You are on page 1of 12

SPE 82231

30 Years of Predicting Injectivity after Barkman & Davidson:


Where are we today?
T. Hofsaess, SPE, Corepep, Consulting in Reservoir and Petroleum Physics; W. Kleinitz, SPE, Preussag Energie GmbH

Copyright 2003, Society of Petroleum Engineers Inc.


predictions [4]. An alternative strategy would be to employ
This paper was prepared for presentation at the SPE European Formation Damage Confer- core samples for measuring and predicting injectivity behav-
ence to be held in The Hague, The Netherlands 13-14 May 2003.
iour. Because of the lack of documented success with this ap-
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
proach, the validity and accuracy of the proposed models is
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to not ascertained. Some reviews on models and experiences can
correction by the author(s). The material, as presented, does not necessarily reflect any posi-
tion of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE be found in Vetter et al. [4] and Todd et al. [5].
meetings are subject to publication review by Editorial Committees of the Society of Petroleum
Engineers. Electronic reproduction, distribution, or storage of any part of this paper for com-
For many years, the models and methods of Barkman and
mercial purposes without the written consent of the Society of Petroleum Engineers is prohib- Davidson have been applied systematically at 'Preussag Ener-
ited. Permission to reproduce in print is restricted to an abstract of not more than 300 words;
illustrations may not be copied. The abstract must contain conspicuous acknowledgment of gie' to the planning and interpretation of reinjection projects.
where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836,
Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
In view of the aforementioned problems and the authors’ own
experiences, the theory of Barkman and Davidson was criti-
cally analysed and advanced. For this purpose we adopted a
Abstract pragmatic approach analysed the behaviour of individual in-
The reinjection of produced water is an important prerequisite jections in detail and compared them to model predictions.
for maintaining pressure and for flooding reservoirs. Thirty The present article summarizes these developments, provides
years ago Barkman and Davidson introduced the first compre- hints for application and indicates unresolved problems. Other
hensive theory for predicting the behaviour of injection wells. approaches to the further development of injection predictabil-
This theory was employed and tested for many years in actual ity and the model of Barkman and Davidson may be found e.g.
injection projects at 'Preussag-Energie', including the develop- in [6] and [7].
ment of a special-purpose instrument. In the present article, a
survey of operational experience is given, and necessary ex- Barkman and Davidson: A short review
tensions of the model are delineated. Furthermore, the article Barkman and Davidson [1] assume that solids suspended in
highlights some important aspects which should be considered the water will be deposited as a filter cake above the sandface
in laboratory investigations. or other filter media. They consider this process, known as
'cake filtration', at constant injection pressure (p = const).
Introduction Then, the temporal development of injectivity is given by the
Water is injected into porous layers for pressure maintenance deterioration of the injection rate Q(t). The authors normalize
in oil reservoirs, in geothermal projects and more recently for this development by the initial rate Q0 and introduce the di-
the disposal of brine from the salt caven leaching process. mensionless decline parameter:
This may lead to several severe problems. One particular Q(t )
source of danger are the solids which are suspended in the α(t ) = ................................................................. (1)
water. If they are deposited above the sandface (cake filtra- Q0
tion) or within the formation (deep bed filtration), the injectiv- As a simple indicator for the injection behaviour they intro-
ity may be reduced dramatically. This leads to increasing in- duce the half-life t½, at which the rate Q has declined to one-
jection pressures and consequently to costly stimulations or half of its initial value Q0: α(t½) = ½. Figure 1 shows an exam-
even to irreversible wellbore damage. ple for the decline curve α(t) and the half-life.
The first comprehensive and apparently most often quoted For injections, four different scenarios are considered:
publication which deals with the requirements for the quality Wellbore narrowing: The filter cake is deposited above the
of injection water and which presents models to predict injec- sandface of an open hole and thus reduces the effective
tivity, was presented by Barkman and Davidson some 30 years wellbore radius.
ago [1]. Although many articles have been published on this Perforation plugging: The filter cake is deposited at the tip
topic since then, the present state of its predictive power is of perforations and thus reduces their effective lengths.
rather poor. It is a frequent experience that predicted lifetimes Invasion: The filter cake builds up inside the formation at
or half-lives of injection wells are orders of magnitude below an unpredictable 'radius of invasion' and grows towards
the actually observed values [2, 3]. The authors' recommenda- the sandface.
tion to determine the model parameters by membrane filtra-
tions has often been considered to be the cause of these false
2 SPE 82231

Wellbore fill-up: Solids suspended in the water undergo a [9]. Hydroguard filters water through cellulose-acetate mem-
gravity separation and thus reduce the thickness open branes at constant pressure. The system is equipped with a
for flow. microcontroller which allows checking of the WQR at freely
These scenarios – in particular perforation plugging and inva- chosen intervals of time.
sion – are discussed further in section 5. As already mentioned above, the model of Barkman and
For each of these modes of deposition, the authors supply Davidson often underestimates the half-lives of injection wells
solutions for the injectivity decline α(t). For illustrational pur- by several orders of magnitude. In the pertinent literature this
poses, this function is reproduced in table 1a for the case of is often claimed to be due to the determination of the WQR
wellbore narrowing. The result is rather complicated, mainly through membrane tests. This overall reasoning certainly
because a closed-form solution can only be provided implic- makes sense in view of the vastly different pore widths and
itly as t(α). pore structures of membranes and formations. Among other
The development of injectivity is determined essentially by things, this topic will be discussed further in the sections
two parameters: The half-life increases along with the perme- to follow.
ability contrast kc/kf between filter cake and formation. The
other essential parameter is given by the water quality ratio Course of injections for arbitrary process control
(WQR). Barkman and Davidson define it as: The results of Barkman and Davidson hold for constant pres-
w sure only. Real injections are usually rate-controlled, however,
: Water Quality Ratio, WQR .....................................(2a) running at constant or piecewise constant rate. The models of
kc
Barkman and Davidson have been generalized to arbitrary
Here w is the weight/weight concentration of suspended sol-
process control in [10]. (The free variability of pressure and
ids, and kc is the permeability of the filter cake. As a rule, not
rate refer to the derivation of the model equations. A possible
all suspended solids are separated by filtration; hence, one
dependence of water and cake parameters on these conditions
should rather refer to w as the concentration of total filterable
has not been considered.)
solids (TFS) [8].
In order to obtain uniform solutions for arbitrary variations
Applying the models of Barkman and Davidson, we found
of pressure and rate, it is necessary to represent the injectivity
a slightly different definition of the water quality ratio to be
decline in terms of the accumulated injection volume. Then
more advantageous (see the remarks below eqns. 3 and 6b):
the injectivity development of equation 1 may be
w ρW generalized to:
Ω= : Water Quality Ratio, WQR .......................(2b)
k c ρc (p F Q )V=0
α= .............................................................. (4)
ρW is the density of water, and ρc is the "bulk density" of the (p F Q)V
cake, that is, weight/volume of solids plus water within
the cake. ρF is the flowing pressure, which is responsible for the trans-
The values of w/kc and Ω are usually expressed in units of port of water (see eq. 5 below). As an illustration, the solution
ppm/mD. Decreasing values of w/kc and Ω signal increasing for wellbore narrowing is provided in table 1b.
injectivity prospects. For the operational design of injections, This representation has the further advantage that the solu-
it is still common to rely on the solids concentration w alone. tions of Barkman and Davidson at constant pressure assume a
This may, however, be misleading: As w/kc and Ω represent simpler mathematical form as well. This is due to the fact that
the ratio of the filterable solids concentration w to the perme- the injection volume is a "natural" parameter for cake filtra-
ability of the filter cake kc, the latter quantity may be just as tion, the amount of deposited cake being proportional to this
important. As an example, the deposition (precipitation) of volume (the concentration of filterable solids being the pro-
ferrous hydroxide will lead to a gel-like cake with extremely portionality constant). Correspondingly, the curves α und 1/α
low permeability. This may easily lead to water quality ratios plotted against the injection volume are independent of the
> 500 ppm/mD even for only slightly charged water [2]. For process control. However, the associated time scale in figure 1
such WQR values extreme injectivity problems ought to holds for constant pressure only. (At constant rate, the time
be expected. scale is of course simply proportional to the volume
To determine the WQR of water, Barkman und Davidson scale: t = V/Q.)
suggest laboratory measurements employing membrane filtra- In analogy to the half-life, the injection volume V½ =
tions. As cake filtration proceeds, the filtration volume V will V(α=½) may be called the half-volume. Both half-life and
grow linearly with √t. Plotting V-√t in a diagram (s. fig. 2a), half-volume refer to the same point of the process. For the
the slope S may be used to calculate the WQR as: special cases of constant flowing pressure pF or constant rate
Q, the rate Q has been halved or the pressure has doubled at
w 1 2ρ c A 2 ∆p 1 2A 2 ∆p the half-volume (see fig. 1).
= ; Ω= .......................(3)
k c S2 ρ W µ S2 µ If one injects at constant pressure, the function α nicely
(A: area of membrane, ∆p: pressure across the membrane, µ: shows the course of injection, as it depicts the decline of the
viscosity). The corresponding equation of Barkman and normalized rate. For an injection at constant rate, however, 1/α
Davidson [1] is misprinted with ρc und ρW misplaced. Defin- is more useful, as it shows the increase of normalized pressure
ing the WQR in terms of Ω in the second equation is advanta- (see fig. 1).
geous, because the density ratio ρc/ρW need not be known. To normalize these functions or the half-life or -volume,
To conduct measurements of WQRs, the special-purpose the pressure law at the start of injection (see eqns. 1, 4) has to
instrument "Hydroguard" was developed by 'Preussag Energie' supplied. A simple form is:
SPE 82231 3

 pF  µ   re   The density ratio ρW/ρc does not appear if the WQR is


  = ln   + S0  (5) specified by Ω (eq. 2b). The evaluation of the WQR by
 Q  V =0 2πh w k f   rw   means of membrane filtrations is independent of ρW/ρc as
Barkman and Davidson [1] always assume a vanishing initial well (see eq. 2b).
skin S0 (see table 1). For perforated or predamaged wells this The increase of pc/Q shows a linear dependence not just on
may underestimate the half-life by far. To avoid normalization the volume, but on the WQR Ω as well.
errors, it may be appropriate to consider the course of absolute The linear approximation (volume of cake << cavity of
pressures or rates, instead of the normalized (relative) course. injection) in equation 6b holds universally for all injection
There is another convincing reason for considering abso- geometries (see section 5).
lute values: As may be inferred from the results given in table Figure 3 shows the evolution of the pressure during injection
1, the half-life and -volume (t½ and V½ at α = ½) are strongly at two different values of the WQR [11]. After the injection of
influenced by the permeability contrast kc/kf between cake and 70·103 m³ the maximal wellhead pressure of 100 bar is
formation. The smaller the permeability contrast, the smaller reached. Subsequently the borehole was stimulated with the
are half-life and -volume as well. It is interesting to consider use of a special solvent. After the second stimulation, the wa-
this for a constant permeability of the filter cake: If the perme- ter quality was improved by employing a flotation tank (IAF)
ability of the formation increases, the half-values decrease! and by decreasing the pH value of the water to 5. This resulted
This behaviour appears to be counterintuitive. However, the in a significant reduction of the solids content and of the re-
reason for this observation is simple: Half-life and -volume sidual oil content in the injected water. The consecutive injec-
refer to the point of the process at which the pressure drop tion of another 70·103 m³ raised the wellhead pressure by 28
across the filter cake is just as large as the pressure drop bar only.
within the entire formation (up to the external boundary radius The pressure increase per unit of injection volume amounts
re in eq. 5). The higher the permeability of the formation is, the to (∆p/∆V)1 ~ 10-3 bar/m³ for the two cases without additional
earlier will this stage be reached. water treatment. With the additional treatment one has
This demonstrates that half-lives and -volumes need not be (∆p/∆V)2 ~ 10-4 bar/m³. In accordance with equation 6b, the
significant for the course of an injection. Injections are often WQR values should differ by a factor of 10 as well: Ω1/Ω2 =
pressure-limited. This may be dictated by plant facilities or by 10. Filtrations with 1.2 µm membranes resulted in WQR val-
a maximal downhole pressure (e.g. to avoid fractures). For a ues of Ω1 = 650 ppm/mD and Ω2 = 6 ppm/mD respectively.
high permeability of the formation, the pressure build-up at Hence, the ratio of these WQR values amounts to approxi-
the half-volume may still be so small that it does not imply mately 100 instead of 10 as observed in the injection. Possible
any significant restriction for the injection. causes for these discrepancies will be discussed further below.
In absolute values, the course of injection may be charac-
terized by the pressure drop pc across the filter cake. For the Build-up of cake during membrane filtration
example of wellbore narrowing given in table 1a, one obtains: Cake filtration represents a well defined special case of solids
pc µ  βV  ρW separation: The solids are separated above the filter medium
=− ln 1 − ;β = w ; Vw = πrw2 h w ..(6a) and thus generate a second porous medium (the cake). Hence,
Q 4πh w k c  VW  ρc the filter cake itself becomes the true separating medium for
A simple calculation reveals that βV is the cake volume (sol- further particle deposition. The mass of the cake grows in pro-
ids plus pores). Hence βV/VW is the fraction of the wellbore portion with the filterable solids that reach the medium. The
volume that is filled by cake. The logarithm in equation 6a porosity and permeability of the cake are constant as long as
implies that the pressure pc will grow without limit as the no (de-)compression occurs as a result of changes in pressure.
borehole is filled completely with filter cake. In membrane filtration with only slightly charged water,
Because of the low permeabilities of filter cake, injections the build-up of a cake occurs only after the filtration of a cer-
become unfeasible even for relatively thin filter cakes. Then tain amount of water. In the early stages other mechanisms,
βV/VW << 1 holds, and equation 6a may be replaced by its such as the narrowing of pores or the partial covering of the
linear approximation: membrane surface, dominate. These mechanisms have been
pc µβV µΩV studied in the context of industrial filtration [13]. A survey
= = (for βV/Vw << 1)............... (6b) was given in [12].
Q 4πh w k c Vw A 2w
The volume at which cake filtration commences depends
(Aw = 2πrwhw is the open-hole sandface area.) For the last on the type of water and the membrane, and it must be deter-
equality in equation 6b, β/kc = (w/kc)(ρW/ρc) = Ω was used mined empirically. For the examples provided by Barkman
(see eqns. 6a and 2b). and Davidson (membrane diameter: 47 mm) it starts after the
The example presented in figure 1 shows a noticeable de- filtration of some 100 ml of water. The water employed for
viation from the (normalized) linear behaviour of equation 6b their investigation had solids contents of approximately 1 %
for injection volumes V > 15000 m³. At this stage, the filling [1]. For injection water, however, the solids content usually
of the injection cavity starts to show up. lies in the range from 10 to 100 ppm and is thus lower by
Four properties of the linear approximation in equation 6b some 2 to 3 orders of magnitude. With all other conditions
should be noted carefully: remaining unchanged, the onset of cake filtration should occur
The absolute course of injection is independent of the per- only after the filtration of some 10 to 100 litres.
meability contrast kc/kf. This is also true for the nonlinear Quite often, the filtration of such large volumes will not be
version in equation 6a. feasible, simply because of the required time. This is true par-
4 SPE 82231

ticularly for filtration at constant pressure and for the surveil- creasing portions of water will flow enter the formation
lance of injection water or of water treatment units at regular through the perforation mantle, generating a filter cake here as
intervals (see below). well. A realistic model for the growth of the cake has to ac-
For checking the occurrence of cake filtration, the method count for this changing distribution of flow at the sandface.
of looking for a straight line in a V-√t plot (fig.2a), as pro- A first estimation for perforations suggests that in the long
posed by Barkman and Davidson, is not particularly well term the cake will cover the sandface with a uniform thickness
suited. In such a plot one may easily obtain a straight line with [10]. Later on, the build-up of filter cake was modelled in a
reasonable statistical precision, even if no cake filtration oc- mathematical framework which amounts to a skin-
curs. A much more sensitive test is obtained from a plot of p/Q approximation of the cake [14]. This approximation assumes
versus the volume V. As equation 6b holds for a membrane as that the permeabilities of cake kc and formation kf obey kc <<
well (with Aw replaced by the membrane area), cake filtration kf. This condition will almost always hold for water
will also produce a straight line in this plot. The slope S' of injections [2].
this line depends on the WQR through: Fig. 5 shows the development of pressure per rate for a flat
A2 cylindrical frac of fixed length, which was calculated in [14].
Ω = S' ........................................................................(7) Its qualitative features also hold for elliptical fracs and for
µ
perforations. Another form of cake formation was investigated
Figure 2 shows an example for these two plots with straight in [15]. This work considers boreholes with a heterogeneous
lines fitted to V ≥ 1500 ml. The V-√t plot (fig. 2a) is almost distribution of flow resistances at the sandface. This may be
linear in this range and thus suggests that cake filtration holds. due to a layered reservoir or result from variations of perfora-
The more sensitive 1/Q-V plot (fig. 2b) casts some doubt on tion properties. Again, the development of the injection has
this conclusion. Moreover, the two diagrams lead to signifi- the same characteristic features as in figure 5.
cantly different values of WQR: 56 ppm/mD in figure 2a and At the beginning of an injection, the filter cake grows pri-
Ω = 111 ppm/mD in figure 2b. marily in the regions with the least flow resistance. As a rule
Hence, to forecast injection behaviour with the aid of this covers only a small fraction of the injection area. Accord-
membrane filtrations, it is important to ensure that cake filtra- ingly, the cake grows rapidly here and causes a disproportion-
tion is observed. Further problems concerning the applicability ately fast increase of pressure (fig. 5). As the injection pro-
of membrane filtration to formations are discussed in ceeds, the flow is distributed more uniformly over the com-
section 6. plete injection area and is accompanied by correspondingly
Apart from injection planning, membrane filtrations offer uniform growth of the filter cake. Thus the development of
another important range of applications: They may be em- injectivity becomes independent of the injection geometry and
ployed to check the quality of injection water, in particular for then depends solely on the total injection area. At this stage
the surveillance of water treatment units or for estimating the the pressure per unit rate increases linearly with the injection
effect of oilfield chemicals on water quality [9, 12]. Opera- volume and is universally given as in equation 6b:
tional experience has proven that the results are significant p c µΩV
even if cake filtration has not yet set in. Although the resulting = ..................................................................... (8)
values of Ω do not represent the WQR of Barkman and David- Q A2
son, they still are a quantitative measure for changes in water Along with the transition to linear development of injectivity,
quality. In figure 4, this is illustrated for the surveillance of a the flow in the formation is redistributed as well. After initial
back-flushable filter. The inflowing water shows an almost peaking at the tips of perforations and fracs, it evolves into a
constant WQR, whereas the effluent values differ signifi- uniform flow across the sandface. This redistribution changes
cantly. In this case, membrane measurements may be used to the total flow resistance and thus leads to the constant portion
optimise the back-flushing times. ∆pc/Q in figure 5.
If the injection can be continued until a substantial part of
Build-up of cake during injection the injection cavity (borehole, frac, perforation, hairline
Cake filtration. Of the four modes considered by Barkman fracs,...) is filled with filter cake, the injection pressure again
and Davidson (see section 2), the cases of wellbore narrowing starts to rise faster (not shown in fig. 5). For the case of well-
and perforation plugging deal with cake filtration. A further bore narrowing, this late nonlinear behaviour is described by
important mode is the build-up of a filter cake in fracs. The the logarithmic term in equation 6a.
geometry of the growing cake influences the way in which the The quadratic dependence on the injection area in equation
injection pressure increases. This shape is nontrivial for perfo- 8 is remarkable. Its origin is easily explained: At fixed cake
rations and fracs, because it depends on the distribution of thickness one has pc/Q ~ 1/A. As water is injected, the thick-
flow at the sandface, which changes as the injection proceeds. ness of the cake grows in proportion with Vc/A. This explains
Barkman and Davidson proposed that perforation plugging the second factor of 1/A in equation 8. The quadratic depend-
starts at the tip of perforations and grows inward, thus shorten- ence on the area has important consequences for the planning
ing the effective perforation length. However, this scenario is and forecasting of injections:
valid only for a very early state of cake build-up. For a clean If the value of WQR is known, and one wants to use equa-
perforation a large fraction of flow enters the formation tion 8 to forecast the injection performance, one needs to
through the tip, leading to an early cake which resembles the know the injection area A. As a rule, this area is only known
form envisaged by Barkman and Davidson. As the resistance rather imprecisely. For perforated holes, for instance, one has
to flow increases at the tip, the flow will redistribute and in- to rely on estimates of the number of flowing perforations,
SPE 82231 5

their lengths and radii, as well as of cemented, crushed, and However, the abundance of published results on core dam-
predamaged zones. The true injection area may easily be mis- age by solids invasion could rather easily be interpreted within
judged by a factor of 2 or 3. The use of equation 8 then leads the concept of Barkman and Davidson's WQR. Results of core
to errors in the rate of pressure increase as large as a factor of measurements are regularly reported by plotting the reduced
4 to 9. As a second illustration for the importance of the 1/A2 permeability k/ki against the injection volume (often in multi-
term, figure 6 shows a chart for the planning of an injection. ples of the pore volume). With the use of the equation for the
For injection into a reservoir of 10 m thickness, a pressure pressure across cores, one may easily show that k/ki is indeed
increase of up to 10 bar per year is considered acceptable. The identical with α (eqns. 1 and 4). Indeed, the shape of α in fig-
figure shows the interdependence between the required WQR ure 1 is similar to typical plots of k/ki. For obtaining WQR,
and the injection area for various rates of injection [14]. however, the reciprocal 1/α is better suited: If one plots the
The nonlinear behaviour in the early stage of an injection reciprocal ki/k = 1/α, a straight line then identifies a mecha-
has to be considered if one wants to check the feasibility of an nism of deposition which corresponds to a unique value of
injection by a short-term injection test. The fast build-up of WQR. The slope S" of this line yields the value Ω of WQR as:
pressure may easily lead to an underestimation of injectivity AL
by orders of magnitude. For an assessment of this mechanism Ω = S" ....................................................................... (9)
ki
one has to estimate the duration of this early stage. Indications
(AL: core volume, ki: permeability of the undamaged core).
for this purpose have been provided in [14, 15].
How much of this damage is caused by cake filtration or by
In the calculations for injections into fracs in [14], fracs of
DBF is not primarily important. The essential point is the fact
fixed length are assumed. For constant-rate injections, this
that this value of Ω may again be used in equation 8 to fore-
length will increase. A first estimate of the resulting evolution
cast the course of injection (assuming that flow velocity and
indicates that the injection pressure will be approximately
pressure do not change the WQR, see section 6).
constant in this case [14]. Other approaches to injection into
The reason why equation 8 may again be involved has
growing fracs have been detailed in [16, 17].
been sketched in [18]: As core measurements show, the depth
filtration of solids occurs in a layer with a thickness of a few
Depth filtration. Solids suspended in injection water are not
cm. For such a thin layer of reduced permeability the mecha-
necessarily separated as a cake by surface filtration. They may
nism described in section 5a leads again to a uniform growth
enter the formation and be deposited in the pores. In industrial
of deposits and hence to the linear behaviour of equation 8.
filtration this mechanism is called deep bed filtration (DBF).
As far as we know, core measurements have never been in-
Of the modes considered by Barkman and Davidson [1],
terpreted in this way. However, this approach could constitute
the "invasion" process (see section 2) shows some similarity
the most reliable method of forecasting the fate of injections.
with DBF. The two mechanisms are different, however, as
Moreover, it could be an important supplement for better judg-
DBF proceeds from the injection face and progresses into the
ing the validity of the simpler membrane filtration tests (see
pore space, whereas Barkman and Davidson assume that the
the next section).
deposition starts at some 'radius of invasion' within the forma-
tion, which is not specified further, and then grows towards
Laboratory investigations versus field data
the injection face.
Three problems in forecasting the course of injections have
Of course, the behaviour proposed by the authors might
already been addressed in the preceding sections. These con-
occur if the pore widths decrease at a certain radius within the
cern the knowledge of the injection area of a well (section 5a),
formation. To the best of our knowledge, however, this
the question of whether surface or depth filtration occurs (sec-
mechanism has not been reported in the literature for homoge-
tion 5b), and whether measured values of WQR really repre-
neous pore spaces. As the model does not specify the radius of
sent cake filtration (section 4) or - if depth filtration in cores is
invasion, it has no predictive power, and there is a danger of
involved - whether they are measured in a region of linear
fitting any course of injection by simply assuming a suitable
dependence of p/Q on V (section 5b).
radius of invasion.
Another important question is whether values of WQR
Two models known from industrial DBF have been inves-
measured in the laboratory still hold under injection condi-
tigated in [18] for radial flow (injection into an open hole). If
tions. In particular, one has to check whether the WQR is in-
the filtration probability grows sufficiently fast for decreasing
dependent of the pore widths of membranes and the real pore
flow velocity, then a maximal deposition may indeed occur
space. This independence was tacitly assumed by Barkman
within the formation. In contrast to the approach of Barkman
and Davidson in [1]. Experience with industrial filtration tells
and Davidson, the deposition still proceeds into the formation,
a different story: The permeability of the filter cake depends
and the radius of maximal deposition may be calculated from
on the pore width of the membrane as well as on the initial
the model parameters.
conditions (pressure, rate) of the filtration [20].
The interpretation of deep bed filtration for the prognosis
To gain initial insight into this question, the values of
of injection has hitherto been hampered by two factors: On the
WQR were measured with the use of cellulose acetate mem-
one hand, the literature on DBF offers a wide range of models
branes with pore widths (diameters) between 0.45 µm and 1.2
from basic research to industrially oriented phenomenological
µm [9]. The results depicted in figure 7 show that the WQR
approaches ([19] and literature in [18]). On the other hand,
varies by a factor of 4 within this range of pore widths. There-
there is no established quantitative method to relate results
fore, the assignment of WQR makes sense only if the pore
from core measurements to injection behaviour.
width of the membrane is specified as well.
6 SPE 82231

The linear slope of the doubly logarithmic plot in figure 7 residual oil. This leads to filter cakes of extremely low perme-
corresponds to power law behaviour: Ω ~ d-1,2. For an injec- ability [2, 12]. They may be squeezed only partially into the
tion, this law may be extrapolated to the pore width of the formation, even at rising pressures, and will cause severe
formation. For this purpose one may use the hydraulic diame- damage there.
ter as a simple representative average pore width. However, This suggests that filtrations at constant pressure yield too
one should be aware of the fact that the dependence in figure 7 pessimistic results, because no sufficient shear forces build up
was obtained experimentally, and there is no reason to guaran- to squeeze solids through the cake. On the other hand, filtra-
tee (not even to expect) that the observed dependence may be tions at constant rate may also yield too pessimistic results:
extrapolated to (significantly) larger pore widths. The material which is squeezed through a thin filter medium
To also measure the WQR for larger pore widths, which (membrane, bag filter,...) at constant rate will not be included
may be more representative of formations, one may also use in the value of WQR. As a rule, however, a portion of this
commercially available cartridge or bag filters (or cores, of material will be deposited in the formation by depth filtration.
course). Figure 8 shows the pressure build-up for cartridge This results in a mixture of surface and depth filtration [15,
filters which may also be employed for water treatment [12]. 18], and both components affect the course of injection.
The nominal pore widths of the two filters are 10 and 20 µm. These considerations do not allow conclusions concerning
As shown in figure 8, different volumes of water have to pass which type of filtration tests (const. pressure or const. rate)
through these filters before cake filtration sets in. The build-up yields more reliable values for the WQR. As sketched in sec-
of a filter cake is signalised by the linear increase in pressure. tion 5b, investigations of core damage may be helpful in this
The value Ω of the WQR may again be determined from equa- respect. Another approach is to analyse real injections in order
tion 7 if the area A of the cartridge filter is known. For this to derive the true observed WQR, which may be compared
water, the slopes and hence the WQRs are almost identical for with laboratory measurements. This procedure is outlined in
both pore widths. the following example:
In figure 9, the WQRs measured on membrane and bag fil- Source water and treated water from production were in-
ters are plotted for various pore widths. For both filters linear jected into a formation at a rate of Q = 10 m³/h each. The in-
behaviour is observed in the doubly logarithmic plot. The crease in injection pressure was approximately proportional to
slopes are approximately equal and again correspond to a the injected volume (∆p/∆V = 1.1·10-2 bar/m³ for the source
power law (this time with Ω ~ d-0.85). However, the two water, ∆p/∆V = 7.2·10-2 bar/m³ for the produced water). These
straight lines are mutually shifted. At fixed pore width, this values may be inserted into equation 8. If the viscosity µ and
shift corresponds to a factor of 30 in WQR. Such factors of 30 the injection area A are specified, one may calculate the true
(and larger) may well be considered as typical for the underes- WQRs for the injections:
timation of injectivity which may result from a direct applica- A 2 ∆p
tion of the Barkman and Davidson models. Ω= .................................................................... (10)
µQ ∆V
How can the difference between membrane and bag filters
that is apparent in figure 9 be explained? As the ranges of pore To illustrate the important role of the injection area A, the
widths do not overlap, it is conceivable that the filtration of WQR-results are shown in table 2 for several scenarios with
solids decreases dramatically for pore widths between 3 µm different injection areas A. The following data were assumed
and 5 µm. This would require a correspondingly sharp size for the well: thickness hw = 20 m, wellbore diameter dw = 5½",
distribution for the suspended solids. The different types of shot density Np = 5/ft = 16.4/m perforated along the full 20 m,
filtration processes employed in the measurements lead to perforation length Lp = 0.15 m, perforation radius rp = 0.01m.
another plausible explanation: The membrane filtration was The modes considered are: wellbore narrowing (A1 = 2πrwhw =
conducted at constant pressure, whereas the bag filtration took 8.8 m²), perforation plugging of the Barkman and Davidson
place at constant rate. For these two processes, the shear type (linear flow through the tip) (A2 = Npπrp2 = 0.1 m2), and
forces acting on the deposited particles are different. The perforation plugging with flow through the whole mantle (ra-
higher shear stress which occurs at constant rate may squeeze dial flow, see section 5a) (A3 = 2πrpLpNp = 3.1 m²). With these
retained particles through the filter (bleeding). In figure 10, values and a viscosity of µ = 1 mPas, equation 9 yields the
this behaviour is demonstrated for a bag filter with a pore WQRs given in table 2. The values obtained by membrane
width of 5 µm. Measurement of the WQR for the influent and filtration with a pore width of 1.2 µm are also listed in
effluent streams with membrane filters (1.2 µm) clearly shows the table.
that the bag filter becomes inefficient after a throughput of The WQR calculated from equation 9 is proportional to the
approximately 70 m³, and that suspended solids then pass square of the injection area. Therefore, the ratios of WQR val-
through the filter without hindrance. ues for the different modes are the same as the ratios of the
Similar behaviour can be expected for injections at con- corresponding squared areas (see table 2). As is obvious from
stant rate. Solids which are squeezed through the cake will table 2, the WQRs for each water type vary over a range of
enter the formation. Gel-like deposits, such as those which almost 4 orders of magnitude.
may originate from ferrous hydroxide, may block the flow if For the injections considered here, perforation plugging
they are injected at constant pressure. During injection at con- during radial flow should be the relevant mode. Hence, the
stant rate, the increasing pressure will easily squeeze these source water was injected with a WQR of Ω = 3.7 ppm/mD,
substances into the formation. The situation is different for whereas Ω = 12 ppm/mD was measured during the membrane
corrosion products and ferrous sulphide in produced water. filtration. To a large extent this deviation can be explained by
Because of their lipophilic nature, they may be covered with the dependence of WQR on the pore width. Figure 9 shows
SPE 82231 7

the WQR of the source water from membrane filtrations WQR of 25 ppm/mD, which was observed in the injection, by
(p=const.) with pore widths of 1.2 µm and 3 µm. The WQR an order of magnitude.
which was recalculated from the injection is indicated as well; This pronounced discrepancy can be attributed to a general
a hydraulic diameter of 1.92 µm, which corresponds to the problem which occurs during membrane filtrations of water
formation permeability of 23 mD, was employed for the calcu- with residual oil. During constant pressure filtrations, the
lation. If one interpolates the two values from membrane fil- membrane often becomes impermeable after a throughput of
trations with a straight line in figure 9, the WQR of the injec- only a few 100 ml, long before cake filtration sets in. There-
tion is predicted to be about 5 ppm/mD. The remaining dis- fore, no reliable measurement of the WQR is possible in
crepancy corresponds to a ratio of 5/3.7 = 1.35. Hence, the this case.
prediction from membrane filtration is too pessimistic, since it Investigations of these membranes with a scanning elec-
underestimates the observed behaviour by 35 per cent. tron microscope reveal that the membrane surface is covered
Of course, the concept of hydraulic radius is only a rough almost completely by an oil film [12] (see fig. 11). An expla-
approximation to the true pore widths of the formation, and it nation for this effect was given in [21]. If the surface pores of
disregards structural differences between membrane and for- the membrane are constricted by deposited particles, capillary
mation. Bearing this in mind, one may conclude that labora- forces may prevent oil from entering the pores. During injec-
tory prediction and observed behaviour agree fairly well. One tion at a constant rate into a formation, the capillary forces are
should, however, keep other sources of possible errors easily surmounted. For thin filter media (membrane, bag fil-
in mind: ters,...) the situation is dramatically different, in particular for
The stage of cake filtration might not yet have been filtrations at constant pressure: As the pores get constricted, a
reached (see section 4). In this respect it is interesting to state in which oil drops are retained on the surface will be
compare the throughput volumes per unit area. For mem- reached, if their diameter is not larger than the membrane
brane filtrations one typically uses a few litres. With an thickness. As soon as this happens, oil films can grow indefi-
area of 15 cm², the throughput of 1000 ml corresponds to a nitely on the surface and thus shut off the flow almost com-
water column of about 0.7 m. For an injection area of 3.1 pletely. Values of the WQR which are measured under these
m² (see table 2) this amounts to an injected volume of circumstances will underestimate the injectivity by orders
about 2 m³. Hence, the membrane filtration covers only a of magnitude.
minute injection volume, and if the cake has not yet As a way out, one may consider membrane filtrations at
reached its final structure, the laboratory predictions will constant rate. However, it is an open question whether realistic
be erroneous. predictions of WQR can be obtained in this way. The results
The true injection area is never exactly known. The factor in [21] suggest that the combination of capillary forces and
of 1.35 between the WQR from membrane tests and injec- thin filter media may still lead to effects which are absent for
tion, which was derived above, corresponds to an underes- injections. If this turns out to be true, one would have to take
timation of the injection area by a factor of √1.35 = 1.16 recourse to core measurements for determining the WQR of
(see eq. 9); this leads to A3= 3.6 m² instead of A3 = 3.1 m² produced water.
in table 2.
As discussed above, the different operational conditions (p Conclusions
= const. for membrane filtration and Q = const. for the in- Even after 30 years the model of Barkman and Davidson [1]
jection) may have an impact on the WQR as well. How- still presents the best known framework for forecasting water
ever, as figure 9 shows, the bag filtration with Q = const. injections. Its most important feature is the concept of the wa-
would lead to a value of the WQR which is much too op- ter quality ratio, which paves the way for
timistic. The question of whether this is due to the extrapo- quantitative assessments.
lation of pore widths from 5 µm to 1.9 µm, or to the However, the indiscriminate application of the model may
aforementioned fact that solids are squeezed through the easily underestimate the true injection behaviour by several
bag filter, remains unclear. In the latter case, the solids orders of magnitude for both half-lives and decline in rate.
squeezed through the filter are lost for the determination of For the case in which the permeability of the damaged
the laboratory WQR value, whereas they may have an im- formation (by surface or depth filtration) is much lower than
pact on depth filtration in the formation. the permeability of the undamaged formation, which almost
For the produced water the difference between the WQRs always holds for injections, wide ranges of injection are gov-
from injection (25 ppm/mD in table 2) and from membrane erned by a universal law of pressure/rate change (eq. 8). Fore-
filtration (1120 ppm/mD) is much larger than for the source casting the course of an injection not only requires a reliable
water. No membrane tests were performed at larger pore value of the WQR; a good estimate of the injection area and
widths for the produced water; hence, we cannot extrapolate its range of uncertainty is needed as well, because the changes
the WQR from 1.2 µm to the hydraulic diameter of 1.92 µm. in injection pressure depend quadratically on this area.
For the source water, this extrapolation leads to a correction of Injections into perforated wells, into fracs, and into forma-
the WQR from 12 ppm/mD to 5.5 ppm/mD. If a correction of tions with heterogeneous permeability distributions may show
similar magnitude is assumed for the produced water as well, an initial pressure increase per unit of injection volume which
one may expect that the laboratory value of 1120 ppm/mD is much larger than that which occurs during long-term behav-
should be reduced by a factor of 2 to 3, with a resulting value iour. If one wants to judge injectivity prospects by test injec-
of 300 to 500 ppm/mD. This value is still higher than the tions, it may be essential to account for this effect.
8 SPE 82231

Several aspects are important for a proper measurement of S' = slope in the p/Q-V- diagram, eq. 7
the WQR: S" = slope in the k/ki-V- diagram, eq. 9
1) Cake filtration starts only after the throughput of a certain t = time
amount of water. A meaningful value of the WQR can be V = volume
obtained only after cake filtration has set in. w = concentration of total filterable solid, w/w
2) The WQR depends on the pore width. To account for this WQR = water quality ratio, eqns. 2a, 2b
dependence, the WQR has to be measured for several pore
widths and has to be extrapolated to the average pore width
of the formation. α = normalized injection decline, eqns. 1 and 2b
3) This extrapolation should not be extended to pore widths β = wρw/ρc (eq. 6a)
which differ significantly from those employed in the labo- µ = viscosity
ratory investigations. Besides cores, commercially avail- ρ = density
able filters (bag filters, cartridge filters,...) may be of use Ω = water quality ratio (WQR), eq. 6b
for selecting pore widths which are close to those of Indices:
the formation. c = filter cake
4) WQRs measured in filter tests may depend strongly on the e = external boundary radius
operational mode. Tests conducted at constant pressure i = initial value
tend to be too pessimistic, those conducted at constant rate p = perforation
tend to be too optimistic. W = water
Even if these rules are observed, several further uncertainties w = wellbore
remain. For example, it is not known whether and to which ½ = half-value
extent measured WQRs depend not only on the pore width,
but also on details of the pore space. The measurement of the References:
WQR for produced water containing residual oil presents a [1] Barkman, J.; Davidson, D.: "Measuring Water Quality
major problem. Employing thin filter media at constant pres- and Predicting Well Impairment", JPT July (1972)
sure may yield values which are too pessimistic by orders of 865-873.
magnitude. The coverage of the filter surface by an oil film, [2] Dietzel, H.; Kleinitz, W.: "Required Quality of Water for
which causes this problem, may be identified as almost com- Injection in Secondary Projects", OIL GAS European
plete blocking of flow before cake filtration has set in. The Magazine 3/1989, 24-27.
question of whether filtrations at constant rate can remedy this [3] Mitchell, R.; Finch, E.: "Water Quality Aspects of North
problem is still uncertain. Sea Injection Water", JPT June (1981) 1141-1152.
Successful measurements of the WQR by means of core [4] Vetter, O.; Kandarpa, V.; Stratton, M.; Veith, E.: "Parti-
tests have not yet been documented. Such investigations are cle Invasion into Porous Medium and Related Injectivity
important for eliminating the remaining uncertainties, in par- Problems", SPE 16255, SPE Int. Symposium on Oilfield
ticular for oil-contaminated water. Apart from their direct use Chemistry , San Antonio, Texas, Feb. 1987.
for forecasting injection behaviour, they may also be helpful [5] Todd, A.; Kumar, T.; Mohammadi, S.: "The Value and
for allowing a better evaluation of the simpler membrane tests. Analysis of Core-Based Water-Quality Experiments as
For this purpose, core tests would have to be analysed within Related to Water Injection Schemes", SPE Formation
the context of the WQR (as sketched in section 5b) and com- Evaluation, June (1990) 185-191.
pared with parallel membrane filtration tests. [6] Pang, S.: "Colloid Retention and Flow in Porous Media",
Measuring WQRs by membrane tests as proposed by Dissertation, University of Texas at Austin, Aug. 1996.
Barkman and Davidson is an excellent means of monitoring Sharma, M., Pang, S.; Wennberg, K., Morgenthaler, L.:
injection water as well as the effectiveness of water treatment "Injectivity Decline in Water Injection Wells: An Off-
plants (see section 4 and fig. 4). The restrictions outlined shore Gulf of Mexico Case Study" SPE 38180, presented
above in 1) - 4) do not apply here. To obtain comparable re- at the 1997 SPE European Formation Damage Confer-
sults, one simply has to keep the pore width and filtration vol- ence, Den Haag, 2-3 June 1997, 341-351.
ume constant. Pang, S.; Sharma, M.: "Evaluating the Performance of
Open-Hole, Perforated and Fractured Water Injection
Notation: Wells" SPE 30127, presented at the SPE European For-
A = area mation Damage Conference, Den Haag, 15-16 May
d = diameter 1995, 489-498.
DBF = deep bed filtration Pang, S.; Sharma, M.: "A Model for Predicting Injectiv-
h = thickness ity Decline in Water Injection Wells" Paper SPE 28489
k = permeability presented at the 69th Annual Technical Conference and
L = length Exhibition in New Orleans, 25-28 Sept. 1994, 275-284.
N = perforation shot density Wennberg, K.; Sharma, M.: "Determination of the Dy-
p = pressure namic Filtration Coefficient and the Transition Time for
Q = volumetric flow rate Water Injection" SPE 38181, presented at the SPE Euro-
r = radius pean Formation Damage Conference, Den Haag, 2-3
S = slope in the V-√t-diagram; skin in eq. 5 June 1997, 353-364.
SPE 82231 9

[7] Roque, C.; Chauveteau, G.; Renard, M.; Thiebault, G.; [17] Gadde, P.; Sharma, M.: "Growing Injection Well Frac-
Boutéca, M.; Rochon, J.: "Mechanism of Formation tures and Their Impact on Waterflood Performance",
Damage by Retention of Particles Suspended in Injection SPE 71614, SPE Annual Technical Conference and Ex-
Water" SPE 30110, presented at the European Formation hibition, New Orleans, LA, 30 Sept. – 3 Oct. 2001.
Damage Conference, Den Haag, 15-16 May, 1995, [18] Hofsaess, T.; Kleinitz, W.: "Infiltration von Feststoffen
329-343. in Wasserinjektionsträger – Infiltration of Solids into
Nazbar, L.; Chauveteau, G.: "Permeability Damage by Water Injection Reservoirs", Erdöl Erdgas Kohle
Deposition of Colloidal Particles" presented at the Euro- (September 1995), 363-370.
pean Formation Damage Conference, Den Haag, 2-3 [19] Sahimi, M.; Gavalas, G.; Tsotsis, T.: "Statistical and
June, 1997, 161-171. Continuum Models of Fluid-Solid Reactions in Porous
Nazbar, L.; Chauveteau, G.; Roque, C.: "A New Model Media", Chem. Eng. Sci. 45 (1990), 1443-1502.
for Formation Damage by Particle Deposition" presented [20] Wakeman, R.; Tarleton, E.: "Solid/Liquid Separation
at the 1996 Formation Damage Symposium, Lafayette, Equipment Simulation & Design – An Expert Systems
14-15 February, 1996. Approach", Filtration & Separation July/Aug. 1991,
[8] Coleman, J.; McLelland, W.: "Produced water reinjec- 268-274.
tion: How clean is clean?", SPE 27394, 1994. [21] Hofsaess, T.; Kleinitz, W.: "Kapillare Eintrittseffekte bei
Paige, R.; Murray, L.; Martins, J.; Marsh, S.: "Optimis- der Injektion von ölhaltigen Wässern", DGMK-Früh-
ing water injection performance", SPE 29774, 1995. jahrstagung 2001, p. 331-344.
[9] Kleinitz, W.: "Kontrolle von Injektionswässern in
Erdölaufbereitungsbetrieben – Controlling the Quality of Tables
Injection Water in Oil Fields", Erdöl Erdgas Kohle (June
1988), 256-262. Tab. 1-Solutions of the injectivity development for
[10] Hofsaess, T.; Kleinitz, W.: "Modelle der wellbore narrowing
Filterkuchenbildung bei Wasserinjektionsbohrungen - On 1a)
the Growth of Filter Cakes during Water Injections", Q
Erdöl Erdgas Kohle (September 1994), 359-364. α= V (α ) = F ⋅ Q0 ⋅ H(α ) t (α) = F ⋅ G(α)
Q0
[11] Flixeder; F.; Kleinitz, W.: "Veränderung der
πh w rw2
Injektionswasserqualität in einem Erdölfeld – Auswir- F=
β Q0
kungen auf Injektionsstrategie und Bohrlochbehandlung
– Variation of the Injection Water Quality in an Oil Field G(α ) = 1 +
1 1
− +
1  2(1−1 α )
⋅Θ
– Consequence on Injection Strategy and Well Treat- 2 ln Θ  α 2 ln Θ 

ment", Erdöl Erdgas Kohle (February 1986), 67-74. H(α ) = 1 − Θ2(1−1 α )


[12] Kleinitz, W.: "Zur Bestimmung der Wasserqualität in
kc kf
Aufbereitungsanlagen für Injektionswässer", Disser- β=w
ρW r
Θ =  e


ρc  rw
tation, TU Clausthal, 1998. 

[13] Hermans, P.; Bredée, H.: "Principles of the mathematical


treatment of constant pressure filtration"; Journal of the
1b)
Society of Chemical Industry, Transactions and Commu-
nications, Jan. 1936, S.1T. 
ln 1 −
βV 

Hermans, P.; Bredée, H.: "Zur Kenntnis der Filtrations- (p Q)V = 0 1 k Vw 
α(V ) = = 1+ f ⋅ 
(p Q)V α(V ) 2k c r 
gesetze"; Receuil des traveaux chimique des Pays-Bas; ln e 
 rw 
54 (1935), 680.
Gonsalves, V.: "A critical investigation on the viscose β=w
ρW
Vw = πrw2 h w
ρc
filtration process"; Receuil des traveaux chimique des
Pays-Bas; 69 (1950), 873.
[14] Hofsaess, T.; Kleinitz, W.: "Zum Wachstum von
Filterkuchen in Perforationen und Fracs – On the Growth
1a) t(α) at p=const. by Barkman und Davidson. From [1]
of Filter Cakes in Perforations and Fracs", Erdöl Erdgas
with corrections from [10]. (Because of a misprint, t(α) is
Kohle (April 2001), 201-206.
displayed as tα [10].)
[15] Hofsaess, T.; Kleinitz, W.: "Injectivity Decline in Wells
1b) α(V) at arbitrary p,Q. From [10].
with Nonuniform Perforation Properties", SPE 39586,
1998 SPE International Symposium on Formation Dam-
age Control, Lafayette, LA, 18-19 February 1998, p.
631-640.
[16] Longuemare, P., Détienne; J.; Lemonnier, P.; Boutéca,
M., Onaisi, A.: "Numerical Modelling of Fracture Propa-
gation Induced by Water Injection/Re-Injection", SPE
68974, SPE European Formation Damage Conference,
The Hague, 21-22 May 2001.
10 SPE 82231

Tab. 2-Comparison of WQRs for two types of

pressure / rate of filtration, bar∗s/ml


injection water 0,4

Measured data

Injection area A Observed WQR 0,3


Fit for V ≥ 1500 ml
A A2 source water produced water

m2 m4 ppm/mD ppm/mD
0,2

A1: open hole 8.8 77.4 29.5 201

0,1
A2: perforations
0.1 0.01 0.004 0.03
(lin. flow)

A3: perforations
3.1 9.6 3.7 25 0
(rad. flow)
0 500 1000 1500 2000 2500

membrane filtration filtration volume, ml


12 1120
(1,2 µm)

Fig. 2b-Pressure / rate versus filtration volume for the membrane


The values were calculated from the injection data for an open filtration of fig. 2a). The WQR may be calculated from the slope of
the line (linear fit for V ≥ 1500 ml) in accordance with eq. 7.
hole and for two types of perforation plugging. The bottom (From [12]).
row shows the values obtained by membrane filtrations. (See
section 6, near eq. 10.) 120

Wellhead pressure pf, bar


Injection days
680 ppm/mD (1,2 µm)
0 100 200 300 400 500 100
1 5 6 ppm/mD (1,2 µm)
t1/2 = 97 d
80
0,8

Stimulation

Stimulation
Inverse decline parameter 1/ α

4
Decline parameter α

60
0,6

3
40
0,4

20
2
0,2

0
3
V1/2 = 15700 m 0 50 100 150 200 250 300
0 1
Injektion volume, 10³ m³
0 10000 20000 30000 40000
Injection volume, m3

Fig. 1-Evolution functions α and 1/α plotted against injection vol- Fig. 3-Pressure build-up for an injection at constant rate. After the
ume. For Injections at constant rate, the time scale is shown on second stimulation the water treatment was improved.
the upper abscissa. The volume scale is independent of the proc- (From [11]).
ess control. Half-life and -volume represent the same point of the
process at which α = ½ and 1/α=2 holds. The curves represent 2000

wellbore narrowing for a wellbore volume Vw = 0.3 m³, re/rw = 5000,


kc/kf = 0.01, Q0 = 10 m³/h, and wρw/ρc = 3 ppm. 1000
Water quality ratio, ppm/mD (0,8 µm)

Influent
500
Effluent
2000
300
200
filtration volume, ml

1500
100

50
1000

20
Measured data 2 6 10 14 18 22 2 6 10 14 18 22
500 Time, h
7h 6h 7h 8h 6h 7h 7h
Fit for V ≥ 1400 ml

Back-flushing intervals
0
0 2 4 6 8 10 12 14 16
Fig. 4-WQR data for a back-flushable filter (From [12]).
total time of filtration, t/s

Fig. 2a-Filtration volume versus square root of time for a mem-


brane filtration. The WQR may be calculated from the slope of the
line (linear fit for V ≥ 1400 ml) in accordance with eq. 3.
(From [12]).
SPE 82231 11

3,5

Differential pressure, bar


0,08 10 µm
pD / QD

3,0
20 µm
2,5

0,06
2,0

1,5
0,04
1,0

0,5
long-time behaviour
0,02
numerical solution 0
∆ 0 10 20 30 40 50 60 70
Filtration volume, m3
0
0 0,2 0,4 0,6 0,8 Ω V 1 Fig. 8-Pressure build-up for cartridge filters of 10 µm und 20 µm
D D
pore widths. (From [12]).

Fig. 5-Evolution of pressure/rate for the injection into a flat cylin- 100

drical frac. ΩDVD and pD/QD: dimensionless volume and pres-


dhydr Reservoir
sure/rate. The exact numerical solution and the behaviour for

Water quality ratio, ppm/mD


large injection volumes are shown. (From [14]). 10

1000 p = const
ppm/mD
Parameter of water quality ratio Ω

3
Q, m /d:
1 mem
brane
100 filter
Bag
100 300 filter
500 0,1
1000 Q = const
10000

10
0,01
1 1,2 3 5 10 25 50 100
Pore width, µm

Fig. 9-WQR from membrane filtrations at constant pressure and


1
from bag filtrations at constant rate. (From [12]).
1 open 10 100 1000 10000
hole sandface area A, m2 1 000
H = 10 m :
Water quality ratio, ppm/mD (1,2 µm)

perforations fracs

Fig. 6-Interdependence between required WQR and injection area


for a maximal pressure increase of 10 bar per year. (From [14]).

40

pconst = 1 bar
Water qualitäty ratio, ppm/mD

Influent
20
Effluent

100
0 20 40 60 80 100 120

10 Filtrate volume, m³

Fig. 10-Ageing (bleeding) of a bag filter for filtration at constant


5
rate. The WQRs measured by membrane filtrations (1.2 µm) for
the influent and effluent streams show that the bag filter becomes
ineffective after the throughput of about 70 m³ of water and then
3 lets the suspended solids pass through. (From [12]).
0,2 0,5 0,8 1 2
Pore size of the membrane filter, µm

Fig. 7-Dependence of the WQR on the pore width in membrane


filtrations at p = 1 bar. (From [9]).
12 SPE 82231

a) b)

Fig. 11-Membrane filtration (nominal pore width 1.2 µm).


a) Source water with deposits on the surface; b) Treated pro-
duced water with surface flooded by oil and some open pore
channels. (From [12]).

You might also like