Water Alternating Gas Injection Studies – Phase 1
Water Alternating Gas Injection Studies – Phase 1
D. H. Tehrani, A. Danesh, G. Henderson and M. Sohrabi
Department of Petroleum Engineering, Heriot-Watt University
Summary
The Department of Petroleum Engineering at Heriot-Watt University has been studying the water
alternating gas (WAG) injection processes, using experimental (micromodels) and theoretical
(network modelling) methods during the past 3 years. Last year we presented the results of the
micromodel studies on water-wet systems. This paper presents the results of the experiments on
oil-wet and mixed wet systems and compares them with those of the water-wet models. A sister
paper by K. S. Sorbie et al. will present the results of the network modelling studies.
Many of the oil reservoirs, including those in the North Sea, are approaching the end of their
waterflooding. At this stage a significant quantity of oil will still remain in the reservoir. It is
known that using the WAG injection some of that oil can be produced. Unfortunately, the
underlying physics of the three-phase flow is not well understood to allow reliable predictions to be
made for economic evaluation. WAG injection will involve drainage and imbibition processes
taking place sequentially. The three-phase relative permeability and capillary pressure functions
are, therefore, extremely complex. It is practically impossible to develop such functions for
realistic reservoir situations, using core displacement methods. The approach we have adopted is to
develop a mathematical network simulator, which covers all the significant physical flow processes
involved in the WAG injection. But to gain confidence that such a simulator can indeed reflect the
physics of the flow realistically, we test it against the actual physical micromodel observations. If
the predictions of the simulator agree with those observed in the micromodel, we can then operate
the simulator with real reservoir fluid data and basic rock properties in 3-D, to calculate the required
pore scale relative permeability functions, Fig.1. Observations made in the micromodel studies
have considerably improved our understanding of the underlying physical principles. This
knowledge is extremely valuable in formulation of our network model simulator.
Introduction
Waterflooding, gas injection and wateralternating-
gas injection (WAG) are wellestablished
methods for improving oil
recovery. In reservoirs that have been
waterflooded, it is still possible to recover a
significant quantity of the remaining oil by
injecting gas alternately with water. Gas can
occupy parts of the pore space that otherwise
would be occupied by oil, thereby mobilising
the remaining oil. Water, injected subsequently,
will displace some of the remaining oil and
gas, further reducing the residual oil saturation.
Repetition of the WAG injection process can
further improve the recovery of oil.
Christensen, Stenby and Skauge1 reported an
excellent review of some sixty fieldapplications
of WAG. Several field trials
have been reported as being successful, e.g., in
Kuparuk2, Snorre3 and Gulfaks fields4. Both
immiscible4-6 and miscible gases7 have been
used. A very large number of coreflood
experiments8-12 and analytical and numerical
simulations11,14 have been carried out. A
recent study has considered the WAG process
for improving the hydrocarbon recovery in
gas/condensate reservoirs13. Most of the
research work, conducted so far, has been on
either core flooding8,9,10 or numerical
simulation11,12, sometimes alongside field
trials. The relationship between the injection
gas/water ratio and oil recovery has been
empirically investigated using core displacement
experiments, often at low pressure and
generally with water wet cores8,10.
Micromodels were used as early as 1960 for
fluid displacement studies15. Some lowpressure
micromodel studies of three-phase
displacement have also been performed16,17.
However, as far as we know, no micromodel
visualisation of the WAG injection has been
carried out to directly observe the physical
processes taking place in the porous media,
using live oil, live water in equilibrium with
injection gas and models with different
wettability. Larsen et al.18 reported some
limited results of their WAG micromodel
studies.
To do reservoir development planning, for
possible implementation of a WAG scheme,
the operator needs reliable performance and
hydrocarbon recovery prediction, to use for
economic evaluation. To achieve this, good
simulation incorporating proper reservoir fluid
and rock description is needed. This requires
accurate sets of relative permeability and
capillary pressure functions in a three-phase
fluid flow regime. But it is impractical to
measure these for all the different rock types
and fluids present in a reservoir and describe
them in terms of IFT which, itself is a function
of fluid composition and pressure. We are
attempting to develop a 3-phase 3-D
mathematical network simulator, which has in
it all the significant physical flow processes
involved in WAG injection, formulated as
accurately as possible. But to gain confidence
that such a simulator can indeed reflect physics
of the flow realistically, we test it against a
series of WAG experiments performed using
micromodels. We can conduct the actual WAG
injection, observe and record the flow
processes and measure the model fluid
saturations and recoveries. To enable us to
magnify and view the pore scale images and to
analyse the fluid flow, we have had to use 2-D
glass micromodels and model fluids with
known properties. Although the results will not
be directly applicable to any real reservoir,
they can be used to verify the accuracy of the
predictions made by our network model
simulator. We will then run the pore scale
simulator to predict the fluid distributions and
the recoveries for a given set of pore geometry,
wettability and fluid properties. If these agree
with those observed and measured in the
micromodel (Figs. 2 to 5), we shall then have
enough confidence, to operate it with real
reservoir fluids and rock properties in 3-D
mode, to calculate the required pore scale
relative permeability and capillary pressure
functions (Fig. 1). These will later be upscaled
for use in the numerical reservoir scale
simulation.
Objective
The objective of the current micromodel
studies is to improve our understanding of the
physical principles underlying such processes
taking place in porous media and to develop a
network model simulator that can produce
complex three-phase relative permeability and
capillary pressure functions. Observing and
recording the fluid flow behaviour within the
micromodel during the WAG injection process
will help achieve this. The video record of the
fluid displacements is used to obtain
qualitative and quantitative information on
three-phase fluid flow during WAG injection.
These will then be used to compare with the
results of the network model, which will
attempt to simulate the same processes, using
the micromodel fluids and geometrical data. If
the simulated results match the fluid
distribution and the recovery data obtained by
the experiments reasonably well, then it can be
confidently used to simulate and obtain the
three-phase relative permeability and capillary
pressure functions, using realistic reservoir
rock and fluid properties in three-dimensional
space, Fig. 1.
Experimental Facilities, Test Fluids and
Procedure
These have been described in our report to the
DTI IOR Seminar of June 200019.
Wettability Alteration
Strongly water-wet micromodelThe procedure for preparation of water-wet
micromodel is described in Ref. 19.Strongly oil-wet micromodel
Wettability of the glass micromodel was
changed from strongly water-wet to strongly
oil-wet by ageing the model in a North-sea
crude oil. After cleaning the model with
copious amounts of distilled water and acetone
successively, the micromodel was completely
saturated with 2 % (wt./wt.) CaCl2 brine. The
model was then allowed to reach the state of
ionic equilibrium for one week at 100oF. Then
the crude oil was injected continuously until no
brine remained in the pores, and proceeded
with ageing for three weeks. To examine the
degree and stability of the oil wetness achieved
after ageing, blue water was slowly injected
into the micromodel. Fig.3a shows magnified
image of a section of the oil-wet micromodel,
after initial waterflooding. Examining the fluid
interfacial curvatures of the pictures and
checking them in sequence of slides once a
day, over a period of one week, proved no sign
of change of oil-wetness. As the effective time
needed to conduct a micromodel experiment
was shorter than a week; we gained confidence
in the stability of the achieved oil wetness.Mixed-wet micromodel
Starting with the oil-wet micromodel, a dark
blue-dyed water solution was prepared.
Having the micromodel completely saturated
with oil; the highly concentrated dye solution
was injected into the micromodel. As a result,
some of the oil-filled pores were displaced with
dyed water. The injection of dyed water
continued for 24 hours. Then the micromodel
was sealed for 48 hours. In the type of mixedwet
model we obtained, some pores were
water-wet while some others were oil-wet or
neutral-wet. Fig.4.b shows a magnified picture
of a small section of a mixed-wet micromodel
in which all three types of water-wet, oil-wet
and neutral-wet pores can be identified from
the shape of the water/oil interfaces.
EXPERIMENTAL RESULTS
Experiments with strongly water-wet model
The results of these experiments were reported
in detail in our June 2000 report19. However,
some of the results will be shown here for
comparison with the oil-wet and mixed-wet
model results. Fig. 2 presents a few pictures of
one section (of ten) from the middle of the
micromodel. It is interesting to note the
distributions of the fluids under ‘initial’
conditions, Fig. 2b, where the ‘connate water’
is not only in form of films on the pore surface
and in small pore throats, but also in clusters of
large pores surrounded by small throats. It is
also interesting to see that after waterflooding,
Fig. 2c, water has risen in the corners and on
the sides of the pores, squeezing the oil out,
rather than displacing it piston wise. These
can, of course, be all explained remembering
that the model is water wet, the velocity is low
(1.2 m/d, as in real reservoirs) and the flow is
capillary dominated. Seeing the videos and
observing the WAG cycles are indeed very
revealing and improve our understanding of the
processes considerably. Finally, the last
picture of a five-cycle WAG, Fig 2d, depicts
how the fluids have been redistributed and
extra oil has been produced. Fig 5 shows that
an extra 10% of additional oil has been
produced after the first three cycles of WAG
injection, in water-wet model. Moreover, the
additional recovery in subsequent cycles has
been very small.
Experiments with strongly oil-wet model
Initially the oil-wet model was saturated with
oil, with no ‘connate water’ present. Water
flooding has displaced a significant quantity of
oil, this time by piston wise displacement,
Fig.3a. Video films and magnified pictures,
not shown here, clearly showed that a thin film
of oil stays on the walls of the pores and apart
from the clusters of pores, which are
surrounded with small pore throats, the other
pores have been very efficiently displaced by
water. Fig.3c shows that even a good quantity
of the oil, which was trapped in the abovementioned
clusters, has been produced. Fig. 5
shows that about 22% of extra oil have been
recovered after three WAG injection cycles.
Experiments with mixed-wet models
A series of experiments were carried out with
mixed-wet models of different degrees of oil
wetness. About 1200 interfaces, each
indicating a different degree of wettability,
were counted and the degree of wettability was
determined, Fig. 4b. Fig. 4 depicts the
experimental results of a section of the
micromodel, in which the number of oil-wet
pores was almost, equal to the number of
water-wet wet pores. Here the water-wet pores
exhibit the behaviour we observed in the
strongly water-wet model and oil-wet pores
behave as in strongly oil-wet model. However,
the interaction between the pores of different
wettability results in a recovery behaviour,
which does not show a completely
intermediate trend. Fig. 5 illustrates that
production of extra oil goes on beyond the
third cycle and eventually surpasses that of the
oil-wet model. In this case the additional
recovery, after five WAG cycles, was about
27% of the initial oil-in-place.
Conclusions
The results of the reported experiments, which
were performed in micromodels of different
wettability, were very revealing. They provide
an excellent basis for checking and verifying
the formulation of our network model
simulator (see a separate paper presented to
this seminar on network modelling results).
The algorithm, shown in Fig. 1, describes the
strategy of this research project leading to
acquisition of an improved tool for
determination of the saturation functions.
These functions are of utmost importance in
making prediction of real reservoir performance
under WAG injection operation.
Nomenclature
Sorw = residual oil saturation to waterflood
SorWAG = residual oil saturation after 5 cycles
of WAG injection.
Acknowledgements
The WAG project at Heriot-Watt U. is equally
sponsored by: The UK Department of Trade
and Industry, BP plc, Marathon International
(GB) Ltd, Mobil (North Sea) Ltd, Norsk Hydro
a.s.a, SAGA Petroleum a.s.a, and Total Oil
Marine plc, which is gratefully acknowledged.