Académique Documents
Professionnel Documents
Culture Documents
SPE 31087
Abstract
A numerical model has been developed that addresses both
plugging of, and sand production through single wrapped
screens. The model was developed on the basis of a fractal
model for the particle size distribution of reservoir sands. A
database of sand types from the North Sea and Haltenbanken
areas was established. Principal component analysis was
used to reduce the number of significant variables in the
database, and to provide a basis for a prediction model for
critical slot widths. A series of laboratory experiments were
performed, and four critical slot widths were identified for
each sand type, defining a safe design interval for screen slot
width. A mathematical model was developed that can be
used to predict the critical slot widths for other sand types
from the area.
Introduction
Single, wire wrapped screens with keystone shaped wire
have been used to control sand production in oil and gas
wells since the 1930's. They have the advantage over
prepacked screens in that they do not become pl~gged as
easily by drilling mud. Furthermore they functiOn as a
surface filter, where the plugging material is easily removed,
whereas prepacked screens are depth filters where plugging
material tends to get trapped inside the prepack.
Single wrapped screens do, however, have a reputation
for being susceptible to plugging and/or sand production
when designed according to the various traditional criteria
(Refs. 1 and 2). This indicates that the design criteria for
single wrapped screen completions should be revised.
Sand control with screens is basically a function of the
relationship between particle size and screen slot width. The
pioneering work was published by Coberly (Ref.3) in 1937.
Coberly concluded that spherical particles could generally be
retained when the slot width was 2.5 times the particle
diameter or smaller. He also stated that in a mixture of
particles of different size, the sand control properties of a
155
.............
SPE 31087
N(d;;,di)=K(
~ir
(I)
Inti
SPE 31087
on PCJ and PC2 are fine and badly sorted, while sands with
high scores on PCJ and PC2 are coarse and well sorted.
It was found that these two principal components were
the most important ones for describing plugging and sand
control of single wrapped screens. The first component,
PCJ, explained 48% of the variation in the data and PC2
explained 24%, in total 72% of the total variance for the
first 2 principal components.
The principal component analysis effectively reduces the
number of variables from 9 to 2.
A total of 5 sand types were chosen for laboratory
testing. The sand types and their database entries are shown
in Table 1. They were chosen on the basis of their scores
on PC1 and PC2 as illustrated in Fig. 2. Particle size
distribution curves are shown in Figs. 3,4 and 5.
The principal component analysis showed that there were
no typical variation in the particle size distribution between
the various North Sea fields. Neither were there any
significant regional difference between the North Sea and the
Haltenbanken.
Experimental procedures
Identification of main experimental parameters.
In a typical North Sea sandstone reservoir, the variation in
both particle size and distribution is large. The permeability
often varies by a factor of 100 within the reservoir. Thus,
design criteria that specifies one single optimum slot width
for each sand type are not very useful because it will be very
difficult to chose which sand to use as a basis for the design.
It would be more relevant to define a range of acceptable slot
widths for each sand type, and then attempt to select a screen
that will fit into this range for all the sand types in the
completed interval.
This approach was adopted in the present study. Four
slot widths were determined for each sand type:
d__
the largest slot size where severe plugging was
frequently observed.
d_
the smallest slot size where no plugging was
observed.
d+
the largest slot size where sand production did not
occur.
d++
the smallest slot size where continuous sand
production did occur.
The d __ and d++ slot widths should be considered as
extreme lower and upper limits that should not normally be
exceeded, while d_ and d+ are lower and upper limits for an
ideal screen design.
The other parameters that were recorded during the
experiments were:
Amount of produced sand and sand production mode
(initial, intermittent, continuous)
Permeability ratio and skin factor for each sand type, slot
width and flow rate
Nature of plugging (permanent or removable)
Particle size distribution of produced sand.
Experimental set up. The screen filtration experiments
were performed in a radial flow cell as illustrated in
Figs. 6 and 7. The experimental set-up consisted of an
adjustable pump, a radial flow cell representing a 22.5
section of a well with a 7 .5" screen, a sand trap and a fluid
reservoir. The radial cell was fitted with 2 differential
Experimental results
The critical slot widths, determined from the experiments are
presented in Table 2.
157
Discussion
General flow properties. A sand control screen will
necessarily restrict the fluid flow into the well to some
degree, even when it is functioning as intended. Intuitively,
one should expect that the degree of flow restriction would
be a function of the screen slot width, the particle size
distribution of the sand, and maybe the rate of flow through
the screen. This turned out not to be the case, however, as
the skin factor varied unsystematically between 0.0 and 0.5
for all the sands and screen slot widths. Even if the
permeability of the sand varied from 0.2 Darcy for All to
20 Darcy forB 19. This can not be considered to be a serious
flow restriction, and it can be concluded that single screen
completions will not significantly restrict well production
as long as they function as intended.
The slot area is typically only 5-10% of the total screen
area. Fluid flow will converge on the slots, and the fluid
velocity will increase by a factor of 10-20 through the
slots, depending on the slot width and the width of the
wrapped wire. The converging flow results in a differential
pressure that is higher than expected from the Darcy
equation, where it appears as the observed skin factor. In
this way the flow properties of the screen is very dependent
SPE 31087
dail
(2)
Here dcrit is the predicted critical slot width, {30, ... /312 is a
set of constants, and t 1 and t2 are the score values, or coordinates, on the first two principal components. The
predicted values are compared with the observed values from
the laboratory experiments in Table 4. The difference
between the observed and predicted values are less than 50
microns which is approximately half the typical step of 100
micron between two consecutive screen sizes. This indicates
that the accuracy of the prediction model is equal or better
than the accuracy of the experiments. The accuracy of the
prediction model cannot be evaluated statistically because of
the limited number of experiments. Two more sand types
have been tested to verify the model, however, and the
observed results are very similar to the predicted critical slot
widths.
The predicted values for the critical slot widths are
plotted as a function of PC 1 and PC2 in Figs. 8 to 11.
From Figs. 8 and 9 one can see that the risk of
screens being plugged is high for fine sands and for coarse
sands with a large fraction of fine material. As expected, the
risk of plugging the screens is low for coarse, well sorted
sands. But the risk of plugging is also reduced for fine sands
with a high fines content. This is maybe surprising, but it
means that the original permeability of these sands are so
low that it is in the same range as the permeability of the
filter cake.
In Fig. 10 one can observe that sand control is a
function both of the particle size and the degree of sorting
and content of small particles. For fine sands, a low score
on PC2, indicating a badly sorted sand with a lot of fines,
will increase the risk of sand production. But for coarse sand
with high scores on PCI, a low score on PC2 will reduce
the risk of sand production.
Once the data from the principal component analysis and
the set of constants (from eq. 2) for the critical slot widths
are known, the prediction model is easily implemented in a
standard spreadsheet. A simple, user-friendly computer
program for the design of screen slot widths is currently
being developed.
158
SPE 31087
K = proportionality constant
f = exponent of particle size distribution
function (and fractal dimension of sand
matrix)
ell= d4ofd90
dcrir= critical slot widths (d__, d_, d+ ord++)
/30 , /312 = constants in the prediction model
ti = score value, or co-ordinate, on principal
component i
Acknowledgements
We would like to thank Saga Petroleum as and Statoil for
the permission to publish the material; Bjarne Aas of RFRogaland Research for helpful review comments; and Jorunn
0vsthus of RF for her accurate laboratory work.
References
1. Penberthy, W. L. and Shaugnessy, C. M.: Sand
Control, SPE, 1992
2. Sparlin, D. D. and Hagen, R. W., Selection and
design of sand control methods, Course Manual,
ICCI, 1991
Conclusions
1. No sand types have been identified during the reported
work that are not suited to screen completions. For all the
sand types tested it has been possible to identify an interval
of screen slot widths that will neither be plugged nor
produce sand. The width of the design interval varies as a
function of the particle size distribution of the sand.
2. A well functioning screen represents a skin factor of
less than 0.5.
3. The risk of sand production is increased in a situation
corresponding to an open annulus, partially filled with sand.
Plugging of screens by formation sand has only been
observed in this situation.
4. The risk of plugging the screen is decreased when the
fluid flow velocity through the screen is increased gradually.
This corresponds to bringing a well on stream slowly.
5. Design criteria for screen slot width based on one
single point on the particle distribution curve can not
accurately predict neither plugging of the screens nor sand
production through the screens.
6. By introducing a fractal description for the particle
size distribution of the formation sand, and using
multivariate analysis, it has been possible to develop a
quantitative method for design of screen slot widths. The
method identifies a safe interval of slot widths where
plugging and sand production are not likely to occur.
7. The prediction model is applicable to sands from the
North Sea area and Haltenbanken, and can easily be extended
to other areas.
8. A method is proposed, where the prediction model
can be used to design screen completions for specific
reservoirs or parts of reservoirs.
Nomenclature
159
SPE 31087
d 10
[micron]
d 40
d 50
d 90
[micron] [micron]
f1
ell
f2
[micron]
lnt1
lnt2
[micron]
[%]
A11
213
109
89
38
2.91
-3.22
-8.86
173
18.47
810
219
136
126
68
1.99
-1.12
-6.49
104
60.87
C31
249
210
197
131
1.61
-0.45
-9.23
152
84.55
815
475
340
316
169
2.01
-2.20
-7.99
379
29.25
819
491
353
329
197
1.79
-0.78
-8.41
306
59.59
d __
d_
d+
d++
A11
0*
100
100
200
810
100
250
250
300
C31
0*
200
400
600
815
200
300
600
800
819
0*
100
500
800
d1o
2d1o
(Gulf Coast)
(Coi:JOOy)
A11
213
810
d+
d++
427
100
200
219
439
250
300
C31
249
498
400
600
815
475
949
600
800
819
491
982
500
800
160
SPE 31087
d..
Sand
d+
d++
values
values
values
values
values
values
values
A11
0*
100
100
100
93
200
185
810
100
79
250
249
250
273
300
351
C31
0*
21
200
201
400
377
600
550
815
200
205
300
300
600
594
800
787
819
0*
-11 **
100
100
500
512
800
827
* d__ were set equal to 0 when severe plugging of the 100 micron slot was not observed.
** Negative slot widths are artefacts of the prediction model.
++C31
++
+*
ll.
-1
+
t+
t"'+
++
/819 +
+
+
:\:
::t ~to
-q.
+
+
++
++
++ + +
++
+
t+
+
+
+
+ +
+ +
++ ++ + +
+
+
+++ + +
+
+815 +
-1
-2
-2
+ +
-3
-3
+ A11
-4
-4
-4
-2
4
PC1
161
SPE 31087
150mm
50mm
100
100
10
10
..
~
z
0.1
0.1
0.01
Isomm
0.001
1000
L---'------'--L-..L.....J......L...L..J....L_----"---'---'-........__......._L.J...J
100
Particle size
10
250mm
(micron)
Directionofflow
~mm
819
100
..
10
10
0.1
0.1
en
i.
.
0.01
:I
0.001
L___
_l____j__L_..J........I.....L...J....Ll...-_
Particle
100
size
0.0001
1000
___..L._L-...JL....L.....J.....J.....LU
(micron)
100
100
-~---~-~-~-~~~~~-----~-
..<:
80
en
a;
~
60
".
.
40
D.
20
0
1
10
100
Particle size
1000
(micron)
0
1 o'
162