Beruflich Dokumente
Kultur Dokumente
This paper was prepared for presentation at the 2009 SPE European Formation Damage Conference held in Scheveningen, The Netherlands, 27–29 May 2009.
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 have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. 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 SPE copyright.
Knowledge of the filtrate profile along the wellbore is models (appropriate for the field scale). We use two
important for well log interpretations and for evaluating the boundaries (wells) constrained by constant rate or constant
damage imposed on the formation by the invading fluid. A pressure. The flow equation is solved using the IMPES
number of potential damaging mechanisms and modelling of method (implicit in pressure and explicit in saturations with
these are described in details by Civan (2000). For water- respect to time). The reader is referred to the literature, e.g.,
based muds, one of the main damage mechanism is Aziz and Settari (1979) for further details on reservoir
particulate plugging of the formation, either by externally simulators.
introduced particles or by in situ fines mobilized by the The phase flow between grid cells is computed from the
invading fluid. The mechanisms of particle transport in pressure solution. Knowing the flow, local concentrations of
porous medium have been investigated by e.g., Gruesbeck oil, water and dissolved or suspended species are updated for
and Collins (1982), Sharma and Yortsos (1986), Wennberg the new time-step. Chemical and physical processes are
and Sharma (1997) and Al-Abduwani et al. (2005). decoupled from the flow equation. Components are first
Traditionally, the computation of fluid loss has relied on moved between cells, and then changes over the time-step,
using experimental filtration rates or, as in the more due to physical and chemical process, are computed locally
sophisticated models, experimental cake permeability and in each grid-cell.
porosity obtained for a specific mud. In our approach, we Rock properties such as absolute permeability k, porosity
bring this one step forward by computing the cake properties φ and also relative permeabilities and clay content are
from its compositions. The properties from which the cake's allowed to vary between layers (sections). Moderate
porosity and permeability are computed are ascribed to the compressibility of both the formation and the fluid
individual components. In Ding et al. (2002) and components is handled by the model. Corey type relative
Suryanarayana et al. (2005) the damage in invaded zones are permeabilities are used:
simulated based on experimental return permeability and an
assumed damage profile. We make no such assumptions, but S j − S jr
k rj = k rje S Ej
jn , S jn = , j = o, w . (1)
model the flow and retention of both particles and polymers 1 − S wr − S or
inside the formation, and compute the permeability reduction
from the amount of trapped material. As stated earlier, this is a multi-component simulator.
Models that are used should be consistent with the Properties of components, like viscosity or their effect on
physics involved at a macroscopic level. Input parameters viscosity, are described in the following sections.
should ideally be of three kinds, (a); properties of involved In simulations with mud filtration, the rate will change
components (e.g., viscosity, density, size of particles,..), (b); very rapidly during the first period when the filter cake is
properties of the formation (e.g., permeability, porosity, ....) built up from zero thickness, and an IMPES-type solution
and (c); model parameters describing events which are will be unstable. We solve this by pre-computing the growth
independent of the above properties (a and b). The benefit of of the cake during the time-step assuming the pressure
this approach is that the amount of empirical input difference between the well (mud pressure) and the
parameters are reduced and that the parameters used are more connection block to be constant. The estimated cake
universal and show less variance among different properties are used in the computation of the new pressure
experiments. Finally, the proposed approach will increase the solution for the total grid, and finally the rates from this
utilization of earlier experience since ideally all experiments solution are used to update the real growth of the filter cake.
should be matched by one single data set, provided adequate
descriptions of the experiments are available. Filter cake
The filter cake model presented here computes cake
Simulation model properties from properties of the individual components
This section describes the various modules that constitute the present in the mud. The model handles an arbitrary number
newly developed simulation model. The two main parts of of polymers and solids.
the Maximize program are the filter cake model handling Static filtration. The relation between cake thickness hc
filter cake buildup and controlling flow into the formation and filtrate volume Vf is obtained from mass balance
and the reservoir flow model handling flow inside the Bourgoyne et al. 1986:
formation. The fluids introduced to the formations contain a
V f C1 c sm
number of dissolved and dispersed components, which in hc = , C1 = , (2)
turn may change the original flow properties of the formation Ac c sc − c sm
through different chemical and physical processes. The
where Ac is the filter cake area and csm and csc are the
retention of solids and polymer are split into a pore throat
concentrations of solids in the mud and the filter cake,
trapping model and pore lining adsorption model. Brine
respectively. The flow rate through the cake expressed by
interaction with the rock surface and clays are described by a
Darcy's law is:
multi-component cation exchange model, which provides
(3)
input to processes like fines migration and clay swelling. ΔpAc k c
q= .
μhc
Flow model
The flow model handles two-phase flow in 1-D linear Setting the rate q = dVf/dt, inserting hc from Eq. 2 and
(typically used for cores), 2-D rectangular and 2-D radial integrating the resulting expression yields the known filtrate
SPE 122241 3
volume versus square root of time relation (a spurt loss The local pressure gradient within the filter cake can be
volume Vsp is also added to the resulting expression): described by Darcy's equation. At a given time, the flow rate
u will be constant through all layers of the cake for an
2Δpk c incompressible fluid. Then, according to the Darcy equation,
V f = Vsp + Ac C st t , C st = . (4)
μC1 the term kcdp/dx must be constant (independent on location in
the cake) and equal to u·μ. (assuming μ is constant):
There are three unknowns in Eq. 4, namely the spurt loss Vsp,
the cake permeability kc and the cake porosity φc (csc =1-φc is dp n dp uμ
=− = . (7)
part of C1). Two of these variables, kc and φc, depend on the dx dx k c ( x)
mud composition and the pressure drop over the cake while
the spurt loss depends on the mud composition and properties Let pn represent the local net overburden pressure in the
of the rock. The cake's porosity and permeability will cake (pn=pm-p) and Δp the fluid pressure drop over the cake.
decrease with increasing pressure drop, due to The overburden pressure pn will range from zero at the cake
compressibility of the cake, and thus reduce the rate increase. inlet to Δp at the cake outlet. The velocity u can be expressed
The main effect of temperature is through its influence on in terms of the average cake permeability kca and Δp. Let kc
the filtrate viscosity; and thereby on the filtration parameter be a function of pn and combine kca with Eq. 7 gives:
Cst. Different polymers used in mud will have different upper Δp hc
k Δp
temperature limits above which they become chemically
∫ k c ( p n )dp n = uμ ∫ dx, u = ca ,
unstable. Below this limit, only a weak temperature
0 0
μhc
dependency is assumed.
Δp (8)
∫ k c ( p n )dp n
Cake permeability. The permeability of a porous medium
can be estimated with the Carman-Kozeny equation (Lake
1989). Because mud is a multi-component system containing k ca = 0 .
different kinds of solids of very different sizes, we choose the Δp
following variant: The effective permeability will be only a function of Δp
∑ S 0i ci
and independent of the cake thickness. To solve the integral
φ3
k= , S0 = , (5) of kc(pn) we need a functional relation between kc and pn. In
2τ (1 − φ ) S 0
2 2
∑ ci the literature, kc and φc are normally treated uncoupled using
the same equation type (Outmans 1963):
The tortuosity parameter τ = (Lt/L)2 accounts for the
effective flow path length Lt, ci is volumetric concentration of k c = k o p − n1 and φ c = φ o p − n2 , n1 , n 2 ≥ 0 . (9)
species i and S0 is the specific surface area of the medium
In our case, we want to calculate the permeability using
defined as the surface to volume ratio. In particular for
the Carman-Kozeny Eq. 5, include the effect of
spheres, S0=6/Dp, where Dp is the particle diameter. The
compressibility through the porosity and integrate kcdpn to
specific surface area in a mixed system is readily obtained by
obtain the average kca. We use the classic definition for
volume weighted averaging. The operator <·> used for S0
compressibility, i.e.,
indicates that the property represents a volume weighted
average. 1 dφ
Another useful expression that relates an effective pore nc = − ⇒ φ = φo e −nc pn , nc ≥ 0 . ( 10 )
φ dp
diameter to permeability and porosity is the following:
Note that the porosity in Eq. 10 converges to the correct
32τk limit φο when pn approaches zero, which is not the case with
Dφ = . (6)
φ Eq. 9. The solution to the average cake permeability, as well
as the computation of the average porosity, is given in the
The Carman-Kozeny approach treats the porous medium Appendix. Here we show only the final expressions. Defining
as a bundle of capillary tubes and Dφ corresponds to the tube z(pn) = 1-φ(pn), the average permeability of a compressible
diameter when all tubes have the same diameter. filter cake is:
Cake compressibility. The compression of the cake is a
result of the local pressure difference between the cake 1 Fk (Δp ) − Fk (0)
k ca = ⋅ ,
matrix and the fluid filling the pore space. The cake matrix 2τS 02 nc Δp
pressure is assumed constant and equal to the mud pressure ( 11 )
1
pm. The fluid pressure decreases through the cake as a result Fk ( p n ) = + 2 ln z − z.
of the viscous flow. Since this pressure difference will vary z
throughout the cake, physical cake properties such as
porosity and permeability, which depend on the “overburden and the average filter cake porosity is:
pressure”, will be functions of spatial position. We are
interested in some effective average properties, which we
will obtain by integrating over the cake thickness.
4 SPE 122241
Fφ (Δp ) − Fφ (0)
Polymer viscosity. The effect of polymer concentration on
φ = , the fluid viscosity is typically expressed at low shear rate
Fk (Δp ) − Fk (0) with a modified Huggins equation Sorbie (1991). We use the
( 12 ) expression:
1
Fφ ( pn ) = + 3 ln z − 3z + 12 z 2 .
z (
μ 0 = μ s 1 + νc + k 'ν 2 c 2 + k ' 'ν 3c 3 , ) ( 15 )
The average filter cake permeability and porosity where c is the polymer concentration (volume fraction), μs is
computed by Eqs. 11 and 12 can be used directly in Eqs. 2 the solvent (e.g., water) viscosity and ν is a dimensionless
and 4. form of the intrinsic viscosity [η] in units [cm3/cm3]. The
The compressibility of the filter cake can be split into a
intrinsic viscosity [η] is defined as the limit of the reduced
reversible (elastic) and an irreversible (compaction) part. A
viscosity when polymer concentration approaches zero. k’ is
simple way to model this is to apply an irreversibility factor, the Huggins constant, which for a range of polymers in good
Firf. Let Δpmax be the historical maximum Δp. If the current
solvents is reported to be equal to 0.4±0.1. Eventual changes
pressure drop Δp is less than Δpmax, then an apparent effective in temperature and in molecular weight Mw due to, e.g., pore
pressure drop Δpa is used: throat trapping is accounted for by the following expression:
Δpa = Δp + Firf(Δpmax - Δpc). ( 13 )
⎛ M ( x, t ) ⎞
a
Partially reversible compression of a filter cake is ν = [η ]ρ p ⎜⎜ w ⎟⎟ (1 − BvT (T − T0 )) , ( 16 )
illustrated in Figure 1. ⎝ M w0 ⎠
Solids. The information needed in the filter cake model
includes a size distribution of the solid particles, the specific where exponent a is reported to be in the range 0.5-1 (higher
surface area S0 and the solids contribution to the cake's value for good viscosifiers). T and T0 are actual and reference
porosity and compressibility. temperatures, respectively, and BνT is an empirical parameter.
The size distribution can be either in the form of a The units for the density ρp must be consistent with the units
lognormal distribution or entered as a table. If a tabular form used for [η] so that ν becomes dimensionless. The shear
is used, the volumetric cumulative distribution function thinning effect of polymers is represented by the Meter's
F=[0,1] is taken to be piecewise linear when plotted against model (Meter and Bird 1964) with the final polymeric
ln(D). S0 is computed by integrating S0(D) over the size viscosity given by:
distribution. If tabular input with n rows is used: μ0 − μ s
μ p = μs + ,
6φ Pα −1
S 0 (D ) = sh , ⎛ γ
1+ ⎜
⎞
⎟ ( 17 )
D ⎜ γhf ⎟
⎝ ⎠
xi
6φ sh ⎛ 1 1 ⎞
ΔS i = ∫ x
ci dx = 6φ sh ci ⎜⎜ − ⎟⎟, where exponent Pα has a value between 1 and 2 (high shear
xi −1 e ⎝ Di −1 Di ⎠ thinning) and γhf is the shear rate γ at which (μ0-μs) is
( 14 )
F − Fi −1
ci = i , x = ln (D ), reduced to one half. The main influence of temperature is
xi − xi −1 through the solvent viscosity, μs. We use the same Arrhenius
n type equation for oil and water to compute the viscosity as a
S 0 = ∑ ΔS i . function of pressure and temperature:
2 ⎡ ⎛1 1 ⎞ ⎤
⎢ BT ⎜⎜ − ⎟+ B p ( p − po )⎥
⎟ ( 18 )
The shape factor φsh is defined as the ratio of the particle's μ = μoe ⎣⎢ ⎝ T To ⎠ ⎦⎥ ,
surface area to the surface area of a sphere with the same
volume, i.e. for spheres φsh=1. Examples for other shapes; φsh where μo is the reference viscosity at temperature To and
is 1.24 for a cube and 2.3 for a flat particle with dimensions pressure po. BT and Bp are empirical constants. The
0.1×1×1 in any units. The density probability function f(D) = temperature is in absolute units (°K). We use the fluid
dF/dD. viscosity in Eq. 18 to describe flow through the filter cake,
For each solid type present in the mud, a reference Eqs. 3 and 4. Yet another correction for the content of solids
porosity φo corresponding to zero overburden pressure and a is done before the effect of polymer is computed:
compressibility parameter nc are required.
μo
Polymer. For polymers we need to know their effect on μs = . ( 19 )
fluid viscosity, which is important in dynamic filtration and (1 − csm )2.5
for flow inside the porous medium. Secondly, we need the
size of the polymer molecules in order to evaluate Polymer size. We assume a lognormal distribution of Mw
entrapment in the filter cake and in the formation. Finally, we in the computations. The size of the polymer molecule in
must estimate the resistance to flow caused by the polymer in solution can be estimated from the intrinsic viscosity [η]
the filter cake. using Flory's empirical equation for the mean end-to-end
distance:
SPE 122241 5
The basic equation used for both solids and polymers (k=s,p) involved. The likely explanation is that small amounts of
allows for both trapping and re-entrance of particles back to trapped material will not be packed close together as a cake,
solution. The trapping rate for sub-component i is computed but rather be spread over a larger area within the narrow parts
by: of the pore space. One way to compensate for this is to
introduce a modification factor for the specific surface area
dσ i
= w (λ1,k ct ,i − λ2,k σ i ) , k=s,p.
u of the polymer, fpm.
( 36 )
dt φS w In the second method we compute a "pore throat
permeability" kpt within the narrow pore space defined as a
The units for trapped material and total mobile fraction fpt of the total pore space. A specific surface S0pt for
concentration of component i is in [pv] units so that over a this pore throat volume is computed by combining surface
time step Δt we have Δσi=-Δct,i. The unit for parameters λ1 area of the original rock, S0r(φ, k) obtained from Eq. 5, with
and λ2 is 1/length. For solids, the de-trapping term λ2 will be the surface area of the deposited solids and polymers. The
(close to) zero. If λ2 is zero, Eq. 36 says that a fraction of all rock porosity is corrected for deposited material, φpt. Then
passing particles equal to λ1 will be retained per length unit kpt(φpt, S0pt) is calculated by Eq. 5 and finally the new
and the expected traveling length E[x] is 1/λ1. To make the effective permeability is obtained by harmonic averaging
equation more general, we allow λ1 to be a function of with k.
particle size and λ2 to depend on fractional flow. For solids,
λ1 and λ2 are given by: ⎛ σ p +σs ⎞
φ pt = φ ⎜⎜1 − ⎟,
⎟
α11s ⎝ a ⎠
⎛ Di ⎞
λ1s = λ10 s ⎜
⎜ Dφ
⎟
⎟
, λ2 s = λ20 s (1 + λ21s f o ) , ( 37 )
S 0 pt =
( )
S 0 r (1 − φ )a + S 0 sσ s + S 0 pσ p φ
(1 − φ )a + (σ s + σ p )φ
⎝ ⎠ ,
4. Determine average filter cake kca=9.3·10-6 mD and filtrate volume is only through the change in the brine
<φc>=0.416 from average cake properties <φo>=0.515, viscosity, from 0.926 in the API test (25°C) down to 0.318 cp
<nc>=0.0244 bar-1 and Δp=58.6 bar, Eqs. 11, 12 and 26. in the core tests (90°C). The effect of composition is more
5. Finally, compute the static filtration parameter complex. In the core experiments we observe a reduction in
Cst=0.0309 ml/cm²s0.5 and fluid loss Vf(30 min) = 8.86 Vf when OCMA-clay is added (FD1 versus FD2). In the
ml from Eq. 4. calculations, the filter cake in FD2 will be thicker, but with a
Experimental and simulated filtrate volumes are plotted higher kca because of the reduced relative polymer content.
versus square root of time in Figure 3 (mud 1) and Figure 4 However, adding OCMA decreases the average solids
(mud 2). We observe that increasing the permeability (pore particle size which may potentially trap more polymer in the
size) of the discs results in a higher Vsp. This is because a cake. In the simulations, approximately 9% of the xanthan
smaller fraction of the solids will be held back at the disc and 30% of the starch is passing through the filter cake in
surface, and in addition the average size of retained particles FD1. In FD2, these numbers are reduced to zero for xanthan
will be larger, so that a larger volume is required to form the and 1% starch. The increased entrapment of polymer in the
initial cake. Vsp is more than doubled for mud 2 compared to filter cake will reduce kca and this is the reason, at least in the
mud 1. This is mainly because of the lower solid simulations, for the observed decrease in Vf between FD1 (12
concentration. All these differences are well reproduced in ml) and FD2 (10 ml).
the calculated curves as clearly illustrated in Figure 3 and 4. The relative viscosity of the filtrate produced from mud
The second observation is that the filtration curves are 3a was measured to be approximately 6, and showed shear
parallel to each other, i.e., the filtration parameter Cst does thinning properties. This verifies that a significant amount of
not depend on the properties of the formation, but only on the polymer was passing through the filter cake in FD1. The
mud properties and the differential pressure over the cake. lower level of simulated polymer migration through the cake
Figure 5 shows computed filtrate volumes with mud 1 for in FD2 agrees well with measurements done by Sánchez et
a wide range of permeabilities. We show both the al. (2003), where 1-2% of the polymer was measured (by
analytically computed Vf and Vsp ignoring the pressure drop TOC) to pass through the cake for a mud system similar to
over the disc, and simulated Vf (30 min) using a 10 cm core. mud 3b. However, the actual values of the model parameters
We observe that Vsp starts to increase above 10 mD, while the used to simulate this phenomenon are highly uncertain. In the
net filtration volume after the spurt loss (Vf-Vsp) is constant in simulation we identify the degree of polymer entrapment in
the analytical calculations. The simulated Vf deviates from the cake as a potential important variable for filtration rates,
the computed value at low permeability (0.01 mD) when the however, the properties used in these calculations are poorly
resistance over the core becomes significant compared to the known. The polymer molecular weight Mw is well known as a
filter cake resistance. Also at very high permeability (100 physical property, but its actual value and its size distribution
Darcy) the simulated Vf is lower than the computed one, are highly uncertain. The computation of molecular hydraulic
because of internal blockage of pores that reduces the size from [η] and the trapping parameter αp are also
effective pore size of the formation. uncertain. Finally, the OCMA properties used are also at best
approximate.
Blaxter sandstone Return permeability. The oil return permeabilities in
The results from three static filtration experiments using Table 10 are obtained after back-flooding the core with oil at
Blaxter sandstone are given in Table 10. The core dimensions increasing rates, 1, 2.5, 5 and 8 ml/min, 30 min at each rate.
were 8 cm length and 2'' diameter with an effective filtration Simulated and experimental return permeabilities are in the
area of 16.8 cm2. The permeability was approximately 150 same range, 80-90% of initial oil permeability koi. The
mD and the porosity 20%. The mud system used in core simulated differential pressure over the core and the
experiment FD1 contained 50 g/l sized CaCO3 with size development of the return permeability RKO during the back
distribution given in Figure 2 (Mud 3a), 4 g/l xanthan and 20 flood of FD1 are shown in Figure 7. We see that RKO is
g/l low reticulated starch. 40 g/l drilling solids (OCMA clay) highly dependent on the length of the back flooding period.
was added in experiment FD2 and FD3 (Mud 3b). All the The profile is governed by relative permeabilities, which are
core floods were conducted at 90°C. An overbalance pressure unknown, and viscosity ratio between oil and water, which
of 15 bar was used in FD1 and FD2, and 35 bar in FD3. API may be reduced because of polymer invasion. The relative
filtration (100 psi and 20°C) and HTHP (500 psi and 90°C) permeabilities used (see Table 7) are typical for water-wet
are also reported in Table 10. medium, with a final tuning to the results in FD1, except for
Properties of the polymers and solids are given in Table 2 the oil end-point permeability.
and Table 3. Figure 6 shows a good agreement between The average Sw shown in Figure 7 is still well beyond its
measured and computed shear thinning mud viscosity. Other initial value 0.2. Also shown is RKOE, an estimated return
model parameters used in the simulations are given in Table permeability at maximum oil saturation. If the flood was
4, Table 5 and Table 6. continued, RKO (0.824) should finally reach RKOE (0.916).
Filtration. All the experimental fluid loss volumes in Note that capillary pressure is ignored (zero) in these
Table 10 are well reproduced in the simulations. This simulations and that additional uncertainty in return
indicates that the effects of pressure, temperature and permeabilities may be caused by capillary end effects.
composition are all well taken care of in the simulations. The Damage profile. Another issue much discussed is how
pressure sensitivity (7-35 bar) is handled by the compressible the damage profile looks like in the invaded zones. The
filter cake model. The simulated effect of temperature on simulated damage profile (RKO) for FD1 is shown in Figure
10 SPE 122241
8 together with profiles of trapped CaCO3 and polymers. upscaling a short dynamic experiment to the field situation,
CaCO3 particles are trapped in the first mms of the core, which will be shown in a later section.
while polymer shows a deep penetration. It seems that The second dynamic case in Figure 14 is a filtration
trapped polymer has reached a plateau extending about half experiment taken from Dewan and Chenevert (2001). A
the filtrate invasion volume. The damage profile (RKO) bentonite mud with solid content csm= 0.18 is used.
follows the trapped polymer profile. The amount of external Appropriate rheologic mud properties were simulated using a
CaCO3 particles introduced during the spurt loss period is polymer component. The solid size distribution was taken
obviously very small, and there is no realistic way to from one of the other muds (Duratherm) used in their paper.
simulate any serious damage from trapped CaCO3 particles The simulated slowness curve in Figure 14 matches the
alone. The observed damage is therefore most likely caused experimental curve very well. The experiment starts with
by the polymer. dynamic filtration (100 s-1) at 500 psi, switches to static
Another experiment FD8 shows very serious plugging conditions and back to dynamic conditions again, and ends
when only the polymer mixture is injected (4 g/l xanthan+20 with increasing the pressure to 1500 psi. Each period was 30
g/l starch) shown in Figure 9. When simulating FD8 with min, which was insufficient to reach steady-state conditions.
polymer trapping parameters given in Table 5, only the first After the pressure increase, we first see a short transient
part of the pressure drop curve was matched. The simulated period where the compressible cake adapts to the new
pressure shown in Figure 9 was obtained using a polymer pressure. Then the curve will slowly progress towards the
modification factor fpm=0.15. We are primary interested in same steady-state limit as before the pressure increase,
the effect of much smaller polymer amounts and find that although at a slower rate because more mass must
using fpm=1 (as in Table 5) is fine. accumulate in the cake at the higher pressure.
In Figure 10, the back flooding of core B14 is extended The uncertainties in the particle size distribution and the
with an extra step where the "final" oil in place is displaced rheologic properties make the value of the model parameters
with a high viscous oil (M82) at reduced rate. Using the in Table 8 less general. The main purpose was, however, to
pressure derivative with respect to the front position of M82, demonstrate the capability of the dynamic model to
the permeability profile along the core can be estimated. We reproduce typical experimental behaviour.
approximate the front position by assuming constant So
throughout the core. In Figure 11 we show the estimated ko Particle retention
profile (Exp. VF) together with initial koi. We also show An experiment (Exp. 1) taken from Al-Abduwani et al.
profiles from simulation of the experiment and an estimated (2003) is used to demonstrate the pore throat trapping model.
ko-profile from the simulated pressure history (Sim VF). Note In this experiment, a dilute 20 ppm hematite (Fe2O3) particle
that the estimated increase in ko towards the outlet end in the suspension was injected into a Bentheimer sandstone core.
oil flood, which is to the left in Figure 11 and the mud side, is The particle size was stated to be in the range 0.1-5 µm with
a result of dispersion of the viscous oil front. The same 65% less than 1 µm. The size distribution used in our
behavior was seen when applying the method on simulated simulations is given in Figure 15 and additional information
data (Sim VF). in Table 3. Core data used in the simulations are L=12.2 cm,
B14 has a length of 4.7 cm, and was subjected to static k=1300 mD, φ=0.22 and Sw=1. The trapping parameters for
mud exposure for 50 hours (mud 3b), during which close to 2 solids tuned to this experiment (Table 5) are also used in the
pv filtrate had entered the core. The results in Figure 9 other simulations presented in this paper.
strongly indicate that ko is reduced throughout the whole By using dx/dt=uw/(φSw) in Eq. 36, setting λ2 to zero and
core, which supports the previous results (Figure 8) integrating, we find the effluent concentration for a given
indicating that polymers may cause deep, and rather evenly particle size to be:
distributed damage.
c = c 0 exp(− λ1 L ) , ( 46 )
Dynamic filtration where λ1(Dp) is given by Eq. 37. The computed size
Two examples of simulating dynamic experiments are dependent relative effluent concentration for the present core
given in Figure 12 to Figure 14. The first example is a core is shown in the size distribution plot (Figure 15). The average
test FD15 (Blaxter sandstone) with mud 3b (described effluent concentration is estimated from the size distribution
earlier). The experiment is simulated with three different sets to be 13.1 ppm when 20.7 ppm is injected.
of parameters (fn, nF, β, τy) given in Table 8. The plot of Vf Simulated and experimental results are compared in
versus time in Figure 12 has a non-uniform development of Figure 16 to Figure 19. If pore throat trapping is considered
its slope which may suggest some experimental problems. as the only retention mechanism, the simulated effluent
This is more clearly seen when plotting the slowness (inverse concentration seen in Figure 16 starts at the 13 ppm
velocity) versus Vf as done in Figure 13. Simulation DS3 puts analytical value and then declines slowly as an effect of
more weight on the last part and assumes a steady-state limit, previously trapped material. The simulation was extended
of approximately 6000 sec/cm, is reached. DS4 puts more beyond the experimental time period, and at some point,
weight on the first part, while DS2 is in between. DS2 and when an external filter cake has been formed, the effluent
DS4 have not reached steady-state within the 4 hours concentration dropped rapidly to zero. The experimental
timeframe. In DS2 no erosion at all is assumed, only a curve is different, it breaks through at a low concentration
reduced attachment rate. This illustrates the problem of and increases rather slowly towards a plateau after
SPE 122241 11
approximately 300 pv. This curve shape may be in part passing caused by the heterogeneous permeability
explained by a gradual change in the trapping pattern due to a distribution.
distribution in pore throat size, but it is also influenced by Figure 23 and Figure 24 show the computed permeability
other retention mechanisms. A closer match to the impairment along the well. The normalized oil permeability
experimental profile is obtained by including pore-lining for each zone is computed by two methods in Figure 23, (a)
adsorption in the experiment. as a productivity index computed from the rate in each zone
A total of 662 pv was injected in Exp. 1. The final and (b) analytically ignoring cross-flow between zones
distribution of hematite and permeability along the core is (RKO). The error introduced by ignoring cross-flow is very
shown in Figure 17 and Figure 18, respectively. In both small in this case, and almost zero for the total well with
cases, the simulated and experimental profiles match. The relative PIoil and RKO equal to 0.794 and 0.792, respectively,
experimental data represents averages over core segments. A in FS4. The same results are plotted in terms of skin
corresponding segmental permeability distribution computed distribution in Figure 24 where S (current) corresponds to
from the simulation results is almost overlapping the RKO. S (kabs) is computed from the reduction in absolute
experimental curve shown in Figure 18. permeability, corresponding approximately to RKOE. The
Figure 19 shows the specific surface area of trapped radial distribution of RKO and RKOE along the well is given
hematite particles throughout the core from the simulation. A in Figure 25.
corresponding effective "average" particle diameter We conclude from the upscaling of laboratory results
computed as Dp=6φsh/S0 is also shown. Obviously, average with mud 3b, that no significant real damage to the formation
size of trapped particles is larger at the inlet end, while should be expected from externally introduced particles in
smaller particles will penetrate deeper. the present case. Other internal damage processes are not
evaluated in this example. The "real" damage as indicated by
Fines migration and clay swelling RKOE in Figure 25 is in line with the observed polymer
Simulations of fines migration and clay swelling upon damage observed in the laboratory experiments. One should,
injection of low salinity brine are demonstrated in Figure 20. however, expect a temporary reduced productivity because of
In the fines simulation, 1% of the rock matrix is defined as incomplete displacement of mud filtrate. The size and
fines, held in place using the pore lining adsorption model. In duration of this decline depend on filtrate penetration depth,
the clay swelling simulations, 2% of the rock matrix is relative permeability and the variance of the permeability
defined as swelling clay with a maximum swelling capacity distribution.
of 10 (Qswmax in Eq. 45). Other parameters in Eq. 45 are The fluid loss volumes in Table 11 are very high, in fact
ksw1=0.0002 meq/(ml·min) and ksw2=0.001 meq/(ml·min). The an order of magnitude larger than expected. The worst case is
effective salinity, Cse, is essential equal to the NaCl FS3, in which a high erosion rate is assumed. Using a steady-
concentration indicated in Figure 20 and 0.7 meq/ml with the state slowness of 5628 sec/cm corresponding to that obtained
initial brine. for DS3 in Figure 13, but corrected for the different shear
rate in FS3, and the filtration area of the well (πDwLw = 1094
Field scale simulations m²), we compute a filtrate volume of 1750 m³ compared to
To test the upscaling to field conditions, we use mud 3b 1717 m³ given in Table 11. Similarly, the static fluid loss for
on a horizontal well from the North Sea. Permeability and the well is estimated to 55.6 m³ using the filtration parameter
porosity distribution along the wellbore is given in Figure 21. Cst from FD3 (0.0032 m³/m²hr0.5), which compares well to
We set the wellbore diameter to 0.24 m, and use a constant the simulated 55 m³ in Table 11. The deviations are because
pressure boundary at 100 m in our 2D radial model. The of differences in overbalance pressure and low permeable
temperature is 90°C. Filtrate invasion is simulated for a zones offering flow resistance comparable to the mudcake
period of 250 hours with a overbalance pressure of 86 bar. resistance. These computations indicate that the field scale
Then we simulate cleanup by back producing the well using a simulations are correct in terms of upscaling the laboratory
30 bar drawdown for 24 hours. Four cases are run, FS1 with results, but our description of the field conditions may be
static filtration and FS2-FS4 with dynamic conditions at insufficient. Of course, our interpretation of the laboratory
shear rate of 170 s-1. Cases FS2-FS4 correspond to various results is very important, and in the case of FD15, the
matches to the dynamic core experiment FD15, and dynamic estimated dynamic parameters are very uncertain. But all the
filtration simulations DS2-DS4 shown in Table 8. interpretations used predicted large fluid loss volumes in the
The simulations are summarized by main numbers in field case. Other explanations should be sought.
Table 11. Typical results from the simulations are One obvious thing is that the formation will be gradually
demonstrated by simulation FS4 in Figure 22 to Figure 25. exposed to the mud as the well is drilled. In the simulations,
Figure 22 shows the saturation distribution around the well the total well was exposed from time zero. Exposing the
after mud exposure for 200 hours and after the 24 hours back formation according to a well penetration rate of 10 m/hr is
production period. The variation in penetration depth is shown in Figure 26 (FS4b). This will, however, only result in
mainly a result of the porosity distribution along the wellbore a moderate reduction of the filtrate volume from 702 to 535
because filtration rate through the filter cake is independent m³. Simulating with periods of 50 hrs dynamic + 25 hrs
on the formation properties. After the back production, we static, and last 50 hrs static, reduces the loss to 251 m³
see that there is still significant amount of water around the (FS4c). In simulation FS2, only reduced mudcake growth and
well, indicating that substantial production is needed to re- no erosion are assumed. Applying the same periodic
establish original fluids saturation. This is a result of by- dynamic/static history on FS2 reduces the fluid loss from 504
12 SPE 122241
γhf polymer viscosity parameter, [t-1] SPE 17617 presented at the SPE International Meeting on
Petroleum Engineering, Tianjin, China.
[η] 3
intrinsic polymer viscosity, [L /m] Aziz, K. and Settari, A. (1979). Petroleum Reservoir Simulation.
λnk n=1, 2, 10, 20, 21, 22: trapping parameters for London and New York, Elsevier Applied Science Publishers.
solids or polymer (k=s, p), [L-1] Bezemer, C. and Havenaar, I. (1966). Filtration Behaviour of
n=3, 4, 40, 41: adsorption parameters for solids or Circulating Drilling Fluids. SPEJ 6 (4): 292-298.
polymers (k=s,p), [L-1] Bourgoyne, A. T., Chenevert, M. E., Millhelm, K. K. and Young, F.
S. (1986). Applied Drilling Engineering, SPE.
μ viscosity [m/Lt] Cerasi, P., Ladva, H. K., Bradbury, A. J. and Soga, K. (2001).
ν dimensionless intrinsic viscosity Measurements of the Mechanical Properties of Filtercakes.
σ trapped material, [pv] SPE 68948 presented at the SPE European Formation Damage
τ tortuosity parameter, (path length/length)2 Conference, The Hauge, The Netherlands.
τy yield strength (filter cake), [m/Lt2] Civan, F. (2000). Reservoir Formation Damage : Fundamentals,
Modelling, Assessment, and Mitigation. Houston, Texas, Gulf
φ porosity
Publishing Company.
Dewan, J. T. and Chenevert, M. E. (2001). A Model for Filtration of
Subscripts Water-base Mud During Drilling: Determination of Mudcake
a average Parameters. Petrophysics 42 (3): 237-250.
c cake (filter cake) Ding, Y., Longeron, D., Renard, G. and Audibert, A. (2002).
cr critical Modelling of Both Near-Wellbore Damage and Natural
d dynamic Cleanup of Horizontal Wells Drilled With a Water-Based Mud.
e end point SPE 68951 presented at the SPE International Symposium and
m mud Exhibition on Formation Damage Control, Lafayette, Lousiana.
Ferguson, C. K. and Klotz, J. A. (1954). Filtration from Mud During
o oil
Drilling. JPT (February): 30-42.
p particle, polymer Fordham, E. J., Allen, D. F. and Ladva, H. K. J. (1991). The
r relative Principle of a Critical Invasion Rate and Its Implications for
s solids Log Interpretation. SPE 22539 presented at the SPE Annual
sp spurt loss Technical Conference and Exhibition, Dallas, Texas, U.S.A.
sw water in swelling clay Gruesbeck, C. and Collins, R. E. (1982). Entrainment and
t trapping Deposition of Fine Particles in Porous Media. SPEJ 22 (6):
w water 847-855.
φ pore property Hirasaki, G. J. (1982). Ion Exchange with Clays in the Presence of
Surfactant. SPEJ 22 (2).
Hirasaki, G. J. and Lawson, J. B. (1982). An Electrostatic Approach
Acknowledgements to the Association of Sodium and Calcium with Surfactant
The major part of this work was done within the Well Micelles. SPE 10921.
Productivity 2002 project supported by the European Union Lake, L. W. (1989). Enhanced Oil Recovery, Prentice-Hall, Inc.
(EU) 5th Frame programme. The authors would like to thank Meter, D. M. and Bird, R. B. (1964). Tube Flow of Non-Newtonian
EU and the industrial sponsors of the project, Agip BP Polymer Solutions: Part I. Laminar Flow and Rheological
Exploration, ChevronTexaco, Norsk Hydro and Shell Models. A.I.Ch.E 10 (6): 878-881.
Norway. We would also like to thank the other partners in the Outmans, H. D. (1963). Mechanics of Static and Dynamic Filtration
project: ResLab, Eni Tecnologi, Rhodia and M-I Drilling In the Borehole. SPEJ (September): 236-244.
Pope, G. A. and Nelson, R. C. (1978). A Chemical Flooding
Fluids. Experimental data used in this paper is mainly Compositional Simulator. SPEJ (Oct.): 339-354.
supplied by ResLab and Eni Tecnologi. Sánchez, E., Audibert-Hayet, A. and Rousseau, L. (2003). Influence
The second phase of this work was supported by Shell, of drill-in fluids composition on formation damage. SPE 82274
BP and StatoilHydro. presented at the SPE European Formation Damage Conference,
Supply of lab test data for model calibrations by The Hague, The Netherlands.
Halliburton is much appreciated. Semmelbeck, M. E., Dewan, J. T. and Holditch, S. A. (1995).
Invasion-Based Method For Estimating Permeability From
Logs. SPE 30581 presented at the SPE Annual Technical
References Conference and Exhibition, Dallas, U.S.A.
Al-Abduwani, F. A. H., Hime, G. and Alvarez, A. (2005). New Sharma, M. M. and Yortsos, Y. C. (1986). Permeability Impairment
Experimental and Modelling Approach for the Quantification Due to Fines Migration in Sandstones. SPE 14819 presented at
of Internal Filtration. SPE 94634 presented at the SPE the SPE Symposium on Formation Damage Control, Lafayette,
European Formation Damage Conference, Scheveningen, The LA.
Netherlands. Sorbie, K. S. (1991). Polymer-Improved Oil Recovery, Blackie and
Al-Abduwani, F. A. H., Shirzadi, A., van der Broek, W. M. G. T. Son Ltd.
and Currie, P. K. (2003). Formation Damage vs. Solid Particles Suryanarayana, P. V., Wu, Z., Ramalho, J. and Himes, R. (2005).
Deposition Profile during Laboratory Simulated PWRI. SPE Dynamic Modelling of Invasion Damage and Impact on
82235 presented at the SPE European Formation Damage Production in Horizontal Wells. SPE 95861 presented at the
Conference, The Hague, The Netherlands. SPE Annual Technical Conference and Exhibition, Dallas,
Arthur, K. G. and Peden, J. M. (1988). The Evaluation of Drilling Texas, U.S.A.
Fluid Filter Cake Properties and Their Influence on Fluid Loss. Wennberg, K. E. and Sharma, M. M. (1997). Determination of the
Filtration Coefficient and the Transition Time for Water
14 SPE 122241
Injection. SPE 38181 presented at the SPE European The denominator is given in A.3 and the numerator integral
Formation Damage Conference, The Hague, The Netherlands. is after using the previous substitution, z=1-φ:
Wu, J., Torres-Verdin, C., Sepehrnoori, K. and Delshad, M. (2004).
Numerical Simulation of Mud-Filtrate Invasion in Deviated Δp z ( Δp )
−1 ⎡1 2⎤
Wells. SPE Reservoir Evaluation & Engineering 7 (2): 143- ∫ φc kdp n = 2n τS 2 ⎢⎣ z + 3 ln z − 3z + 12 z ⎥
⎦ z (0)
. (A. 8)
154. 0 c 0
By using
1
Appendix Fφ ( pn ) = + 3 ln z − 3z + 12 z 2 , (A. 9)
Computation of average filter cake properties z
The compressible cake permeability, kc(pn) can be we get the average porosity
Fφ (Δp ) − Fφ (0)
expressed using Eqs. 5 and 10:
φ = .
φo3e −3nc p n
(A. 10)
Fk (Δp ) − Fk (0)
k c ( pn ) =
( )S
2
. (A.1)
2τ 1 − φo e − nc p n 2
0
Substituting Eq. A.1 into Eq. 8 yields: Table 1: Mud compositions (g/liter).
Δp Δp Component Mud 1 Mud 2 Mud 3a Mud 3b
1 φo3e −3nc p n
∫ kc dpn = 2τS 2 ∫
Polymer (P1,P2) 25 25
dp n .
(1 − φ e )
(A.2)
− nc p n 2 Xanthan 4 4
0 0 0 o Starch 20 20 20
CaCO3 100 100 50 50
Defining z = 1 − φ = 1 − e − nc pn in Eq. A.2 yields: Drilling solids (OCMA) 40
Δp Barite 155
1
z2
(1 − z )2 dz,
∫ k c dp n = 2τS 2 n ∫ z2 Table 2: Polymer properties; numbers in italic do not influence
0 0 c z1
(A.3) static fluid loss computation.
Δp z ( p = Δp )
−1 ⎡1 ⎤ 2
∫ k c dp n = 2τS 2 n
Name P1 P2 Xanthan Starch
⎢ z + 2 ln z − z ⎥ .
0 0 c ⎣ ⎦ z1 ( p =0) Viscosity, low shear
[η] (ml/g) 1500 3200 5500 600
Define:
k' 0.4 0.4 0.4 0.4
1
Fk ( p n ) = + 2 ln z − z .
k'' 0.01 0.01 0.07 0.07
(A. 4)
z Mw exponent, a 0.8 0.8 0.8 0.5
BvT [1/°C] 0 0 0.006 0.001
The average filter cake permeability is then:
T0 (°C) 20 20 20 20
1 Fk (0) − Fk (Δp ) Shear thinning
k ca = ⋅ . (A. 5)
2τS 02 nc Δp Pα 1.7 1.7 1.8 1.5
γhf (s-1) 2 2 1.5 5
Let kca(x)=kca(pn(x)) be the average permeability from the Size parameters:
cake surface to depth x within the cake. Because the rate and 7 7 6 6
Mw (g/mol) 10 10 3·10 5·10
viscosity of the permeate flow are constant, the product
σ ln(Mw) 1 1 2 2
kca(pn(x))·pn(x)/x must be constant. If we define x as the
relative position within the cake as a function of pn then the Dch (µm) 0.0008 0.0008 0.001 0.0012
distance from the cake surface to pressure pn(x) is: ρ (g/cm³) 1 1 1 1
Δhc Fk ( p n ) − Fk (0)
Computed properties:
x( p n ) = = . (A. 6) S0 (m²/cm³) 5000 5000 4000 3333
hc Fk (Δp ) − Fk (0) Mean Dhp (µm) 0.414 0.498 0.428 0.242
In order to calculate the flow rate, we also need to know the Filtercake properties:
thickness (dc) of the cake. This is readily obtained from mass φo 0.85 0.85 0.85 0.85
balance once the average porosity of the cake is found. Using nc (bar )
-1
0.095 0.095 0.08 0.08
Eqs. 7 and 8 we can write the average porosity as:
Δp
1 ∫ φk c dp n
φ = ∫ φ c dx = 0
. (A. 7)
Δp
∫ k c dp n
0
0
SPE 122241 15
0.7 16
10
Vf (ml)
0.5
8
0.45
6
0.4
4
0.35
n c =0.04
2 20 µm 3 60 µm 3
0.3
20 µm sim 60 µm sim
0 5 10 15 20 25
0
Cake overburden pressure (bar) 0 1 2 3 4 5 6
100
Vsp 1 Vf-Vsp (1)
0.8
Vf 1 Vf sim (1)
Vsp 2
0.6
Filtrate volume (ml)
10
0.4
0.2 Mud 1 1
Mud 2
Mud 3
0
0.1 1 10 100 1000
1000
Viscosity (mPa·s)
4 100
2
20 µm 1 60 µm 1 10
20 µm sim 60 µm sim
0
0 1 2 3 4 5 6
1
Sqrt (min) 1 10 100 1000 10000
Figure 3: HTHP filtration at 85°C and 850 psi, experimental and Shear rate (1/sec)
simulated for 20 and 60 µm filter discs, Mud 1. (Ac=45cm²).
Figure 6: Mud 3a, experimental (converted Fann viscosities) and
calculated, 25°C.
SPE 122241 17
0.5 0.5
0.6 0.8
Figure 7: Simulated back production, FD1. Differential pressure PVinj B14: back production
(dp), current (RKO) and maximum (RKOE) oil return permeability,
and average Sw. Figure 10: Back production with low viscosity oil (CLS370) and
high viscous oil (M82) in experiment B14.
160
1 0.001
140
Trapped CaCO3×1 and
RKOE 100
0.6
Ko
0.0006
Ko/Koi
RKO 80
CaCO3pt 60
0.4 0.0004 Sim. Sim VF
xanthan 40
Exp. VF Kio
0.2 starch 0.0002 20
Ko max
0
0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Core position, x/L B14
x/L
Figure 11: Interpretation of oil permeability profile (Exp. VF) from
Figure 8: Simulated oil permeability reduction, current (RKO) and
viscous oil flood, B14.
estimated final (RKOE), and pore throat trapped solids (CaCO3)
and polymers.
50
45 exp.
35
40 DS2
Sim
Filtrate volum (ml)
DS3
30 Exp. FD8 35
DS4
30
Pressure drop (bar)
25
25
20 20
15
15
10
10 5
FD15
0
5 0 50 100 150 200 250
0 Time (min)
0 1 2 3 4 5
Figure 12: Dynamic filtration Blaxter sandstone, Mud 3b.
PV injected Simulation of experiment FD15.
Figure 9: XC+Starch injection, 150 mD Blaxter sandstone.
18 SPE 122241
0.008
10000
DS3
7000 0.004
DS4
6000 0.003
5000 0.002
0.001
4000
0
3000 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/L
2000
1000 Figure 17: Experimental (Al-Abduwani et al. 2003) and simulated
FD15 hematite distribution in core after 662 pv injection.
0
0 10 20 30 40 50
Vf (ml)
1
Normalised permeability
Figure 13: Slowness vs. Vf in dynamic filtration experiment FD15.
0.8
0.6
60000
Exp. Mud 98067 0.4
Sim Sim
50000
0.2 Sim. segment
Slowness (sec/cm)
Exp. segment
40000 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/L
30000
Figure 18: Experimental (Al-Abduwani et al. 2003) and simulated
20000 permeability profile in core after 662 pv injection.
10000 12 3
500 psi 500 psi 500 psi 1500 psi
0
8 2
0 30 60 90 120
(m²/cm³)
6 1.5
Time (min)
4 1
Figure 14: Experimental, replotted from Dewan and Chenevert S0
(2001), and simulated dynamic filtration experiment. 2 0.5
Dp (µm).
0 0
F Hematite
Ceff
particle size of hematite in core after 662 pv injection (simulated).
0.6 0.6
Ceff (Dlim)
0.4 0.4
1000
0.2 0.2
Fines
0 0 Swelling clay
0.1 1 10 100
Particle dimeter (µm)
Pressure drop (bar)
16 1
Effluent hematite (ppm)
Sim
14
Exp.1 0.5 N 0.1 N 0.01 N 0.001 N
12 Exp.2 Brine Brine
NaCl NaCl NaCl NaCl
10 Exp.4
Sim A=0 0.1
8 0 5 10 15
6 PV injected
4
Figure 20: Simulation of fines mobilization and clay swelling.
2
0
0 200 400 600 800 1000 1200 1400 1600 1800
PV injected
PI (relative)
lg(mD) fraction
0.6
2.686 0.248 0.5
0.4
200 200 0.3
0.2
0.1
1.765 0.196 0
0 200 400 600 800 1000 1200 1400
400 400 Wellbore (m)
0.845 0.145
Figure 23: Sim. FS4. Relative productivity index for oil, from
simulated well rate and computed ignoring crossflow (RKO). After
Wellbore (m)
Skin
1000 1000 0.1
0.01
0.001
1200 1200
0 200 400 600 800 1000 1200 1400
Wellbore (m)
600
Wellbore (m)
600
400 400 0.250 0.850
0.521 0.520
800 800
Wellbore (m)
Wellbore (m)
1000 1000
800 800
0.182 0.180
1200 1200
1000 1000
1400 1400
1200 1200
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Radial direction (m) Radial direction (m)
1400 1400
Figure 25: Normalized oil permeability after 24 hours production
RKO (left) and RKOE computed at maximum So (right), sim. FS4.
0.0 1.0 2.0 0.0 1.0 2.0
Radial direction (m) Radial direction (m)
Figure 22: Oil distribution around well after 200 hours mud
-1
exposure at 86 bar, 170 s (left) and after 24 hours back
production at 30 bar drawdown (right).
20 SPE 122241
0.696
400
0.532
Wellbore (m)
600
0.369
800
0.205
1000
1200
1400