeability and porosity fields.
Examples of different parameter types that may be
estimated by this approach include: (a) individual block
permeabilities and porosities; (b) geological objects such as
channels and faults; (c) pilot points that form the basis of a
kriged distribution; and (d) seismic attenuation values from 3D seismic images.
An important conclusion of this work is that the value of
each piece of information does not reside in its isolated use but
in the value it adds to integrated analysis of the complete set of
information. Thus data that traditionally was considered to be
of low information content for reservoir characterization
SPE 38653
Problem Statement
An example of the type of problem we want to solve has been
summarized in Fig. 1, where (a) shows a two-dimensional
SPE 38653
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
Theory
The process of inversion to determine values of reservoir
parameters, such as permeability and porosity, from indirect
measurements is referred to as a parameter estimation
problem (also referred to as an inverse problem). The usual
approach to solve the parameter estimation problem is by
going through three major steps, not only for the specific case
of this work but for any general problem. These steps are as
follows:
1.
2.
3.
4.
5.
Mathematical Model
The physical system under study is represented by a
mathematical model that is constructed by applying the
fundamental physical laws that are relevant to the problem.
The purpose of the mathematical model is to predict with
reasonable accuracy the behavior of the system under different
conditions. The problem of computing the response of the
mathematical model to an external perturbation is referred to
as the forward problem. The physical properties that remain
invariant for different problems are referred to as parameters
of the system. The ones that change are referred to as
variables. The opposite problem, the inverse problem, consists
of finding a set of parameters for a given model such that the
predicted behavior of the system replicates the true behavior
r
r r
d cal = d cal ( )
(1)
r
where R npar is the vector of the parameters of the
mathematical model. Most or all of the parameters are directly
related to the distribution of permeability and porosity in the
reservoir.
r
ki = ki ( )
r
i = i ( )
(2)
(3)
r
r
r
r
T
E = d obs d calc W d obs d calc
E=
) (
r
1 r
d obs d calc
1 r r
prior
2
) C (dr
(4)
d
calc +
d
T
r
1 r
C prior
obs
(5)
r
H GN = E
(7)
H GN = 2G TWG
(9)
r
where G is the matrix of the first derivatives of d cal , also
referred to as sensitivity coefficients.
r
d cal
G= r
(10)
E
E = r
SPE 38653
r
r
r r
E = G T C d1( d obs d cal ) + C1( prior )
1
d
H GN = G C G + C
T
(11)
(12)
r
E( * ) 1
r
E( * ) 2
(13)
(14)
r
where 1 and 2 are small positive numbers. * is referred
to as an optimal point and should provide a set of parameters
that results in a good match of the data.
(6)
r
p ( t )
si p ( t ) = i r
r wc
wci ( t )
si ( t ) =
r
S w j ( t )
r
s jS ( t ) =
r
(15)
(16)
(17)
SPE 38653
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
r
r
r
JZk +1 = DZk + Wk +1
i
with
(18)
r
r
Z0 = 0
r
r
yk
Zk =
i
r k +1
f
J = r k +1
y
r
f k +1
D = rk
y
rk
k
y = p1k , S wk _1 , p2k , S wk _ 2 ,K , pnbloc
, S wk _ nblock
(19)
(20)
(21)
(22)
(23)
r
r
d cal = G + constant
r r
where the sensitivity matrix G is computed at = *
(24)
G = USV T = U p S pV pT
(25)
G g = V p S p1U Tp
(26)
R = V pV pT
(27)
r
r
Cov{ * } = G g Cov{d obs }G gT
(28)
The same data were used for another inverse problem, this
time without using the information about the existence of a
channel. We kept the 40 30 simulation grid but now we
parameterized with 100 parameters, each parameter
representing the permeability of 12 adjacent cells. We refer to
this method of parameterization as the large pixel model
because of the resemblance to the pixel approach. The initial
guess was a homogeneous reservoirs. Fig. 5 (b) shows the
calculated permeability distribution. Fig. 5 (c) shows also the
boundaries of the channel that was used originally to generate
the data. This problem was much harder to solve because of
the larger number of parameters. It took approximately 400
iterations to obtain a good match of the data. Thus a first
impression is that the object modeling approach was 10 times
faster, but it must be remembered that each iteration of the 100
parameters requires much longer CPU time. Hence the object
model approach was actually 1000 times faster than the large
pixel approach.
Black and White 4-D Seismic Data
One observation in the previous example is that we were using
exact data. This can be considered unrealistic, especially in
the case of the 4-D seismic data since with the current
technology it is not possible for the geophysicists to prepare an
exact map of change of saturation. Most likely they will be
able to prepare a map where they can assert the areas of the
reservoir where there were changes in the saturation but the
magnitude of the changes will be unknown. Thus we used the
data shown in Fig. 1 but instead of using the exact change of
saturation maps in Fig. 1 (f) and (g) we used the coarse maps
shown in Fig. 6 (b) and (d). Fig. 7 (a) shows the true
permeability field, (b) the calculated permeability when
exact data is used, and (c) the calculated permeability when
the 4-D seismic data is used in a coarse black and white
format. We see from (c) that the black and white data can be
used in the approach we developed here and still provide a
reasonable description of the reservoir.
Fault Model
We can use the object model approach to find the location of
faults in a reservoir by using a diversified data set similar to
the one depicted in Fig. 1. As a first approximation, we can
model a sealing-fault as a rectangle, where the permeability
inside is very low (10-5 md) and the width is small compared to
the dimensions of the reservoir. We set free all the other
parameters that define the rectangle object. Thus the rectangle
can change its length, rotate and translate in space. Fig. 8
shows an example of finding the location of a single fault. This
figure shows the location of the fault and the water saturation
distribution in the reservoir at the end of the simulated time.
We do not use the saturation information in the match. The
first frame shows the first guess for the fault. The first guess is
substantially different than the true case (last frame) from
the point of view of water saturation (see the well closest to the
right bottom corner). The procedure developed in this work
SPE 38653
kj =
npar
i,j i
(30)
i =1
SPE 38653
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
k j = k 0j + i , j ki
(31)
i =1
SPE 38653
k
point of view, they are seem to have
SPE 38653
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
10
SPE 38653
References
Nomenclature
r
= parameter model vector
r
* = optimal parameter model vector
r
prior = a priori parameter model vector
k=
k0 =
=
=
2 =
E =
Cd=
C =
Cov{ } =
DST =
r
d =
E=
r
f =
G=
H=
J=
R=
Sw =
S=
t=
U=
V=
W =
r
y =
r
Z =
permeability
permeability from realization
kriging weight
porosity
variance
gradient of E
data covariance matrix
a priori parameter covariance matrix
covariance matrix operator
drill stem test
data vector
objective function
material balance equations vector
matrix of sensitivity coefficients
Gauss-Newton approximation to the Hessian matrix
of objective function
Jacobian of material balance equations
resolution matrix
water saturation
singular value matrix
time
factor from singular value decomposition
factor from singular value decomposition
weighting matrix
vector of pressure and saturation in the simulation
mesh
sensitivity vector
Subscripts
obs = observed value
calc = calculated value
npar = number of parameters
nblocks = number of blocks in simulation grid
p = number of nonzero singular values
T = transpose of matrix
Superscripts
= vector
-g = generalized inverse of a matrix
k = iteration index
p = pressure
wc = water cut
Sw= change of water saturation
SPE 38653
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
11
12
SPE 38653
6000
4000
2000
log(k)
Permeability
8000
(h)
(a)
- Porosity
Pressure
Water Cut
DST
Well # 1
c
(i)
(b)
(c)
DST
Well # 2
Production History
Well # 2
(j)
(n)
DST
Well # 3
Production History
Well # 3
(k)
(o)
c
(g)
(m)
c
(f)
Production History
Well # 1
c
(d)
0.6
0.4
c
(e)
Saturation
DST
Well # 4
Injector
Injection History
Well # 4 - Injector
0.2
(l)
(p)
SPE 38653
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
Iter #1 E= 197,097
Calculated Reservoir
10000
Permeability
Field
Iter #2 E= 113,596
Iter #3 E= 41,512
15000
5000
Permeability
True Reservoir
13
Iter #7 E= 15,617
Change of
Saturation #1
0.15
DST
Well # 1
aaaa
aaaaaaaaaa
aa
a
a a aa
aa
aa
True Pressure
True Water Cut
Calculated Values
a
aa
a
aaaaa
aa a
a aa aaaa aaaaaaa
a
aaaaaa
a aaa
aaa
a a a
aaaaa a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a
a
a
aaaa
aa
a
a
a
a
a
a
a
a
a
a a a a aa a
a a aa a a a a
Pressure & Water-CutaHistory
a a a a
aa
a aa a
a
a
a
Well # a2
a
a a a
a
a a a
a
a a a a a
a
a a a a a a
a a a a
a a
aaaaa
a a a a a a a a a a a a a a a a a a a a a
aaaa
aa
a
aaaaaaa
Pressure & Water-Cut History a a a a a a a a
a
a a a
Well # 3
a a
DST
Well # 3
Iter #38 E= 71
DST
Well # 4
Injector
aa a a a a
a a a a a a a
a
a a a a a a a a a a a a a a a a a a a a a a a a a a a a
Pressure History
Well # 4 - Injector
a aaaa
a a aa
a a aaa
15000
10000
5000
a
a
a a a a aaa
a a a a a
a a a
a a a a
aaaaa a a a a a a a a a a a a a a a a a a a a a a a a a a
a a
aaa a a a a a a a a a a a a a a a a a a a
a
aa
a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a
DST
Well # 2
aa
True Permeability
aaaa
aaa
a
(a)
(c)
Calculated Permeability
15000
a
10000
5
2 = kmatrix
5000
a
0
(b)
1 =
kch
-1
4
Calculated Permeability
anne
Saturation #2
md
a aa aaaaa aaaaaa
0.20
Change of
md
Sw
Match
of the Data
0.30
(c)
(b):
(c):
14
SPE 38653
Iter #1
True Sw #1
Filtered Sw #1
Iter #2
Iter #3
0.4
0.2
Sw
0.3
Iter #7
0.1
Iter #11
Iter #15
0.0
(a)
(b)
True Sw #2
Filtered Sw #2
0.4
Iter #20
Iter #25
Iter #30
0.1
0.0
(c)
Iter #35
(d)
Iter #42
Iter #48
Linear Search
True Permeability
Iter #45
Iter #51
0.6
0.4
Sw
0.2
Sw
0.3
0.2
a
(a)
6000
103
5000
md
4000
2000
10
md
3000
1000
(b)
(a)
(b)
a
(c)
(c)
(d)
INTEGRATING WELL TEST DATA, PERFORMANCE HISTORY AND 4-D SEISMIC INFORMATION
10000
5000
c
102
10
(a)
(c)
k/k - True k
k/k - Calculated k
10
(a)
1
Calculated Permeability
c
(d)
Fig. 12 Variance maps: (a): True reservoir . (b): k/ map for true
reservoir. (c): Calculated reservoir. (d): k/ map for calculated reservoir
(b)
Fig. 10 3-D seismic model. (a): True reservoir. (b):
Calculated reservoir.
Map of -1
c
1/ md-1
10-1
10-2
c
Fig. 11 Map of . Homogeneous reservoir.
-1
10-1
(b)
c
c c
k md
k/k md/md
c
103
c
c c
104
15
Calculated k
True k
True Permeability
k - md
SPE 38653
well location
16
SPE 38653
k/ scale
Data: Pressure
c
1
k/k md/md
10
10-1
Data: Pressure + Water Cut
Data: Pressure + Sw
Data: Sw
k/ scale
1
10-1
10-2
k/k md/md
10
10-3
Data: DST + Shut-in Pressure