.—-.. ..—-.. ——-.

Society of Petroleum E;gineers

SPE 39586

Infectivity Decline in Wells with Nonuniform Perforation Properties

T. Hofsaess, Consultant; W. Kleinitz, PREUSSAG ENERGIE GMBH

Copyright 1998, SWlety of Petroleum Engineers, Inc. and perforation-properties generate a nonuniform flux
This paper was prepared for presentation af the 1998 ‘SPE “International Synif.iosium on–- ‘- distribution along the wellbore. As an example Fig. 1 shows
Formation Damage Control held in Lafayette, Louisiana, 18-19 February 1698.
the partial rates at the beginning of an injection and 9 months
~s paper was seleoted for presentation by en SPE Prcgram Commiffee folloting review of later. The initial flow is strongly peaked in the upper 2 meters
information contained In en abstract aubmittad by the author(s). Contents of the paper, aa
presented, have not been reviewed by the Saiety of Petroleum Engineers and are sub~ct to of the perforated zone. 9 months later it looks quite different
correction by the author(s). The material, as presented, does not necessarily reflect any
poaitlon of the Seciely of Petroleum Engineers, its officers, or members. Papera presented at
and is much flatter than before. The basic mechanism behind
SPE meetings are subject to publication review by Editorial Committees of the Society of this is rather obvious: The large partial volumes injected into
Petroleum Enginaers. Electronic reproduction, distribution, or storage of any parf of fhis paper
for commercial purposes without the written consent of the Soolely of Petroleum Engineers is the highly conducting parts carry with them correspondingly
prohibited. Permission to reproduce In print Is restricted 10 an abstract of not more than 3W large amounts of solids, which lead to high damage rates in
word% illustrations may not k copied. The abstract must contain conspicuous
acknowledgment of Mere and by Mom the paper was presented. Write Librarian, SPE, P.O. these regions. Obviously, the injection-damage has a tendency
Sox 833836, Richardson, TX 75083-3836, U.S. A., fax 01-972-952-9435,
to level off the local rates and one may speculate that
injections end up with uniform rates along the wellbore. We
Abstract will show in the sequel that this is not the case. However, our
Nonuniform perforation properties and permeability variations main focus will not be on such local properties, but on the
close to the wellbore introduce heterogeneities which affect the question which influence the heterogeneities have on the
infectivity decline in water injection wells. We present a global injection properties such as the half-life and the
method to calculate these effects and derive results for a evolution of the overall damage rate.
simple model which incorporates the dependence of the water-
quality-ratio on the entrance porewidths. Common degrees of The Method
heterogeneity may lead to significant changes in the infectivity Generally, near wellbore damage may be described as a skin
decline, which cannot be described by conventional models of which depends on injection time or on the accumulated
uniform wells. Early damage rates may be orders of magnitude injection volume V. For steady state conditions one has:
larger than expected and wells with large initial skins are prone

‘“=‘i{ln[il+iv)}
to having their half-lifes within this early region. Due to the
Apwf (1)
resulting nonlinearities, extrapolations from test-injections may zz~~f

lead to wrong conclusions and it is usually not possible to

predict the half-life by the conventional use of a single value The heterogeneities near the wellbore generate a complicated
for the water-quality-ratio. flow pattern which depends on the details of the reservoir and
the completion. To arrive at a generic working model, we have
Introduction to simplify this to its essential features. Thus we introduce an
The injection of water is important for pressure maintenance, effective radius rh beyond which the flow is assumed to be
water-flooding projects and for the disposal of produced homogeneously distributed, independent of the coordinate
waters. One of its major limiting factors is the filtration of along the wellbore. Disregarding any crossflow for distances
suspended solids at or near the injection faces. As a convenient smaller than rh one may rewrite Eq. I as:
measure for the performance of an injection, Barkman and
Davidson [1] introduced the concept of its half-life. Despite of
(2a)
continuos efforts to improve the associated theory and
‘Pwf=%{’n(:)+p”(v)}
measurements, the prediction of these half-lifes remains very
uncertain, ranging between rather exact values and order-of- with the dimensionless damage pressure PD given by:
magnitude errors. This uncertainty is not restricted to
estimations based on laboratory measurements, but also holds
for injection-tests, which may show early damage rates that
pD(”)=&Rnt(”n*>
qnl>...) (2b)

differ substantially from the long-time behaviour.

With m = 1,..,N enumerating the perforations or layers, this is
Here we address the question, to which extent the
a set of N equations which describe the pressure drop along
inevitable heterogeneities of the injection faces may influence
parallel conducting layers with partial flow rates q~ and flow
the performance of water injections. Variations of permeability

631
2 T. HOFSAESS, W. KLEINITZ SPE 39586

resistances R~. The average rate per perforation is <q>=Q/N. concencentration for w instead of Barkman and Davidson’s
In general the flow resistances Rm may depend on the w/w concenmation.) Notice that the factor 2zhkf/N2 is entirely
accumulated injection volume V~ which has passed through due to the normalization employed in Eqs. 2 and 3. Thus the
the perforation, on the rate q~ and even on the full rate-history. quantities which lead to heterogeneous behaviour are the
The models which are currently employed in conventional perforation surface area AP,~ and the water-quality-ratio
uniform damage calculations lead to a resistance R~ which (w&)m.
depends linearly on V~ only: Due to the strong dependence on the injection area, even
(3) relatively small variations in perforation radius and length may
R“, = an, + bn,NVnt
give rise to substantial variations in b~. More importantly,
(The explicit factor N has been introduced for notational perforating may induce fractures around the perforations [6].
convenience, see Appendix A.) The growth parameters b~ Those which participate in fiow will effectively increase the
(dimension: l/vol) characterize the damage rate of individual area AP,~ [2] and lead to large variations in the growth
perforations, and the dimensionless constants am represent the parameters bm.
distribution of initial flow resistances. They are related to the It is well known that the water-quality-ratio w~, a: introduced
initial skin S.=S(0) and damage pressure PD(0) through: by Barkman and Davidson, does not only depend on the solids
concentration w and the type of solids, but also on the pore-
P~(0)=ln
()
~
w
+SO =fian,
(9)

A few remarks about the heterogeneity

(4)

radius rh are
size and -structure of the injection face [7, 8]. In fact, it is one
of the major problems for laboratory measurements of this
quantity, to use representative filter media or cores. One often
finds that W* varies like some power of the porewidth [8].
appropriate: For layered formations without crossflow For a produced water in southern Germany, for example,
(commingled), rh may become as large as re. But as a rule, it w~-( l/d)*”2 was found to hold over the complete investigated
will be of the order of a few wellbore radii r~ or perforation- range of porewidths from 2 to 50pm. Using the hydraulic
lengths LP.If the initial skin So is not too small, hS preCiSe radius concept to relate this to the formation permeability (at
value has little effect on the final solution. If So is negative, rh constant porosity), this corresponds to:
must be large enough to ensure a positive value of the
dimensionless pressure PD(0). For densely perforated wells it
will often be sufficient to use rW+Lp as the effective wellbore (6)
radius. If the initial skin is still negative, one has to incorporate
the improved permeability close to the wellbore in Eq. 1. In with n=O.6. So the structural and size variation of the entrance
some cases it may be necessary to use the full steady-state pores is the second major source which determines the
pressure solution to account for particular near-wellbore distribution of the growth parameters.
properties and to calculate the effective heterogeneity radius rh.
Sometimes it will be necessary to include crossflow effects as Internal filtration.
well (e.g. partially completed wells). Several authors have used theories of deep bed filtration to
The solution of the equations is given in Appendix A and is model the internal damage in water injections. Brief reviews
discussed later in the section on results. Our next step is to from different perspectives and further literature may be found
consider the initial resistances am and growth parameters b~ in [5, 9-11].
and to investigate how heterogeneity enters.
The filtration coefficient k (capture probability per length
of the filter medium) is a complicated function of many
Sources of heterogeneous behaviour
parameters for which only semiempirical expressions of
Cake Filtration
limited applicability exist [11 ]. In the simplest model one
As shown in [2] the large permeability contrast between
filter cake and the adjacent formation rapidly leads to the assumes L to be independent of the flow velocity and the
development of uniform flow across the perforation surface deposit that has already been formed. The solution for radial
with the filter cake growing in sheets of uniform thickness. For flow has been presented in [10, 12]. The important point to
elongated perforations this results effectively in a radial flow notice is that this solution holds for arbitrary flow rates, which
model [3]. One may then use Barkman and Davidson’s model is needed for the heterogeneous damage approach (Eq. 2b).
for open-hole wells [1], by simply replacing wellbore radius Using the results of [ 10] the growth parameter b~ is:
and height by the perforation radius and length. For ratios of 2tikf c
cake volume/perforation volume <0.5, a linear approximation bn, = — F(A.[rl),nJ (7)
N* A:,n, kn,Kn, )
is valid and the growth factor is:
Here km is the permeability in the filtration region and Km is a

[)
2tikf ~
b,,, = — — (5) dimensionless constant which parametrizes the reduction of
NzA;,n, kc “,
permeability per deposited particle concentration. The inlet
concentration c being the same for all perforations, it is again
The ellipsoidal model proposed in [4, 5] leads to the same
the injection surface AP,n, which has a major impact on the
result in its linear approximation. (We use a vol/vol

632
SPE 39586 INFECTIVITY DECLINE IN WELLS WITH NONUNIFORM PERFORATION PROPERTIES 3

distribution of the growth factors. The other two possible are generated both through the perforation geometry (area in
sources of heterogeneity are F(LrP,~) and krnKrn. b~ and penetration length and damage radius in am) and
l/L is the length scale for the depth of the deposit. It through their conductivity (w&, L, kK in b~ and crushed zone
changes with the geometrical and physiochemical properties and wellbore damage in the am).
of the porous medium. For shallow filtration, l/LcrP,~, the One certainly needs a better understanding of these
function F(x)=xexEz(x) [13] varies only between 0.4 and 1. For relations to properly account for the details of heterogeneous
regions in which the suspended solids may pass through the injection damage. Unfortunately, according to Brooks [20], the
commercially available simulators are not suited to study
rock without much filtration, 1/& is much larger and F
nonuniform perforation properties.
becomes small (F=&rP,~). Although the model of Eq. 7 cannot
really be applied to this case (having assumed filtration within
A Simple Model
radial flow around the perforations), it remains of course Although the general properties of the theory do not depend on
trivially true that the damage growth factor vanishes if the
the specific distributions of the initial resistances am and the
solids pass unfiltered through the rock. growth parameters b~, the use of a definite model is
The third contributing factor - k~K~ - is similar to the convenient to obtain an impression about the degree of
water-quality-ratio of cake filtration: Small values of Km heterogeneity at which significant deviations from uniform
strongly reduce the conductivity just like small cake behaviour occur.
permeabilities K do. As Km itself depends on the nearby native For simplicity we assume that the heterogeneous injection
permeability km as well as on the injection water, the factor behaviour is dominated by the local variations of the pore
C/kmKm in Eq. 7 should behave similar to wk in Eq. 6. structure according to the empirical relation given in Eq. 6. All
We should also mention that another common deep other possible sources of heterogeneity are ignored and the
filtration model uses a filtration coefficient with the dimension entrance-permeabilities are assumed to dominate over other
l/time. This corresponds to L-wv in our notation with v being factors which might contribute to the initial resistances. In
the flow velocity. Indeed, although the velocity dependence of particular, the perforation geometry is treated as being
perfectly uniform without any size-variations.
% varies from case to case, A is often found to depend on some
Introducing appropriate scales, the distributions of the a’s
inverse power of v [11]. However, the solutions of these
and b’s are given by:
models for radial flow, which have been presented in [4, 5, 9,
14, 15], and for more general models in [10], are only valid at k k“
constant flow rates. They must not even be applied to the a=a~ , b.p & (8a)
conventional uniform damage models at constant flowing
k () k
pressure, not to mention the locally varying flow rates and the water-qualtity-ratio is:
considered here.
(8b)
Initial Flow Resistances.
The initial flow resistances may roughly be separated into
the contributions of the near-wellbore permeabilities and those For mathematical convenience we have switched to the
of the perforations. continuous case, assuming that the number N of perforations is
A convenient measure for the heterogeneity of the very large (see Continuum Limit in Appendix A). Then the
permeability close to the wellbore is the Dykstra-Parsons index m is replaced by a continuous argument x, which is
coefficient used in waterflooding [16-18]. Depending on the dropped in the following presentation (a(x), b(x), k(x) in Eq.
specific field and well this alone may introduce a large spread 8a). Appendix C shows how to fix the overall size scales ct and
in the distribution of the am.
P.
Whereas quite detailed information is available from The entrance-permeabilities are assumed to be log-
numerical work on uniformly perforated wells [19, 20], the
normally distributed which is a simple yet realistic
flow resistances of nonuniform perforations are not well approximation to actual vertical permeability distributions, Its
understood today. One should however expect that the most log-mean permeability is k~ and the degree of heterogeneity is
important contributions retain their role also in the
conveniently characterized by the Dykstra-Parsons coefficient
heterogeneous case. At fixed uniform shot density and phasing
KDP [16, 18]. Its value is in the range 0.5 SKDPS0.9 for most
this would give rise to a flow resistance which depends only
reservoirs [17].
weakly on the perforation diameter, is roughly proportional to
In this simple model there is no difference between the
the inverse perforation length and has the usual additive term
entrance permeabilities and the permeabilities of the formation
for the crushed zone and corrections for wellbore damage
close to the wellbore. In real cases they are of course different
depending on wether the perforation terminates inside the
and this will relax the strict correlation between the initial
damaged zone or penetrates through it [ 19].
resistances and the growth parameters in Eq. 8a. We have also
Apart from giving an important contribution to the initial
tried out completely uncorrelated distributions of the a’s and
flow resistances, the perforation properties will also introduce
b’s (not presented here) to check that the general results do not
correlations between the distributions of the am and b~. These
depend on this feature.

633
4 T. HOFSAESS, W, KLEINITZ SPE 39586

The decreasing sensitivity to heterogeneous behaviour with

Results stronger wk-dependence on permeability reaches its
The course of the injection is given by the evolution of the minimum at n=2 and then turns around to increase with n. This
dimensionless damage pressure PD as a function of the ‘critical’ value of n also holds for other distributions, but we
injection volume V in Eq. 2a. It is valid both at fixed rate and will not pursue this point any further here.
at fixed flowing pressure. The general shape of PD(V) may be The large initial darnage rate which may result from
inferred from the solution given in Appendix A and is shown heterogeneous wellbore properties is particularly important if
in Fig, 2. Unlike the homogeneous case it is no longer linear in one wants to estimate the injector half-life from a test-
the injection volume, but it approaches linear behaviour for injection. The linear extrapolation of a large initial damage
large volumes. Stinting from its initial value pD(0) it moves rate may be completely misleading. If heterogeneity is the
within an inclined strip of width A which has the final large- cause, results will be much too pessimistic and in some cases a
volume slope. The expression for A and the limiting forms for later turnaround to smaller damage rates might still render the
small and large volumes are supplied in Appendix A. injection feasible. On the other hand, if infectivity has already
Notice that the damage rate (the slope in Fig. 2) may be very declined too much, and stimulations are considered, proper
large at the beginning of the injection and decreases diversion techniques may be needed to avoid a simple
monotonically to its final value at larger volumes. The reason repetition of the same rapid decline.
for this behaviour has already been mentioned in the
introduction: It occurs if initially a large portion of the flux is Half-Life Volume.
directed into a highly conductive region. The associated small Having ascertained that heterogeneity can lead to very
injection area leads to a high damage rate. As the conductivity large initial damage rates, we have to estimate how long the
of this region decreases, parts of the flux are diverted to other early nonlinear behaviour may last. To achieve this, it is
perforations. As a result the effective injection area grows, necessary to locate the half-life of the injector on the
resulting in a smaller damage rate. The final partial flow rates dimensionless damage pressure curve P~(V).
q~ at large volumes are solely determined by the growth The half-life is defined as the injection time (or volume) at
parameters b~. They do not become uniform, but still depend which the rate declines to 1/2 of its original value (for
on the individual b~-values (Eq. A-1 I ). constant-pressure-injection) or the flowing pressure reaches
The degree of nonlinearity in P~(V) depends on the twice its initial value (for constant-rate-injection). According
properties of the distributions. Fig. 3 illustrates this for the to Eqs. 1, 2 the half-life volume Vln may be expressed as:
model of Eq. 8 with n=O.6 and some values of the Dykstra-
‘D(~/2)– ‘D(o)= ln(re/rw)+sO (lo)
Parsons coefficient KDP. To clearly expose the shape of the
curves, the axes have been scaled to unit initial resistance and An approximate criterion for separating the early nonlinear and
unit slope at large volumes, It is evident that for KDP above late linear parts of P~(V) may be read from Fig. 2. For
approximately 0.7 the effects of heterogeneity will become
PD(V1n)<P~(0)+A the half-life volume is in the nonlinear
important, u~dess they are restricted to small injection
region and for significantly larger values it will be close to the
volumes.
linear region. Then with PD(0) from Eq. 4 and the results of
Before we turn to discuss this question, we will have a look
Appendix A, the condition for Vln to be located in the early
at the implications of early nonlinear versus late linear
nonlinear region of the dimensionless damage pressure curve
injection behaviour.
is:
Ratio of early to late Damage Rates. ln(r./rw) + SO < _=A
With the results of Appendix A, the ratio of early to late W_, (11)
ln(~, /rW) + so - PD(()) I b 2
damage rates is: ( IT)

E dpD/dVlv=o (/)
b as 1 2 Again the right side of the equation is independent of the
—~ (9) absolute sizes of the a and b and depends on the distributions
L dpD/dv[v+. ‘~ ()~ only. Fig. 5 shows A/PD(0) for our simple model. If the value
This ratio is independent of the absolute sizes of a and b, such on the left side of Eq. 11 can be estimated for a specific well,
as the a and ~ in Eq. 8a and depends on the properties of the one can read from Fig. 5 at which degree of heterogeneity KDP
and at which exponent n of the water-quality-ratio the half-life
distributions only. Fig. 4 illustrates this for our simple model.
volume will be located in the early nonlinear regime.
The ratio may become very large at common values of the
For small values of the skin this may require rather
Dykstra-Parsons coefficient. At fixed degree of heterogeneity
accurate etsimations of rh, but the equation may be readily
(K~p), smaller values of n lead to larger early damage rates.
employed for wells with a high initial skin, such as those
This is due to the fact that a water with a weak dependence of
w/~ on the entrance permeability (small n) cannot profit much presented by Sharma and Pang in [12] with values of SO as
large as 200 and 400. Then the left side of Eq. 11 is close to its
from the high permeability tail of the distribution. Thus the
large flux which is initially directed into this region, minimal value 1 and a relatively small degree of heterogeneity
is sufficient to satisfy the inequality. Thus wells with a high
deteriorates rapidly with a high damage rate.
initial skin will be prone to reach their half-lifes in an early

634
 -.——

SPE 39586 INFECTIVITY DECLINE IN WELLS WITH NONUNIFORM PERFORATION PROPERTIES 5

period of rapid infectivity decline. approximation of Eq. A-9 holds. Then VID is given by:
If the half-life volume is located in the early nonlinear
regime, how does the large initial damage rate affect the value ~,2=($={ln(~)+So-A} ,V+~ (15’)
of Vln? Inserting the small-volume expansion of Appendix A
into Eq. 10 this may be estimated as:

&,2=
{()} In $
w
+SO w~ ~j (12)
with A from Eq. A-12. Thus the effective growth parameter at
large volume is b.f~ l/<b-*n>2, If it is admissible to disregard
variations of the perforation-areas in Eq. 5, this tells us, how
(/)
the average water-quality-ratio has to be calculated:
This depends on the volume-scale inherent in b. To expose it
in a scale-independent way, one may divide it by the half-life
, V+ca (16)
volume of a uniform injector to obtain:

(v/2),e,,rog.(1/~)3 ~b

The average has to be taken with respect to the properties of

(~lz)unyoml ‘ba3
m (13)
the injection faces in the wellbore. If these can be estimated,
measured relations for the water-quality-ratio like e.g. Eq. 6
Notice that this only applies if the inequality of Eq. 11 holds. may be used to calculate the effective w%. However, it is
Here b. is the growth factor of the uniform injector. With the important to check the validity of the large-volume limit before
simple model of Eq. 8 one has b“=~. The result is shown in employing it. As the example in the next section shows, its
Fig. 6. The reduction of the half-life volume may be dramatic indiscriminate use may lead to large errors in the calculated
and reach an oder of magnitude already at rather modest half-life.
degrees of heterogeneity.
The effects of heterogeneity change if the half-life volume A Simulated Example
of the injector is located on the late linear portion of PD(V). Fig. 8 shows the development of the flowing pressure for a
Inserting the large-volume approximation of Appendix A into simulated injection at constant rate with the model of Eq. 8 at
Eq. 10, the ratio of half-life volumes of heterogeneous / n=O.6 and KDP=().8. The remaining data are supplied in Table
uniform injectors is: 1. An open-hole well was used, because the simple model
includes no distribution of the perforation area. Hence the
(v/z)h,tero,,
~b ~ 2 growth parameter bin Eq. 5 depends only on the total injection
(14) area and one may switch to a perforated well simply by using
(y/2)unfornl u ()w perforation data with the same total injection area. Appendix C
The results for our simple model are presented in Fig. 7. explains how the model parameters a and ~ are calculated
Pronounced effects occur only at much larger degrees of from the supplied data.
heterogeneity. Actually the effects may be larger than shown in We used 2 values for rh (4rW and 8rW) to check that the
the figure, as we have neglected a correction term proportional results are not very sensitive to this value. The dotted
horizontal line in Fig. 8 shows twice the initial flowing
to A in Eq. 14 (see Eq. 15 below). Notice that now the half-life
pressure. The half-life volumes resulting for rh/rW=4 and 8 are
volumes are larger than in the uniform case. Also, at fixed
16100 and 14800 m3. Although the pressure curves look
degree of heterogeneity (KDP), a strong dependence of the
almost linear here, they are still very far from the large-volume
water-quality-ratio on the inlet permeability (large n) is more
regime. This is evidenced by their distance from the
beneficial than a small one and there is practically no change
asymptotic large-volume lines in the upper part of the figure.
in VID for n<O.2. The reason for this is the same as the one
Actually, the dimensionless flowing pressure at the half-life is
given below Eq. 9: An injection with uniform inlet_
close to PD(0)+A, the value which was used above as an
permeability (kM in our model) will only experience the fixed
water-quality-ratio at this permeability y. Compared to this, a approximate criterion to locate Vln in the early part of the
water with large n will benefit from the permeability damage pressure curve (see Eq. 11). Thus an estimation of the
distribution and cause less overall damage. half-life by the large volume limit is not possible. Using Eq. 15
One should notice that this mechanism is a consequence of this would result in Vln=3530m3 for rh/rW=4 and even in a
our model-assumptions and may not be operative in general. In negative value for rh/rW=8 (see Fig. 8). Likewise the
the simple model presented here, the regions of large initial approximation for small volumes of Eq. 12, shown as the
permeability coincide with the regions of large entrance pore steeply rising straight line in Fig, 8, produces the much too
width. A different situation arises if regions with large near- small value Vln=1700m3. Hence one has to use the full
wellbore permeabilities are fed by perforations with crushed nonlinear form to determine the proper half-life.
zones of narrow entrance pore-throats or if the variations of the Fig. 9 shows the half-life volumes as they might be
effective perforation sizes are taken into account. predicted from measuring the water-qualitity-ratio using
“representative” cores. As these we chose several different
Averaging the Water-Quality-Ratio average permeabilities: the harmonic mean kH= 1/< Ik, the
For sufficiently large half-life volumes, the large-volume log-mean kM=eln(k)>, the arithmetic-mean kA=Ck> and the

635
 6 T. HOFSAESS, W. KLEINITZ SPE 39586

average permeability kf of the formation. The measurements a = initialflo w resistance, dimensionless

would produce w&-values as in Eq. 8b, so we calculated them b = growth factor ofjlow resistance, llvol
by inserting the different values for k. The harmonic- and log- c = solids concentration, vollvol
mean cases give the best results. However, this is likely to h = wellbore height
change from example to example, because the calculation of K~P = Dykstra-Parsons coefficient
the half-life with any single (or averaged) water-quality-ratio k = permeability
assumes that the damage pressure depends linearly on the k~= average formation permeability
injection volume. So, although the resulting straight line in the L,, = pe~oration length
pressure-volume-diagram may occasionally hit the actual N = number ofpeforations or segments
pressure curve at the correct V,n, it cannot account for the n = constant, Eqs. 6, 8
parts at smaller and larger volumes. Thus unless the wellbore PD = dimensionless damage pressure
is sufficiently uniform or the large-volume-limit of Eq. 15 p = pressure
applies, one has to consider the nonlinear behaviour of the Q= volumetric rate
infectivity decline. q = partial volumetric rate
r = radius
Applicability rl, = effective radius of heterogeneity
All input-values which were used in the simulated example S = skin
may be measured or estimated for a real well. The only SO= initial skin
nonconventional values are: 1) The exponent n in the model- V = volume
equation 8b, which may be obtained from measurements of the w = solids concentration, vol/vol
water-quality-ratio. 2) The permeability distribution close to a = constant, Eq. 8a
the wellbore with its mean value and width (kM and the ~= constant, Eq. 8a
Dykstra-Parsons coefficient KDP). Approximations to this can ApWf = wellbore$owing pressure
be obtained with logging methods.
A = filtration coeflent, I/length
The simple model used here was introduced mainly for
illustrational purposes. The necessity to distinguish between p = viscosi~
entrance porewidths and the near-wellbore permeabilities, and <> = average, expectation value
to include the size-variations of the perforations has already
been pointed out. In some cases the simple model may be Subscripts
applicable, though. This is likely to be the case, if initially the e= drainage (radius)
local flowrates in the well are strongly correlated with the M = log-mean value
permeability distribution close to the wellbore. However, one m = enumerates pe~oration or wellbore segment
certainly has to investigate more detailed models to arrive at p = pe~oration
reliable predictions of injection performance for heterogeneous u = uniform
wells. w= wellbore
Y2 = half-l$e
Conclusions
1. Variations of the permeability close to the wellbore and References
nonuniformity of perforation properties are sources of 1, Barkrnan, J.H.; Davidson, D.H.: “Measuring Water Quality and
heterogeneous injection behaviour. Predicting Well Impairment” JPT (July 1972), 865-873.
2. Hofsaess, T,; Kleinitz, W.: “Modelle der Filterkuchenbildung
2. Common degrees of heterogeneity (e.g. values of the
bei Wasserinjektionsbohrungen - On the Growth of Filter Cakes
Dykstra-Parsons coefficient) may change the injection during Water Injections” Erddl Erdgas Kohle (September 1994),
behaviour significantly. 359-364.
3. Early damage rates may be orders of magnitude larger 3. Hofsaess, T. “Uberarbeitung und Weiterentwicklung von
than expected from conventional estimations. hrjektionstheorien und von Methoden zur Qualit~tsbestimmung
4. The half-life may be reached within the period of large von Injektionswtissern” Preussag Energie GmbH, internal report
early damage rates. Wells with high initial skins are I990.
particularly prone to show this behaviour. 4. Pang, S.; Sharma, M,: “Evaluating the Performance of Open-

5. For the same reason the linear extrapolation of test- Hole, Perforated and Fractured Water Injection Wells”
SPE30 127, presented at the SPE European Formation Damage
injections may lead to much too pessimistic estimates of the
Conference, the Hague, 15-16 May 1995,489-498.
injector performance.
5. Pang, S.: “Colloid Retention and HOW in porous Media”,
6. Due to the nonlinearity induced by heterogeneous Dissertation, Univ. of Texas at Austin, Aug. 1996.
behaviour, it is usually not possible to predict the half-life by 6. Pucknell, J.; Behrmann, L.: “An Investigation of the Damaged
the conventional use of a single value for the water-quality- Zone Created by Perforating”, SPE228 11, presented at the 66’h
ratio. Annual Technical Conference and Exhibition in Dallas, TX,
Oct. 6-91991,511-522,
Nomenclature 7. MitchelI, R.; Finch, E.: “Water Quality Aspects of North Sea
A = area Injection Water” JPT (June 1981), 1141-1152,

636
 ..-— . —.—. .— .——.

SPE 39586 INJECTIVITY DECLINE IN WELLS WITH NONUNIFORM PERFORATION PROPERTIES 7

8. Kleinitz, W,: “Kontrolle von Injektionsw5ssem in PD(V)= l/(1/R) (A-3)

Erdblaufbereitungsbetrieben - Controlling the Quality of
Injection Water in Oiltields”, Erd61 Erdgas Kohle (June 1988),
256-262.
9. Pang, S,; Sharrna, M.: “A Model for Predicting Infectivity
Decline in Water Injection Wells” Paper SPE28489 presented at
v=
()
;(R-a) (A-4)

the 69*hAnnual Technical Conference and Exhibition in New With <..> denoting the averages:
Orleans, 25-28 Sep. 1994,275-284.
10. Hofsaess, T.; Kleinitz, W.: “Infiltration von Feststoffen in
Wasserinjektionstrtiger - Infiltration of Solids into Water
=+x...
(...) nr
(A-5)

Injection Reservoirs” Erd61 Erdgas Kohle (September 1995),

363-370. G plays the role of an auxiliary variable, which may be used to
11. Wennberg, K,; Sharma, M.: “Determination of the Filtration generate the R~ in Eq. A-2 and calculate the corresponding
Coefficient and the Transition Time for Water Injection” values of PD and V. It enters explicitly in the equation for the
SPE3818 1, presented at the SPE European Formation Damage time dependence of injections at constant wellbore flowing
Conference, the Hague, 2-3 June 1997,353-364,
pressure: Using Q=dV/dt, the integration of Eq. 2a gives:
12. Sharma, M.; Pang, S.: “Infectivity Decline in Water Injection
Wells: An Offshore Gulf of Mexico Case Study” SPE381 80,
presented at the 1997 SPE European Formation Damage
Conference, the Hague, 2-3 June 1997,341-351.
13. Abromowitz, M.; Stegun, I.: “Handbook of Mathematical
Functions” Dover publications, New York, 1965,
‘=2~h4Apwf{1n(:)+ (A-6)

14. Van Velzen, J,; Leerlooijer, K,: “Impairment of a Water Continuum Limit. For analytical calculations it is convenient
Injection Well by Suspended Solids: Testing and Prediction” to assume a very large number of perforations or wellbore-
SPE23822 presented at the SPE Intl. Symposium on Formation
segments and to take the continuum limit N+oo. Ignoring
Damage, Lafayette, LA, February 26-27, 1992,
15. Van Oort, E,; van Velzen, J.; Leerlooijer, K,: “’’Impairment by correlations, the positions of the perforations in the wellbore
Suspended Solids Invasion: Testing and Prediction” SPEPE play no role. In particular, angular positions do not enter and
(August 1993), 178-184, one may use a dimensionless depth x=fi. The spacing of
16. Dykstra, H.; Parsons, R.: “The Prediction of Oil Recovery by perforations or segments is 1~ and the continuum limit is
Waterflood”, Secondary Recovery of Oil in the United States, obtained by:
2nd cd., API, New York (1950), 160-174.
17. WiI1hite,G.: “Waterflooding”, SPE Textbook, Dallas (1986). Nq~, + q(x), NV~, + V(x), (...)+ ~;&..,
18. Lake, L.; Jensen, J.: “A Review of Heterogeneity Measures Used
in Reservoir Characterization”, Unsolicited paper SPE20156,
1989.
19. Karakas, M.; Tariq, S.: “Semianalytical Productivity Models for (A-7)
Perforated Completions” SPEPE (February 1991), 73-82,
20. Brooks, J.: “A Simple Method for Estimating Well Productivity” The growth rates b~ given in Eqs. 5, 7 contain a factor
SPE38 148, presented at the 1997 SPE European Formation (NAP,~)2. To take the continuum limit one may replace:
Damage Conference, the Hague, 2-3 June 1997,57-64.
21. Paul, G.; Lake, L.; Pope, G.; Young, G,: “A Simplified NAP,., + As(x)
Predictive Model for Micellar-Polymer~Flooding” SPEj0733
presented at the 1982 California Regional Meeting, San with the total injection area A=Ncb, and AP,~/cA>+a(x)
Francisco, March 24-261982. being the relative injection area per perforation.

Appendix A - Solution of the Model Equations Small and large volume solutions, Expanding Eqs. A-2 to A-
General. We consider the flow resistances Rm of Eq. 3. The 4 for small and large G, the leading orders for small and large
explicit factor N (m= 1,,.,N) has been introduced for notational injection volumes are obtained as:
convenience, as it allows to keep the same a, b in the limit
(/)
v,
3
N+CO (see below). Using qml<q> = Nqm/Q = dV~ldV, Eq. 2b
may be integrated to give: ‘D+ (I;a)
+
——
;,:)’ V+o (A-8)

G(V) = ~OvdV’P~(V’ ) = NVn,(an, + ~ bn,NVn,) (A-1) And for V+CQ;

The same method of integration may be employed for any R~ PD + * , {V+(a/b)} (A-9)
which depends on V~ only. Solving Eq. A-1 for V~: 1
( /m)
R.,(V) = ~2G(V)bnl + a:, (A-2)
(A-1O)
Inserting this into Eq. 2, dividing by Rm and taking the
average, one obtains:

637
 —. —=

8 T. HOFSAESS, W. KLEIN!TZ SPE 39586

Table 1: Numerical values for simulated example

‘n’ “* (A-11)
General data: Watec
w/kc = 1 ppm/md at 50 md
Q = 300 mu/d n = 0.6
The quantity A referred to in Fig. 2 and Eq, 11 is obtained
r~ = 200 m p = 1 mPas
from Eqs. A-8, A-9 as:
rW= 0.1 m Distribution of inlet permeabilities:
A_ (a/~) .— 1
(A-12) h=lOm k~ = 50 md
(l/m)’ (Ifs) kf = 250 md K~P = 0.8

I
Appendix B - Log-Normal Distribution
The probability density of the log-normal distribution is

-.
—
Gaussian with respect to x=In(k/kM). Hence the expectation
values are:

(~(k))’*

(k”)= k~exp(n202/2)
The Dykstra-Parsons
~~~(k~eX)eXP(-X2i202)

coefficient
a

This gives the log-mean-value Cln(k)>=ln(kM). Using Eq. 8a

the small and large volume limit of Eqs. A-8, A-9 is easily
obtained from:
(B-1)

(B-2)

KDP is related to the standard

~-1u’
~
“-. “-” “;‘“
.:
d
w ,~
- “- . .
-. ..—,—
--—..-_
Aug. 91
, . I
~ti
deviation o by [21]: E 0.2
KDP = I – ~-” (B-3) Fig.1: Flow distribution at start and after 9 months of a water
Injection (40,000m3) into the Dogger-beta sandstone of the
To obtain PD and V, Eqs. A-3, A-4 are inserted into B-1 and Gifhorn trough.

the integrations are performed numerically for different values

of G.

Appendix C - Simulated Example

To calculate the values of a and ~ for the model-equation 8a,
one may proceed in the following way: One rewrites b~ of Eq.
5 as:
2nhkj w
b=——
A2 kc

with the total injection area A. Using the values of Table 1, b

may be calculated at k=kM, Using Eq. 8a this value is
identified as ~. With Eqs. A-8 at V=O and 8a, B-2:

‘&=& =aexp
(-“22
/)
‘D(o)
To obtain u, &is calculated from KDPvia Eq. B-3 and P~(0) is
obtained from the first Eq. 4.
n PD(());

Fig.2: Shape of the dlmensionleas pressure

v
damage curve
(general solution). The dotted line shows the approximate
criterion for separating small- and [erge-volume behaviour.

638
SPE 39586 INFECTIVITY DECLINE IN WELLS WITH NONUNIFORM PERFORATION PROPERTIES 9

20

16

12

0
0.2 0.4 0.6 0.8 1

o -v 1 Dykstra-Parsons coefficient K~p

Fig.3: Dimensionless pressure damage curves for the model of Fia.4: Ratio of early to late damage ratea for the model of Ecr; 8.
Ea. 8 with n=O.6. Axes scaled to unit Pn(0) and unit sloDe at large Fr~m left to righk n= 0.2,0.4,..., 1~
V. “From top to bottom: KDP = 0.95, 0.9, 0;8; 0.7, 0.5, 0. -

0.5 0.7 0.9 0.5 0.75 1

Dykstra-Parsons coefficient KDP Dykstra-Parsons coefficient”
Flg.6: Reduction of half-life volumes by heterogeneity for the case
Fig.5: Indicator for half-life being located in the early nonlinear
that the heterogeneous half-life is reached in the early nonlinear
part of the pressure damage curve. Model of Eq. 8. From left to
damage phase. Model of Eq. 8. From top to bottom: n = 1, 0.8, ....
right n = 0.2, 0.4,..., 1. See text below Eq. 11 for details.
0.2.

639
 ——

10 V. HOFSAESS, W. KLEINITZ SPE 39586

Vln in late part

I

.
7::
r ,,
r
,,
,,
,,
,,
I r ,,
,,
II F ,,

0.7 0.8 0.9 1

v .!

1j V/1000m3

Dykstra-Parsons coefficient KDP o 30

Fig.7: Increase of half-life volumes by heterogeneity for the case Fig. 8: Pressure versus accumulated injection volume for the
that the heterogeneous half-life is reached in the late linear simulated example. rtir~ = 4 and 8. The straight inclined lines are
damage phase. Model of Eq. 8. From top to bottom: n = 1, 0.8, .... the large volume solutions which are reached asymptotically. The
0!2. steeply rising line at the left is the small-volume approximation.
Dotted lines show the half-life-values.

I VI12/1000 m3 x“

Fig.9: Half-life volumes for the simulated example. k. indicates

the permaabillty which was used to estimate VIH in the
conventional uniform model. H = harmonic mean, M = log-mean,
A = arithmetic mean, f = avarage formation permeability.

640