Professional Documents
Culture Documents
net/publication/266671387
CITATIONS READS
58 520
4 authors, including:
Elizabeth Spiteri
Oil Search Alaska LLC
4 PUBLICATIONS 838 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Multicomponent, multiphase flow in porous media with temperature variation View project
All content following this page was uploaded by Elizabeth Spiteri on 12 June 2015.
parameter model, and constitutes the basis for a number questration in heterogeneous saline aquifers. In the final
of relative permeability hysteresis models. Other trap- section of this paper we gather the main conclusions and
ping models are those of Jerauld12 and Carlson.13 These anticipate ongoing and future work.
models are suitable for their specific applications but, as
we show in this paper, they have limited applicability to Summary of model equations
intermediate-wet and oil-wet media.
In this section, for clarity, we present the trapping and
Land11 pioneered the definition of a “flowing satura-
hysteresis model. A detailed description of the tools used
tion”, and proposed to estimate the imbibition relative per-
and a physical interpretation of the results is provided in
meability at a given actual saturation as the drainage rel-
the subsequent sections.
ative permeability evaluated at a modeled flowing satura-
tion. Land’s imbibition model11 gives accurate predictions A trapping model attempts to capture the residual
for water-wet media,14 but fails to capture essential trends (trapped) oil saturation Sot after waterflooding, given an
when the porous medium is weakly or strongly wetting to initial oil saturation Soi achieved during primary drainage
oil. The two-phase hysteresis models that are typically of oil into a water-filled medium. Therefore, the trapping
used in reservoir simulators are those by Carlson13 and model may be viewed as a curve on a diagram of initial oil
Killough.15 A three-phase hysteresis model that accounts saturation versus residual oil saturation (initial–residual or
for essential physics during cyclic flooding was proposed IR curve). The development of a new trapping model is
by Larsen and Skauge.16 These models have been eval- motivated by the fact that, in contrast with most com-
uated in terms of their ability to reproduce experimental mon trapping models,11, 12 initial–residual curves obtained
data17, 18 and their impact in reservoir simulation of water- from pore-network simulations are not monotonic when
alternating-gas injection.18, 19 Other models are those by the medium is not strongly water-wet. This non-monotonic
Lenhard and Parker,20 Jerauld12 and Blunt.21 More re- trend has been confirmed experimentally.10
cently, hysteresis models have been proposed specifically The simplest functional form that allows us to reproduce
for porous media of mixed wettability.22–24 a non-monotonic behavior is the quadratic expression:
All of the hysteresis models above require a bounding 2
drainage curve and either: (1) a waterflood curve as input; Sot = αSoi − βSoi . (1)
or (2) a calculated waterflood curve using Land’s model.
The task of experimentally determining the bounding wa- Parameters α and β correspond to the initial slope and the
terflood curves from core samples is arduous, and the de- curvature of this curve, respectively. They must satisfy the
velopment of an empirical model that is applicable to non following requirements:
water-wet media is desirable. In this paper, we introduce
a relative permeability hysteresis model that does not re- 0 ≤ α ≤ 1, β ≥ 0. (2)
quire a bounding waterflood curve, and whose parameters
may be correlated to rock properties such as wettability The model contains two parameters, α and β, which must
and pore structure. be calibrated for each rock–fluid system. Therefore, they
Since it is difficult to probe the full range of relative will depend on the pore-space morphology and the wetta-
permeability hysteresis for different wettabilities experi- bility characteristics of the medium. Typical dependence
mentally, we use a numerical tool –pore-scale modeling– of these two parameters on the wettability of the medium
to predict the trends in residual saturation and relative is discussed and explained later in the paper.
permeability. As we discuss later, pore-scale modeling is Inspired by Land’s model,11 we propose to calculate the
i
currently able to predict recoveries and relative permeabil- waterflood relative permeability kro at the actual oil sat-
d
ities for media of different wettability reliably.25–30 We will uration So as the drainage relative permeability kro at a
use these predictions as a starting point to explore the be- different flowing saturation:
havior beyond the range probed experimentally.
i d
In summary, this paper contains two main results: kro (So ) = kro (Sof ). (3)
1. A new model of trapping and waterflood relative
The flowing saturation is evaluated by means of the fol-
permeability, which is able to capture the behavior
lowing expression:
predicted by pore-network simulations for the entire
range of wettability conditions. ·
1
Sof = (α − 1)
2. A demonstration of the importance of nonwetting- 2β
phase trapping and relative permeability hysteresis for p
¸
2
+ (α − 1) + 4β[So − Sot + γ(So − Sot )(So − Soi )] ,
the assessment of geological CO2 sequestration pro-
cesses.
(4)
In the next section we present a summary of the equa-
tions of the proposed model. We follow with an overview of where Sot is the trapped oil saturation given by Eq. (1),
pore-scale modeling of trapping and relative permeability and γ is a parameter related to the presence of film flow.
hysteresis. We then present a new model of trapping and It is dependent on the wettability characteristics of the
waterflood relative permeability. As an application of the medium, and is typically negative for strongly water-wet
importance of trapping and hysteresis we perform three- media and positive for weakly and strongly oil-wet media.
dimensional simulations of realistic scenarios of CO2 se- This behavior will be discussed in detail later.
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 3
Pore-scale modeling of trapping and hysteresis a cube of 3×3×3 mm3 containing 12,349 pores and 26,146
In pore-network modeling the pore space is described by throats. The absolute permeability of the rock is 2287 mD
a network of pores connected by throats with an idealized and the net porosity is 0.183, where 0.0583 is clay-bound
geometry. A set of physically-based rules describe the con- or micro-porosity.
figuration of the fluids within each pore and throat, as well We investigated sequences of two consecutive displace-
as the mechanisms for the displacement of one fluid by an- ments: primary drainage (oil invasion) and waterflooding.
other. This approach was pioneered by Fatt31 and has re- During primary oil drainage the network, which is initially
ceived increasing attention over the past decade. Blunt32 filled with water, is assumed to be strongly water-wet with
and Blunt et al.33 provide a detailed description of the a receding contact angle θr = 0◦ . As the oil invades the
fundamentals and applications of pore-network modeling, largest pores first in piston-like displacement, the water re-
together with an extensive literature review. One of the cedes and is squeezed to the crevices and pore throats until
successful application areas of pore-network models is the a very high capillary pressure or a target oil saturation is
prediction of multiphase flow properties, such as capillary reached. At this point, the surface of the rock in con-
pressure and relative permeability. This success hinges on tact with oil will undergo wettability alteration, while the
the following: corners and elements that still contain only water remain
strongly water-wet. Wettability alteration is accounted for
1. The ability to reproduce the essential geometric fea- by changing the contact angle. In principle, one could
tures of the pore space of real rocks. A realis- change the advancing contact angle θa and the receding
tic three-dimensional pore-space characterization may contact angle θr independently. In this work, however, we
be obtained in a variety of ways: assembly of two- used a correlation proposed by Morrow43 to link both the
dimensional sections to form a three-dimensional im- advancing and receding contact angles with an intrinsic
age;34 direct X-ray microtomography of the three- contact angle θi . This relationship is shown in Fig. 1.
dimensional pore space;35 stochastic 3D modeling
with statistics inferred from two-dimensional thin sec-
tions;36 process-based reconstruction in which grain
deposition, compaction and cementation are mod-
eled.37, 38
2. The ability to capture wettability effects. Most pore-
network models used today introduce wettability ef-
fects based on the pore-level scenario of wetting pro-
posed by Kovscek et al.6 Their model mimics the
saturation change typical of a hydrocarbon reservoir.
The medium is initially filled with water, and the rock
surfaces are water-wet. During oil migration, the oil
invades the pore space, altering the wettability of the
solid surface in contact with the oil. In this fashion,
the network displays mixed wettability: a fraction of
a pore or throat may be oil-wet, while the corners and
crevices not in contact with the oil remain water-wet.
The combination of realistic pore geometry/topology and Fig. 1— Relationship between receding and advancing contact
correct characterization of displacement and trapping angles on a rough surface, as a function of intrinsic contact
mechanisms has allowed pore-network models to pre- angle measured at rest on a smooth surface.43 [Figure from
dict hysteretic capillary pressure and relative permeability Valvatne and Blunt.28 ]
curves under a wide range of wettability characteristics.
Experimental measurements of hysteretic relative per- During waterflooding there are several physical mech-
meability for mixed-wet and oil-wet media are scarce.10, 39 anisms by which the water can displace the oil in place.2
Given the success of pore-network models to reproduce ex- These mechanisms include piston-type displacement, coop-
perimental data,26–28 in this work we have adopted the use erative pore-body filling and snap-off. The predominance
of pore-network modeling as a way to investigate the full of any given displacement mechanism is strongly depen-
spectrum of wettability conditions. Pore-network simula- dent on the wettability (specified by the advancing contact
tions results are taken as “data” to develop and validate angle). These displacement processes and their implemen-
empirical trapping and hysteresis models. tation are described in detail in the literature.28, 29, 40, 41
After individual displacement events the transport proper-
Description of the pore-network simulations. We ties are calculated. The equations for absolute permeabil-
used the two-phase flow pore-network simulator developed ity, relative permeability and other transport parameters
by Valvatne and Blunt.28 The model has similarities with can also be found in the literature.28, 40, 41, 44 Following
other network models.29, 40, 41 A full description of the this procedure, pore-network simulations have been shown
model is given in the references above and will be omit- to reproduce experimental capillary pressure and relative
ted here. permeability curves both in primary drainage and water-
We used a three-dimensional pore-network of a Berea flooding40 and for a variety of wettability conditions.26–28
sandstone developed by Bakke and Øren.42 The model is We investigated the full range of wetting conditions after
4 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
S
Trapped oil saturation (Sot)
o,max
0.6 0.6
S
oi i
0.4 0.4 k
ro(w)
d
k
ro(w)
0.2
0.2
0 S
ot
S
ot,max
150 0
0 0.2 0.4 0.6 0.8 1
Co 0.8 Oil saturation (S )
nta100 0.6 o
ct S )
an
gle 50 0.2
0.4
r a t ion ( oi
satu
(θ) 0 l oil Fig. 4— Waterflood relative permeability curves that cross
Initia due to the nonmonotonic trapping relationship.
imbibition
0.8
Contact angle (180)
1
0.6 drainage
0.6
0.2
0.4
0
0 0.2 0.4 0.6 0.8 1
Gas saturation (S ) 0.2
g
Oil saturation (S )
ro
Imbibition o
0.8
Fig. 7— Oil relative permeabilities generated from pore-
0.6 network simulations in strongly oil-wet media.
0.4
Development and validation of a new model of
relative permeability hysteresis
0.2
In the previous section we used pore-network modeling to
highlight the following features of wettability effects on the
0 waterflood relative permeability:
0 0.2 0.4 0.6 0.8 1
Oil saturation (S )
o 1. The initial–residual (IR) curves are not monotonic for
media that are not strongly water-wet (Fig. 2).
Fig. 5— Relative permeability curves of the nonwetting phase
for water/gas (top) and water/oil (bottom) systems from
Oak’s46 experimental data in water-wet Berea sandstone. 2. For intermediate-wet and oil-wet media, the scanning
curves of oil relative permeability may cross (Fig. 4).
)
meability models that incorporate hysteresis12, 15, 16, 20–22
ro
are based on this trapping model. It was developed to 0.8
0.8
curves since any shifting will make the endpoint trapped
ro
S
o,max
0.6 d
kro Jerauld trapping model. Jerauld’s trapping model12
is an extension of the Land trapping model that accounts
ki
for the “plateau” observed in the initial–residual (IR)
0.4 ro
curves for mixed-wet rocks.50 The trapped nonwetting-
S phase saturation is given by:
oi
0.2
Soi
Sot = 1+b/C
. (8)
S (S )
ot oi
Sot,max 1 + C Soi
0
0 0.2 0.4 0.6 0.8 1
Oil saturation (S ) In the original publication,12 the expression of the trapped
o
saturation Sot was given in terms of the residual satura-
tion achieved when Soi = 1, Sor = 1/(C + 1). Jerauld
Fig. 8— Parameters required in the evaluation and application introduced a second tuning parameter b in addition to the
of the Land trapping model. Land coefficient. If this parameter is set to zero, Jerauld’s
model reduces to the Land trapping model. When this pa-
rameter is equal to one, the trapping curve has a zero slope
Carlson trapping model. A simplified hysteresis at Soi = 1. Although Jerauld argued that the IR curves
model proposed by Carlson13 implicitly defines a trapping should not have a negative slope, his model allows for such
model. The Carlson model requires the bounding drainage behavior if b ≥ 1.
and waterflood curves. The trapped oil saturation is de- Although the fit of Jerauld’s model to the pore-network
termined by shifting the bounding waterflood curve to in- data was good for water-wet and intermediate-wet condi-
tersect the intermediate initial oil saturation at the flow tions (for which the model was designed), it was not as
reversal. The idea behind Carlson’s interpretation is to satisfactory for strongly oil-wet media. One of the reasons
use the model of the waterflood relative permeability scan- is that Jerauld’s model assumes that the IR curve has a
ning curves as being parallel to each other. This geometric unit slope near the origin. This behavior does not conform
extrapolation procedure is illustrated in Fig. 9. to pore-network predictions (see the curves corresponding
The trapped nonwetting-phase saturation is computed to θ = 110◦ and 160◦ in Fig. 2).
as
Sot = Sot,max − ∆So , (7) A new trapping model. We notice that the shapes
where ∆So is the shift in the waterflood scanning curve of the trapping curves (Fig. 2) may be fit to a parabola.
with respect to the imbibition bounding curve (see Fig. 9). We establish the following simple quadratic relationship
8 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S )
oi oi oi oi
Contact angle 100 Contact angle 120 Contact angle 140 Contact angle 160
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S )
oi oi oi oi
Fig. 10— Performance of the new trapping model: initial–residual curves calibrated against pore-network simulation data.
between the trapped oil saturation Sot and the initial oil
saturation Soi :
2
Sot = αSoi − βSoi . (9)
2
The parameters α and β correspond to the initial slope α
and the curvature of this curve, respectively. These pa- β
rameters were tuned to minimize the least squared error
1.5
between the model prediction and the pore-network sim-
ulation data. The optimization is constrained by the fol-
lowing restrictions:
1
0 ≤ α ≤ 1, β ≥ 0. (10)
trapping model reproduces the observed trapping behavior we propose is an extension of Land’s model to account
for all wettability conditions, even if it may slightly over- for the different pore occupancies at different wettability
estimate the trapping of the bounding waterflood curve conditions.
(Soi = So,max = 1 − Swc ). We should also mention that
the pore-network simulator is likely to underestimate the Land waterflood model. As a prelude to the devel-
trapping for the bounding curves due to an overly opti- opment of the new waterflood relative permeability model
mistic criterion for the stability of oil layers. If a new, proposed in this work, we revisit the derivation of Land’s
free-energy based stability criterion is implemented,45 we relative permeability model. The basis of Land’s formula-
expect a better agreement between the trapping model and i
tion is to express the waterflood relative permeability kro
the pore-network predictions. at a given oil saturation (So ) as being equal to the drainage
d
When the parameters calculated from the optimization permeability kro evaluated at a flowing oil saturation Sof
are employed, the resulting trapping surface is shown in (see Fig. 13):
Fig. 12. This surface should be compared with the one i d
kro (So ) = kro (Sof ). (11)
obtained from pore-network simulations (Fig. 3).
0.6 d S
kro(w) oi
0.6
0.4
0.4 ki
ro(w)
0.2
S
of
∆S S
0.2 o o
0
150 S (S )
ot oi
Co 100 0.8 0
nta 0 0.2 0.4 0.6 0.8 1
0.6
ct
an 0.4 n( S oi) Oil saturation (S )
o
gle
50 0.2 ratio
(θ) satu
0 l oil
Initia
Fig. 13— Trapped and flowing saturations used to determine
the waterflood relative permeability from the drainage relative
permeability.
0.6
0.5 ∆So = So − Sof . (12)
0.4
Land makes the assumption that the trapped saturation
0.3 ∆So is the cumulative trapped saturation at a given point
0.2 in the waterflood process and that this quantity increases
0.1
as more of the flowing saturation becomes trapped. He
assumes that the maximum amount of cumulative trap-
0
0 ping, equal to the trapped saturation determined by his
50 0.2 ) trapping model (Equation (5)), occurs when the flowing
Con 0.4 n S oi
(
tac t an100 0.6 t u r atio saturation becomes zero (So = Sot (Soi )). It is important
a
gle (
θ 150 0.8 ial oil s
)
Init to note that in Land’s formulation it is necessary to obtain
the maximum trapped oil saturation Sot,max from a core-
flood experiment in order to extract the appropriate Land
Fig. 12— Trapping surface determined with the new trapping
trapping coefficient C.
model.
The intermediate trapped saturation ∆So is equal to the
cumulative trapped saturation Sot minus the amount of oil
that is still flowing and will eventually be trapped:
The waterflood relative permeability model.
Most existing relative permeability hysteresis mod- ∆So = Sot − Sot (Sof ), (13)
els12, 15, 16, 20, 21 either require a bounding waterflood curve
or model this curve according to Land’s11 waterflood rel- where
ative permeability model. The development of his model Soi
is described below. The new relative permeability model Sot ≡ Sot (Soi ) = (14)
1 + CSoi
10 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
o o o o
θ = 20 , C = 1.703 θ = 40 , C = 1.812 θ = 60 , C = 1.807 θ = 80 , C = 26.88
1 d 1 1 1
k (data)
rg
i
ro
rg
0.8 o i
k (data) 0.8 0.8 0.8
rg
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (S ) Oil saturation (S ) Oil saturation (S ) Oil saturation (S )
o o o o
o o o o
θ = 100 , C = 293.7 θ = 120 , C = 408.5 θ = 140 , C = 415.3 θ = 160 , C = 426
1 1 1 1
Oil relative permeability (kro)
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (So) Oil saturation (So) Oil saturation (So) Oil saturation (So)
Fig. 14— Comparison of Land’s trapping model for the bounding relative permeability curves with pore-network simulation data.
where the trapped saturation is given by the new trapping In Fig. 17 we compare the new waterflood model with
model: pore-network simulated data for a set of intermediate scan-
2
Sot (Soi ) = αSoi − βSoi . (20) ning curves. Model predictions were obtained using the
We substitute Equation (20) in Equation (19) and solve same parameters determined from the trapping curves and
for Sof , to obtain a new model for the flowing oil satura- the bounding waterflood curves.
tion:
· Wettability correlations. In this paper we have con-
1 sistently used the intrinsic contact angle θi as a measure
Sof = (α − 1)
2β of wettability. This parameter is almost impossible to de-
termine with any certainty in the laboratory since most
¸
p
+ (α − 1)2 + 4β[So − Sot + γ(So − Sot )(So − Soi )] . rocks are characterized by a large range of contact angles.
(21) Moreover, we have used a particular model43 that links
the intrinsic contact angle with the receding and advanc-
This expression of the flowing saturation is then used in ing contact angles.
Equation (11) to evaluate the waterflood relative perme- Ideally, one would correlate the trapping parameters α
ability. and β and the waterflood parameter γ with a measure
The tuning parameter γ can be obtained from fitting of the overall wettability characteristics of the rock that
the model to experimental bounding waterflood curves. In can be determined in the lab. In fact, previous investiga-
this investigation, we used bounding waterflood curves ob- tions26, 28 have shown that pore-network models are able to
tained from pore-network simulation. The dependence of perform quantitative predictions of laboratory wettability
the parameter γ on the intrinsic contact angle is shown in measurements.
Fig. 15. The trends in this relationship are the ones ex- Common measures of wettability are the Amott wetta-
pected. For water-wet media, the parameter γ is negative, bility indices,9 Iw and Io . A strongly water-wet medium is
indicating that the Land trapping model overestimates the associated with Iw = 1 and Io = 0, whilst values of Iw = 0
trapped saturation and subsequently underestimates the and Io = 1 correspond to a strongly oil-wet medium. The
relative permeability. For oil-wet media this parameter Amott–Harvey index Iwo is probably the most popular
takes positive values, which allows for the model water- measure of wettability and is defined as:
flood curve to be below the one predicted by the Land
model. Iwo = Iw − Io , (22)
o o o o
θ = 20 , γ = −1 θ = 40 , γ = −0.75 θ = 60 , γ = −0.1 θ = 80 , γ = 1.4
1 d
1 1 1
k (data)
ro
ro
k (model)
ro
0.8 i
0.8 0.8 0.8
o k (data)
ro
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (S ) Oil saturation (S ) Oil saturation (S ) Oil saturation (S )
o o o o
o o o o
θ = 100 , γ = 1.65 θ = 120 , γ = 1.3 θ = 140 , γ = 1.3 θ = 160 , γ = 1.2
1 1 1 1
Oil relative permeability (kro)
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (So) Oil saturation (So) Oil saturation (So) Oil saturation (So)
Fig. 16— Comparison of the new waterflood model with pore-network simulation data for the bounding relative permeability curves
(initial saturation Soi = 0.76).
o o o o
θ = 20 , γ = −1 θ = 40 , γ = −0.75 θ = 60 , γ = −0.1 θ = 80 , γ = 1.4
1 d 1 1 1
kro (data)
Oil relative permeability (kro)
i
ro
kro (model)
0.8 i
0.8 0.8 0.8
o kro (data)
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (S ) Oil saturation (S ) Oil saturation (S ) Oil saturation (S )
o o o o
o o o o
θ = 100 , γ = 1.65 θ = 120 , γ = 1.3 θ = 140 , γ = 1.3 θ = 160 , γ = 1.2
1 1 1 1
Oil relative permeability (kro)
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (So) Oil saturation (So) Oil saturation (So) Oil saturation (So)
Fig. 17— Comparison of the new waterflood model with pore-network simulation data for scanning relative permeability curves
(initial saturation Soi = 0.65).
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 13
0.8
Wettability index
0.6
0.4
I
0.2 o
I
w
0
50 100 150
Fig. 20— Schematic of the trail of residual CO2 that is left be-
Contact angle (θ) hind due to snap-off as the plume migrates upwards [courtesy
Fig. 18— Amott oil and water wettability indices for all intrin- of Marc Hesse].
sic contact angles.
0.6
0.4
0.2
Sgt=0.4
0
0 0.2 0.4 0.6 0.8 1
Gas saturation (S )
g
1
Swc=0.31
)
rw
0.8
1. Hysteresis and trapping, by comparing the results of Effect of hysteresis and trapping. We begin by illus-
Case 2 (in which hysteresis is modeled) and Case 1 (in trating the dramatic effect of relative permeability hystere-
which hysteresis is not modeled). sis on the predictions of the fate of the injected CO2 . We
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 15
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08 INJ-08
INJ-06 INJ-06
INJ-04 INJ-04
INJ-05
INJ-05
INJ-01
INJ-01
Fig. 24— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 1 (no hysteresis).
Right: results from Case 2 (with hysteresis).
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 25— Evolution of CO2 saturations at the three “observation” gridblocks for Case 1 (no hysteresis) and Case 2 (with hysteresis).
are shown in Fig. 27. As seen from the plots of grid- to Case 3 and Case 4 are of course identical. Once wa-
blocks 7, 21, 1 and 11, 11, 1, the system undergoes trapping ter injection starts (in Case 4) the system experiences a
of the CO2 in both cases. However, the case with low drastic change in the flow dynamics. Water starts displac-
injection rate shows that CO2 reaches a saturation value ing the CO2 radially away from the wells. This can be
of 0.7 (connate water) at the top of the anticline (grid- seen as a forced imbibition process that leads to almost
block 13, 18, 1), whereas a value of just 0.15 is achieved in immediate trapping of the CO2 phase in large portions of
the high-rate case. the reservoir. It explains why we see a sharp decrease in
CO2 saturations at the observation gridblocks—the CO2
Effect of subsequent water injection. Finally, we in- is being pushed elsewhere—down to some residual satura-
vestigate how the performance of the CO2 sequestration tion. It is worth noting that this saturation value is lower
project is affected by the injection of a slug of water af- than what would be predicted from Land’s trapping model,
ter injection of the CO2 into the aquifer. The motivation Sgt = Sgi /(1+CSgi ), because part of the CO2 is being dis-
is to enhance the imbibition process that naturally occurs placed through existing connected paths and is therefore
at the trail of the nonwetting CO2 plume as it migrates not trapped.
upwards.
In Fig. 28 we compare the fluid distributions of Case 3 Summary. We offer the following brief summary of the
and Case 4. In both cases we inject 0.15 pore volumes findings from the simulations of CO2 sequestration in
of CO2 for the first ten years. In Case 4 this is followed saline aquifers:
by injection of 0.05 pore volumes of water during a 1-year 1. Accounting for trapping and relative permeability hys-
period. These results confirm the expected response: sub- teresis of the nonwetting CO2 phase is essential in or-
sequent water injection induces more trapping and reduces der to correctly characterize the migration and final
significantly the amount of CO2 that accumulates at the distribution of the injected CO2 . Trapping occurs as
top of the aquifer. the trailing edge of the plume ascends (and is replaced
The evolution of the CO2 saturation at the observation by water) after injection stops.
gridblocks (Fig. 29) offers additional insight into the be-
havior of the displacement process. During the CO2 injec- 2. Trapping of the CO2 leads to more favorable scenarios
tion period—the first ten years—the curves corresponding for sequestration purposes: a large fraction of the CO2
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 17
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08
INJ-08
INJ-06
INJ-06
INJ-04
INJ-04
INJ-05
INJ-05
INJ-01
INJ-01
Fig. 26— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 2 (injection
over 1 year). Right: results from Case 3 (injection over 10 years).
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 27— Evolution of CO2 saturations at the three “observation” gridblocks for Case 2 (injection over 1 year) and Case 3 (injection
over 10 years).
is trapped and immobile for practical purposes, and the observed behavior: (1) non-monotonicity of the initial–
is more spread out throughout the aquifer, thereby residual curves, which implies that waterflood relative per-
increasing the interfacial area for subsequent dissolu- meabilities cross; and (2) convexity of the waterflood rela-
tion. tive permeability curves for oil-wet media due to film flow
of oil.
3. High injection rates result in more effective seques- We have also investigated the impact of trapping and
tration of the CO2 . A shorter injection period leaves relative permeability hysteresis in the context of CO2 se-
less time for the buoyant CO2 to reach the top of the questration projects in saline aquifers. We have provided
formation, from which it is difficult to immobilize. conclusive evidence that hysteresis is an order-one factor
in the prediction of the migration and final distribution of
4. Injection of a water slugs after CO2 injection increases the CO2 . Trapping of the CO2 occurs during the upwards
the effectiveness of the sequestration project. The in- migration of the CO2 plume, but only after injection has
jected water forces break-up of large connected CO2 stopped and the trailing edge of the plume is naturally be-
plumes, enhancing trapping and immobilization of the ing displaced by water. This imbibition process leads to
CO2 . A similar behavior would be observed if there trapping of the CO2 . A trail of residual CO2 is left be-
were a regional groundwater flow. hind as the plume migrates upwards. We have shown that
hysteresis effects have a very favorable effect on the effec-
Conclusions and outlook tiveness of CO2 sequestration. From the point of view of
the design of a CO2 project, we have shown that trapping
We have presented a new model of trapping and water- of the CO2 can be enhanced by: (1) operating at high in-
flood relative permeability. Development of the model jection rates, and (2) alternatively injecting water and CO2
is motivated by the inability of existing models to cap- in the spirit of classical WAG for enhanced oil recovery.
ture the trends observed for intermediate-wet and oil-wet
media. Due to scarcity of reliable experimental data, we
have used pore-network simulation as a means to predict
the trends in trapping and relative permeability hystere-
sis. The new model is able to capture two key features of
18 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08 INJ-08
INJ-06 INJ-06
INJ-04 INJ-04
INJ-05 INJ-05
INJ-01
INJ-01
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 28— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 3 (no water
injection). Right: results from Case 4 (water injection after CO2 injection).
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 29— Evolution of CO2 saturations at the three “observation” gridblocks for Case 3 (no water injection) and Case 4 (water
injection after CO2 injection).
Nomenclature Acknowledgements
EJS, RJ and FMO gratefully acknowledge financial sup-
Roman letters port from the industrial affiliates of the Stanford University
b exponent parameter in Jerauld’s trapping model Petroleum Research Institute for Gas Injection (SUPRI-C)
C Land trapping coefficient and support from the Global Climate and Energy Project
Io Amott oil wettability index (GCEP). MJB would like to thank the sponsors of the
Iw Amott water wettability index Imperial College Consortium on Pore-Scale Modelling and
Iwo Amott–Harvey wettability index the NERC for funding.
krw water relative permeability
d
kro drainage oil relative permeability References
i
kro waterflood oil relative permeability [1] T. M. Geffens, W. W. Owens, D. R. Parrish, and R. A.
Sw water saturation Morse. Experimental investigation of factors affecting lab-
So oil saturation oratory relative permeability measurements. Petrol. Trans.
So,max maximum oil saturation AIME, 192:99–110, 1951.
Sof flowing oil saturation [2] R. Lenormand, C. Zarcone, and A. Sarr. Mechanisms of
Soi initial oil saturation the displacement of one fluid by another in a network of
Sot ultimate trapped oil saturation capillary ducts. J. Fluid Mech., 135:123–132, 1983.
∆So intermediate trapped oil saturation [3] I. Chatzis, N. R. Morrow, and H. T. Lim. Magnitude and
detailed structure of residual oil saturation. Soc. Pet. Eng.
Greek letters J., 23:311–330, April 1983.
α initial slope of IR curve in new trapping model [4] E. C. Donaldson, R. D. Thomas, and P. B. Lorenz. Wet-
β curvature of the IR curve in new trapping model tability determination and its effect on recovery efficiency.
γ film flow parameter in new waterflood model Petrol. Trans. AIME, 246:13–20, 1969.
θ, θi intrinsic contact angle
[5] W. W. Owens and D. L. Archer. The effect of rock wetta-
θa advancing contact angle bility on oil–water relative permeability relationships. J.
θr receding contact angle Pet. Technol., pages 873–878, July 1971.
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 19
[6] A. R. Kovscek, H. Wong, and C. J. Radke. A pore-level [23] J.-C. Moulu, O. Vizika, P. Egermann, and F. Kalaydjian.
scenario for the development of mixed wettability in oil A new three-phase relative permeability model for vari-
reservoirs. AIChE J., 39:1072–1085, 1993. ous wettability conditions. In SPE Annual Technical Con-
ference and Exhibition, Houston, TX, October 3–6 1999.
[7] H. T. Kennedy, E. O. Burja, and R. S. Boykin. An investi-
(SPE 56477).
gation of the effects of wettability on the recovery of oil by
water flooding. J. Physical Chem., 59:867–869, September [24] P. Egermann, O. Vizika, L. Dallet, C. Requin, and
1955. F. Sonier. Hysteresis in three-phase flow: experiments,
modeling and reservoir simulations. In SPE European
[8] T. F. Moore and R. L. Slobod. Displacement of oil by wa- Petroleum Conference, Paris, France, October 24–25 2000.
ter — Effect of wettability, rate and viscosity on recovery. (SPE 65127).
Prod. Monthly, 20(10):20–30, August 1956.
[25] A. B. Dixit, S. R. McDougall, K. S. Sorbie, and J. S. Buck-
[9] E. Amott. Observations relating to the wettability of ley. Pore-scale modeling of wettability effects and their in-
porous rock. Petrol. Trans. AIME, 216:156–162, 1959. fluence on oil recovery. SPE Reserv. Eval. Eng., 2:25–36,
[10] P. P. Jadhunandan and N. R. Morrow. Effect of wettability 1999.
on waterflood recovery for crude-oil/brine/rock systems. [26] P. E. Øren and S. Bakke. Reconstruction of Berea sand-
SPE Reserv. Eng., 10:40–46, February 1995. stone and pore-scale modelling of wettability effects. J.
[11] C. S. Land. Calculation of imbibition relative permeabil- Pet. Sci. Eng., 39:177–199, 2003.
ity for two- and three-phase flow from rock properties. [27] M. D. Jackson, P. H. Valvatne, and M. J. Blunt. Prediction
Soc. Pet. Eng. J., 8(2):149–156, June 1968. Petrol. Trans. of wettability variation and its impact on flow using pore-
AIME, 243. to reservoir-scale simulations. J. Pet. Sci. Eng., 39:231–
[12] G. R. Jerauld. General three-phase relative permeability 246, 2003.
model for Prudhoe Bay. SPE Reserv. Eng., 12(4):255–263, [28] P. H. Valvatne and M. J. Blunt. Predictive pore-scale mod-
November 1997. eling of two-phase flow in mixed wet media. Water Resour.
Res., 40:W07406, doi:10.1029/2003WR002627, 2004.
[13] F. M. Carlson. Simulation of relative permeability hys-
teresis to the nonwetting phase. In SPE Annual Technical [29] A. Al-Futaisi and T. W. Patzek. Impact of wettability
Conference and Exhibition, San Antonio, TX, October 5– on two-phase flow characteristics of sedimentary rock: A
7, 1981. (SPE 10157). quasi-static description. Water Resour. Res., 39(2):1042,
doi:10.1029/2002WR001366, 2003.
[14] C. S. Land. Comparison of calculated with experimen-
tal imbibition relative permeability. Soc. Pet. Eng. J., [30] A. Al-Futaisi and T. W. Patzek. Secondary imbibition
11(4):419–425, December 1971. Petrol. Trans. AIME, 251. in NAPL-invaded mixed-wet sediments. J. Contaminant
Hydrol., 74(1–4):61–81, 2004.
[15] J. E. Killough. Reservoir simulation with history-
[31] I. Fatt. The network model of porous media I. Capillary
dependent saturation functions. Soc. Pet. Eng. J.,
pressure characteristics. Petrol. Trans. AIME, 207:144–
16(1):37–48, February 1976. Petrol. Trans. AIME, 261.
159, 1956.
[16] J. A. Larsen and A. Skauge. Methodology for numeri- [32] M. J. Blunt. Flow in porous media—pore network models
cal simulation with cycle-dependent relative permeabili- and multiphase flow. Curr. Opin. Colloid Interface Sci.,
ties. Soc. Pet. Eng. J., 3(2):163–173, June 1998. 6:197–207, 2001.
[17] D. J. Element, J. H. K. Masters, N. C. Sargent, A. J. [33] M. J. Blunt, M. D. Jackson, M. Piri, and P. H. Valvatne.
Jayasekera, and S. G. Goodyear. Assesment of three-phase Detailed physics, predictive capabilities and macroscopic
relative permeability models using laboratory hysteresis consequences for pore-network models of multiphase flow.
data. In SPE International Improved Oil Recovery Con- Adv. Water Resour., 25:1069–1089, 2003.
ference, Kuala Lumpur, Malaysia, October 20–21 2003.
(SPE 84903). [34] R. M. Holt, E. Fjoer, O. Torsoeter, and S. Bakke. Petro-
physical laboratory measurements for basin and reservoir
[18] E. J. Spiteri and R. Juanes. The impact of relative per- simulation. Marine and Petroleum Geology, 13:383–391,
meability hysteresis on the numerical simulation of WAG 1996.
injection. In SPE Annual Technical Conference and Exhi-
[35] P. Spanne, J. F. Thovert, C. J. Jacquin, W. B. Lindquist,
bition, Houston, TX, September 26–29 2004. (SPE 89921).
K. W. Jones, and P. M. Adler. Synchrotron computed mi-
[19] C. A. Kossack. Comparison of reservoir simulation hys- crotomography of porous media: topology and transports.
teresis options. In SPE Annual Technical Conference and Phys. Rev. Lett., 73(2001–2004), 1994.
Exhibition, Dallas, TX, October 1–4 2000. (SPE 63147). [36] M. A. Ioannidis and I. Chatzis. On the geometry and
[20] R. J. Lenhard and J. C. Parker. A model for hys- topology of 3D stochastic porous media. JCIS, 229:323–
teretic constitutive relations governing multiphase flow, 334, 2000.
2. Permeability-saturation relations. Water Resour. Res., [37] S. Bryant and M. J. Blunt. Prediction of relative perme-
23(12):2197–2206, 1987. ability in simple porous media. Phys. Rev. A, 46:2004–
[21] M. J. Blunt. An empirical model for three-phase relative 2011, 1992.
permeability. Soc. Pet. Eng. J., 5(4):435–445, December [38] P. E. Øren and S. Bakke. Process based reconstruction of
2000. sandstones and prediction of transport properties. Transp.
[22] R. J. Lenhard and M. Oostrom. A parametric model for Porous Media, 46(2-3):311–343, 2002.
predicting relative permeability-saturation-capillary pres- [39] M. J. Oak. Three-phase relative permeability of
sure relationships of oil-water systems in porous media intermediate-wet Berea sandstone. In SPE Annual Tech-
with mixed wettability. Transp. Porous Media, 31:109– nical Conference and Exhibition, Dallas, TX, October 6–9,
131, 1998. 1991. (SPE 22599).
20 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
[40] P. E. Øren, S. Bakke, and O. J. Arntzen. Extending pre- [59] F. J. T. Floris, M. D. Bush, M. Cuypers, F. Roggero,
dictive capabilities to network models. Soc. Pet. Eng. J., and A. R. Syversveen. Methods for quantifying the un-
3(4):324–336, December 1998. certainty of production forecasts: a comparative study.
[41] T. W. Patzek. Verification of a complete pore network Petrol. Geosci., 7(Sp.Iss.):S87–S96, 2001.
simulator of drainage and imbibition. Soc. Pet. Eng. J., [60] Netherlands Institute of Applied Geosciences. PUNQ Case
6(2):144–156, June 2001. Studies. http://www.nitg.tno.nl/punq/cases/index.shtml.
[42] S. Bakke and P. E. Øren. 3-D pore-scale modelling of
sandstones and flow simulations in the pore networks. Soc.
Pet. Eng. J., 2:136–149, 1997.
[43] N. R. Morrow. Effects of surface roughness on contact
angle with special reference to petroleum recovery. J. Can.
Pet. Technol., 14:42–53, 1975.
[44] M. J. Blunt and P. King. Relative permeabilities from
two- and three-dimensional pore-scale network modeling.
Transp. Porous Media, 6:407–433, 1991.
[45] M. I. J. van Dijke, M. Lago, K. S. Sorbie, and M. Araujo.
Free energy balance for three fluid phases in a capillary of
arbitrarily shaped cross-section: capillary entry pressures
and layers of the intermediate-wetting phase. J. Colloid
Interface Sci., 277(1):184–201, 2004.
[46] M. J. Oak. Three-phase relative permeability of water-
wet Berea. In SPE/DOE Seventh Symposium on En-
hanced Oil Recovery, Tulsa, OK, April 22–25, 1990.
(SPE/DOE 20183).
[47] C. R. Holmgren and R. A. Morse. Effect of free gas sat-
uration on oil recovery by water flooding. Petrol. Trans.
AIME, 192:135–140, 1951.
[48] J. R. Kyte, R. J. Stanclift Jr., S. C. Stephan Jr., and L. A.
Rapoport. Mechanism of water flooding in the presence of
free gas. Petrol. Trans. AIME, 207:215–221, 1956.
[49] S. G. Dardaganian. The displacement of gas by oil in the
presence of connate water. MS Thesis, Petroleum Engi-
neering, Texas A&M University, 1957.
[50] G. R. Jerauld. Prudhoe Bay gas/oil relative permeability.
SPE Reserv. Eng., 12(1):66–73, February 1997.
[51] P. M. Cox, R. A. Betts, C. D. Jones, S. A. Spall, and
I. J. Totterdell. Acceleration of global warming due to
carbon-cycle feedbacks in a coupled climate model. Nature,
408(6809):184–187, 2000.
[52] K. S. Lackner. A guide to CO2 sequestration. Science,
300(5626):1677–1678, 2003.
[53] G. Hitchon, W. D. Gunter, T. Gentzis, and R. T. Bailey.
Sedimentary basins and greenhouse gases: a serendipitous
association. Energy Conv. Manag., 40(8):825–843, 1999.
[54] S. Bachu. Sequestration of CO2 in geological media: cri-
teria and approach for site selection in response to climate
change. Energy Conv. Manag., 41(9):953–970, 2000.
[55] S. Holloway. Storage of fossil fuel-derived carbon dioxide
beneath the surface of the earth. Annu. Rev. Energy En-
viron., 26:145–166, 2001.
[56] G. S. Bromhal, W. N. Sams, S. Jikich, T. Ertekin, and
D. H. Smith. Simulation of CO2 sequestration in coal beds:
The effects of sorption isotherms. Chem. Geol., 217(3–
4):201–211, 2005.
[57] K. Pruess and J. Garcı́a. Multiphase flow dynamics during
CO2 disposal into saline aquifers. Env. Geol., 42(2–3):282–
295, 2002.
[58] S. Bachu. Sequestration of CO2 in geological media in re-
sponse to climate change: capacity of deep saline aquifers
to sequester CO2 in solution. Energy Conv. Manag.,
44(20):3151–3175, 2003.