Você está na página 1de 135

Miscible Flooding : Introduction

Displacement of crude oil with drive agents such as water or natural gas leaves behind an immiscibly-trapped oil
saturation, which is a large fraction of the initial oil saturation. This is the case even when the amount of drive agent
is equal to a very large number of pore volumes. It would seem advisable, therefore, to use a displacing agent that is
miscible with the crude oil. By "miscible," we mean that there is only a single non-aqueous phase present when any
proportions of the crude oil and displacing agent are mixed together and allowed to come to equilibrium. The phrase
"miscible in all proportions" is often used to describe this condition.
While the above definition of "miscible" is generally true, there is a distinction between a displacing agent that is
immediately miscible with the crude oil and one that develops miscibility after a series of equilibrium contact stages.
In the first case, the agent is first-contact miscible with the crude oil. In the second case, the agent is multiple-
contact, or conditionally miscible. Another phrase used to describe this second case is developed miscibility.
Miscible Flooding Processes
There should clearly be an advantage in using a displacing agent that is either first-contact miscible or multiple-
contact miscible with the crude oil. Under secondary recovery conditions (i.e., at the original or connate water
saturation), it should be possible to displace all of the crude oil, since there is no trapped or residual oil saturation.
(This is almost, but not completely, true). However the pore volume that was occupied by crude oil will then be
occupied by the displacing agent. For such a displacement to be economical, it is necessary either to recover this
displacing agent or to use an agent that is much less valuable than crude oil.
This constraint on the displacement presents a problem. Few displacing agents are both miscible with and worth less
than crude oil. Natural gas, flue gas, and even nitrogen separated from air in an air liquefaction plant are likely to be
cheaper than crude oil on an "equal reservoir volume" basis. (If air itself is injected, in situ combustion occurs rather
than miscible displacement.) Until about 1970, LPG (propane and butane) was also much cheaper than crude oil, but
it has since become worth almost as much as the crude oils that it might be used to displace.
Carbon dioxide is also cheaper than crude oil when it is recovered from natural deposits or from gas mixtures richer
in carbon dioxide than flue gas. These carbon dioxide-rich gas mixtures are found, for instance, with natural gas
(e.g., in the gas fields of the Delaware/Val Verde Basin in west Texas), with hydrogen in the effluents from steam
reforming of natural gas and from the combined partial oxidation and steam reforming of natural gas to produce
ammonia synthesis gas. These carbon dioxide-rich gas mixtures are usually processed with amine-type solvents,
which dissolve the carbon dioxide under pressure in an absorption tower, leaving the valuable natural gas, hydrogen
or ammonia synthesis gas in the exiting gas phase. The carbon dioxide is removed from the solvent in a second
tower by heating at a lower, usually near-atmospheric, pressure. Most of the cost of the removed carbon dioxide is
for dehydration and compression from near atmospheric pressure to a pressure sufficient to inject in a reservoir, or to
transport it to a reservoir and then inject it.
Carbon dioxide is also found in large natural deposits (like natural gas), often containing small amounts of methane.
Levorsen (1954) believes that these deposits result from an intrusion of hot magmatic rock into sedimentary
carbonate rocks, such that the temperature of the carbonate rock rises sufficiently to calcine it-that is, to convert it to
calcium oxide, which dissolves in the magmatic rock-and free carbon dioxide. The carbon dioxide then rises until it
encounters a trap containing pore space. Such carbon dioxide deposits exist in northeastern New Mexico, in south
central Colorado, in southwestern Colorado, in eastern Utah, in southwestern Wyoming and in north central
Mississippi. Carbon dioxide at injection pressures can be obtained more cheaply from these natural sources than
from the carbon dioxide scrubbing processes discussed above, because much less compression cost is involved.
A large fraction of an injected miscible agent can be recovered from a reservoir by waterflooding after the crude oil
has been displaced. However, since we could have recovered about the same fraction of the original oil in place by a
waterflood, the economic gain by using the miscible drive agent is equal to the difference in value between the
waterflood residual oil and the same residual saturation of the cheaper displacing agent. It is common practice in
most miscible floods, therefore, to follow the miscible drive by a water drive, and to evaluate the results in terms of
the value of the extra oil recovered (beyond the waterflood-recoverable oil) minus the cost of performing the
miscible flood prior to the waterflood. This cost must include the cost of the maximum amount of the displacing
agent which must be purchased, minus any credit for the displacing agent that is produced back from the reservoir.
In most cases, a waterflood will already be in progress by the time a miscible flood is considered. In fact, the
reservoir may be approaching the economic limit of water cut (currently considered to be about 98%). When the
waterflood has progressed beyond the point where water has broken through at production wells, a subsequent
miscible flood is considered to be a tertiary recovery process. In this case, the miscible drive agent must immiscibly
displace water in order to contact the crude oil. Much of this oil will be trapped and immovable, extending through
several pores but cut off at the pore throats by water films bridging the pore throats. Therefore, the amount of
displacing agent required is considerably greater than the volume of crude oil remaining in the pores.
As a rough guide, it is generally necessary to displace at least half as much water as the residual crude oil saturation
to remobilize all of the oil in the water-swept regions of the reservoir. It is then possible, with a final waterflood, to
drive out the excess miscible displacing agent, down to the nonaqueous phase residual saturation. Again, the process
must be debited with the maximum amount of the miscible agent that must be used, and credited with the value of
the excess miscible agent driven out by the final waterflood.
Both secondary and tertiary miscible drive processes are therefore judged on the same basis: the value of the extra
oil recovered over that which could be obtained by waterflooding, and the cost in terms of the gross miscible agent
used less credit for excess miscible agent produced. Of course, the miscible process also entails extra operating
costs.
Furthermore, all known miscible displacing agents that are cheaper than crude oil are also much less viscous and
usually less dense than crude oil. This means that these agents will have a mobility ratio much greater than one, and
viscous fingering of the miscible agent through the oil can be expected, as well as gravity tonguing (override) due to
the lower density. The same is true during the immiscible displacement of water in excess of the connate water
saturation during a tertiary miscible flood. The sweep efficiency at any given throughput will therefore be less than
that of a waterflood. Also, miscible drive agents are generally more expensive than water, so is not possible to
overcome the poor sweep efficiency just by injecting large number of pore volumes (as can usually be done in a
waterflood when the sweep efficiency is poor because of unfavorable mobility ratio). Generally, the amount of
miscible agent will be limited to considerably less than a pore volume, and must itself be driven through the
reservoir by a following waterflood. Despite these drawbacks, miscible floods have been found to be economically
advantageous. There are two main variations of miscible drives: horizontal drives and gravity-stabilized vertical
(downward) drives. The process efficiency is strongly affected by whether or not the drive is assisted by gravity.
The results of these projects, in terms of the percentage of original oil-in-place (OOIP) recovered, range from 3-15%
greater than water-flooding for horizontal drives to 15-25% for gravity-assisted drives.
There are a variety of miscible recovery processes, which are characterized by the type of miscible agent used, and
also, in the case of multiple-contact miscibility, by the procedure through which miscibility is attained. The main
mechanismsworking in a miscible process are extraction, solubilization, vaporization, condensation and dissolution.
These processes trigger other mechanisms such as oil swelling, viscosity reduction and solution gas drive.

Miscible Flooding with LPG
Natural gas is first-contact miscible with LPG, and LPG is first-contact miscible with most crude oils under most
reservoir temperatures at pressures of 600 to 1,700 psia. This suggests a process in which a slug of LPG is injected
and then natural gas is injected to displace as much crude oil and LPG as possible before the producing gas/oil ratio
rises to an economic limit.
We can expect recovery of both oil and LPG to be incomplete, since the LPG fingers through crude oil and the
natural gas fingers through the LPG and then through the crude oil. In a homogeneous, single-layer reservoir, a
simple estimation using the five-spot correlation I developed in 1972 gives 52% oil recovery for an 18% (basis
hydrocarbon-filled pore volume) LPG slug followed by dry gas. In a heterogeneous, multi-layer reservoir, this
recovery would be reduced by a factor of 1/8 to 1/4, depending on the degree of heterogeneity, because of gravity
segregation in each layer and varying slug sizes in different layers. Some further reduction would result from areas
outside of well patterns, near the edges of the reservoir, not being significantly swept.
Miscible Flooding with Natural Gas
At sufficiently high pressure, lean natural gas (natural gas containing less than 10% of components other than
methane), flue gas, or pure nitrogen becomes multiple-contact miscible with most crude oils. The pressure required
depends upon the temperature and the nature of the crude oil. At temperatures in the 100-130 F range, and with
crude oils of 35-100 API gravity, the miscibility pressure is about 4,000-5,000 psia. At elevated temperatures (200-
250 F) the pressure is about 7,000-8,000 psia. For lower API gravity crude oils, the pressure rises by about 500 psia
for each 5 degrees lower API gravity, and vice versa.
The cost of natural gas has historically been lower in the U.S., relative to its fuel value, than the price of crude oil,
and still lower based on the cost of a barrel of gas at reservoir conditions compared to the value of the oil that it
displaces. Regulation by the Federal Power Commission of the price of natural gas which was transported in
interstate commerce began in 1954. Due to the initial decision which placed interstate gas under federal control, and
subsequent FPC rulings (particularly the "Phillips Case" in 1960, and the "Permian Basin" ruling in 1965), producers
of natural gas have until recently only been able to charge a price for natural gas which was based on its generally
low production cost, not on its competitive fuel value in terms of heat of combustion per standard cubic foot relative
to the cost of the same amount of heat when derived from crude oil. The latter ratio would require about the same
price for 5.5 thousand standard cubic feet of natural gas as for a barrel of crude oil. When crude oil was $3 per
barrel, this would have called for a price of 3.00/5.5 or about 55 cents per thousand standard cubic feet (mscf).
Instead, the price was set by law at levels ranging from 15 to 20 cents at the wellhead.
The volume of the gas at reservoir conditions can be estimated roughly by the ratio of the pressure to atmospheric
pressure. Thus, at 4,500 psi relative to 15 psi (atmospheric pressure) the volume is about 300 times smaller, so that
1,000 cubic feet of gas at atmospheric pressure would occupy about 3 1/2 cubic feet at reservoir pressure. In most
cases, about 1.5 to 2.0 mscf would equal the volume of one barrel of crude oil.
It is evident, therefore, that natural gas qualifies as a miscible drive agent which is cheaper than crude oil and can
therefore be left in place of crude oil in a reservoir.
It is possible to obtain multiple-contact miscibility at a pressure that is lower than that required for first-contact
miscibility. The maximum pressure that can be applied in the process is generally limited by the fracturing pressure
of the reservoir. This fracture gradient is seldom more than 0.7 psi per foot of depth and is sometimes as low as 0.5
psi/ft. For example, the maximum injection pressure at 5,000 ft vertical depth will be from 2,500 to 3,500 psia.
The pressure required for miscibility depends on the reservoir temperature. The temperature of the reservoir depends
on the natural rate of rise of temperature with depth from the surface temperature, which averages about 70 F. If
this geothermal gradient (as this rate of rise is called) is as low as 0.6 F per 100 feet of depth, as it is in some parts
of west Texas, then the reservoir temperature at 5,000 feet depth is only 100 F. Such a relatively low temperature
(for the reservoir pressure which can be employed) is advantageous; it is easier to obtain miscibility with any given
crude oil. Typical geothermal gradients are 1.02.5 F per 100 feet (1.84.60 C per 100 m).
The small amounts of ethane and propane normally present in natural gas that has been processed to remove LPG
make the miscibility pressure slightly lower than that for pure methane; high pressure miscible gas drives do, in fact,
use processed ("dry" or "lean") natural gas rather than pure methane, which would be hard to obtain.
The pressure not far from an injection well is considerably lower than the injection well bottom-hole pressure, which
is limited by the fracturing pressure. Therefore, it is usually necessary to take advantage of the lower pressure
requirement of the conditionally-miscible or multiple-contact miscible process. In this process, the equilibrium
contact of the lean gas with the first crude oil it encounters does not lead immediately to miscibility; but it does
produce a gas containing an increased proportion of the low molecular weight hydrocarbons normally present in the
crude oil. This addition of ethane, propane, butane and pentanes to the methane brings the gas closer to miscibility
with the crude oil. As this enriched gas moves away from the injection well, it acquires, by a series of such contacts,
a sufficiently high content of these intermediate hydrocarbons as to be immediately miscible with the next crude oil
which it encounters. This progressive enrichment of the gas as it moves away from the well is considered a
vaporization process, and so this high pressure gas miscible drive process is called a vaporizing miscible drive.
Both nitrogen (available from the atmosphere) and flue gas (which can be obtained by burning a hydrocarbon fuel to
obtain about ten times as much flue gas volume as the hydrocarbon fuel in a gaseous state) can be used instead of
lean natural gas, often at an economic or logistic advantage.
Miscible Flooding with Enriched Gas
It is possible to achieve multiple-contact miscible displacement at much lower pressures than that required for the
vaporizing gas drive process by using natural gas to which a moderately high proportion of intermediate
hydrocarbons (LPG and natural gasoline) has been added. These intermediates are transferred from the injected gas
to the crude oil that it first encounters, and by a series of such transfers (the stripped gas moving on out into the
reservoir), the crude oil near the well is so enriched in the intermediates that it is immediately miscible with the next
increment of injected gas. Since the injected gas must be externally enriched in intermediate hydrocarbon content,
this is often called an enriched gas drive process. Since the process of transfer of intermediates from gas to oil may
be considered a condensation process, it is also called a condensing gas drive.
Despite the apparently clear distinction between the vaporizing and the condensing gas drive processes, it was
learned that, in quite a few cases, the miscible displacement mechanism is not just one or the other of these
processes; rather it is a combination of both. Zick (1986) first described this process, while Stalkup (1987) and
Novosad and Costain (1989) further developed the subject. In a triangular diagram, where a pseudo-critical point or
plait point is normally shown on a two-phase envelope having vapor phase on one side of the plait point and liquid
phase on the other side, and which is often used to explain either vaporizing or condensing gas drive mechanisms,
instead of the plait point, a neck appears and the two-phase envelope then widens into a new region in which both
phases are dense super-critical fluids.

Miscible Flooding with Carbon Dioxide
Carbon dioxide flooding is currently a popular form of multiple-contact miscible flooding. It has some
characteristics which are significantly different from those of the hydrocarbon miscible solvents, and others which
are generally similar. The principal differences are:
1. It is significantly soluble in reservoir brine or injected flood water, whereas hydrocarbons are not (however,
it is not miscible with water under any conditions where water is a liquid phase).
2. At typical reservoir conditions, it has a density near that of reservoir crude oils, while the density of the
light hydrocarbon miscible drive agents is considerably lower. In some cases, carbon dioxide is slightly
more dense than the crude oil; however, it is always considerably less dense than liquid water. Hence, there
is much less gravity segregation of carbon dioxide than there is of hydrocarbon solvents from crude oil,
though it still occurs relative to a mobile water phase.
3. At any given pressure, more carbon dioxide than methane will dissolve in crude oil. The swelling of crude
oil due to dissolved carbon dioxide is about the same as the swelling due to the same ratio of methane,
hence, since more dissolves at a given pressure, it swells crude oil more at a given pressure than does
methane.
4. At any given pressure, it reduces the viscosity of crude oil more than does the amount of methane which
will dissolve at the same pressure. A consequence of its solubility in water is that mobile carbon dioxide
can transfer through water phase to dissolve in and swell waterflood-trapped oil ganglia.

Phase Behavior in Multiple-Contact Miscible Processes
Phase behavior refers to the existence of one or more phases at given conditions of temperature, pressure and
composition, and the relationships between the phase compositions.
Equilibrium Phase Compositions
It is common engineering practice to deal with processes where components present in two or more phases (solid,
liquid or gas) change from one equilibrium set of phase compositions through a series of other such compositions
until they attain a desired degree of transfer from one phase to another. Usually, these phase composition changes
result from changes in temperature, pressure, or the total composition of all of the phases present in a volume region
small enough that the phase compositions throughout the region, as well as the total composition, are constant. This
uniformity in composition must be brought about by some mixing process.
In chemical plants and refineries, this uniformity can be attained by powerful mixing devices in pipelines or vessels.
Even more common is the use of an elongated, horizontal or vertical, vessel containing mixing devices such as
distillation trays or zones filled with loose packing material. Different phases, such as liquid and gas, or two only
partially miscible liquids, are often injected into these vessels at different locations so that the phases move in
opposite directions through the vessel. The desired result is that the exiting phase compositions are very different
from the entering phase compositions.
While the exchange of components between the phases as they pass through the system may be continuous
(especially in the case of the packed vessel or tower), the exchange process can still be accurately described as being
equivalent to some number of stages of equilibrium component exchange, or alternatively as a larger series of steps
that partially approach equilibrium compositions (e.g., in each plate space of a distillation tower). In the component
exchange which may take place between an injected fluid and in-place fluids in porous, permeable rock surrounding
an injection well in an oil reservoir, the mixing effects are due to molecular diffusion and, to a much larger extent, to
the variations in flow path lengths and velocities of the fluid traveling through interconnected pores and to the
dividing and rejoining of these flow paths. The latter mixing effects are proportional to the velocity of flow and are
stronger in the direction of flow than in directions transverse to the flow direction (Perkins and Johnston, 1963).
This mixing is similar to molecular diffusion but is distinguished from it by calling it dispersion instead of diffusion.
Because of these mixing effects, exchange of components between an injected stream and the in-place fluids takes
place, and these exchanges can be interpreted as comprising a series of equilibrium steps. In such a series, after a
given equilibrium is reached, either one or both of the separate phases move into other regions, where they come
into contact (and mix) with another, different portion of the other phase. After a series of such sequential contacts,
one of the phases may approach a composition which is miscible with subsequent portions of the other phase.
This desired result can be studied in terms of the series of equilibrium steps that can bring it about. It is helpful to
consider how equilibrium compositions are determined, based on the pressure, temperature and overall composition
in a given mixed region, and how the desired result of miscibility can be brought about for crude oil and possible
solvents.
Gibbs Phase Rule
In general, phase behavior is governed by Gibbs Phase Rule (see Findlay, 1951), which states that the number of
degrees of freedom (i.e., the number of system conditions such as temperature, pressure and composition which can
be altered independently of each other) is given by:
F=C-P+2 (2.1)
where
F=the number of degrees of freedom
C=the number of components (different chemical compounds or uncombined chemical elements)
P=the number of phases present in the system (a phase being defined only relative to another phase, such
that between the phases there exists a surface which completely separates one phase from the other with a
surface tension force consequently existing in that surface).
The system is defined as a region in which the conditions of temperature, pressure and composition are uniform, and
equilibrium is defined as the condition in which the properties of the system and the phase distribution of
components within the system do not change as time progresses, and will return to the same state if an independent
variable (such as temperature) is altered slightly and then returned to the initial value.
This rule is very helpful when the system has only a few components. For example, in the case of steam in
equilibrium with liquid water, there are two phases but only one component, so there is only one degree of freedom.
Thus, if the pressure is changed, the temperature adjusts automatically, and vice versa. The tables of saturated steam
properties thus have only one independent variable.
However, whenever crude oil is involved in a phase equilibrium, there are so many different chemical compounds
present in the crude oil that the degrees of freedom may seem to be almost infinite. Nevertheless, even in this case,
the distribution of each chemical component between such phases still follows definite physical and chemical rules.
In general, the behavior can be specified in terms of the ratio of the concentration (mole fraction) in a given phase
relative to that in another phase; these values are called K values.
Flash Calculations From K Values
If K values are given for the components of a mixture, then the composition of the liquid and vapor phases can be
rigorously determined using an equilibrium flash calculation (named for the rapid or "flash" vaporization of gas
from a liquid phase when pressure is lowered). For more than two components, though, this procedure requires a
trial-and-error calculation involving the variation of the liquid or the vapor mole fraction until the correct value is
found.
The behavior of these vapor/liquid K values as temperature and pressure vary has been studied extensively for
compounds present in crude oils. The results are presented in the Gas Processors Suppliers Association (GPSA)
Engineering Data Book (1972) as a series of K-value charts for each of the hydrocarbons from methane through iso-
and normal pentane, and for the mixed hexanes through decanes. This publication is presently the most authoritative
compilation of these data.
Calculations From Equations of State
As an alternative, the equilibrium K values can be calculated by means of an equation of state, plus constants
appropriate to each of the chemical components present in any given mixture (see Reid, Prausenitz and Poling,
1987).
The simplest of such equations is the ideal gas law, PV = nRT, where n is the number of moles present (thus, PV =
RT when there is only one mole present). In that law, there is only one constant (R), in addition to the state variables
of system volume (V), system temperature (T), and system pressure (P). R is called the universal gas constant, and
values are given for different sets of units in handbooks such as the CRC Handbook of Chemistry and Physics.
Unfortunately, real compounds and elements do not obey the ideal gas law at temperatures and pressures very far
from typical ambient temperatures (e.g., 15-38 C) and atmospheric pressure. It is therefore necessary to make
suitable corrections.
For a single component, it is necessary only to add an empirical adjustment factor Z which varies with temperature
and pressure, so that the non-ideal gas law PV= ZRT gives an accurate description of the behavior (per mole) of the
given component as the pressure and temperature vary, with one set of Z values applying to the component in the
gaseous state and another set of Z values for the liquid state. This need for an empirical correction factor Z has made
it evident that a more elaborate gas law is needed, and efforts have been made to supply more correct gas laws.
The first moderately successful elaboration of the ideal gas law was made in 1873 by van der Waals. More recent,
more successful ones are those of Redlich and Kwong (especially in the version as modified by Soave), and Peng
and Robinson. These are described in detail in physical chemistry texts and in Reid, Prausnitz and Poling (1987).
The original articles in which these improved equations of state were presented by their authors are reprinted in SPE
Reprint Series Book No. 15, Phase Behavior (1981).
In most cases, the engineer will use an equation-of-state computer program to calculate phase equilibria, or will use
a flash equilibrium computer program which requires K values to be supplied (e.g., by taking them from the GPSA
Engineering Data Book) In either case, it is helpful if the engineer knows what is taking place. For this reason, some
simple examples of flash calculations and of equation-of-state calculations are given in the Appendix.

Phase Equilibrium Diagrams
Equilibrium compositions may be calculated for a given mixture at a variety of pressures and temperatures, or, if a
single temperature is more appropriate (such as a fixed reservoir temperature), for a variety of pressures. This will
give direct calculations of bubble-point pressure, and of phase proportions and compositions at pressures below the
bubble point. It is of interest in some cases to calculate the behavior (versus pressure at a reservoir temperature) of a
series of mixtures of a crude oil with solvent at increasing mole fractions (x) of solvent. This is called a p-x diagram.
If, at a given reservoir temperature and pressure, the phase behavior of various mixtures of solvent with a given
crude oil is calculated in a region of mixtures in which a vapor and a liquid phase result, then these pairs of vapor
and liquid compositions may be plotted on a triangular equilibrium phase diagram, and the liquid and vapor
compositions trace out a curve which contains all of the two-phase compositions in that diagram. Such a diagram is
helpful in explaining the behavior of the multiple-contact miscible flooding processes mentioned previously.
An equilateral triangle is usually used to construct the diagram, though it is also possible-and sometimes
advantageous-to use a right triangle. The discussion below assumes an equilateral triangle.
The key assumption of triangular equilibrium diagrams is that a given mixture of components is composed of only
three pure components, or that the mixture behaves as if it were composed of three components. (The latter may be
called pseudocomponents.) It is, in fact, often possible for hydrocarbon systems containing a light (low viscosity)
crude oil, LPG and natural gas to be adequately characterized as composed of just three pseudocomponents. The
first component is methane, the second component is LPG, and the third component is C7+
(i.e., heptanes and all
higher mole weight components lumped together). The natural gas is considered to be composed of methane plus a
slight amount of LPG, and the crude oil is considered to be composed of C7+
, LPG and methane. The mixture
referred to as "LPG" in this case is actually a mixture of C2-to-C6 hydrocarbons; this is a somewhat broader carbon
number range than we usually include in the term LPG. The C2-to-C6 range is commonly termed "intermediate
hydrocarbons," but we shall use the term "LPG" to designate this wider range.
It is normal practice in the ternary representation to place C1 at the top vertex, C7+ at the lower left vertex, and C2-C6
(or LPG) at the lower right vertex, as shown in Figure 1 .


Figure 1


Each vertex represents the point of 1.0 mole fraction, or 100 mole percent, of the component named at that vertex,
and the opposite side is 0.0 mole fraction, or zero mole percent, of that component.
Any point within the triangle represents a mixture of the three components, and the amount of each component is
determined by the distance of the point from each of the three sides. The sum of these distances is equal to the
distance of any vertex from its opposite side.
That is, the sum of the three mole fractions always equals 1.0 (the sum of the mole percentages equals 100).
Furthermore, with regard to any three points in a straight line within the diagram (including a point or points on the
boundaries), the intermediate point represents a mixture of the compositions represented by the two points at the
extremities. Such a set of points represents a material balance, and the amounts present are determined by the lever-
arm principle. That is, if the intermediate point is the fulcrum, the amount at one end times its distance from the
fulcrum is equal to the same product for the other end.
The diagram usually contains at least one two-phase region, and may contain more than one such region and/or a
three-phase region. The two-phase region is bounded by a "two-phase envelope," and contains tie lines that end on
the envelope at each end. These two endpoints represent the compositions of the two phases which co-exist at
equilibrium as a result of de-mixing of any mixture on the tie-line (always a straight line) joining the two ends. The
proportions of the two phases which will result from a given mixture point on the tie-line are determined by the
lever arm principle stated above.
There are an infinite number of tie-lines present in the two-phase region, just as there are an infinite number of
points on the two-phase envelope. Where the twophase envelope terminates at one side of the diagram, a tie-line
exists which lies in that side of the diagram. As the tie-lines diverge from this limiting tie-line on the side of the
diagram, their slope usually changes, and they usually get shorter when the two-phase envelope is convex toward the
opposite vertex as shown. The tie-lines may eventually get so short that they converge into a point, which is called
the plait point or pseudocritical point. The tangent to the curve at the plait point is called the critical tie-line, and is
an important determinant with regard to attainment of multiple-contact miscibility, as will be shown.
The area of the two-phase region varies with temperature and pressure. In reservoir engineering applications, the
temperature is normally determined by the depth and the geothermal gradient, such that there is a reservoir
temperature which does not vary much from the top to the bottom of oil-bearing formations of moderate thickness.
In this case, the two-phase area is primarily determined by pressure, which does vary as a consequence of
production operations in the reservoir. The two-phase area decreases with increasing pressure. It is often possible, by
raising pressure, to attain miscibility when it is impossible at a lower pressure. The area lying outside the two-phase
envelope is a region where only a single phase exists. If a displacing fluid comprised of methane and LPG (and
therefore represented by a point on the right side of the triangle) can be connected with a point representing the
composition of reservoir crude oil by a straight line which does not cross the two-phase region, then the displacing
fluid is first-contact miscible with the crude oil. This is illustrated by Figure 2 and Figure 3 .


Figure 3





Figure 2


The variation in size of the two-phase region with change in pressure is illustrated by the sequence in Figure 4 ,


Figure 4


Figure 5 ,


Figure 5


and Figure 6 ,


Figure 6


with the pressure decreasing from Figure 4 to Figure 6 .
As can be seen, the relative size of the two-phase region grows as the pressure falls. Eventually the two-phase region
bulges out of the right side of the triangle, and the result is seen in Figure 6 . If the pressure were lowered
considerably more, the two-phase region would also bulge out of the bottom of the triangle. Then there would be left
three one-phase regions close to the apexes of the triangle. This last occurrence is at such a low pressure (near
atmospheric) that we are not concerned about this case in miscible drive processes. There is a case where a neck
develops instead of a plait point; this is discussed later. It is a case of incomplete miscibility, but high oil recovery is
possible.
If the reservoir pressure were high enough that the two-phase region disappeared out of the left side of the triangle,
then all of the components shown would be miscible in all proportions, and any injected fluid, including pure
methane, would be first-contact miscible even with a crude oil entirely lacking in C2-C6 intermediate hydrocarbons,
i.e., just C7+. This situation is not a real case, however. The usual case is one similar to Figure 5 .

Phase Behavior in Hydrocarbon Miscible Floods
High Pressure or Lean Gas (Vaporizing) Drive
In the past the relative value of the components has generally been: methane lowest, C2-C6
next, and C7+ highest. It is
therefore desirable to use a drive agent containing as little LPG and as much methane as possible to attain
miscibility. The case of almost pure methane (natural gas from which as much LPG has been removed as is
economically possible) is that which we call the high pressure gas drive, or lean gas (vaporizing) drive. But the
possibility of using this lowest cost drive agent depends on the composition of the crude oil (as a point in the
triangular diagram) relative to the envelope of the two-phase region. If the line connecting the crude oil composition
and the drive gas composition lies outside the two-phase region, then we have a first-contact miscible high pressure
gas drive situation (highly desirable). Note that for this to happen, the crude oil must contain a rather high content of
both methane and LPG, not very much C7+.While this does happen, it is not very common. A more common case is
shown in Figure 1 , where the line between the crude oil point and the drive gas (nearly pure methane) passes
through the two-phase envelope.


Figure 1


In Figure 1 , however, note that the crude oil composition still lies to the right of the critical tie line. This is most
important, and if it is not the case, as shown in Figure 2 , then multiple-contact miscibility is not attainable.


Figure 2


It is of course possible to move the critical tie line by changing the pressure level in the reservoir, but it is not
possible to change the location of the crude oil composition (other than to reduce its methane content by pressure
depletion, which is usually unfavorable from this standpoint). If the pressure can be raised sufficiently, either by gas
injection or by water injection or both, a high pressure miscible flood might become possible even if it had not been
at the time when miscible flooding was first contemplated. Otherwise, consideration must be given to the possibility
of enriched gas as the miscible displacing agent.
If nitrogen or flue gas is the high-pressure miscible displacing agent, either of these would take the place of methane
as the lowest molecular weight component. A complication arises when methane is present as a major component in
the crude oil. In this case, a three-dimensional phase diagram may be needed (see Stalkup 1984) with nitrogen (or
flue gas) at one apex of a tetrahedron, methane at a second apex, and LPG and C7+ at the other two apexes. The two-
phase region becomes a mound within the tetrahedron, and every tie-line has ends at specific points, which together
comprise a curve on the surface of the mound. There is still a critical tie line which is tangent to the surface of the
mound at the pseudo-critical point. To achieve multiple-contact miscibility, it is still necessary that a line from the
crude oil composition be able to reach a gas-phase composition, which can be achieved by multiple contacts of the
injected gas with the crude oil. Such a gas phase composition will therefore be a point on the tie-line curve on the
surface of the mound, and will lie on the gas-phase side of the pseudocritical point. If a straight line connecting the
crude oil composition with such a point is not possible, then a high pressure miscible gas drive is not possible with
the given system, but if it is possible, then such a drive is possible.
The reason for this difference is as follows: In Figure 1 , a mixture of drive gas and crude oil falling in the two-phase
region (for example at the point marked "x" on the tie-line nearest to the pseudocritical point P
c
) will separate into a
liquid phase lying at the lower end of the tie line and a gas phase with the composition of the upper end of the tie
line. In this case, note that the dotted line connecting the crude oil composition with this gas phase composition lies
entirely in the single-phase region, i.e., the first (immiscible) contact of the crude oil and drive gas forms a gas phase
which is much richer in LPG (and incidentally in C7+ ) so that this gas is miscible with crude oil on ensuing contacts.
This illustrative case, in which only one immiscible contact was sufficient to produce a gas phase miscible with
crude oil, is not really representative. It normally requires quite a few stages of immiscible contact before the line
between the richest gas so produced and the crude oil composition grazes the two-phase envelope. Thus, the last
stage required is where the gas phase is at the point where a tangent line from the crude oil composition touches the
two-phase envelope. This point must lie above the pseudocritical point, since all points above are gas phases and all
points below are liquid phases. The tangent to the two-phase envelope at a point above the pseudocritical point must
have a slope such that it crosses the critical tie-line at a point above the pseudocritical point. Therefore, all possible
compositions for crude oil lying on the tangent line must lie to the right of the critical tie line extended downward.
We might ask, why can the crude oil composition not lie to the left of the critical tie line if it is also above the
pseudocritical point? The answer is that, in that case, we would call it a gas instead of a crude oil. We may define a
gas in the single-phase region to be a composition which lies above the pseudocritical point and to the left of the
critical tie line. The region to the right of the critical tie line is ambiguous. At most of the pressures of interest, two
phases may exist here as dense supercritical fluids, but this occurs only when there are more than three components
present, and when some of them are of relatively high molecular weight.
Enriched Gas (Condensing) Drive
Now, what can be done if the crude oil composition lies to the left of the critical tie line ( Figure 2 )? In this case we
must simply use a drive fluid composition that lies on the other side of the critical tie line. This is shown in Figures 3
and 4.
In Figure 3 , the gas has not been sufficiently enriched in LPG to meet this criterion, and so all mixtures of the drive
gas and crude oil remain immiscible.


Figure 3


The first contact cannot produce a liquid richer in LPG than the lower end of the tie line, where the line connecting
crude oil with the drive gas meets the upper part of the two-phase envelope. The contact of drive gas with that
enriched liquid cannot produce any mixtures lying to the right of the tie line which, when extended, passes through
the drive gas composition (the second dotted line in the diagram).
In Figure 4 ,


Figure 4


however, the first contact of the enriched gas with crude oil, forming the mixture marked "x" on the first tie line to
the left of the pseudocritical point, separates into a gas (at the top end of the tie line) that contains less LPG than the
drive gas and a liquid (at the lower end of the line) which contains more LPG and methane than the crude oil. Note
that the tangent to the two-phase envelope from the enriched gas drive composition touches the envelope at the
lower end of this tie line. The enriched drive gas is miscible with that liquid phase, since the dotted line between
them lies entirely in the single-phase region.
In most cases, many more stages of contact would be required to reach a liquid composition that would be miscible
with the drive gas (for example, suppose that the enriched drive gas were only slightly to the right of the point where
the critical tie line intersects the right side of the triangle). Nevertheless, so long as the drive gas and crude oil
compositions lie on opposite sides of the critical tie line, multiple-contact miscibility can be achieved.
Notice that we have only two degrees of freedom in adjusting conditions to make this possible: pressure and drive
gas composition. We cannot readily (or economically) change the reservoir temperature from its initial value just for
the purpose of attaining miscibility; hence, we may consider the temperature a constant. The crude oil composition
is likewise fixed at whatever composition exists when the miscible flood is being considered. Note, however, that if
the reservoir has undergone pressure depletion, the pressure may vary over different regions of the reservoir, and so
the dissolved gas content and thus the composition may vary (in the terms we have been considering) on the
triangular diagram. This introduces complications which we shall not consider further here; they are best taken care
of by reservoir simulation, using a compositional simulator in which the initial composition may be varied as
required over the area and depth of the reservoir.
The LPG Slug Process
There is, of course, one more case which has been mentioned-that of using an LPG slug to drive crude oil and then
using dry gas (nearly pure methane) to
drive the LPG. This case is shown in Figure 5 , where the LPG is shown as containing a slight amount of methane
and C7+ in order to move the point from the apex into the single-phase region close to the apex.


Figure 5


This process allows the lowest pressure of all of the processes discussed, while still maintaining miscibility.
The minimum pressure that may be allowed in the LPG slug process is that at which the two-phase region begins to
bulge out of the right side of the triangle. This means that the dry drive gas is then immiscible with the LPG. This
happens at a higher pressure than that at which LPG and crude oil become immiscible. Hence, the critical condition
for attaining miscibility in this method is to maintain a pressure high enough for the drive gas and LPG to remain
miscible.
In Chapter 5 of Miscible Displacement (SPE Monograph No. 8), Stalkup (1984a) gives plots from which this
pressure may be determined. Stalkup also describes circumstances where the two-phase region bulges out of the
bottom of the triangle before it bulges out of the right side; this occurs primarily when the reservoir temperature is
near or above the critical temperature of the LPG. The principal hazard involved in the LPG slug process (in regard
to miscibility) occurs during the process rather than at the beginning. Diffusion and dispersion cause mixing at the
front of the LPG slug between the crude oil and the LPG; on the triangular diagram ( Figure 5 ), these mixtures are
on the line joining the LPG and the crude oil. At the back of the slug, diffusion and dispersion create mixtures along
the line joining the LPG and the drive gas compositions. When these dispersion zones overlap, mixtures are created
that lie between these two lines-and that is, in the two-phase region. When the peak LPG concentration falls below
the peak of the two-phase envelope, miscibility will be lost.
Viscous fingering contributes strongly to this last-mentioned effect. LPG fingers through the crude oil, and the dry
gas follows the LPG fingers, usually occupying the middle. The result is that crude oil, LPG and gas are flowing in
parallel rather than in sequence. An LPG slug size calculated on the basis of dispersion rates to be adequate to avoid
loss of immiscibility when the fluids travel in sequence may be inadequate when they travel in parallel. This
depends on the magnitude of the transverse dispersion rate relative to the longitudinal dispersion rate used in the
sequential calculation, and also on the thickness of the sheath of LPG between the gas fingers and the crude oil.
In areal terms, as determined by study of such viscous fingering in a Hele-Shaw (parallel-plate) model, it appears
that the band of "LPG" surrounding fingers of "gas" is about one-fifth as thick as the annular band would be if
fingers did not occur. If dispersion is in the low range where molecular diffusion is the controlling factor in both
longitudinal and transverse mixing, then the slug would need to be approximately five times as large as would be
required if fingering did not occur. In three dimensions, this translates to 5
3/2
, or about eleven times as large.
However, if both longitudinal and transverse dispersion are in the velocity-controlled region, where transverse
dispersion is about one-tenth as great as longitudinal dispersion, then the hazard is not loss of miscibility in the
transverse direction but in the longitudinal direction. In this case, the slug is thinned by fingering just as much (if not
more) at the tips of the fingers as at the sides. If the displacement becomes immiscible at the tips of the fingers, then
penetration of crude oil by immiscible fingers of gas occurs. This immiscible displacement of oil is much less
efficient (on the microscopic scale) than the miscible displacement, and it may be expected that much more oil
would be left behind in the paths of the fingers than would be the case if immiscible gas fingering did not occur
The Combined Vaporizing/Condensing Process
At the 1986 SPE Annual Meeting, Aaron Zick of ARCO presented a paper (SPE 15493; see the References) which
showed that in experimental contacts of moderately heavy crude oils (at temperatures from 160-205 F and
pressures from 3,100-3,600 psia), with solvent compositions representing enriched gas such that a condensing
multiple-contact miscibility mechanism was expected (with pressures 500 psi or more above the multiple-contact
miscibility pressure as found by slim-tube tests), miscibility did not actually occur. The gas and liquid phases
produced by multiple contact approached each other in composition, but then held constant or diverged; the K
values approached 1.0, but the values for the lightest components did not get as low as 1.0, and those of the heavier
components rose close to, but not as high as, 1.0. The densities of the two phases became close but not equal.
Simulations with the Peng-Robinson equation of state using only one C7+ pseudocomponent did not show this
behavior but rather showed the typical condensing mechanism with miscibility attainment. On the other hand, more
extended analyses of the crude oils were available, and when three or more pseudocomponents were used with the
highest one being a C30+ pseudocomponent, the P-R equation of state was able to match the experimental behavior.
The triangular diagram calculated in the latter case is compared with that of a true three-component (C1, C4, C10)
condensing multiple-contact miscibility system in Figure 6 and Figure 7 (sketches similar to figures in Zicks paper).


Figure 6





Figure 7


Later papers (Stalkup, 1987; Novosad and Costain, 1989) verify Zicks conclusions, both with respect to the
experimental behavior and to the need for several pseudocomponents, ranging up to a relatively high molecular
weight pseudo-component, in order for an equation of state to match the experimental behavior.
Stalkup and other authors have pointed out that the simple three-component representation, which is so convenient
for describing the miscibility mechanisms as discussed above, is not an adequate representation for crude oils. Only
in the case of a light crude oil practically a condensate-is this representation valid. For real crude oils, the near
approach to miscibility permits a high oil recovery, but
in essentially every case, a small amount of high molecular weight oil is left behind by the displacement process. In
effect, this is always shown by the slim-tube tests commonly used to determine "miscibility" pressure. The oil
recovery is never 100%, either at breakthrough or at a throughput of 1.0 or 1.2 pore volumes of solvent (these
throughputs are recommended by different authors). The oil left behind in the tube is cleaned out using an aromatic
solvent such as benzene or toluene; these are good asphalt solvents. If the solvent is evaporated, the remaining oil is
found to be of much higher molecular weight, density and viscosity than the crude oil, and it contains more asphaltic
components. This kind of heavy oil residue is predicted by the vaporizing/condensing process. Even propane solvent
will leave such a heavy oil precipitate. The amount of this heavy oil residue is somewhat greater with CO2 (3%-8%)
than it is with propane or LPG-enriched natural gas (2%-4%). This is in accord with what is known about the
relative precipitation behavior of these solvents from their behavior in lubricating oil refining processes and cat
cracker feed preparation by solvent precipitation of heavy ends. This does not preclude high oil recoveries by
solvent displacements in oil reservoirs, but it means that a small residual saturation of heavy, asphaltic oil will
remain in almost every case. We still use the phrase "miscible flooding" for this process, as a matter of convenience.
In all of the processes discussed, if a waterflood has taken place before the miscible flood is carried out, then an
immiscible displacement of water by the miscible drive agent must occur in order to mobilize all of the oil, at least
part of which will exist in a water-trapped saturation.

Phase Behavior in Carbon Dioxide Miscible Floods
The solubility of CO2 in water is determined by the local CO2 partial pressure at any given temperature and by the
water salinity (solubility decreasing with increasing salinity and with increasing temperature, but increasing with
increasing pressure). The CO2 partial pressure at equilibrium is equal to the mole fraction of carbon dioxide in the
non-aqueous phase times the pressure in the that phase. If a phase consisting of 100 mole percent carbon dioxide is
introduced, at a given pressure, into a system containing rock, brine and a waterflood-trapped oil phase, then before
the CO2 phase becomes diluted by dissolved oil or hydrocarbon gas, that pressure is the CO2 partial pressure, and an
according amount of carbon dioxide dissolves in the water phase to arrive at equilibrium with the pure carbon
dioxide. Any other non-aqueous phase which is present (though not in direct contact with the CO2) must arrive at
equilibrium with this water phase, and can do so only by arriving at 100% carbon dioxide content.
Of course, this means that if oil is anywhere present, it will continue to acquire carbon dioxide from the water phase
until its mole fraction of carbon dioxide matches that of any other non-aqueous phase within mass-transfer distance.
The swelling effect would make the trapped oil phase become at least partly mobilized, and thus would dilute the
mobile carbon dioxide phase with oil. Equilibrium could only be achieved when the non-aqueous phase has the
same content of carbon dioxide and of oil, and the CO2 content in the water is the appropriate solubility for the
pressure and carbon dioxide mole fraction in the non-aqueous phase.
The phase behavior of pure carbon dioxide resembles that of ethane. Its equilibrium K value in mixtures with
hydrocarbons, from methane to the higher molecular weight components of crude oil, is intermediate between the K
values of methane and ethane; the GSPA Engineering Data Book recommends using the square root of the product
of the values for methane and ethane as the K value for carbon dioxide.
In mixtures with crude oil at temperatures up to about 130O F; CO2 swells the oil while raising the bubble-point
pressure until a pseudocritical point is reached. The results are often plotted on a p-x (pressure/mole fraction of
solvent) diagram. If further carbon dioxide is added beyond the pseudocritical point, the system then exhibits a small
amount (by volume) of hydrocarbon liquid dew phase. Both phases contain a high mole fraction of carbon dioxide at
this point, but the upper phase has a higher mole fraction of carbon dioxide than the lower dew phase. The molecular
weight of the hydrocarbons in the dew phase is considerably higher than that of the hydrocarbons in the upper phase.
If the pressure on the system is lowered, a bubble of gas phase appears, which is lighter than the upper phase
previously present (so it was also a liquid phase). Further reduction in pressure leads to division of the middle phase
between the gas phase and the heavier liquid phase until the middle phase disappears, leaving only a gas phase and a
liquid phase. In addition to the three phases mentioned, a tiny amount of asphaltene particles may be precipitated
from solution as a semi-solid material. If the experiment described is performed in a gauge glass, the asphaltene
particles make black specks on the glass.
At temperatures above 130 F, when a dew phase appears, the upper phase is a vapor phase rather than a second
liquid phase, and on lowering pressure below the two-phase boundary the amount of liquid (dew) phase first
increases, then at considerably lower pressure decreases again (similar to retrograde condensation).
Due either to the liquid dew phase, or to the asphaltenes, or to both in varying degrees, the mobility of carbon
dioxide in an oil reservoir is much lower than would be expected from its viscosity. This has been observed in both
secondary miscible floods (Pontious and Tham, 1978) and in tertiary miscible floods (Hansen, 1977; Youngren and
Charlson, 1980).
Gardner et al. (1981) indicate that a lower relative permeability curve, which applies in a partly oil-wet rock in the
presence of both crude oil and carbon dioxide (as well as brine), is responsible for this lower mobility of the carbon
dioxide. Brannan and Whittington (1977) attribute lowered injectivity for water following enriched gas injection at
the Levelland field to similar relative permeability effects (in their case, to three-phase relative permeability effects).
According to Stalkup (Chapter 8 of his monograph, 1984) carbon dioxide achieves multiple-contact miscibility via
the vaporization of hydrocarbons. It appears that carbon dioxide is somewhat unusual in that it primarily takes up
hydrocarbons extending up through the gasoline, kerosene and gas oil molecular weight range, rather than the light
hydrocarbons ethane through pentane, as in the case of natural gas vaporizing gas drives.
In common with propane and butane, CO2 tends to precipitate high molecular weight fractions such as asphalt from
crude oils when the crude oil is in small proportions-for instance, in 2:1 to 5:1 ratio of the solvent to crude oil. Most
of the crude oil dissolves in the solvent, and in the second phase containing the asphaltic components, a considerable
proportion of solvent is present (although less than in the less dense phase).
Thus, according to the available evidence, CO2 is a poorer solvent for crude oil than either propane or butane.
Miscibility pressures are therefore generally higher than for LPG, though far less than the miscibility pressure for
dry natural gas (95% methane).
The different behavior of CO2 observed between its critical temperature of 88 F and 130 F as compared to that
observed above 130 F is shown by the p-x diagrams of Figure 1 and Figure 2 .


Figure 1


The triangular phase diagrams for these systems contain, in the case of the systems between 88 F and 130 F, a
triangular three-phase region, in addition to two or three two-phase regions.


Figure 2



Miscible Flood Phase Behavior: Combined Effects of Fingering and Heterogeneity
In the gas-driven LPG slug process, the fingering of LPG through crude oil is accompanied by fingering of gas
through the LPG fingers. When the gas fingers break through the forward ends of the LPG fingers, an immiscible
contact of gas with crude oil occurs. When we consider the transition in composition through the sides of the fingers,
it is also possible that the maximum LPG concentration will fall below the value needed to maintain miscibility.
This could result from the thinning of this transition region as the fingers lengthen, and from the accompanying
transverse dispersion, which causes crude oil to move inward and the miscible agent(s) to move outward.
These phenomena indicate the need for a much larger slug size than would be needed in the absence of fingering.
Even a larger slug, however, may not entirely prevent immiscible regions from developing, because the finger
thicknesses are determined by formation heterogeneity as well as by the mobility ratio and the velocity. For any
given combination of these variables, there is a natural spectrum of finger size in a homogeneous formation, but
these finger sizes are even more strongly affected by rock heterogeneity.
Formation heterogeneity has many forms, which may be generally distinguished between layered and non-layered
rock formations. Non-layered rock heterogeneity is mostly of weathered reef origin-the weathering creates irregular
vertical channels in coral or stromatolite reefs, which do not have strong layering before the weathering. There are
also fractured shales and breccias which have relatively little layer character. Most rock formations, however, have
been formed by deposition of coarse-to-fine particles of rock (either of sandstone or calcium carbonate shell
fragments) and these are called clastic formations. They typically form layers, because long periods of time were
involved in laying down successive deposits of significant thickness (several inches to a foot or more), and the
conditions of transport and deposit of the particles changed over such time periods. The size range of the particles in
successive layers varies and this results in layers of different porosity and permeability. The variation of layer
permeabilities is often described by measures such as the Dykstra-Parsons coefficient of permeability variation.
Changes in formation properties subsequent to the original deposition are also important. The processes causing
these change are called diagenetic (rebirth) processes. They include dissolution and precipitation of various rock
minerals, and transformations such as that of limestone (calcium carbonate) to dolomite and transformations such as
that of limestone (calcium carbonate) to dolomite (calcium/magnesium carbonate). Between periods when fairly
coarse sand or limestone particles were being deposited, very fine particle beds, which were mud-like in texture,
were often deposited; after compaction these turned into shales in sandstones or their equivalent in limestones and
dolomites. These shale beds, and the permeable beds between them, have a variety of thicknesses, ranging from the
order of a centimeter to multiple-meter scale In the course of the climate and weather changes during deposition of
sequences of layers, the edges of rivers, lakes and ocean beaches on which much of the layer deposition occurred
shifted from side to side and back and forth; often, the water level changed significantly over time. This resulted in
both permeable and impermeable layers ending over some finite distances, so each layer may not continue over
typical well distances.
From the changes in weight of formations above them (including alternate deposition and erosion), and regional
tectonic movements, many rock beds have undergone faulting, in which a broad slab of rock many tens or hundreds
of meters thick moves up or down relative to rock on each side. The friction as the rock faces slide past each other
causes local grinding and melting so that the faults are often impenetrable to fluid movements. In some cases, the
rock cracks in many more places, creating fractured rock formations. These heterogeneities control fluid
movements within the permeable formations to such a degree that fluids can be transported into contact with each
other (for example in adjacent layers) that are not in phase equilibrium. The processes which lead only gradually by
multiple contacts to a miscible state are not able to prevail in such three-dimensional heterogeneous circumstances,
rid contacts between immiscible phases can occur in many places during a displacement that would be miscible in a
one-dimensional system. This leads to incomplete displacement of some of the crude oil. Even in a homogeneous
formation or layer, besides well-pattern sweep efficiency, there are fluid-related mechanical effects-viscous
fingering and gravity tonguing-which limit the volume fraction of the layer being swept by the miscible fluids,
resulting in incomplete displacement of the crude oil. Even when individual layers are relatively homogeneous, the
variation of properties among the different layers often results in some layers being incompletely swept, while in
other, more permeable layers, high flow rates result in the economic limit of gas/oil ratio or of water cut being
reached at the production wells.
Physical Dispersion of Miscible Fluids
The basic literature on this subject is an article by Perkins and Johnston (1963), which is also reprinted in SPE
Reprint Book No. 8, Miscible Processes (1965). Other significant articles are in the same reprint book, namely those
by Blackwell (1962) and van der Poel (1962). The articles by Pozzi and Blackwell (1963) and by Blackwell, Rayne
and Terry (1959) are also recommended.
In SPE Reprint Book No. 18, Miscible Processes II (1985)) see the two articles by Koonce and Blackwell (1965), as
well as those by Warren and Skiba (1964), Coats and Smith (1964) and Baker (1977). The articles by Coats and
Smith and by Baker deal with the contribution of dead-end pores to the overall dispersion behavior. Fatt (1959) had
suggested that such dead-end pores would affect pressure transient behavior. Coats and Smith, and later Stalkup
(1970), thought that the effect of dead-end pores on dispersion would be significant on the laboratory scale but not
on the field scale; the article by Baker concludes that this is not the case; it is also important at field scale, although
much less so than in the laboratory.
The articles by Warren and Skiba and by Blackwell and Koonce introduce the effects of heterogeneity on dispersion.
Warren and Skiba deal with random three-dimensional heterogeneity, and conclude that its effects on dispersion can
be described adequately by the standard treatment of dispersion in one dimension (with a much larger dispersion
coefficient). Blackwell and Koonce also deal with the broadening of a dispersion zone by transverse dispersion
between adjacent layers. This is taken up in more detail by Lake and Hirasaki (1981), who give criteria for deciding
when adjacent layers will have sufficient overlap in dispersion zones to be considered a single layer with a greater
longitudinal dispersion coefficient.
Mahaffey et al. (1966) show that using laboratory sand packs or slabs of natural rock for laboratory miscible floods
results in relatively few viscous fingers, because the dispersion rate is high compared to the convection rate, while
flooding on a field scale results in a relatively high degree of viscous fingering, because the transverse dispersion
rate in the field is almost at the molecular diffusion level, and is much lower than the convection rate. They
recommend the use of Hele-Shaw (parallel plate) models for experimental study of miscible areal sweep efficiency
because, using an analysis similar to that of Taylor (1953), they found that the dispersion coefficient is proportional
to the square of the plate spacing, enabling models with very close spacing to simulate field behavior more closely
than sand-packs or rock slabs.
Perkins and Johnstons (1963) equation for the longitudinal dispersion coefficient KL for clastic rocks (such as
sandstone) is:
KL/D
0
= 1/F
c
+ 0.5(u d
p
/D
0
) (3.1)
The similar equation for the transverse dispersion coefficient KT is:
K
T
/D
0
= 1/F + 0.0157(u d
p
/D
0
) (3.2)
where
D0 = molecular diffusion coefficient for the fluid pair involved
F = formation electrical resistivity factor (ratio)
= porosity (1/F = 1/ , = tortuosity), fraction or ratio
u = superficial interstitial velocity = q/A+, cm/s
= pore shape factor based on average particle size of grains
d
p
= average grain size of the clastic rock, cm
Table 1 of their paper gives values of ad for several outcrop sandstone samples; the average value is 0.36. Below (u
dp/0) values of about 1.0 for longitudinal dispersion, or about 10 for transverse dispersion, the second term on the
right is not to be counted. The ratio K/D0 is just equal to 1/F = 1/ below that point. In oil reservoirs, the value of
the second term is often low enough to place the dispersion value in that range where it is equal to 1/ , while in
laboratory test equipment it is usually in a much higher range of the second (velocity) term.
The quantity is the average tortuosity of the flow path through the pores; it usually has a value on the order of 1.5
to 2.5 (length of the crooked path divided by the straight distance).
Most miscible flood calculations involve numerical methods utilizing computer simulations. In such computations,
numerical dispersion can actually overshadow the physical dispersion effects discussed here.
Numerical Dispersion in Computer Simulations
What is called "truncation error" by mathematicians causes a dispersion of fronts in reservoir simulations done with
finite-difference computer displacement simulation programs, which is similar to physical dispersion in some ways
and different in others. Lantz (1971) gives formulas for the equivalent dispersion due to different forms of
discretization of space and time and of calculation of transmissibility between blocks.
For the case of a simple backward difference and the IMPES (Implicit Pressure, Explicit Saturation) method of
solving the simultaneous flow equations, the diffusion coefficient is:
D(total) = D(physical) + (u
L
) * [_x/L - uAt/L]/2,
Or, dimensionlessly,
[K
Leff
./D
0
] = [KL/D
0
] + (u
L
/D
0
) * [1 - AN
i
/(PV)
n
]/2
n
(3.3)
where
u = the linear velocity
L = the length in the general coordinate direction x
n = the number of grid blocks in the coordinate direction x
AN
i
= the fraction of a mobile pore volume injected per time step At
(PV)
n
= the mobile pore volume per grid block.
In general, an attempt to reduce the quantity [1 - AN
i
/(PV)
n
] below about 0.90 to 0.95 (for the IMPES method of
solution) results in a significant increase in error of the mass balance with increasing number of time steps. For a
given L and u, the only effective way to cut dispersion is to increase the number of grid blocks (n). Note that for a
given flow velocity (u), the number of blocks (n) must be increased along with an increase in size (L) of the total
system, just to keep dispersion constant.
In the case of immiscible displacements, physical dispersion can be introduced into the computer simulation
equations either by means of capillary pressure or by adding a second-order term containing KL to the first order
terms of the Buckley-Leverett differential mass-balance equation. Adding the second order term is the only way to
account for physical dispersion in miscible displacement simulations. Only in unusual circumstances would it be
necessary to add this type of term, because the numerical dispersion due to the second term in Equation 3.3 is
usually so large compared to physical dispersion that it is difficult to make it small enough to equal the physical
dispersion expected in a field process.
For example, for a flow velocity u of one centimeter per hour (0.000278 cm/s), an L of 10,000 cm, and a D0 of
0.0003 cm2/s, and n = 10 grid blocks (10 meters length per grid block), we get about 400 for the second term of
Equation 3.3. That is, the effective dispersion is about 400 times the ratio KL/D0. This is why it is difficult in
computer simulation to match the sharpness of the fluid fronts that actually occur in oil reservoirs. Furthermore, it is
impossible to display features such as viscous fingers with a limited number of grid blocks, such as the ten just used
in the example.
Numerical dispersion in a finite-difference simulator is similar to the dispersion due to discretization of a linear
reaction system, as expressed by a series of ideally-mixed reaction vessels or "CSTRs" in the field of reaction
engineering. This was discussed by Kramers and Alberda (1953). In a single ideally-mixed vessel, dispersion is so
great that an injected tracer emerges immediately from the exit, while as the number of vessels approaches infinity,
the displacement behavior of an injected tracer approaches plug flow (i.e., zero dispersion). The dimensionless
number u
L
/D0 is approximated by (1 + n)/2, where n is the number of vessels in series.

Miscible Flood Sweep Efficiency
Injected fluids not only disperse into fluids already present in an oil reservoir, but also fail, to some extent, to contact
the fluids in place and drive out the desired oil. There are a variety of reasons why this sweep-out of crude oil fails
to take place.
When water is the displacing fluid, capillary forces cause it to travel more rapidly than oil, through some of the
smaller rock pores. Water thereby reaches pore throats ahead of the retreating oil phase and forms tiny films across
the mouths of these pore throats, preventing any further flow of oil through not only the small pores, but through the
larger ones as well. Oil is trapped in isolated multi-pore "ganglia" several millimeters in length and less in width.
The enormous multitude of these fine ganglia adds up to a significant fraction of the crude oil encountered by the
water-from one third to one half-not being produced at the production wells. Thus, while waterflooding is relatively
inexpensive and sweeps most of the area of well patterns, the efficiency of displacement of oil on a microscopic
scale is disappointing.
On the other hand, when miscible flooding agents are used, the efficiency with which they sweep out well patterns is
not very good either. This is partly because of gravity segregation and viscous fingering, but also because the well
patterns for reservoir flooding processes have relatively stagnant regions, where oil can be driven out only by using
many volumes of injected fluid relative to the volume of the reservoir. Water is low enough in cost so that many
volumes can be used, but all of the other flooding agents, including those miscible with the crude oil, are relatively
expensive; when using these agents, we can afford only a fraction of the reservoir volume. The sweep efficiency of
these well patterns thus becomes a matter of major concern.
Sweep efficiency varies not only with the amount of agent injected, but also with its mobility, or relative ease of
flow through a resisting porous medium. For a single fluid, mobility is proportional to the reciprocal of the fluid
viscosity, which measures the resistance to flow of the fluid. When one miscible fluid displaces another, we refer to
the mobility ratio (M), which is the ratio of the mobility of the displacing fluid to that of the displaced fluid; the
mobility ratio is the reciprocal of the viscosity ratio. For the miscible drive fluids currently available, this always
turns out to be a number considerably greater than one, and these mobility ratios give relatively poor well-pattern
sweep efficiency ( Figure 1 ).


Figure 1


Note: The equations referenced below may be found in the section titled "Well Pattern Sweep Efficiency."
Muskats Equations
Muskat (1937, 1949) derived equations for the sweep efficiency of ideal well patterns for unit mobility ratio and in a
single homogeneous layer, with no gravity segregation effects.
Deppes Approximation Method
Deppe (1961) gives important extensions of Muskats work. (There is a closely related article by Prats et al. (1959)
that is also recommended reading.) Deppe gives the equations referred to above, but in slightly different form,
expressing permeability in millidarcys instead of in darcys: hence the numerical constant at the front of the right
hand side is a thousand times smaller. He also uses base 10 logarithms rather than natural base logarithms, so his
constants are also reduced by the factor of the natural logarithm of 10. His constant of 0.0011538, when multiplied
by 1,000 and then by 2.303, gives the value 3.54 used in the section titled "Well Pattern Sweep Efficiency."
The constant in the denominator as Deppe gives it is 0.2688; when this is multiplied by 2.303 we get the value of
0.619 as given by Muskat. Among other information, Deppe gives equations for patterns or parts of patterns on the
edges of fields, which were not previously available. He shows that both regular well patterns and irregular well
patterns can be considered to be composed of injection circles around injection wells and production circles around
production wells, with the sum of the areas equal to the total area of the well pattern.
Claridges Equations
Claridge has developed a set of equations (not published elsewhere) for the influence of mobility ratio and areal
heterogeneity on areal sweep efficiency (E
a
) at injected fluid breakthrough in ideal well patterns.
Applying Claridges Equations to Deppes Method
Claridges breakthrough sweep efficiency equations, mentioned above, may be used to modify Deppes Method.
Deppe states that at breakthrough of injected fluid, the flow changes from being radially inward over the entire
perimeter of the circle representing the interface between the injected and displaced fluid to being radial flow inward
within a sector of fixed radius but increasing central angle, while the displaced fluid flows inward in the remaining
sector of the circle. Deppe places this circle generally at a radius less than the radius of his production circle, and
bases this lesser radius on the breakthrough sweep efficiency.
Claridge proposes the modification of making the production circle radius just equal to this radius at which the flow
changes from being concentric to parallel flow in sectors. This implies that the area of the injection well circle
divided by the total pattern area should be made equal to the areal sweep efficiency E
a
. The production circle area
would then be proportional to (1 - E
a
), and flow inside it would always be sectorwise. When the injected fluid
reaches the periphery of the injection circle by radial flow, it then begins to flow into the production well through a
sector which begins with a zero central angle and increases in central angle while the displaced fluid continues to be
produced from the production circle in a sector which is correspondingly narrowing.

Irregular Well Patterns in Miscible Flooding
The shapes of actual well patterns often do not closely correspond to those of ideal well patterns. They may be
distorted versions of ideal patterns, or they may have no obvious resemblance. Furthermore, in multiple well
patterns, injection and/or production wells generally do not have the same injectivity or productivity for all wells of
each type, as required by ideal pattern relationships. The reasons for irregular pattern shapes are (1) it is difficult to
drill wells to a series of exact predetermined bottom-hole locations, and (2) the number of wells that can be drilled to
cover an irregular reservoir is limited. Differences in injectivity and productivity are caused by variations in the
thicknesses and permeabilities of the layers penetrated by each well.
When planning oil recovery projects under such circumstances, it may be helpful to get some indication of the flow
behavior as determined by actual well locations and flow rates. In the cases discussed, we need to assume that only
one fluid is flowing (or that the mobility of displacing fluid and displaced fluid are the same), that the density is
constant, and that the formation thickness, porosity and permeability are constant over the area of interest.
It is possible to make a set of points track the streamlines; the curves connecting these points at any given time
outline the flood front. They also show, by assigning a given volume of flow to be represented by each point, how
much of the flow from a given injection well arrives at different producers.
Finally, it is possible to define the space between adjacent streamlines in a given well pattern as "stream tubes," and
to calculate the flow through these stream tubes by Darcys Law and the Buckley-Leverett method. It is necessary to
allocate the injected fluid to the different stream tubes, and to calculate the rate of flow based on the total pressure
drop and the length and average cross-sectional area of each stream tube. With this method, it is further possible to
stack one layer on top of another, and (by computer) carry out the accounting of flows of the components such as
oil, water and gas from each stream tube and in each layer, to obtain a total outflow of each component. This permits
assigning different rock properties to the layers. The development of this method by Higgins and Leighton (1962,
1963), just before the wide-spread appearance of finite-difference computer simulators (Peaceman et al., 1959,
1962), came the closest of all prior oil recovery calculation methods to providing a means of calculating oil recovery
without a variety of ideal restrictions. However, it still did not include the effects of viscous fingering, of gravity
segregation of phases or miscible fluids within each stream tube, or of crossflow between layers.

Viscous Fingering in Miscible Floods
"Viscous fingers" are actually branch-like protuberances of a less viscous fluid penetrating into a more viscous fluid,
when the less viscous fluid pushes the more viscous fluid through some container (including through the pores of
rock in an oil reservoir). When some of these protuberances reach an outlet of the system, additional fluid flows
mostly through the fingers and does not displace the fluid between the fingers at a very rapid rate. This also amounts
to an inefficient displacement in the path of the miscible fluid, as well as inefficiency in reaching the relatively
stagnant areas of well patterns, so that both linear displacement efficiency and areal sweep efficiency are impaired.
The degree of impairment becomes worse as the mobility ratio increases.
Estimating Displacement and Sweep Efficiencies
It can be helpful, when first evaluating miscible flood potential, to have a simple graphical correlation for a five-spot
pattern, which relates the combined displacement efficiency and areal sweep efficiency (in the presence of viscous
fingering) at breakthrough and for subsequent throughput to the mobility ratio between displacing and displaced
fluids (Claridge, 1972). If the proposed flood is uneconomic in such a simple ideal case, then it will also be
uneconomic in a real-world, heterogeneous reservoir. If it looks good in the ideal case, then more complex
evaluations may be warranted.
This correlation was used, for example, in an 1984 NPC study as part of the procedure for calculating the future
potential recovery of crude oil by miscible methods; Esso Canada also used this correlation as part of the procedure
for designing their Judy Creek hydrocarbon miscible flood.
Gravity Stabilization
The density difference between an injected fluid and the oil it is supposed to displace can cause a vertical separation
of the fluids, which can be either helpful or harmful.
If the oil reservoir has a significant tilt, then a heavier injected fluid can be injected at the bottom and the oil can be
driven upward. Likewise, if the injected fluid is lighter, as is usually the case for miscible flooding fluids, the fluid
can be injected at the top of the reservoir and the oil can be driven downward.
If the reservoir is relatively flat, then the injected fluid must travel horizontally, and in this case it can either
gradually move to the top or to the bottom of the horizontal layer (depending on whether it is lighter or heavier than
the crude oil), leaving undisplaced oil behind. This unfavorable gravity segregation can be calculated by the same
method for miscible or immiscible fluids (e.g., Dietz (1953) or Hawthorne (1960) for single layers, or Stiles (1949
)or Dykstra-Parsons (1950) for multiple layers).
In the favorable case of gravity being used to improve displacement efficiency and areal sweep efficiency, there is a
maximum rate of displacement which must not be exceeded if the gravity (or relative buoyancy) effect is to prevent
viscous fingering. This was quantified by Hawthorne (1960) in the form of an equation for the critical velocity of
movement of the displacement front to avoid viscous fingering:
u
c
= q/A = k * A * g * sin /(
o
/k
ro
- u
g
/k
rg
) = [M/(M-1)j * k * k
ro
* A * sin /
o

(3.4)
The critical rate is thus [M/(M - 1)] times the drainage rate of oil when it flows under the influence of gravity alone
(no additional driving force). This equation applies to either miscible or immiscible fluids. However, in the case of
miscible fluids, Dumore (1964) pointed out that because of mixing of the two fluids by dispersion, the concave-
upward viscosity blending behavior could lead to less stability than if the pure fluid viscosities were maintained. He
therefore defined a "stable rate" which is less than the critical rate defined by the above equation. His initial solution
for the stable rate was:
u
st
= u
c
* (M -1)/(M * LnM) (3.5)
In deriving this equation, he used a viscosity blending relationship. based on the antilog of the sum of volume
fraction times the logarithm of viscosity This is not as good a relationship as one given by Peaceman and Rachford
(1962), where the volume fraction of each fluid is divided by the volume fraction of solvent plus a factor "a" times
the volume fraction of oil. When the latter was used by Dumore, the equation for the stable rate became
ust=a*u
c
* (M-1)/(M*LnM) (3.6)
and a good fit was obtained to Dumores experimental data using a value of "a" = 0.52.
It is recommended that Equations 3.4 and 3.6 be used in designing gravity-stabilized miscible floods. The value of
"a" for the given fluids can be obtained by fitting experimental viscosity blending data with Peaceman and
Rachfords blending relationship.
Gravity Tonguing
A basic reference on this subject is Dietz (1953). Dietz derived a method for calculating the development of a
gravity tongue in a homogeneous formation, either of heavier fluid than the in-place crude oil along the bottom of
the formation, or of a lighter fluid along the top of the formation. The derivation is given and discussed in some
detail in the section titled "Vertical Sweep Efficiency Models." His method is useful in considering the behavior of
such gravity tongues in line drive well patterns.



Effects of Layer Heterogeneity in Miscible Flooding
To make estimates of the effect of layer heterogeneity on sweep efficiency, we need to express the heterogeneity in
quantitative terms. This can be done by use of a Lorenz diagram, discussed by Pirson (1958), or by use of the
method of Dykstra and Parsons (1950). The varieties of reservoir heterogeneity are discussed below.
Reservoir Heterogeneity
Linear displacement efficiency, pattern sweep behavior, gravity tonguing, viscous fingering and dispersion all act
within each layer of an oil reservoir to limit the oil recovery obtainable from that layer. However, all of these
combined would not make the oil recovery in a given layer fall much below 70% of the movable oil saturation at a
throughput of one or two pore volumes. But overall recovery is usually considerably less than that-more like 50% of
the movable oil at best, and often much lower. There is evidently still another major factor limiting oil recovery.
This factor is reservoir heterogeneity.
While heterogeneity can (and usually does) manifest itself in an areal variation of permeability within a layer, this
areal variation is usually less significant than the permeability variations among the layers. This is a natural
consequence of the sedimentary process. Over a typical interwell distance of 300 meters (a city block length), for
example, the same sedimentary process is likely to be taking place in most of the sedimentary environments.
(Exceptions are high-energy deposits like braided streams, and channel cuts such as those that occur in deltas.) This
does not preclude the termination of some layers over the interwell distance, or some of the barriers between layers.
The pinch-out of permeable layers or of barrier-forming shale layers is a fairly common occurrence, somewhat more
frequent in lagoonal carbonates and in alluvial sandstones than in barrier sandbar sandstones or pinnacle reef
carbonate reservoirs. Channel cuts of more permeable sandstone through less permeable delta sands and multiple
slip faults are common in delta environments.
In the case of layers that terminate between wells, the question of whether fluids injected into these layers can reach
other wells occurs. In some cases, the layer in question extends to some producer (or producers) but not to other
producers; this limits the areal sweep mostly to the direction of the connected producers. However, while the
permeability of intervening shale layers may be low, it is not zero. The vertical distance between a layer that pinches
out and the ones above and below it that do not pinch out is small, and the horizontal area (at right angles to the
vertical direction of flow) is much larger than the vertical area perpendicular to the direction of horizontal flow.
Typically, the area ratio is 100 to 1,000, and the vertical distance is 0.01 to 0.001 times the horizontal interwell
distance. This combination of A/L allows the same flow rate in the vertical direction as in the horizontal direction
with a vertical permeability only 0.0001 to 0.000001 times as great as the horizontal permeability, when the pressure
gradient is the same. When the adjacent layers are both connected from well to well, the inter-layer pressure
gradients are small-typically much smaller than the horizontal pressure gradient, except for possible capillary
pressure differences. Hence, interlayer crossflow is moderate, or may be prevented entirely by high capillary
pressure difference between the permeable layers and the shales or silts between the permeable layers.
If, on the other hand, the flow in a given layer is stopped due to pinch-out, the pressure in that layer builds up to
approach the injection well pressure. This results in a large pressure difference between that layer and the adjacent
layers above and below. Even with very low-permeability shale barriers, if the capillary pressure difference between
the shale layer (usually at 100% water saturation) and the oil-bearing sand layers is not too great to be overcome by
the pressure difference mentioned, leakage of oil through the shale layers from the pinched-out layer to the adjacent
connecting layers can occur at a significant rate. A pinched-out sand layer of medium to high permeability may
therefore be equivalent to a connected layer of considerably lower permeability.
The Lorenz Diagram
If the k
h
values for individual layers are arranged in descending order and divided by the sum of the k
h
values, a
scale is obtained from 0 to 1 in which the layers comprise the parts. If the corresponding h values of the layers are
divided by the sum of h, a scale from 0 to 1 is again obtained.
If the cumulative k
h
fractional values are plotted against the cumulative h fractional values, there is a curve
obtained, starting from 0 and ending at 1, which is convex-upwards and occupies the region above and to the left of
the diagonal across the square from 0 to 1. If the area between the curve and the diagonal is divided by one-half of
the area of the square, the result is called the Lorenz coefficient of heterogeneity, and the plot is called a Lorenz
diagram (see discussion in Pirson, (1958)). Such a diagram may be used for reservoir modeling purposes, as
explained later.
The Dykstra -Parsons Coefficient
The distribution of permeability is generally fairly well represented by simply plotting the logarithm of the
permeabilities (in descending order) versus the cumulative fraction of h on a cumulative probability scale. In most
cases this gives an approximately straight line, and the permeability is then said to be log-normally distributed. In
cases where different facies occur in the same reservoir, having permeability ranges which overlap only to a limited
extent, then the plot mentioned may show two or more straight-line segments. This situation will also be exhibited in
a plot of log permeability versus ; this plot will contain line segments of different slopes. However, because of the
large scatter typical of this latter type of plot, it is more difficult to detect the presence of two or more different
relationships, whereas it is relatively easy to see on the ordered log permeability versus cumulative h plot. The
latter type of plot is used to calculate an index of heterogeneity called the Dykstra-Parsons (1950) coefficient of
permeability variation, v.
On such a plot, the value of the dependent variable-permeability, in this case-is read at the midpoint (50% of
cumulative h) and at 84.1%. Then the Dykstra-Parsons coefficient v is defined as:
= [k50% - k84.1%]/k50% (3.7)
On the cumulative probability scale, the values of 15.9% and 84.1% represent a minus one and plus one standard
deviation from the mean value at 50%. This latter range contains 68.3% of the area under the standard (or Gaussian)
error curve. The latter curve (for this case) is the percentage of the reservoir volume having a given logarithm of
permeability versus that value of the logarithm of permeability; the range of the logarithm is from minus infinity to
plus infinity as the permeability varies from zero to plus infinity. In the Dykstra-Parsons plot, we are plotting the
cumulative percentage versus the logarithm of permeability rather than percentage at a given log permeability, but
the same principles apply. It would have been more logical for Dykstra and Parsons to use the logarithm of
permeability instead of permeability itself in their index, . To make the argument of the logarithm dimensionless,
the permeabilities may be divided by the value at 50% (km), and the logarithm of the ratio may then be plotted
versus the cumulative h on the probability scale. The 50% point value of the logarithm would then always be at a
value of zero (logarithm of 1.0), and the logarithm at the 84.1% point would have a negative value. The standard
deviation a(log k/km) would then be:
(log k/km) = (3.8)
The range of would be from zero (no variation of permeability; a single value of permeability for the entire
reservoir) to infinity (permeability ranging from zero to infinity). However, the way that Dykstra and Parsons
defined their coefficient of permeability variation causes the range of to run from zero to one as the permeability
range varies from zero to infinity. In some ways, this is more convenient, but it must then be remembered that high
values of -those approaching 1.0-represent a very great degree of heterogeneity.
It may be helpful to quote representative values of . The writer has not seen a case of reservoir whole-core
permeability/porosity data where the value of the Dykstra-Parsons is less than 0.6. This means that the
permeability at the 84.1% point is 0.4 times the permeability at the 50% point. At the 95% point, for this slope, the
value of the permeability would be about 0.22 times the midpoint permeability. Correspondingly, at the 5% point the
permeability would be equal to the midpoint permeability divided by 0.22, or about 4.53 times the k50% value.
Thus, the range of permeability for 95% - 5% or 90% of the rock would be from the lowest value (at 95%) to the
highest value (at 5%) with a ratio of (1/0.22)2 about 20 to 1. If the lowest value were 10 md, the highest would be
200 md, for 90% of the rock. For 99% of the rock, the ratio of highest permeability to lowest permeability is about
110 to 1 (still for a Dykstra-Parsons of 0.6). So if our data are sufficiently detailed to see the first and last 1% of
the h, we would expect a permeability range on the order of 100 to 1. In my own experience, I have seen a value
of the Dykstra-Parsons as high as 0.93. This means that the value of the permeability at the 84.1% point is only
0.07 times the value at the 50% point. The permeability ratio for 99% of the total rock is then 800,000 to 1, or from
0.01 md to 8 darcys. This is, in fact, about as great as the reliable range of measurement of permeability for
equipment able to measure both the highest and lowest values of this range.
The more common range of the Dykstra-Parsons is from 0.7 to 0.8. At a typical value of 0.74, for example, the
permeability at 84.1% is 0.26 times the value at 50%. The range for 99% of the rock is then about 1,000 to 1. If the
cut-off value of permeability for rock considered to be part of the reservoir is 1.0 md, then some thin layer will have
a permeability as high as 1.0 darcy.
Cross-Sectional Flow Calculations
Most direct studies of this matter, before the advent of computer simulation, were based on simple cross-sectional
models of the formation-that is, two-dimensional vertical models. Layers were arbitrarily arranged in order of
descending permeability. Then, disregarding crossflow of fluids between layers, flow of an invading fluid through
these layers was calculated and the outflow from the layers was added to obtain oil recovery versus the amount of
displacing fluid injected. While the results of these calculations illustrate the significant effect of heterogeneity on
oil recovery at a given throughput of displacing fluid, they are not an adequate substitute for computer simulation of
the given process. Prior to the widespread use of computer simulation it was recognized that these methods do not
give good matches to actual waterflood performance (see Craig, 1971, Chapter 8); this served as one of the
incentives for the adoption of computer simulation.
Combatting Heterogeneity Effects
Major efforts by reservoir engineers to combat the adverse effects of heterogeneity on oil recovery have led to the
use of water-soluble polymers (as a substitute for plain waterflooding), water-alternating-gas (WAG) injection
procedures for miscible projects, and the use of small amounts of surfactants in water mixed with miscible or
immiscible gas injectants, so that a foam is generated in the rock pores. Each method yields pronounced benefits. It
is not always as effective as desired due to gravity segregation of the injected water and gas. The foam method is
intended to offset this tendency by holding the water and gas in intimate contact as bubbles and their intervening
water films-surface tension forces essentially prevent the gravity segregation.


Immiscible and Miscible Flow Processes Accompanying Miscible Flooding
One-dimensional oil recovery calculations can be made for both immiscible and miscible reservoir displacement
processes. These are helpful in understanding flow behavior.
For those who are not fully familiar with the principles involved, this Section explains the subject of immiscible
simultaneous flow of two phases over a cross section in which the volume fractions (saturations) of the phases in the
pore space are uniform. Assuming that this uniformity exists, we can treat the mixed flow as one-dimensional. With
this limitation, we can obtain a direct mathematical solution for various reservoir processes.
Fractional Flow Calculations
In simple one-dimensional treatments, we generally assume that the properties of the porous rock system are
uniform in the single direction, x. Diffusion and the more general process of dispersion are not considered. The
process is time-dependent, giving us two independent variables, x and t. It is therefore necessary to describe the
given process by partial differential equations, based on the material balance in a differential interval, x (of
submicroscopic length), in the length dimension and in a differential interval, t, in the time dimension. Such an
equation is often called the continuity equation.
Various authors have wrestled with the fact that permeable media such as natural rocks contain immobile solid
particles and moving, or movable, fluids within the pore space between the solid particles. In microscopic terms,
therefore, the pore space containing the fluids whose movements we want to describe with a "continuity equation"
does not, strictly speaking, constitute a "continuum"; therefore, a continuity equation based on a smaller-than-
microscopic differential slice is not, strictly speaking, applicable. The conclusion of each of these authors has been
that a continuity equation is nevertheless applicable if we want to describe the overall process, even though it does
not describe all of the motions on the microscopic scale of individual pores. The overall process contains the
assumption of volume averaging over a multi-pore scale. These and other assumptions of most one-dimensional
treatments are listed by Pope (1980).
A continuity equation must be written for n - 1 components, where n is the number of components; it is assumed that
a total material balance holds over all n components and so the equation for the nth component is redundant. In
addition to the differential material balance or continuity equation, it is essentially always necessary to specify
relationships giving physical and/or chemical properties of the materials undergoing the movements or changes
described by the continuity equation. These equations, specifying properties and relationships between properties,
are known as constitutive equations. Constitutive equations are used to describe the flow process in general, as well
as physicochemical equilibria or rate processes.
The simplest case is two-phase flow in a one-dimensional (homogeneous) porous and permeable medium. The
phases may be gas and water, gas and oil, or water and oil. The constitutive equations used are Darcys Law applied
to each phase. A general integrated solution can be obtained which, for translation to an individual case, requires
that an empirical coefficient in each of the two Darcys Law equations (namely, the relative permeability) be
specified as a function of the volume fraction (saturation) of one of the two components. An analytic solution
therefore requires that an analytic (algebraic) equation for each of these relative permeabilities be provided.
The two Darcys Law equations are combined into one equation for substitution into the general integrated solution
of the continuity equation. This combination of the two Darcys Law equations is called the fractional flow equation,
since it gives the fraction of one phase flowing relative to total flow as a function of the relative permeabilities and
of the phase viscosities.
If the relative permeabilities are given as analytic functions of the saturation of one of the phases, then the fractional
flow equation expresses the flowing fraction of a given phase in terms of the saturation of that phase and of phase
viscosities (which are assumed to be constant).
The basic importance of the constitutive relationship in describing various processes is exemplified in a paper by
Pope (1980): The Application of Fractional Flow Theory to Enhanced Oil Recovery. His paper describes a variety of
cases, of which the first and most basic is the simple two-phase linear flow case we have described above; but he
then proceeds to describe polymer and chemical flooding, carbonated water flooding and miscible flooding by this
method, showing its wide applicability.
The mathematical treatment of two-phase flow in permeable media by this general method was first described in the
petroleum literature by Buckley and Leverett (1942). Their treatment, assuming that the two-phase flow process is
that of oil and water, is given in "Miscible Processess: Oil Recovery Calculations." Welge (1952) showed Buckley
and Leveretts method to be advantageously portrayed in graphical form, on a plot of the fractional flow of
displacing phase versus the saturation of that phase. From these one-dimensional calculations, we can readily obtain
a plot of fractional oil recovery versus pore volumes of throughput, as well as a plot of saturation of the displacing
phase versus the fractional distance from inlet to outlet of the system. Besides water and oil, the case of polymer
flooding and of hot water flooding in a one-dimensional system are shown by example calculations. These results
apply to core floods in the laboratory, and can be extended to calculate the behavior in a single-layer homogeneous
well pattern by Deppes method, as discussed next.
Claridges Modification of Deppes Method
If Claridges modification of Deppes method is used for immiscible flow calculations (by Buckley-Leverett and
Welge procedures), then the injection and production circle areas are different and are determined by the areal
sweep efficiency. This sweep efficiency (E
a
) can be estimated by means of the equations Claridge has given for
different patterns, or by any other analytic or graphical correlation which may be applicable.
A series of fluids may be represented in the injection and production circles as a series of annular banks in the
injection circle and a series of sectors in the production circle. For a favorable mobility ratio waterflood, it is often
the case that the saturation and fractional flow change at the water flood front is quite high, so that the flood may be
approximated by assuming a sharp change from 100% oil ahead of the flood front to a small, fixed percentage after
the flood front passes. This leads to some small errors, but the results are almost correct and give a rather clear
picture of the behavior of the flood in a well pattern.
Immiscible Flow in Primary Miscible Flooding
Often, little attention is paid to the immiscible flow processes preceding, accompanying or following the miscible
part of the overall process. These immiscible flow aspects are discussed below, for a case in which the miscible
flood takes place as a primary process, or at least prior to a waterflood, and the reservoir pressure is sufficient to
attain miscibility. The water is assumed to be immobile, as connate water.
Gas/Oil Immiscible Flow
Lets look first at the behavior of immiscible gas and oil as it occurs in several displacement processes.
Vaporizing Multiple-Contact Miscible Displacement
Some immiscible flow precedes the attainment of miscibility in multiple-contact miscible processes. In a vaporizing
gas drive, the immiscible region exists close to the injection wells. The immiscible displacement of crude oil by
injected gas is inefficient because of the high mobility of gas relative to oil. In fact, the attainment of miscibility
depends on the relative flow of gas past the oil, for if the displacement were piston-like, there would be insufficient
contact to allow the vaporization of intermediate hydrocarbons required to make the gas miscible with the crude oil
farther out in the reservoir.
When crude oil near the well has been stripped of its intermediate hydrocarbons, then any further injected gas will
have to travel (immiscibly) farther out into the reservoir to find crude oil from which to vaporize the intermediates.
Since there will be viscous fingering as well as dispersion of the enriched gas into the crude oil with which it has
become miscible, the process could not survive these fluid-mechanical effects if we had only a first, limited
increment of injected gas enriched in intermediates by exchange with crude oil near the well. Fortunately, as
indicated, the gas will flow immiscibly past the crude oil from which the intermediates have been removed and will,
at some further distance from the well, encounter crude oil that is still rich in intermediates. The early thinking about
this type of miscible process was that piston-like displacement of the crude oil would occur starting at the radius
from the well at which miscibility first occurred. Hence, the region in which crude oil would remain from which
intermediates had been stripped would be only within a short radial distance from the wellbore. However, when the
phenomena of dispersion, viscous fingering and gravity layover or tonguing, as well as continued interphase mass
transfer are realized to be part of the real process, it is obvious that there will exist a growing volume within which
crude oil containing relatively little intermediate hydrocarbons will be moving slowly in simultaneous flow with a
faster gas phase which is immiscible with it.
The volume versus time can be estimated by material balance, considering
- the rate of gas injection
- the amount of hydrocarbons that must be vaporized into it to achieve miscibility
- the volume of crude oil required to provide that amount of intermediate hydrocarbons, based on the crude
oil analysis
The balance can be expressed as a gas rate (e.g., in moles per day), times the moles of intermediates which must be
added per mole of injected gas, divided by the moles of such intermediates present in the crude oil per unit of
reservoir volume (e.g., per cubic foot). The result is reservoir volume of immiscible flow versus time (for the units
assumed, in cubic feet per day). Of course, because of gravity tonguing and fingering, the volume will be of
irregular shape, but the calculation gives a reasonable estimate of the total volume denuded of intermediates-versus
time. Within this volume, the gas and oil flow can be described by fractional flow relationships. These relationships
cannot be made exact, because of the gradual changes in oil and gas phase viscosities with distance and time, but an
approximate calculation with average properties will suffice to estimate the displacement of crude oil out of this
growing volume versus time, due to the immiscible gas drive. Of course, once a sufficient bank of intermediates-
enriched gas has been generated to provide an effective barrier between the gas following and the oil bank ahead of
the enriched gas, then there will be no further growth of the immiscible zone.
Enriched (or Condensing) Miscible Gas Drive
In a multiple-contact miscible process, intermediates are transferred out of the injected gas and into the crude oil in
place around the injection well, until the oil is sufficiently enriched in intermediates to be miscible with the next
increment of injected gas. Accordingly, the gas, which has lost much of its content of intermediates, may be
immiscible with the crude oil in place as it travels farther away from the well. For this to be the case, the crude oil
must be saturated with respect to "dry gas" in solution (i.e., it must be at its bubble-point pressure). There is no
chance of gas that has lost its intermediates being able to achieve miscibility as it goes further, because both it and
the crude oil which it encounters lack sufficient intermediate hydrocarbons in either phase to achieve miscibility.
Consequently, it is possible for this gas, which has been stripped of its intermediates, to travel all the way to
producing wells as a separate, immiscible gas phase. The appearance of such a stripped gas phase at producers does
not, therefore, indicate that miscibility has not occurred.
It is possible that the stripped gas may not proceed all the way to producers. If the crude oil is undersaturated with
dissolved gas at the pressure existing over most of the reservoir volume, then stripped gas would also dissolve in the
under-saturated crude oil as the gas enters the reservoir. This behavior can be calculated if the appropriate data are
available (Rs versus pressure). Of course, this means that not only intermediates but also the "dry gas" in the
injected gas stream will transfer to the crude oil near the injection well, until the diluted crude is also saturated with
"dry gas." This total dissolution will proceed as a front moving away from the well, and the diluted crude oil will be
displaced miscibly by further enriched gas (with dispersion, viscous fingering and gravity layover).
CO2 Flooding
Since this is considered a vaporizing gas drive, the description of the process under that heading applies. Crude oil
that has been stripped of its intermediate hydrocarbons (in this case, extending up into the diesel fuel boiling range)
will be left behind, and because it is much more viscous than the initial crude oil, it will move much more slowly
than the miscible displacement front. However, this stripped crude oil is being generated at the displacement front
(which is irregular in shape because of viscous fingering and gravity layover), and so it extends all the way back
from the miscible displacement front to the injection wells. Slow-moving heavy oil is further stripped of higher
hydrocarbons by the CO
2
immiscibly flowing past, so its volume (saturation) decreases as this stripping proceeds.
If the heavy oils saturation is already below (or later falls to) the residual saturation to a gas drive, it does not move,
but rather continues to be stripped of its lowest molecular weight hydrocarbons. Hence, the existence of a dew phase
at the CO
2
displacement front results in a liquid heavy hydrocarbon phase which either flows simultaneously and
immiscibly with the CO
2
, or remains immobile and immiscible with the CO
2
(though undergoing mass-transfer with
it).
Consequently, we normally find immiscible gas/oil flow behind the CO
2
flood front. The amount is difficult to
calculate by the fractional flow method, because the oil phase has strongly varying properties (viscosity) as it
continues to be extracted by CO
2
flowing past it. Nevertheless, it may again be helpful to make approximate
estimates of the rate of flow of the immiscible oil phase toward the producers. In one field case, evidence was found
that the well effluent consisted of a CO
2
-rich stream containing most of the relatively light crude oil, combined with
a small stream of a heavy oil consisting of the rest of the crude oil; the remaining oil found in cores taken well
behind the CO
2
flood front was a heavy oil that had been even further stripped (C
28+
).
Gas/Oil Immiscible Flow Due to Gas Penetrating an LPG Slug
The immiscible part of the displacement occurs when the dry gas following the LPG slug penetrates the tips of the
LPG fingers. This process should be designed with a sufficiently large LFG slug so that this penetration does not
occur until the LPG fingers themselves break through at the producers. My work (see Claridge, 1978 in the
References) includes a method for calculating the LPG slug size required as a function of the mobility ratio in front
and in back of the LPG slug.
Water/Gas and Water/Oil Immiscible Flow
Now we will focus on the behavior of immiscible combinations of water with either gas or oil as they occur in
several displacement processes.
Vertical Segregation During Immiscible Flow of Water Phase and a Simultaneously Injected Miscible Solvent
Stone (1982) has presented an important method for estimating the vertical sweep efficiency of the water phase and
the miscible gas phase during the simultaneous injection approximation of WAG operation. In contrast to the Dietz
and Hawthorne, Stiles and Dykstra-Parsons analyses, in which flow is assumed to take place only in the bedding
plane, Stone explicitly considers the vertical downward movement of water and the upward movement of miscible
gas during each phases travel in the direction of bedding-plane displacement. Details of Stones method are given
in the section titled "Oil Recovery Calculations."
Water/Gas Immiscible Flow During WAG Operation
By WAG, we refer to the process of alternating water and gas injection. In this version of miscible flooding, which
is most commonly practiced in hydrocarbon-enriched gas floods and CO
2
floods, the miscible process in the non-
aqueous phase is accompanied by a continuous immiscible process. There may be a trapped non-aqueous saturation,
which is regularly re-mobilized (at least in part), and then re-trapped by the water travelling through the formation.
The non-aqueous phase often travels faster than the aqueous phase. Both oil and solvent may become partially
trapped during this process, so that the flowing fractions will depend on the mobile parts of their respective
saturations. Because the solvent does not mix well with the oil (due to viscous fingering and gravity tonguing), there
will be some places in which water contacts nearly pure solvent, others in which it contacts nearly pure crude oil,
and still others in which it contacts mixtures of the solvent and crude oil.
Under these circumstances, it is not proper to treat the immiscible flow as if the saturations were constant over any
given cross section, as is required by the fractional flow method of calculation. Instead, we must resort to computer
simulation programs. Even simulation, however, may not take into account all of the known physical phenomena (as
we will discuss later). It is therefore helpful to have recourse to physical model studies, in which Nature itself
ensures that all of the physical phenomena which should occur do, in fact, take place. While it is difficult to scale
physical models to accurately match all of the ratios of forces and rates that occur on a reservoir scale, approximate
matches can provide helpful results (see Jackson et al., 1985 and Claridge et al., 1988).
Immiscible Flow During the Final Waterflood
The size of a miscible solvent slug is limited by economics. In the case where a slug has been injected continuously
(rather than by the WAG process discussed above), it is necessary to push this slug, and the oil that is still ahead of
it, to the production wells by a final waterflood. Of course, the final waterflood is immiscible with both the solvent
and the oil.
Since the water will probably first encounter the solvent slug (e.g., enriched gas or CO
2
) and trap it just as it would
an oil phase that is, in the same residual saturation-the back side of the slug will be consumed. It is further likely that
the water will come into contact with the oil before the displacement is complete. As the water displaces oil, it will
trap the normal residual waterflood saturation of oil. Hence, we may expect to find most of the trapped solvent in
those parts of the well patterns nearest to the injection wells; in the rest of the pattern, oil will be trapped in the
normal waterflood residual saturation.
Since the economic evaluation of the solvent flood depends on the amount of waterflood-trapped solvent that
replaces the crude oil which would be trapped by waterflooding only, it is essential that the solvent slug be of
sufficient size to provide for this residual saturation over most of the well pattern. This fact alone sets a minimum
solvent slug size of about half of the pore space that we expect to flood with the solvent. Because of vertical and
areal sweep efficiencies, this is less than the residual saturation to waterflood.
Because of viscous fingering and dispersion of solvent through the crude oil, in part of the reservoir the waterflood-
trapped fluid will be a mixture of solvent and oil. However, the trapped saturations will be about the same as we
have just indicated.
As a round figure, if the combined sweep efficiency is 0.5 and the waterflood residual saturation is 0.4, then the
minimum solvent slug size would be 0.2 pore volume. Of course, because of dispersion, gravity layover and viscous
fingering, some solvent will be produced early in the process, before the average flood front arrives and before the
majority of the producible crude oil is produced. It is necessary to provide extra solvent to allow for this early
production; therefore, the slug size should be larger than the calculated minimum. Even if very poor sweep
efficiency is estimated (e.g., 0.25), the slug size should still be approximately double the product of the estimated
sweep efficiency and the waterflood residual saturation. Thus, the solvent slug size should seldom be less than 0.2
pore volume, with sizes up to 0.4-0.6 pore volume apt to be necessary for obtaining the technical optimum oil
recovery we expect from the process. Economic considerations may still require a lower slug size than the technical
optimum estimated in this way, but this estimate should be the starting point in calculating the best flood design.
This immiscible part of the process can be calculated for homogeneous layers by methods presented here, but
computer simulation is the preferred method of calculating combined miscible/immiscible oil recovery processes.
Immiscible flow During Tertiary Miscible Flooding
A tertiary miscible flood is one taking place after a secondary flood (a waterflood, in most cases) has been
performed. A high water saturation at the start of miscible flooding ensures that immiscible flow will occur
throughout the flood, regardless of the miscible process type.
Gas/Oil Immiscible flow Preceding Miscibility
In the non-aqueous phase, the multiple-contact process leading to attainment of miscibility will proceed as described
in the section titled "Miscible Processess: Equilibrium Phase Behavior," but accomplishing each step of the process
in a vaporizing gas drive will involve a larger reservoir volume, because less crude oil is available per unit volume
to supply intermediate hydrocarbons to the solvent (be it dry gas or CO
2
). In the condensing gas drive process, less
intermediates are required per unit reservoir volume to enrich the crude oil to the miscible composition. In either
case, the injected gas must immiscibly displace sufficient water to re-mobilize the oil trapped by the waterflood.
With hydrocarbon solvents, the required mass-transfer of intermediate hydrocarbons in either direction between gas
and oil cannot take place through the water films isolating the water-flood-trapped oil; this trapped oil must be re-
mobilized before the multiple-contact process can take place. The wetting state (water-wet, intermediate wetting, or
oil-wet) is pertinent here, with oil or intermediate wetting being preferred.
Ahead of the miscible flood front, the prior waterflood will continue. The water cut may be expected to increase
steadily, as it did prior to the start of miscible solvent injection, until the oil bank and accompanying solvent break
through at producing wells.
Slug/Water Immiscible Flow
While the prior waterflood is continuing ahead of the miscible flood front, the immiscible slug/water simultaneous
flow begins to re-mobilize trapped crude oil, thereby creating a solvent/oil mixture. The viscosity of this mixture is
greater than that of the pure solvent, making the displacement of water more efficient.
At the front, the non-aqueous phase composition approaches that of solvent-free crude oil. Usually, however, a
mixture persists at the front; a pure oil bank is seldom observed, and if so, it is small. One reason for this is viscous
fingering; as soon as a pure oil bank appears, the solvent behind it fingers it. The existence of an oil bank, then,
depends on the velocity of the solvent fingers versus the velocity of the front. It is commonly observed that solvent
and oil appear together at breakthrough of non-aqueous phase; these two velocities must therefore be nearly equal.
Thus, the solvent-water front is actually a front of solvent mixed with oil displacing water.
If the oil remobilization curve is available, along with the oil/water relative permeability curves, then a stepwise,
trial-and-error procedure using a fractional flow diagram will give an approximate value for the mixture of solvent
and oil at the flood front. For this purpose we may use the re-mobilization curve of Raimondi and Torcaso, as
modified by Todd (1982):
S
ot
= S
orw
/[1.0 + * (k
ro
/k
rw
)] (4.1)
where S
ot
is the trapped oil saturation at the water saturation corresponding to that at which kro and k are both
evaluated, and is an empirical parameter which varies with the rock wetting state and which enables an approximate
fit to the measured curve of trapped oil saturation versus water saturation. Figure 1 shows a plot of reported data
points for strongly water-wet sandstone samples (with probably close to 1.0),


Figure 1


and Figure 2 shows how the curves change with the value of a.


Figure 2


The trial-and-error procedure referred to above involves starting with the fractional flow curve for a pure solvent
displacing water. The tangent to the curve starting at the initial state in the reservoir-which near the injection well
will be at a water saturation near to 1 - S
orw
(and f
w
1) is at a tangent point having a water saturation at which the
trapped oil saturation can be calculated. S
orw
- S
ot
is then the pore volume fraction of oil which would be re-
mobilized by pure solvent injection. The reciprocal of the slope of the tangent is the fraction of a pore volume of
solvent injected to reach the front, based on the pore volume behind the front. At this front, then, we have a certain
pore volume fraction of solvent and a corresponding pore volume fraction of crude oil. The viscosity of this blend
can be calculated (e.g., by the quarter-power mobility blending rule). This leads to a modified fractional flow curve
based on this calculated viscosity, rather than the viscosity of the pure solvent. This curve will lie between the
fractional flow curve for pure crude oil and pure solvent, each displacing water. The tangent from f
w
= 1, S
w
= 1 -
S
orw
to this fractional flow curve will give a water saturation at the tangent point which is lower than that at the
previous tangent point. The calculated trapped oil saturation will be lower, and so the pore volume fraction of re-
mobilized oil will be greater.
As this procedure is repeated, however, we soon find that the proportion of oil and of solvent at the flood front does
not increase until the crude oil/water fractional flow curve is attained (this would mean that a pure oil bank has been
achieved); rather, it approaches a certain mixture which depends on the relative permeability curves and on the value
of a. It is not hard to find this mixture by trial and error. The corresponding fractional flow curve is then one which
can be used to describe the displacement at the solvent flood front.
Stones method may be used to determine the distance at which injected water and solvent, during WAG operation
in the tertiary situation, become completely segregated. Beyond that point, the preceding discussion applies only to
the top layer, where segregated solvent encounters mobile water phase and waterflood trapped oil. In the zone
beneath this segregated solvent zone, only water is flowing, and no oil is being mobilized.
Fractional Flow Calculations: Kovals Method
For miscible fluids, the same type of fractional flow calculations may be made by adopting Kovals assumption
concerning the relative permeabilities of solvent and oil, which states that the relative permeability of solvent is
equal to its saturation fraction, and that of the oil is the oil saturation. This amounts to saying that the permeability of
the single non-aqueous phase is divided among the miscible components in proportion to their volume
concentrations in that single liquid phase. If the flood is secondary, the initial oil saturation is (1 - S
wc
) and the
water is immobile.
When there is some immiscible solvent gas/crude oil behavior (as discussed later), that behavior may be calculated
by the Buckley-Leverett method, using gas/oil relative permeability curves. For the present case, the behavior may
be considered to be that of a first-contact miscible flood. In addition to the simple assumption of relative
permeabilities equal to saturations of solvent and oil, Koval uses an effective mobility ratio (M
e
).
The fractional flow equation is then:
f
s
=1/[1 + (1 - S
s
)/(S
s
M
e
)] (4.2)
df
s
/dS
s
= M
e
/[S
s
(M
e
-1) + 1]
2
(4.3)
The Buckley-Leverett solution then reads
(x/T)S
s
= (df
s
/dS
s
)S
s
(4.4)
If the fractional distance X is set equal to 1.0 (i.e., at the outlet end of the system), then T = N
i
, which is the pore
volume of solvent input required to make a given solvent saturation, S
s
, appear at the exit. This relationship applies
only to the mixed-flow region, so when S = 0, it marks the beginning of mixed flow after pure crude oil has flowed
out up to this breakthrough point of solvent. The solvent throughput necessary to yield solvent saturation S at the
outlet is thus
N
i
= 1/(df
s
/dS
s
)S
s
= [S
s
(M
e
- 1) + 1] 2/M
e
(4.5)
So when S = 0, N
i
= l/M
e
, and when S = 1.0 (sweep-out of oil phase), N
i
= Me. During the mixed flow period, we
can write an equation for the oil recovery Np (as a pore volume fraction), remembering that up to break-through of
solvent the oil produced is just equal to the solvent injected = 1/M
e

N
p
= 1/M
e+
(4.6)
In order to perform the integration, (1 - f
s
) must be expressed in terms of N
i
, so Equation 4.5 is solved for S
s
, and the
result is substituted into Equation 4.2; we then solve for 1 - f
s
, then substitute the result into Equation 4.6 and
perform the integration between the limits of 1/M
e
and N
i

S
s
= [(MN
i
)
1/2
-1]/(M - 1) (4.7)
(1 - f
s
) = [(M/N
i
)
1/2
- 1]/(M
e
- 1) (4.8)
N
p
=1/M
e
+ [2 (M
e
N
i
)
1/2
- N
i
- 2 + 1/M
e
]/(M
e
1), or
N
p
= (2(M
e
N
i
)1/2 - N
i
- l]/(M
e
- 1) (4.9)
This is the same recovery equation as those derived by Dietz for a gravity tongue and by van Meurs and van der
Poel for viscous tonguing in a water-flood. The like derivations are a consequence of the same basic assumption:
that the different fluids flow in parallel but with different velocities. No provision has been made in these derivations
for the segregation of fluids due to vertical flow in opposite directions.
Extension to Well Patterns: Deppes Method
It is a straightforward extension of Kovals version of the Buckley-Leverett method to apply it to well patterns by
Deppes method. The fractional flow calculations are applied to the injection and production circles as one-
dimensional radial systems. The pore volume fractions are represented by fractions of the circle areas, times an
assumed constant thickness. This extension, however, is limited; it is not readily possible to carry out an analytical
calculation for a tilted reservoir by this method, since the gravity term (G) in the numerator of the complete
fractional flow equation contains the factor sina, and the dip angle (a) varies with the angular direction in the
injection and production circles. In some cases this limitation may not be serious, but in others it is. Where it is, we
must have recourse to a computer simulation program, such as a "black oil" simulator to simulate a waterflood, or a
suitable miscible flooding simulator for miscible flooding.
Computer Simulation
Real, heterogeneous reservoirs require three-dimensional solutions, and so it is impossible to learn about the
interactions of processes with reservoir heterogeneities by using one-dimensional solutions. Because of our inability
to obtain exact, analytic solutions for three-dimensional cases, particularly in the presence of heterogeneities, it is
necessary to use numerical methods-incorporated into computer programs which we call reservoir simulators-to
solve the full-scale reservoir problems. Nearly every such simulator is capable of containing in the input data a
detailed description of a given reservoir, with heterogeneities specified in as much detail as allowed by program
dimensions and the memory capability of the computer on which the program is to be run. However, each given
simulator is capable of simulating only certain reservoir processes; there is no one simulator capable of simulating
any and all processes.
The most common simulator is the type known as a "black oil" simulator. This means that the crude oil (which is
typically black in color) is treated as a single component, rather than as a mixture of components, which it really is.
Water and natural gas are also treated as single components in such a program, although each may in reality contain
a variety of chemical species. Thus, this type of simulator is a three-component, three-dimensional simulator.
Despite the severe limitation of the number of components, it is able to accurately describe primary recovery by
solution gas drive, gravity drainage, or by a natural water drive, and also the simple secondary recovery process of
waterflooding. It does not properly simulate the secondary process of gas drive, nor is it able to simulate any of the
well-known enhanced oil recovery processes of the chemical, miscible or thermal types. Specialized simulators have
been constructed to handle certain types of EOR processes; these are generally proprietary programs.

Estimating Oil Recovery in Miscible Flooding Processes
The analytical methods and correlations that have been developed to describe miscible flood behavior all have built-
in limitations with respect to ideal flow pattern shapes, or of two-to-three components, or of one-dimensionality or
(at best) two-dimensionality, and typically no more than two flowing phases. Unfortunately, in the real world there
are always three dimensions, non-ideal flow patterns, many components, often three or four phases (not counting the
rock) and reservoir heterogeneities of many kinds.
In deciding which oil reservoirs to consider in more detail for various possible recovery processes, the engineer must
look to the current process taking place in each reservoir. The first step is usually to apply a screening procedure for
EOR processes, based on reservoir and oil properties. While a series of such screening guides have been published,
the best one is probably that of Taber and Martin (1983). They present the criteria for miscible floods given in Table
1 .


Table 1


A screening guide such as this can encourage an engineer when a given reservoir and crude oil meet the stated
criteria. We often find, however, that the fact that a given reservoir and crude oil do not meet the stated criteria does
not necessarily mean that the process cannot be successfully carried out.
Conversely, if a given reservoir and crude oil meet the criteria, it is not necessarily true that a miscible flood would
be economically successful, even though it might be technically feasible. It is thus helpful to make some recovery
estimates by simple methods, and to evaluate the results with the economic data appropriate to the particular field.
Several methods of making estimates of increasing sophistication are discussed below.
Analytical Estimates for Single Well Patterns
Calculation methods for describing fluid mechanics efects and flow porcesses in miscible flooding generally (or
most easily) apply to single well patterns, and conceptually involve either a homogeneous five-spot pattern or a
heterogeneous vertical cross-section layer model between an injector and a producer. In some cases, the vertical
cross-sectional models can be converted to five-spot calculations by applying Deppes method (or Claridges
modification of Deppes method) to each layer. Another method is to generate a layered model, and then apply
Claridges five-spot correlation to the set of layers. These methods are presented in detail in the section titled "Well
Pattern Sweep Efficiency"
A very generalized method is simply to plot some of the very limited field estimates derived from pilot tests, usually
for a single well pattern, and read a first estimate of recovery from this plot. Figure 1 shows an example.


Figure 1


Areal Sweep Correlation Plus Estimated Vertical Sweep
For this purpose, Claridges five-spot sweep efficiency correlation is suggested. This correlation combines the
sweep efficiency of a miscible displacement average front (for a given pure fluid mobility ratio) with the incomplete
sweep behind that average front (as described by Kovals equations for viscous fingering) at the effective mobility
ratio. This correlation gives not only the breakthrough sweep efficiency, but also the recovery at further throughputs
of injected fluid. It assumes that the miscible displacement starts at maximum oil saturation. The correlation may
also be used for a water-flood. Therefore, the oil recovery calculated for displacement at maximum oil saturation,
minus the calculated waterflood oil recovery, can be used as the extra oil recovery of the miscible flood over that by
waterflooding.
One might think that this calculated recovery could be used in the case where the miscible flood is carried out after a
waterflood, but this is not so. The mobility ratio in the post-waterflood case is different from the pre-waterflood
case, and complications exist due to the necessity of remobilizing waterflood-trapped oil in order to have any extra
recovery. The latter flood may be estimated using the correlation with the post-waterflood mobility ratio, and Todds
modification of the Raimondi-Torcaso equation (Todd et al., 1982) may be used to discount the estimated oil
recovery by the remaining trapped oil saturation based on the average water phase saturation behind the solvent
flood front. This involves fractional flow calculations where the oil/solvent mixture at the displacement front is
found by a simple trial-and-error method. This provides an immiscible effective mobility ratio and a measure of the
average trapped oil saturation behind the front, back to the injection well. From the effective mobility ratio, an
equivalent end-point mobility ratio can be calculated by back-solving Kovals equation for effective mobility ratio.
The five-spot correlation can then be applied using the endpoint mobility ratio.
In case WAG is used, either in the pre-waterflood or post-waterflood case, then Stones method should first be
applied to determine which fraction of a homogeneous layer being treated will undergo a miscible flood followed by
a waterflood and which fraction will undergo a waterflood only. This will occur in each layer of a multi-layer
reservoir
Following these calculations, it is then necessary to apply an estimate of the vertical sweep efficiency of a series of
layers representing a given actual reservoir. This can be done either by applying a vertical sweep efficiency estimate
based on the Dykstra-Parsons , v or by repeating the calculations for each of the layers in a multi-layer model. The
latter procedure involves calculating different slug sizes for each layer, as discussed below for the Deppe method,
except that in this case we use my correlation instead of Koval-type calculations in each layer.
Deppes Method , Applied to Layered Systems
For miscible floods, this involves an extension of Deppes method which accounts for viscous fingering by Kovals
method. It also involves using Claridges areal breakthrough sweep efficiency equations. Because the value of Me in
these equations depends on Kovals relation between Me and the pure fluid mobility ratio, as well as on a calculated
value of the heterogeneity factor, H, for each layer, the estimated areal sweep efficiency for each layer can be used
to determine the radius of the injection well circle for each layer. If H is the same for each layer, then this radius will
be the same for all layers. However, the size of the solvent slug that enters each layer will vary among the layers,
and so the process behavior will vary significantly.
The simplest way to allocate the amount of solvent between layers is on the basis of the k
h
and the |h of each layer.
The slug size relative to the pore volume of each layer is the amount allocated, based on k
h
, divided by h for the
given layer; the slug sizes are thus proportional to the k/|) values. Then the slug size for each layer is found by
dividing each k/|) value by the average k/| value, and multiplying the quotient by the overall slug size (as a fraction
of the original hydrocarbon-filled pore volume). The high permeability layers will, in general, receive large slug
sizes and will be exhaustively swept, while the lowest permeability layers will receive small slug sizes and will be
swept only to a small degree by solvent.
Assuming that the miscible slug will be followed by a waterflood, in the final state the trapped non-aqueous phase in
the most permeable layers will be all solvent, while in the least permeable layers only a small proportion of the
trapped non-aqueous phase will be solvent; the rest will be oil. For typical overall slug sizes, more solvent will flow
through the high permeability layers than is needed to displace all of the oil, and excess solvent will be driven out by
the following waterflood, while there will be a shortage of solvent in the low permeability layers to replace the
waterflood residual oil. Hence, such a layered model emphasizes the inefficiency of solvent usage caused by the
heterogeneity described by the layer properties.
Reservoir Simulation Applications in Miscible Flooding
Factors Effecting Costs of Simulation
The refinement of structure definition by seismic, geologic and petrophysical methods has led to increasingly
detailed reservoir descriptions. Two examples are the work of George and Stiles (1978), and Sneider, Tinker and
Meckel (1978). Reservoir engineering studies of EOR process design and optimization have settled on mathematical
simulation of the processes using, for the most part, finite-difference representations of the material balance and
flow equations between intersection points on a grid of lines covering the extent of the reservoir. Each grid
intersection at which fluid properties, saturations, pressures and flow vectors are specified represents, for
mathematical purposes, the values of these quantities throughout a local block of the reservoir volume, with the
block having its center at the grid intersection. The blocks are adjacent to each other and their total volume is equal
to the reservoir volume.
All details of reservoir structure and rock properties must be discretized and supplied to the computer program as
data concerning the rock and fluid properties in each block, and the fluid injection and withdrawal points at given
block locations. This may require a large number of blocks and also of computed process time-steps (time is also
discretized) to represent the detailed knowledge of reservoir structure and flow behavior postulated.
For example, a twenty-layer model with an areal grid of 20 by 20 blocks each way contains 8,000 grid blocks. If as
many as 200 time-steps are required to represent a process (a conservative estimate), then the product of grid blocks
and time-steps is 1.6 million. To obtain cases for economic evaluation and determine an optimum project design, it
may be necessary to make at least 10 or 20 runs in order to (a) obtain a history match, and (b) make future
projections using the given process with some variations in controllable design variables (such as slug size and
WAG ratio). Considering that costs of real computational time may range from $50/hour to $300/hour, the expense
of developing this grid may be a large fraction of the total funds available for the computing part (in addition to
engineers salaries) of the engineering design studies for the given reservoir.
Even this expensive grid may not be as fine as the geologists and petrophysicists (or the reservoir engineers, for that
matter) would like. It is also obvious that a 20 x 20 areal grid could hardly be adequate for a reservoir containing, for
example, as many as twenty or thirty wells (most candidates for EOR processes have more).
In a given case, some data that might influence the process performance may not be known, and could cost
considerable time and money to obtain. In such a case, a study of the sensitivity of the process to variations in the
uncertain data will often indicate whether the data are important or not. It is therefore even more desirable to have a
relatively simple reservoir model in which to carry out the variable sensitivity studies.
Model Simplification Techniques
In using finite-difference simulators to study the design and performance of EOR processes, it has often been found
necessary to compromise between a detailed representation of reservoir structure and a detailed study of operating
conditions and overall strategy, within the limitations of project engineering costs. In these simulations, a very fine
grid would often be required in order to incorporate all of the variations in permeability, porosity, formation
thickness and fluid saturations in a reservoir known or inferred by the geophysicists, geologists and petrophysicists.
A fine grid would also be required to represent such fluid-mechanics phenomena as viscous fingering and gravity
tonguing within each layer in a reservoir. Use of such fine-grid models would make single simulations (with current
large-capacity, high-speed computers) cost so much as to preclude making comparative runs to study variations in
operating conditions and process strategy.
To retain their capability of performing optimization studies, engineers have sought reasonable compromises
between accuracy in representation of reservoir structure and fluid-mechanical behavior, and accuracy in
determining optimum process design and operation. Among the techniques which have been used are pseudo-
relative permeability curves, pseudo-representation of viscous fingering through the mixing-parameter method, and
a criterion (Lake and Hirasaki, 1981) to determine when adjacent layers have sufficient transverse dispersion to
allow us to consider them a single layer.
Pseudo-Relative Permeability Curves
Hearn (1971) and others have shown that the effect of layers with different properties on immiscible displacement
behavior can be taken into account in single-layer simulations by employing calculated relative permeability curves
in place of measured relative permeability curves. This makes it possible to simulate an entire field, if the field is not
too large. However, the result of the calculations is to produce a relative permeability curve for the invading phase
which is generally convex upward, as compared to the usual shape of rock curves which is concave upward. The
curve for the displaced phase is generally still concave upward. In processes where both aqueous and non-aqueous
fronts proceed through the reservoir at different places at the same time and at the same place at different times, it is
not generally feasible to use these pseudo curves.
Using Multiple Well- Pattern Prototypes
When pseudo-relative permeability curves are not applicable, we need to revert to the rock curves and to a layer
structure that reflects the actual degree of heterogeneity. It is then often impossible to represent an entire field.
Single well patterns are simulated instead, and results multiplied by the number of similar patterns in the field. If
there are several such pattern types (e.g., internal well patterns and external or edge patterns), then a prototype of
each pattern is run and the sets are assembled to give total field results (Bilhartz et al., 1978; Rester and Todd,
1984). There are obvious errors arising from interpattern flows, but on a field-wide basis they tend to balance out.
Layer Combination Based on Transverse Dispersion
Lake and Hirasaki (1981) have described a method for determining when adjacent layers have low enough contrast
in permeability and sufficient transverse dispersion that they will act as a single layer. One specific case involved
using their method to reduce a 23-layer reservoir description to a nine-layer description. The resulting model
preserves most of the heterogeneous features of the original model and is much less expensive to run.
Using 2-D (Vertical) Variable-width Models
Two-dimensional variable width models have been used by various persons, but the first publication in which its
use is reported is that by Warner (1977). A five-spot pattern is represented by a series of blocks in the x-direction,
running from injector to producer. The blocks are assigned widths in the y-direction to approximate the shape of a
one-quarter five-spot, as shown in Figure 1 and Figure 2 , which also show that layers are represented in the z-
direction.


Figure 1


In this case, an odd-number series (1, 3, 5, 7, 9 .


Figure 2


. . 9, 7, 5, 3, 1) represents the relative block widths; the total area is made the same as that of a field quarter five-
spot. The transmissibility between blocks is proportional to the average of successive odd numbers and is thus an
even number series, except at the middle where the two adjacent odd-numbered blocks are identical. Vertical
transmissibilities between layers are in accordance with the odd-number series.
In the actual simulator input data, all of the blocks in the x-direction have the same width; instead of having greater
widths, the pore space for each block is multiplied by the same factors as mentioned for widths.
The nine-spot pattern presents a problem, because in an inverted nine-spot there are four half-wells on the sides and
four quarter-wells in the corners. Assuming equal well productivity, a side well would produce twice as much as a
corner well from within the pattern. Furthermore, the side well is closer to the injection well, and thus there is a
steeper potential gradient to it than to a corner well. A two-dimensional variable width model must represent both of
these facts.
A solution to this problem is to represent the symmetry element of one-eighth of a nine-spot, as shown in Figure 3 .


Figure 3


The right triangle is divided by a line from the injector to the stagnation point on the opposite side of the triangle.
For equal well productivity, this line would be at a point one-third the distance from the corner well to the side well.
The internal area of each portion of the triangle is divided into blocks by circular arcs drawn around the injection
and production wells as shown, so that the total area is preserved. The blocks are equally spaced in radius (length),
and the annular block areas are then transformed into rectangular blocks of the same length and area. The two
triangular portions may be unfolded to lie in a straight line, with the side well at one end, the corner well at the other
end, and the injection well in between (not at the midpoint), as shown in Figure 4 .


Figure 4


This linear system of blocks may then be converted into a two-dimensional, variable-width, vertical model by
adding layers of the same shape. Of course, in the final model the varying widths are changed into correspondingly
different pore space by multipliers equal to the width ratios, so the actual block widths are all the same.
Pseudo-Layer Models
A method for reducing a large number of layers to a small number while preserving the same Dykstra-Parsons
coefficient of permeability distribution, total k
h
and total |
h
has been devised and is described below. This reduction
presents some pitfalls; one of them is also described below.
Obtaining the Pseudo-Layer
In some cases, we need to reduce the number of layers to a degree beyond that attainable by applying the Lake-
Hirasaki method to an actual sequence of layers. Claridge has devised a method for doing this which preserves the
slope of the Dykstra-Parsons permeability/porosity distribution. It is described here with reference to a hypothetical
example, starting with 54 layers defined by a log-obtained porosity profile, and ending with a three-layer model.
Figure 5 shows the log-obtained porosity profile.


Figure 5


With a porosity cut-off of 5%, there are 15 layers out of the 54 which may be assigned the cut-off porosity and
permeability values of 5% and 1 millidarcy; these layers constitute 100 feet out of the total 218 feet.
The rest of the porosity/permeability data were plotted on semi-log paper to give the plot of Figure 6 .


Figure 6


The equation of the best line through the data is:
log k= 15*|-0.75
where porosity is a fraction rather than a percent, and k is in millidarcys.
The porosity values of Figure 5 were arbitrarily jumped into ten categories, including the cut-off value of 5%, and
the thickness associated with each average porosity level was calculated. For each average porosity level, the
permeability was read from the line of Figure 6 (or calculated by the equation). This gives a ten-layer model as
shown in Table 1 .


Table 1


Total thickness = 218 ft, total |h = 16.426 ft (average porosity = 0.07535), and total k
h
= 1,000.0 md-ft (average
permeability = 4.59 md).
(The fact that this is a hypothetical example is shown by the last column: each average porosity level has associated
with it ten percent of the total k
h
. In reality, the fractions of k
h
would vary.)
When the cumulative fraction of k
h
is plotted against the cumulative fraction of |h, a Lorenz diagram is obtained, as
shown in Figure 7 .


Figure 7


The Lorenz coefficient in this case is about 0.46.
Another way of representing the heterogeneity is by means of the Dykstra-Parsons diagram, as shown for these data
in Figure 8 .


Figure 8


The Dykstra-Parsons coefficient v for this case is 0.71.
A procedure for obtaining a three-layered model from these data is as follows:
1. For equal k
h
fractions of 1/3, read the |
h
fractions 0.085, 0.24 and 0.675 from the Lorenz diagram.
- The corresponding k
h
for each of the three layers is 333.3 md-ft, since the total k
h
was 1,000.
- Corresponding |
h
s for the layers are 1.396, 3.94 and 11.09 feet, since the total |
h
was 16.426.
2. Then |= [|
h
/ k
h
](k)
3. Substitute each value of | into the equation for log k,

log k - 15[|
h
/ k
h
](k) + 0.75 = 0

4. Solve this equation by trial and error to obtain the k for each layer, using the corresponding h and k
h
for
each layer. This gives k
1
, k
2
and k
3
.
5. Invert the log k equation, | = (log k + 0.75)/15. Inserting the values of k
1
, k
2
and k
3
gives the values of
|
1
, |
2
and |
3

6. Since
h
= (|
h
)/ |, substitute the three values of |
h
and | to set the values of h
1
, h
2
and h
3
. For this case, the
layer data are given in Table 2 .


Table 2


Total k
h
= 1,004 md-ft, total |
h
= 16.42 ft (both very close to the 10-layer model). Note, however, that the total h
=195 ft instead of 218 ft. This is apparently because the straight-line log (k) vs. | relationship is not compatible with
the Lorenz plot of k
h
vs. |
h
(nor with a straight line on the Dykstra-Parsons diagram of log(k) vs. cumulative |
h
).
A check of the Dykstra-Parsons coefficient, given in Figure 9 , shows the same value of v for the three points of this
three-layer model as for the ten points of the ten-layer model.


Figure 9


There is more deviation of these three points from the line through them than we found with ten points, however
(probably because of the problem just mentioned). No effort has been made to resolve these discrepancies; we
realize that this simplified model is only an approximation and that at the end of sensitivity studies, a few runs
would be made with a 3-D model using the correct reservoir data.
For a five-layer model, values of |
h
would be read from the Lorenz plot for five equal increments of k
h
, and the
series of calculation steps outlined above would be followed. In general, choosing equal increments of k
h
is
preferable to equal increments of |
h
. In the latter case, the deleterious effect of thin, high transmissibility layers on
process efficiency tends to be grossly underestimated. We can see that in the three-layer case above, the highest
permeability has become 37 instead of the highest value of 100 millidarcys in the ten-layer model. If equal
increments of |
h
are used, the highest permeability is 15.78 md, the intermediate value is 3.57 md, and the lowest is
1.01 md. This procedure therefore magnifies the effect of the lowest permeability layers rather than the highest. The
thickness of the highest permeability layer is 42 feet instead of 9 feet. The same calculation procedure was used to
obtain these values.
Arranging the Pseudo-Layers
Since all of the original layers have been cast into an ordered set from which the new pseudo-layers were derived,
this method of derivation leaves no information about the vertical order in which they should be placed for use in
simulations. We must derive this order from the original data. In one case, Claridge (1982) used well-log profiles to
decide on an order. There are six possible arrangements of three layers; designating the layer permeabilities as H
(high), M (medium) and L (low), the six orders are, from the top down, (1) HML, (2) HLM, (3) MHL, (| LHM, (5)
MLH and (6) LMH.
Todd et al. (1982) examined the effect of layer order using orders (1), (3) and (6) of the orders given above and the
set of layer properties given in Table 3 .


Table 3


Total h = 157 ft, total |
h
= 14.06 ft, total k
h
= 434.5 md-ft, and the Dykstra-Parsons coefficient of
permeability variation v = 0.67.
These layer properties were used in a two-dimensional (NX = 9, NZ = 3) variable-width quarter-five-spot model
representing one-fourth of a 40 acre five-spot in all of the simulations discussed below. A base case run was made of
a waterflood to 98% water cut (44% oil recovery, based on OOIP), followed by an 0.4 HPV slug of CO2 and then by
another waterflood to 98% water cut (incremental oil recovery 14.4%, based on OOIP). In this run, the ratio of
vertical permeability to horizontal permeability (k
v
/k
h
) was zero, so there was no interaction between layers and
hence no effect of layer order. Then a series of runs were made with a (k
v
/k
h
) ratio of 0.01 and, as mentioned, with
the layer orders HML, MHL and LMH. They gave nearly identical waterflood recoveries and extra oil recoveries
due to the tertiary process, and they were the same as the values with no vertical transmissibility. The (k
v
/k
h
) ratio
of 0.01 gives a vertical transmissibility (kA/L) which is about 20% of the horizontal transmissibility. A further test
was made in which the (k
v
/k
h
)ratio was made 0.1, with layer order HML; the results were the same.
However, reference to the well logs made it evident that actual layer thicknesses were on the order of one to three
feet; many of these relatively thin layers with all of the different possible orders have been grouped into the three
layers of the model to give the same total thickness. When the actual layer thicknesses are, for example, 2 feet, then
a (k
v
/k
h
) ratio of 0.01 gives a vertical-to-horizontal transmissibility ratio at least ten times that of the thick-layer
model of Table 2 . It is evident from the first series of runs with the thick model that the vertical driving force for
fluid transfers must generally be much weaker than for horizontal transport, hence, to get significant cross-flow
effects (and thus an effect of layer order), a ratio of vertical transmissibility to horizontal transmissibility
considerably greater than 1.0 will generally be required.
Therefore, Todd converted his first three-layer model into a second one by dividing all layer thicknesses by 12, so
that the thickness of the high-permeability layer was only one foot (2.5 feet for the M layer, 9.6 feet for the L layer).
A check run with the MHL sequence and zero (k
v
/k
h
) gave the same 44% waterflood recovery and 14.4% EOR
recovery as before. However, with a (k
v
/k
h
) ratio of 0.01, the EOR recovery was cut from 14.4% to 10.7%.
If the CO2
flood was run in the secondary mode (with no prior waterflood) and the typical waterflood recovery was
subtracted from the gross oil recovery to get the equivalent EOR recovery, it was only 9.3%. It was found, however,
that the EOR recovery of 10.7% mentioned could be raised to 13.3% by injecting the 0.4 HPV CO2
slug in 1:1 ratio
with water. Using a 2:1 ratio of water to CO2
lowered the EOR recovery back to 12.0%, showing that the 1:1 ratio
was better than 0:1 or 2:1.
A study of the mechanism of the EOR recovery loss from 14.4% to 10.7% showed that the oil bank generated by the
CO2
slug in the L and N layers crossflowed into the already CO2
-swept H layer, and that during the subsequent
waterflood this lagging oil was trapped by the water. When water was injected along with the CO2
slug, the mobility
of the combined fluid was much less than when CO2
was injected alone, and the H layer did not become as much of
a pressure sink as it had been with CO2
alone.
Since the L-layer was still 9.6 feet thick, it was decided to split it into three identical layers, each 3.2 feet thick. At
the same time, the M-layer was split into two layers, each 1.25 feet thick. This made six layers, which were placed
in the order (from the top down) of LMHLML. A simulation similar to the one that gave 10.7% EOR resulted in
9.3% recovery; it was concluded that this was more realistic than the 10.7% value. On the other hand, Todd wanted
to use only three layers. All of the prior runs had been made with a mixing parameter value of omega = 0.6 (where a
value of 0.0 indicates no mixing of oil and solvent, and a value of 1.0 indicates complete mixing; he decided to see if
a small reduction in this parameter would reduce the EOR recovery from 10.7% to 9.3% while still using the (thin)
three layer model. A run with omega = 0.5 gave 9.6%, and a run with omega = 0.45 gave 8.8%. It was concluded
that a value of omega of about 0.48 would have been about right to give 9.3% EOR recovery desired. This is
equivalent to an increase in from 0.1 to 0.3 on Figure 9 , at M = 20. Thus, the reduction in apparent heterogeneity
from six layers to three layers can be compensated by reducing omega from 0.6 to 0.48.
In another study performed by Claridge (not previously published), a reduction of 24 layers to 10 layers could be
compensated by using an omega of 0.55 in place of 0.6, thus obtaining the same EOR recovery (in a similar two-
dimensional variable-width model and with k
v
/k
h
= 0.01). It is clear, however, that a reduction in omega does not
reduce EOR recovery by the same mechanism as cross-flow. Therefore, it should not be used as a way of avoiding
multiple layers altogether.
Avoiding the Layer Thickness Pitfall
Based on the evidence just presented, we may state that if the procedure given here is used to reduce a multi-layer
system to a small number of layers, three layers is the minimum number that should be used in a complex process
requiring the use of rock-derived relative permeability curves. The properties of the three layers should be such as to
preserve the kh/|h relationship of the entire system, but the thickness of the layers should be reduced so that they
approximately match the real thickness of the layers in the reservoir. Only then, with an appropriate value of the
kv/kh ratio, will cross-flow effects be realistically evaluated. If the k
h
/|
h
relationship is such that the L-layer and
possibly the M-layer are thicker than actual reservoir layers of similar k and |, when the H layer has been made as
thin as the actual high-permeability reservoir layers, it is appropriate to make a moderate reduction in the mixing
parameter to compensate for the greater-than-normal thickness of the L and M layers. The amount of this reduction
may be determined by a simple test, as described above. Results from the three-layer model may then be scaled-up
to the total reservoir thickness.
Comments on Model Simplification
The approaches described have been very helpful to reservoir engineers in carrying out their tasks of EOR project
design and performance prediction, given the current conditions of mainframe computer capabilities and the costs of
computing services. The simplifications made in reservoir models and in representations of fluid mechanical
behavior are, for the most part, realistic, though not entirely satisfying from the standpoint of truly accurate
representation; that accuracy can only be achieved by major advances in computing capabilities and reductions in
cost. Even then, the manpower required to create the large amount of simulation input for these realistic, detailed
models will need to be augmented by computer assistance in order to acquire the data and reduce it to the form
required for reservoir simulation input.
Process Models for Miscible Flooding
Two general types of computer simulator programs are used to simulate miscible flooding: the modified black oil
type and the compositional type. Each has certain advantages and disadvantages. The modified black oil simulator is
appropriate if we can assume miscibility to prevail throughout the reservoir. It is modified to handle viscous
fingering by using what is called the mixing parameter method.
The Mixing Parameter Method
Koval was the first to find a way to represent the effects of viscous fingering without actually having the fingers. He
resolved the fingers of solvent in oil into a saturation gradient of solvent, treated the simultaneous flow of solvent
and oil as a Buckley-Leverett fractional flow problem (with the relative permeabilities equal to the saturations), and
obtained a solution for the oil production curve. In order to match experimental data, however, he found it necessary
to weaken the mobility ratio compared to the pure solvent/oil viscosity ratio, as has been previously described.
While his solution was successful in describing a wide variety of experimental laboratory model results, particularly
when transformed to a radial instead of linear basis (see Claridge, 1979), his treatment was one-dimensional, and it
was not clear how it could be extended to three-dimensional flow representation of the effects of viscous fingering.
Todd and Longstaff (1972) devised an alternative, but similar, approach: the mixing parameter method, which can
be applied to two miscible, but only partly mixed, fluids within a block in a computer simulation model of an oil
reservoir. The part that is mixed is assumed to have solvent and oil proportions equal to their overall mobile
saturations within the block (excluding any waterfloodtrapped oil or solvent which may be present), and the quarter-
power mobility blending rule is applied to obtain the mixtures viscosity.
The effective viscosity of solvent and of oil are calculated as the product of the blend viscosity to the power omega
and the pure fluid viscosity to the power (1 -omega). The ratio of these effective viscosities is therefore equal to the
pure-fluid viscosity ratio raised to the power (1 - omega), and we may equate this to Kovals effective mobility ratio.
The mixing parameter omega is an empirical parameter (like Kovals empirical blend), which has to be found by
matching experimental data. At first, laboratory data were used (as they were by Koval), but more recently, enough
field data have been matched to allow us to settle on a value in the range of 0.5-0.7; 0.6 is a reasonable compromise.
Kovals effective mobility ratio (Me) and the mixing parameter omega are related as discussed above and as shown
by Figure 10 , which includes the effect of the Dykstra-Parsons variation, .


Figure 10


We can see that in the usual range of mobility ratio M of 5-50 based on pure fluid viscosities, an omega value of 0.5
to 0.7 corresponds to a range of from about 0.1 to 0.2. We may assume this to be the degree of heterogeneity not
accounted for when history-matching with a layered system of areally uniform properties, as was used in the field
project history matches.
The mixing parameter method is designed to handle the physical situation in which the solvent and oil are
considered capable of mixing in a single phase in any proportions (and thus they share the non-aqueous phase
relative permeability between them in proportion to their mobile concentrations in a given block of the reservoir
simulation mathematical model), but are only partially mixed in the reservoir. Some of the solvent lying in the
middle of viscous fingers is unmixed with oil, and some oil lying between fingers of solvent is unmixed with
solvent. A mixed zone of solvent and oil surrounds each viscous finger. These fingers are considerably smaller than
the size of the grid blocks which can be economically used in reservoir simulations. Hence, the fluids inside these
blocks are not completely mixed, and so are not of constant proportions throughout the block volume. The mixing
parameter method calculates transmissibilities of oil and solvent out of each block on the basis of partial mixing
within the block. The mixing parameter scale is from no mixing (with a mixing parameter of zero) to complete
mixing (with a mixing parameter value of 1.0).
The principal virtue of the modified black oil model is its ability to represent, with considerable fidelity, the major
aspects of the miscible flooding process that control the linear displacement efficiency and the areal and vertical
sweep efficiency, and thus the overall oil recovery. It does not represent phase behavior, except for some simulators
which use the mixing parameter method above the miscibility pressure, but calculate phase equilibria when the
pressure falls below the miscibility pressure (or bubble-point pressure). This may happen when, for example, solvent
and oil approach a pumped-off production well at low bottom-hole pressure. Over most of the reservoir, however,
phase equilibrium calculations are assumed to be unnecessary and are not performed.
Compositional Simulators
A compositional simulator is able to match the step-wise approach to miscibility that actually occurs in most
miscible floods. It assumes that the average composition in a grid block of the mathematical reservoir model is
actually uniform throughout the block. The phase behavior is calculated by means of equilibrium phase distribution
K values, which are computed for the conditions in each block using a multi-component equation of state. After a
number of equilibrium contacts, if multiple-contact miscibility is indeed going to occur, the phase equilibrium
calculation will result in a single phase. This actually happens in a linear flow system. In a three-dimensional
system, however, the flow behavior is such that compositions do not stay in the miscible region. Mixing of fluids
which are not part of the same prior mass-transfer process can occur, and then the process tends to enter the two-
phase region of a phase diagram. The compositional simulator will show this behavior.
This approach to reservoir simulation matches phase behavior much more closely than the modified black oil type of
simulator. We can get a reasonable approximation of the greater dispersion of the miscible solvent (which is the
result of incomplete mixing and consequent viscous fingering) by introducing second-order dispersion terms into the
transport equations, which are cast into numerical form for solving the pressure distribution and transport between
grid blocks during each time step. The best compositional simulators for predicting miscible flooding have this
dispersional approach built into them to account for fingering effects.
However, we should realize that calculating phase behavior for large grid blocks on the basis of average
composition is theoretically unsound. To obtain a reasonably correct representation, the grid-block size must be
small enough so that the composition within it can truly be considered almost uniform. In my judgement, this
requires grid block lengths of no more than 10-20 cm in each linear dimension. With larger grid blocks, the
assumption of complete mixing leads to incorrectly calculated transmissibilities out of the grid blocks which, in turn,
leads to incorrect representation of areal and vertical sweep efficiencies. It is possible to modify this behavior by
empirical approaches. Some simulators use the mixing parameter method together with prior compositional
calculations. In my opinion, this application of the mixing parameter method is not justified by physics or
engineering; however, since the mixing parameter method is itself empirical, it is possible to consider it as an
adjustable parameter in a compositional simulator; this makes it possible to match real-world process performance
with a more simplistic physical and mathematical model.
It is evident that a better combination of the modified black oil simulator and the compositional simulator would be
desirable. A new approach is needed to solve this problem.
Recovery Estimates in Multi-Pattern Areas or Entire Fields
There are several approaches available for estimating EOR over entire fields or in areas where data on a number of
patterns within a field are known.
Combining Single Pattern Estimates
In preparing recovery estimates for a CO2 flood in the Willard Unit of the Wasson Field, ARCO engineers (Bilhartz
et al., 1978) made estimates of waterflood recoveries by single patterns (cross-sectional models fitted to the
individual performance of several patterns in a region of the unit), with allowance made for different behavior in
several different regions of the field. They then combined these estimates to obtain a history match of the waterflood
in the entire unit. The fact that a good match was obtained was taken as an indication that the same method would
work for the CO2 flood process.
This procedure of combining single pattern estimates-in this case, of starting CO2 flooding at different points in the
course of waterflooding and for different parts of the field at different calendar times, as well as for five-spots, nine-
spots and several different types of field-edge patterns-was used by Todd, Dietrich and Chase in work performed for
several different clients. They wrote a separate computer program for this purpose, called FIELDPROJ, which uses
as its data the previously simulated CO2 flood histories of the single patterns when CO2 flooding was started at
different points in the waterflood process (Rester and Todd, 1984). This allows higher quality simulations of the
individual well patterns (by having a sufficient number of grid blocks per pattern) to be incorporated into the whole
field estimate.
Direct Simulation of an Entire Field
Computer capabilities are steadily improving, and simulations involving as many as 40,000 grid blocks are currently
being performed. In the case of very large fields such as Prudhoe Bay, this still allows only single-layer simulations,
and the well density is high compared to the number of grid blocks between wells. Nevertheless, this field-wide type
of model allows us to observe major trends and flows which cannot be handled by assembling single-pattern
simulations. In Prudhoe Bay, for example, there is a gas cap which encroaches on the oil column as oil is withdrawn
from the field. Tendencies for interpattern flows (as in the case where a waterflood starts in one region of a field and
moves toward another region where water is not being injected but oil is being produced), become apparent in the
full-field simulations.
In the El Dorado (Kansas) chemical flooding pilot tests, the prior waterflood was causing a general pressure gradient
from west to east which was shown to severely distort the streamlines in the pilot area; this had to be taken into
account in the simulations of the pilot area.
Yet doubt remains as to how good these full-field simulations are with respect to oil recovery predictions, simply
because of the coarseness of the grid and the inability to represent layer heterogeneity effects.
Help is on the way, however: there are clear indications that computer capabilities are improving by orders of
magnitude. At present, it appears that this will be accomplished by using parallel processing with the processors
linked in a multidimensional array. As of 1990, experiments were underway not only linking multiple processors in
one machine, but also linking a considerable number of separate machines together. The efficiency of various
control algorithms and linking arrangements is measured by dividing the increase in computing speed by the number
of processors. Some relatively high efficiencies have been reported. Of course, the increased computing speed has to
be accompanied by a greatly increased memory capacity and corresponding speed of access (input/output
operations).
The steady decrease in the physical size of computer hardware has been a tremendous aid in increasing computer
complexity and corresponding capabilities. Current desktop computers have higher capabilities than the mainframe
computers of the 1950s, which occupied a rooms worth of floor space.
Considering the evolution that has taken place in the past thirty years and the accelerating pace of development, it
seems likely that it will be soon be possible to simulate large fields with a thousand wells or more, with an adequate
number of grid blocks per pattern and with multiple layers.
Imput Data For Complex Situations
At or before that point, we will need to develop computerized methods of providing input data for fine-grid,
heterogeneous simulators. The task will obviously be too large for individual engineers to do by hand; this is already
apparent in the case of large fields and in the case of any field where geostatistical methods are being applied to
include block-by-block porosity and permeability variations in the reservoir model and shale breaks of varying
length, width and location. Evidence indicates that these geostatistical models are more faithful to the reality of
reservoir variations (or at least to their effect on oil recovery) than the provisions we have been able to make in the
reservoir simulation models of the past.
At the same time that computer capabilities are improving markedly, it thus appears that much of this greater
capability may be used simply to increase the fidelity with which the mathematical model of the reservoir matches
the actual heterogeneity existing in any given cross section between pairs of wells. For very large fields, an increase
in the number of grid blocks on the order of 1,000 to 10,000 times may be required (which is not yet seen as
possible), rather than the improvements of 10 to 100 times presently envisioned. In the meantime, there are many
more medium-sized and relatively small fields for which the currently envisioned computer enhancements may well
be sufficient to perform adequate full-field simulations.
Even in this case, the methods discussed here for simplifying reservoir models to allow multi-run studies of the
sensitivity of oil recovery to various flood design options and to provide economic comparisons of process options
may remain helpful, since full-field runs using the maximum number of grid blocks may have to be kept to a
minimum for cost and manpower reasons.

Review of Miscible Flooding Pilot Tests and Commercial Projects
At two-year intervals, in an early April issue, the weekly periodical Oil and Gas Journal publishes a complete
listing of enhanced oil recovery projects, including miscible and immiscible gas injection projects. The listing
primarily covers U.S. projects, but also gives some tables for foreign projects. Since the review covers thermal
recovery (steam soak and drive, and in situ combustion) and chemical flooding (surfactant/polymer, caustic and
polymer) as well as gas injection projects, it contains a great deal of information that is outside the scope of this
module. However, the Oil and Gas Journal listing also is more superficial. While it does contain a certain amount of
reservoir data such as depth, permeability, area and thickness, it does not discuss individual reservoir differentiating
characteristics and process behavior.
The Oil and Gas Journal review includes a summary table of EOR production in barrels per day, categorized by
EOR process. As of 1990, the oil recovery attributed to gas injection projects amounted to about 190,000 barrels per
day, or about 30% of U.S. EOR production. This was still only about 2 1/2 percent of the total 1990 U.S. production
of about 8.2 million barrels per day. Thermal production methods still accounted for roughly two-thirds of U.S. EOR
production. Miscible flooding in the U.S., primarily CO
2
flooding, is growing rapidly; the resulting production is
expected to peak at 1/2 to 1 million barrels per day, probably around the year 2005, before declining.
Because of the strong U.S. demand for LPG and natural gas as fuels, miscible floods based on these hydrocarbon
solvents have not been started for about two decades, except on the north shore area of Alaska, including the
Prudhoe Bay field and the Kuparuk River field, and in the Sand Dunes field in Wyoming. Projects involving
hydrocarbon solvents began in the U.S. primarily during the period from 1955 through 1970. These older U.S.
projects are in their decline or have been terminated. On the other hand, in western Canada and other parts of the
world, the demand for LPG and natural gas has not yet grown to match the supply, so opportunities still exist for
new projects outside the U.S. Within the U.S., the future apparently lies in CO
2
flooding. With these trends in mind,
principal examples of the various miscible flooding projects are reviewed in the following pages.
High Pressure Gas Projects
An outstanding example of this process (in which we will include projects using N2 and flue gas) is the project
carried out in the University Block 31 reservoir in Texas (see Warner et al., 1979). Gas recycling began in this
8,500-foot deep reservoir in 1949, four years after its discovery. The reservoir temperature is 140 F but the oil is
very light (48 API) so that the multiple-contact miscibility pressure with lean gas is 3,500 psia. Injection under
multiple-contact miscibility conditions began in 1955, hence the process had been in progress for about 35 years as
of 1990. Experimental work showed that flue gas was almost as good as lean ("dry") gas for achieving a multiple-
contact miscible drive in this case. Consequently, in 1966 a flue gas plant was installed to make flue gas from a
much smaller quantity of natural gas, in order to reduce costs for purchased natural gas. By 1978, over 43% of the
oil originally in place had been recovered, and the ultimate projected oil recovery is 65%.
Stalkup (1984) discusses this field and gives information on other field examples of the high pressure vaporizing
process (Fairway field, Raleigh field, Hassi/Messaoud field, Neale field, Bridger Lake field, and University Block 9
field).
Nitrogen obtained from air reduction plants has found some use as a miscible drive agent. The outstanding example
is the Jay/Little Escambia Creek field in southern Alabama/northern Florida, operated by Exxon (see Christian et al.,
1981). The initial pressure in this deep reservoir was 7,850 psia, and the oil is 510 API, with 0.18 cp viscosity.
Clearly, this is an unusual example but one in which nitrogen could be used as a miscible displacing agent because
of the light oil and high pressure, and in spite of the high temperature (285 F). The field was depleted to 4,800 psi,
then a waterflood was started to maintain pressure (and to recover waterflood-movable oil). The pressure at the start
of miscible nitrogen injection was 6,085 psi. The minimum multiple-contact miscibility pressure with this crude oil
at 285 F was found to be 3,600 psia with either nitrogen, methane or CO
2
. Nitrogen was chosen for economic
reasons, and also because it is less dense than the oil and could lift the oil from low layers when injected low in the
formation (while CO
2
, by contrast, is more dense than the oil and tended to sink in the reservoir). Computer
simulation studies showed better vertical sweep efficiency for nitrogen than for CO
2
under these unusual reservoir
conditions.
LPG/Dry Gas Miscible Drives
The marked superiority of gas drives (either miscible or immiscible) when they can be carried out in the downward
direction in either a vertical reservoir or a steeply dipping reservoir, has already been mentioned. Some of the more
interesting miscible projects have been vertical or steeply dipping types. Some examples are discussed below.
The Golden Spike reservoir in Canada is a pinnacle reef. A miscible flood was planned as an LPG slug injected at
the top of the reservoir, with crude oil being produced at the bottom, and with dry gas used to drive the LPG slug
(and the crude oil below it) down to the production wells. It was originally estimated that oil recovery would be
close to 98%. However, early breakthrough of dry gas indicated that something was wrong with this estimate. A
geologic re-examination of the reservoir structure found a number of shale barriers, extending only part way across
the reservoir. These barriers provided surfaces where oil layers, surmounted by LPG layers, rested. These layers
would only slowly drain off by gravity flow. The rate of withdrawal at the bottom was higher than the rate of gravity
drainage off of these shale layers; as a result, the dry gas passed down through the openings between the incomplete
shale barriers and reached the wells at the bottom long before the oil and LPG had drained off of the partial barriers.
Re-estimation of the oil recovery indicated a value of about 65%. This problem of passage through partial shale
barriers has been investigated by Prats (1972) and by Richardson et al. (1978).
Another vertical downward miscible displacement is in the Intisar 103D field in Libya. This project has been very
successful; current production rate is about 40,000 barrels per day. There has been no mention of partial shale
barriers such as those in the Golden Spike case.
The majority of gravity-stabilized miscible floods in steeply dipping reservoirs have been in fault blocks next to salt
domes (though this is not the only type of steeply dipping reservoir). The Black Bayou LPG slug/dry gas project was
one of these, operated by Shell. It was designed to maintain a peak propane concentration in the middle of the slug
sufficient to maintain miscibility until the drive reached the wells near the bottom of the reservoir, taking into
account the dispersion coefficients on the top and bottom sides of the slug. Observations at intermediate wells
indicated that the dispersion was proceeding according to calculation. The displacement was operated slowly enough
that gravity stabilization prevented viscous fingering. The process was successful in the first fault block, and the
LPG slug was largely recovered. The slug was used again in a second fault block, and recovered most of the oil
there also. In a third fault block, the prior loss of slug material necessitated using some natural gasoline components
(C5 through C8 ) to make up the volume required. The amount of higher molecular weight components was not
sufficient to affect the miscibility on the back side of the slug.
In a similar flood in the Neale field, the dispersion behavior again took place as predicted. In this and the Black
Bayou case, the reservoir consisted of a sandstone layer that was fairly homogeneous, in contrast to the Golden
Spike case.
There have, of course, been many slug/dry gas projects carried out in horizontal or only slightly dipping reservoirs.
They have been fairly successful, recovering from 5 to 15% of the original oil in place above estimated or actual
waterflood recovery.
Enriched Gas Drives
In projects where the slug was driven by dry gas, it is difficult to distinguish between LPG slug/dry gas and enriched
slug/dry gas floods. This is because it was common practice to dilute the LPG slug with as much dry gas as possible
without losing miscibility; in this form, the slug became simply an enriched gas slug, followed by dry gas. However,
as WAG became commonly used, the WAG procedure of alternating slugs of water and gas was extended to include
these alternate slugs of water and enriched gas; when the design slug had been injected, the flood was continued
simply by continuing to inject water. This is the most common form of this process today.
The majority of these enriched gas slug/WAG/final waterflood projects are being carried out in western Canada.
There are a series of large reservoirs containing fairly light (low viscosity, low specific gravity) crude oils in Alberta
and consisting of bumps atop the Beaverhill Lake formation, which contains an aquifer common to all of them. Most
of these reservoirs are now undergoing hydrocarbon miscible floods using slug material containing a high proportion
of ethane as well as propane and butane. These intermediates are collected from most of the fields at a common
gathering point, and a pooled slug material is delivered back to the fields for injection. Dry gas, in varying
proportions according to each fields need, is blended into the slug at each field. These fields include Swan Hills,
South Swan Hills, Judy Creek, Ante Creek, Kaybob North, Goose River and Virginia Hills. Additional fields which
are being similarly flooded (though they are not part of the Beaverhill Lake group) are the Leduc, Acheson and
Redwater fields (all in the Leduc formation), the Pembina group in the Nisku formation, the Caroline and Willesden
Green fields in the Cardium formation, and several pools in the Rainbow field area of the Keg River formation.
Together, these hydrocarbon miscible floods account for five-sixths of Canadian EOR production. A nitrogen
miscible flood project is planned in the Carson Creek North field, which is a member of the Beaverhill Lake group.
CO
2
Floods
Shell operated a CO
2
pilot flood in a relatively small section of the "S" sand of the Weeks Island field. This is a salt
dome piercement type of reservoir structure, with fault blocks of sandstone layers surrounding the salt dome and
tilted up by the dome. There are many such layers, with the "S" sand fairly deep among them. It is one of the
thickest of the layers, and had a small gas cap and a long oil column above a bottom water aquifer. The natural water
drive had produced most of this oil column while the gas cap was kept intact.
The plan for the full-scale flood is to inject a slug of CO
2
followed by dry gas and drive the oil/water contact back to
its original position, while collecting part of the oil bank developed from the residual oil saturation as the bank
passes wells along the way.
Shell tested this plan in a small piece of the "S" sand which had been broken off and carried upward about 2,000 feet
by the salt dome as it rose. This piece was arc-shaped, with one end of the arc extending down much farther than the
other end. A gas cap also existed in the top of the arc, including all of the short end. The longer end contained only
oil and connate water (no bottom water). Water had been injected into the bottom of the long leg, and the oil bank
was produced by a higher well until the oil layer became rather thin. In the pilot test, a CO
2
slug was injected
through the upper well on top of this thin oil column while water was pumped back out of the well at the bottom.
This gravity-stabilized project performed as simulated, and most of the residual oil, plus the small oil bank which
was left at the end of the upward waterflood, was recovered as the water was pumped out, using a new well
completed about one-half to two-thirds of the way down plus the well at the bottom. This served as a pilot for the
main project in the much larger portion of the "S, sand lying deeper in the reservoir. For that project, CO
2
is being
obtained from Jackson Dome in Mississippi, via a pipeline across the Mississippi River. This CO
2
source is also
being used to supply the Little Creek, Olive and West Mallilieu fields in Mississippi.
Another steeply dipping CO
2
flooding project is in the Sims sand of the Sho-Vel-Tum field in Oklahoma, operated
by ARCO.
By far the largest group of fields undergoing CO
2
floods are in west Texas/eastern New Mexico, most of them in the
San Andres/Grayburg dolomite formations. These are relatively low permeability (1-10 millidarcys), multi-layer,
thick formations (300 feet or more). They often have a gas cap in their top-most layer, which is usually separated
from the other layers by a low-permeability shale layer. Most are underlain by bottom water, and in 1988-90 it
became known that an additional target for EOR is provided by a waterflood residual oil saturation in this bottom
water zone in some of the reservoirs; the residual oil saturation may be present for several hundred feet below the
original water/oil contact.
The fields mentioned above and a variety of others are listed in Table 1a ,


Table 1a


Table 1b , and Table 1c which includes all of the known CO
2
flooding projects in the U.S.


Table 1b


as of 1991.


Table 1c


Some CO
2
flooding in the early 1980s (a relatively small number of barrels per day) was done by using single-well
injection and backflow, similar to steam soak stimulation. The cost of CO
2
delivered in small batches is
considerably higher than that delivered by pipeline to large fields and, correspondingly, the cost of the extra oil by
this method was higher. When the sharp price fall for crude oil occurred in early 1986, most of these projects were
discontinued. New ones were being started in 1990 in Wyoming due to the availability of pipeline CO
2
from the
LaBarge/Big Piney CO
2
field.
There are some field-wide immiscible CO
2
floods: Chevrons Timbalier Field in Louisiana and Phillips Lick Creek
field in Arkansas. They are successful, too; the oil recovery is surprisingly similar to miscible projects, and the ratio
of CO
2
to extra recovered oil is equal to or lower than that of most miscible floods. The WAG ratio is about 1:1, but
because of much lower pressure, a much lower ratio of CO
2
to water measured as scf/bbl is equal to a 1:1 volume
ratio
There are three CO
2
flooding projects in Hungary, with CO
2
supplied from a deposit at a deeper level under one of
the fields (the Budala field), as well as a project in Turkey.
The subsurface deposit sources of CO
2
in the U.S. are listed in Table 2 .


Table 2



Laboratory Testing and Miscible Flood Design
Information from special laboratory tests is used principally for miscible flooding computer simulation, and is not
normally obtained for other purposes. This limitation on special tests notwithstanding, it should be emphasized that
data normally obtained for other purposes can be of great importance for miscible simulations.
Data Normally Gathered for Other Purposes
Miscible flooding simulation and project design requires data pertaining to the structure of the reservoir as obtained
(for other purposes) from seismic and well data, including
- shale barriers and their continuity (or lack of it)
- faults and their degree of continuity, both in the bedding plane and perpendicular to it
- porosity and permeability profiles at each well, including the location of perforations in cased holes
- production equipment, as it may affect bottom-hole pressure and limit injection or production rates
In addition to porosity and permeability data covering the range likely to be present in the reservoir, we require other
measurements, including water/oil and gas/oil relative permeability data, capillary pressure data from samples
representative of the major rock types (facies), and a determination of the most permeable, least permeable, and
average permeability layers encountered in coring some of the wells. At least one measured value of pore-space
compressibility would be helpful.
It is important to make an extra effort to secure core material that is as close as possible to the reservoir condition for
the relative permeability and capillary pressure tests. In the past, two coring methods have been used: pressure core
barrel and sponge-lined core barrel coring equipment, with either stock-tank crude oil or special low-surfactant
content aqueous drilling muds as coring fluid. Recently, a rapid-coring diamond drilling tool with a specially
designed low-invasion drilling fluid has become available; it seems to give core material that is unaltered, except for
depressuring effects and for the outermost 1/4 - to 1/2 -inch layer around the core, where some invasion by filtrate
still occurs. The wetting state of the inside of the core is believed to be very close to its condition in the reservoir.
The core loses pressure as it is brought to the surface, but if it is quickly and carefully wrapped at the surface with
plastic and foil and sealed against air invasion, small cores may be cut transversely through it in an inert atmosphere
at a core-handling laboratory. This can provide a series of cores (with the filtrate-invaded part cut off each end) on
which we may run tests for such measurements as relative permeability tests and capillary pressure.
This extra effort is important, because the wetting state of the rock is significantly changed by the old coring and
core handling and storage methods, in which the cores are deeply invaded by drilling mud filtrate, depressured and
then exposed to atmospheric oxygen over a long period of time. Efforts have been made to clean core samples
derived from old well cores, and then to restore the wetting state by driving reservoir crude oil through the sample
and then storing it at reservoir temperature and pressure for several weeks. Often, it is thought, the sample achieves
an intermediate state between water-wet and oil-wet approaching the original reservoir state, but there is no way to
be sure of this. For quite a few years, most oil reservoirs were thought to be strongly water-wet; it was also thought
that old cores could be restored to this state by using powerful cleaning agents and then heating the cores to about
1,000oF. The cores were then indeed strongly water-wet, but this does not mean that the rock was initially so in the
reservoir after millions of years of exposure to crude oil as well as subsurface water.
There are two common procedures for measuring relative permeability: the steady-state method, and the transient
method of Johnson, Bossler and Naumann. While the steady-state method is much more time-consuming and
difficult than the transient method, its results are generally more reliable if it is done properly. For the steady-state
procedure, the constant-volume apparatus described by Braun and Blackwell (SPE Reprint Series Book No 27,
Reservoir Characterization, Vol. 1) is recommended. Methods that do not preserve material balance, as this closed
system does, have more opportunity for error. With the constant-volume system, the attainment of steady state is
apparent from a constant level in a small-bore part of the phase separator, where the oil/water interface location
(visually determinable at low pressures, and obtained by pressure transducer difference between the top and the
bottom of the separator at high pressures) indicates how much oil and water is in the core. The centrifuge method is
recommended for capillary pressure curve determination.
When reservoir-state core material is used in these tests, the pore space is not known until after the tests, when the
core can be cleaned so as to obtain the porosity (e.g. by the Boyles Law method). The saturations can be calculated
after this is done.
Water composition is not significant (apart from its relatively small effect on the water viscosity and density) so far
as hydrocarbon solvent miscible flooding is concerned. However, the water salinity does have a significant effect on
the solubility of carbon dioxide in the water phase. It is therefore necessary to determine the total dissolved solids
content for estimating the CO2 solubility. Since this solubility is relatively low, a sufficiently accurate estimate of
the solubility of CO2 in the aqueous phase can be made with only this measure of water composition, using
correlations determined for water containing sodium chloride at varying temperatures and saturation pressures (see
Goodrich, 1980).
With regard to crude oil, it is again important to get as near as possible to the crude oil composition present in the
reservoir. It will probably be difficult to determine the dissolved natural gas content accurately. The major problem
here is that the reservoir crude oil will seldom be at its initial condition when a miscible flood begins. It will usually
have been produced until the reservoir pressure is well below the bubble-point pressure, and then water will have
been injected to raise the pressure. This pressure increase will have caused just some of the gas to redissolve in the
crude oil, with other gas being driven to producing wells as a free gas phase. The dissolved gas content of the crude
oil may therefore vary over the area and depth of the reservoir, making it necessary to simply adopt some average
value of dissolved gas content. Samples of separator oil and gas, taken from the exit lines of the separator at a
known pressure and temperature, can then be recombined in the laboratory at this average gas/oil ratio to obtain a
sample of current reservoir crude oil.
In the past, crude oil composition was generally determined by low-temperature distillation of such recombined
crude oil, using a device known as a Podbielniak column. This apparatus consists of a bottom kettle, heated by an
electric heating jacket, and a fractionating column rising to a liquid-air cooled condenser and reflux return system,
with a thermometer at the top.
In this distillation method, a high proportion of the condensed overhead product was returned to the top of the
column to flow down as reflux, and net product was taken from the top of the column only slowly. With this system,
the temperature curve versus top product volume would show definite steps corresponding to the content of
methane, ethane, propane, iso- and normal-butane, and iso- and normal-pentane. Of course, the liquid air cooling at
the condenser was required only for these lighter fractions; pentane and higher fractions were cooled using cold
water.
Beyond normal pentane, product volumes could be measured between the boiling temperatures corresponding to the
gaps between the groups of different isomers of the hexane isomers, and could have continued in this fashion to
higher carbon number groups. At the time the system was devised, however, it was considered sufficient to distill
only the hexanes overhead, then let the bottom kettle cool and measure the remaining amount of the original feed as
heptanes and higher boiling compounds (called C7+
).
In addition to the measured volumes of the hydrocarbon fractions, reduced to mole fractions and to weight fractions,
the compositional analysis also reported the density and molecular weight of the C7+
fraction. Originally, they were
measured independently, the density with a pycnometer bottle (net weight of the contents of a small glass bottle of
known volume, measured on an analytical balance), and the molecular weight by either the cryoscopic method
(freezing point depression of, for example, naphthalene due to a small known weight of added C7+
fraction) or the
ebullioscopic method (boiling point elevation of, for example, benzene due to addition of a small known weight of
C7+
fraction). Each of these measurements could be converted to a molecular weight using the known freezing point
depression or boiling point elevation due to one mole of another compound added to a mole of a specific compound
such as benzene or naphthalene. (This worked only when the added substance was a small fraction of a mole.) In
recent years, some laboratories have adopted the practice of estimating the molecular weight from the density, and
reading the density with a small hydrometer. This practice, however, is not very accurate.
The total laboratory analysis contained other important information, including
- the bubble-point pressure (at the reservoir temperature)
- the volume of the liquid phase versus pressure as the pressure was lowered in steps from the initial
reservoir pressure (or above) down to atmospheric pressure
- the volume of gas phase evolved at each step down in pressure, as well as the specific gravity, relative to
air, of each increment of gas.
These data were used to generate formation volume (or B0 ) factor curves showing relative crude oil volume (based
on the volume at atmospheric pressure and 60 F) versus pressure, and solution gas/oil ratio (or Rs ) curves showing
the standard volume of dissolved gas per standard volume of crude oil as a function of pressure. The viscosity of the
crude oil was also measured (typically with a rolling ball viscometer), during the stepwise pressure depletion, and
plotted as a function of pressure. The minimum viscosity on this plot occurred at the bubble-point pressure.
The compressibility of the crude oil could be calculated from the B0 curve slope above the bubble point. If
compositional simulation is being used, the calculated compressibility of the crude oil as determined from the
contributions of its components should check the value derived from measurement.
The pressure sensitivity of the viscosity could be calculated directly from the viscosity curve above the bubble point.
Again, for compositional simulations, the calculated pressure sensitivity of viscosity, as well as the viscosity as
calculated from the crude oil composition, should match this measured behavior. Also, the calculated depletion
behavior should match the measured depletion behavior, with respect to incremental volumes of evolved gas and its
gravity as well as oil relative volume, for each step in the pressure reduction.
This standard method of analysis persisted for about fifty years, and was known as a PVT analysis (Pressure,
Volume, Temperature). In recent years, the compositional analysis has been performed by chromatography.
Chromatography provides good data for the light hydrocarbon weight fractions and mole fractions. The method can
be extended to successively higher-boiling hydrocarbon fractions (denoted by the carbon number of the paraffin
hydrocarbon having its boiling point at the cut point between the fractions in the chromatogram) up to about C35.
This is done using a chromatographic apparatus that is capable of gradually raising the temperature of the
chromatographic column according to a pre-set program.
The shortcoming of this method is that most crude oils have hydrocarbon components extending up to C100 and
higher, and these components cannot be driven out of the chromatographic column without exceeding thermal
cracking temperatures; if this cracking occurs it not only impairs the analysis but also leaves a coke-like residue in
the column. Hence, no attempt is made to drive these higher hydrocarbons out of the column. Some of the higher
hydrocarbons can be driven back out the entrance end, and the amount measured with the same detector used at the
exit. The amount still remaining in the chromatographic column can be estimated by subtracting the total measured
products from the amount injected. Unless this difference is small, it is difficult to estimate the effect of this
unmeasured portion on the density and overall molecular weight of the crude oil. In this one respect, the low-
temperature distillation method was superior, in that the density and molecular weight of the total C7+
was measured
directly. For compositional simulation purposes, however, and in particular for matching of the phase behavior when
the combination condensing/vaporizing behavior occurs, the extended analysis provided by the temperature-
programmed chromatographic analysis method is almost essential. This is because the input for the compositional
simulator requires the use of at least three components heavier than C7+
-at least two of them heavier than C10, and
one heavier than C30. From the density and molecular weight of these fractions, other properties needed for the
compositional equation-of-state calculations can be estimated.
The C7+
analysis can be extrapolated to higher molecular weights by the method described in Lorenz, Bray and
Clark (1964). They set the upper carbon number limit at 40 in the integrations to match the mole percent and
molecular weight of the C7+
fraction; this is usually sufficient because the exponential at a carbon number of 40 is
very small.
The writer has often used only one constant in the exponential function used to extrapolate mole fractions, and then
it is easy to set an upper limit of 100 or even infinity. Then, the single constant "c" is obtained by:
C
n
= C
6
e-c(n - 6) , C
7+
= C
6
* = C
6
/c, c = C
6
/C7+
(7.1)
However, these procedures are still an extrapolation and are not likely to give mole fractions of each carbon number
as accurate as the data provided by the extended chromatographic method. The densities of the chromatographic
fractions must be estimated. This estimate can be done using a curve of density versus boiling point or carbon
number derived from a distillation of the crude oil (e.g., the data already provided by the U.S. Bureau of Mines
method for a large number of crude oils (see Coleman et al., 1978)). The calculated overall density should match
that measured for the total crude oil or for the C7+
fraction.
When a reservoir crude oil analysis has been obtained or calculated, an equation-of-state calculation computer
program should be used to simulate the pressure depletion behavior and to compare the calculated Rs and B0 curves
with the measured ones derived from the laboratory PVT analysis. If these do not match closely, either the
interaction parameters in the equation-of-state data input must be adjusted, or the analysis itself must be adjusted (if
reasonable interaction parameters do not help to achieve a match). It is not reasonable to proceed with compositional
simulations without these checks on the input data.
Data Specific to Miscible Flooding
Laboratory tests include miscibility determination, the oil remobilization curve, volume blending relationships
between solvents and crude oil, and CO2 solubility in water as a function of pressure and water salinity.
Minimum Multiple-Contact Miscibility Pressure
Procedures for estimating multiple-contact miscibility within 1200 psi, based on the simple C7+
analysis, include the
correlation for enriched gas drive solvent mixtures by Benham, Dowden and Kunzman (1960), and the correlations
for CO2
by Holm and Josendal (1974) and by Yellig and Metcalfe (1980).
The Benham, Dowden and Kunzman correlation consists of a series of separate plots for miscibility pressures of
1,500, 2,000, 2,500 and 3,000 psia. Each plot contains sets of parametric curves showing maximum C1 content of
the enriched gas mixture versus temperature. For 1,500 and 2,000 psia, there are two parameters on the curves: the
molecular weight of the LPG part of the enriched gas mixture, and the molecular weight of the C5+
part of the crude
oil. At 2,500 and 3,000 psia, separate plots are given for different C5+
crude oil molecular weights; the sole
parameter on the curves is then the molecular weight of the LPG.
The Holm-Josendal correlation for CO
2
assumes that the CO
2
is equivalent to a mixed solvent containing 59% C1
and 41% C3, and enters the Benham, Dowden and Kunzman correlation for this maximum C1 content in the series of
plots.
Yellig-Metcalfe present a correlation for CO2
which, for practical purposes, can be approximated by:
P
misc
=12.6*T (7.2)
where T is the temperature in degrees Fahrenheit and P
misc
is in psi. If this is less than the bubble-point pressure, the
latter should be used as the estimate of minimum miscibility pressure.
These correlations may be useful when no experimental data concerning the miscibility pressure are available.
However, it is much better to obtain experimental data by means of a slim-tube test, or by means of the rising bubble
method.
Slim Tube Test
The slim-tube test is really a program of tests at a series of pressures above and below the minimum miscibility
pressure. The slim tube is, in most instances, a coil of high-pressure stainless steel tubing packed with fine sand
(160-200 mesh). It is filled with recombined reservoir crude oil, and the oil is displaced with the desired solvent
while maintaining a back-pressure on the outlet at the specified level; the temperature is set at the reservoir
conditions. The oil recovery when the solvent breaks through at the outlet is determined by observing the fluid in a
sight-glass capillary tube at the end of the coil, before the back-pressure regulator, either by eye or with a photocell.
This recovery is plotted as a function of pressure. The oil recovery is low (60% or less) for immiscible
displacements, but as the minimum miscibility pressure is approached, the breakthrough oil recovery rises to a level
of about 96% for hydrocarbon solvents, or about 92-94% for carbon dioxide. Some experimenters determine the oil
recovery at 1.0 pore volume of throughput, while others use a throughput value of 1.2 PV. Oil recoveries are higher
at the higher throughputs. For LPG, the recovery approaches 98-99%, while for CO2, it rises to 95-96%. The lower
level for CO2 reflects the fact that CO2 leaves more insoluble asphaltic residue than propane or butane fractions do.
The transition region from immiscible to miscible recovery covers a wider range of pressure for CO2 than it does for
the hydrocarbon solvents.
Rising Bubble Test
The rising bubble method (Christiansen and Haines, 1987) offers a different way of finding the multiple contact
miscibility pressure. A bubble of the solvent is inserted under a water level in a high-pressure gauge glass; the
bubble rises into the stagnant crude oil above the water, and passes upward through the crude oil. Well below the
minimum miscibility pressure, the solvent bubble maintains its identity but shrinks gradually in size as it rises to the
top of the oil column. At or slightly above the minimum miscibility pressure, the bubble maintains a spherical top,
but the bottom interface soon vanishes (within 2-5 Cm); then the bubble disintegrates, with pronounced Schlieren
effects (varying refractive index) below, in the path of the bubble. Well above the minimum miscibility pressure, the
bubble smears out quickly, as it rises only 2-3 cm. It is possible to change the pressure level fairly quickly and thus
complete a series of tests at rising pressure levels in a relatively short time compared to a series of slim tube
displacement tests. This advantage was considered sufficient to justify a patent (Christiansen and Kim, 1986).
Results from this method were compared to slim tube results and showed good agreement.
The Oil Remobilization Curve
When solvent is injected into a previously waterflooded reservoir, it may remobilize waterflood-trapped oil. This
remobilization can be measured by displacing a tagged hydrocarbon oil from a reservoir core sample. The relative
permeability curves for the core will have already been determined, using water with a salinity similar to that of the
reservoir flood water. This displacement is continued until the residual oil saturation is closely approached. Then a
series of mixtures of untagged oil and the same water are injected, and the tagged content of the effluent oil is
measured until no more oil of significant tagged content emerges. This series should comprise mixtures of
increasing untagged oil content, until at last the injected fluid is solely the untagged oil. Both the tagged and the
untagged oil should have negligible solubility in the flood water.
The previously determined relative permeability curves allow the calculation of the water saturation corresponding
to the fractional flow of water in each of the mixtures of the series when the core has arrived at steady state. The
succeeding increments of displaced tagged oil can be added together, and the sum subtracted from the initial residual
oil saturation. This gives the remaining trapped oil saturation at each of the fractional flow levels and corresponding
water saturations. Of course, when the cores water saturation has been driven to the irreducible, or connate, water
saturation, all of the tagged oil should have been recovered. It is important that this material balance be achieved, to
verify that the data are correct.
The tagged and untagged oils should have close to the same viscosity. If a viscous untagged oil is used, it is too easy
to displace water, making it hard to get data at high water saturations. Conversely, if the viscosity is as low as that of
the actual solvent, it is difficult to drive the water saturation down as low as desired. I have used the pair heptane-
iso-octane. These hydrocarbons are readily available in their pure state and in useful quantities because they are
gasoline octane number standards. Tagging is unnecessary because the effluent hydrocarbons can be analyzed
chromatographically to determine the content of the initially trapped hydrocarbons.
A possible problem with using this pair of hydrocarbons is that they might cause a significant change in the wetting
state of the core. This can be checked by measuring the pressure drop through the core during each step of the
remobilization procedure and comparing these values with the pressure drop calculated from the previously
measured relative permeabilities at each fractional flow level, using the appropriate oil viscosity. If they are
different, it may be necessary to start over, with preserved-state core material and with reservoir crude oil, using
some method of tagging the reservoir crude oil which will comprise the initial waterflood-trapped oil saturation.
Of course, if it is found that heptane and iso-octane change the wetting state of the core, then practically any
miscible flooding solvent is likely to do so also. This requires determining the relative permeability curves for the
solvent/water pair as well as for the crude oil/water pair. Tagged crude oil as the initial residual oil should be
displaced in steps, with a series of mixtures of the miscible solvent with water. The solvent/water relative
permeability curves would then be used to calculate the water saturation corresponding to the fractional flow of
solvent and water comprising the mobile fluids at steady state. Again, it would be desirable to check calculated
pressure drops with measured pressure drops. In this case, however, there is the possibility that at the beginning of
using a higher fraction of solvent in the injected water/solvent mixture, the remobilization of some of the trapped
crude oil will cause the oil phase to have a higher viscosity than the pure solvent. This can result in driving the water
saturation down farther than intended, and farther down than will be the case after the solvent/water mixture reaches
steady-state fractional flow. Thus, some solvent may replace some of the trapped crude oil, and the amount of
originally-trapped crude oil which is released in each step may be larger than it should be. This would lead to an
optimistically low curve of trapped oil saturation versus water saturation. For this reason, the use of a light oil pair
such as heptane-isooctane is suggested, with the lower viscosity heptane as the initial trapped oil saturation.
As we can see from this discussion, determining the curve of trapped oil saturation versus water saturation is not
easy, and is open to uncertainties. Nevertheless, it can be an important factor in limiting oil recovery when WAG is
desirable for reasons of improving areal sweep efficiency, and it cannot be dismissed as unimportant without first
determining the curve and then determining by simulation that, with this measured curve, the oil recovery is not
reduced when WAG is used. This, of course, would be the case if nearly all of the waterflood-trapped oil is
remobilized by only a small reduction in water saturation.
If relatively few points are obtained for the curve, it may be useful to calculate curves with the modified Raimondi-
Torcaso (1964) equation, using different values of the wetting parameter a, and find a value of a which gives the
best fit to the data available. Then we can calculate a sufficient number of points to provide a table of values for
trapped oil saturation versus water saturation to use as simulator input data.
Raimondi and Torcaso hinted, and Shelton and Schneider (1975) concluded, that it should be possible to determine
the trapped oil saturation vs. water saturation curve directly from the difference in saturations at the same relative
permeability level for the drainage and imbibition relative permeability curves for oil phase. The reasoning is
impeccable, but it turns out experimentally to be wrong. The reason is probably that different pore size ranges
provide oil transmissibility at the same relative permeability level on both curves, hence the proportions of mobile
and trapped oil at the indicated water saturation on the imbibition curve may be such that the immobile part is not
indicated by the saturation difference between the two curves. Therefore, we should not suppose that their simple
method of deriving the trapped oil saturation curve can be used in place of experimental measurement.
Carbon Dioxide Solubility in Water
Goodrich (1980 has provided curves that involve temperature and water salinity parameters as well as pressure.
Interpolation (and sometimes extrapolation) is required for any given case. If a simulator sensitivity study shows a
significant effect of the water solubility of CO2 on the oil recovery at a planned CO2 slug size, then we may want to
check the value derived from Goodrichs curves by an experimental measurement at proposed reservoir conditions.
This is especially likely to be the case when the size of the planned CO2 slug is relatively low-for example, 10% of a
hydrocarbon pore volume-because the first few pore volume percent of the CO2 slug dissolve in water and have
relatively little ability to produce additional oil recovery. Thus, when the curve of extra oil recovery versus CO2 slug
size is projected toward zero oil recovery, it occurs not at zero slug size but at several percent above zero. The exact
amount of dissolved CO2 is then important, because it determines this intercept.
Volume Blending Relationships
We normally assume that volumes of hydrocarbons blend linearly, except in the case of the lightest hydrocarbons.
For the latter, experimentally determined liquid-phase blending densities (as given, for example, in the GPSA
Engineering Data Book ) are used. The resulting phase densities (volume for a given weight or for a given number
of moles) appear to be satisfactory. In the case of CO2 , linear blending of densities (and thus of volumes as well as
weights) in the oil phase seems to be relatively acceptable. Some engineers have had oil-swelling data determined
versus mole fraction of CO2 in oil/CO2 mixtures for their specific crude oils. This degree of swelling is important in
determining the effective degree of mobilization and the resulting mobile oil/solvent phase saturation and density.
The blending density of CO2 in water is markedly different from that of pure liquid or supercritical fluid CO2. It has
almost the same density as water when dissolved in it. There are some data which indicate a slightly higher blend
density than that of water. For most purposes, this slight elevation of the density can be neglected, and in aqueous
phase the CO2 density can be assumed to be the same as that of the water.
With regard to viscosity blending, the data are more sparse. For hydrocarbons, the quarter-power mobility blending
rule, based on volume fractions, appears to be adequate. In refinery practice, another viscosity blending method
which works well is to use the ASTM charts for plotting viscosity vs. temperature lines in which the viscosity scale
is strongly distorted-almost log-log-while the temperature scale is slightly distorted. On this type of chart, the higher
viscosity can be placed on the 100 F vertical line and the lower viscosity on the 200 F vertical line, and a straight
line drawn between these two points The 1000 difference on the temperature scale can then be treated as a volume
percentage blend composition scale, and the viscosity of the blend at a given percentage can be read from the
straight line. For a series of components, the total blend would have to be calculated as a series of blends, with each
previous blend combined with the next one in turn, until all components have been accounted for.
The quarter-power blending rule is easier. The general formula applies to any number of components with volume
fractions V
i
and viscosities
i
, with the viscosity of the mixture
m
, as follows:

m
= [_(V
i
/
i
1/4
)
-4
] (7.3)
This rule was used by Koval (1963), and by Todd and Longstaff (1972). Another rule which is suggested for
mixtures in Reid, Prausenitz and Poling (1987) is:

m
= Log-1(_f
i
Log
i
) (7.4)
in which f
i
was stated to be mole fraction of component i. However, checking a few cases shows using volume
fractions produces a result much closer to that obtained by the quarter-power rule, which is considered to be more
reliable.
For blends of CO
2
with hydrocarbons at elevated pressures, we generally use the quarter-power rule; however, the
viscosity of CO
2
/oil blends is often observed to be lower, for the same proportion of CO
2
, than blends of natural gas
with oil. It appears, however, that this is not so much an error in blend calculations using the pure, supercritical CO
2

fluid viscosity as it is an error in the natural gas/oil blend calculations caused by using the gas phase viscosity of the
natural gas component in calculating these blends. Just as the blending density of natural gas (or of its major
component, methane) is much greater than the gas phase density, the liquid phase blending viscosity of natural gas
appears to be much higher than the gas phase viscosity. A blending viscosity of methane somewhere between 0.1
and 0.4 cp seems to be about right, based on reservoir crude oil composition, and the measured viscosity of the blend
and the other liquid components.
Given these uncertainties in viscosity and density blending data, it may be appropriate to include a few key
measurements in a laboratory program aimed at providing the data for a miscible project design through computer
simulation.
A good source of data for CO
2
density and viscosity is the D.O.E report by Goodrich (1980) . It lists literature
sources for the data shown in the report. Figure 1 is a useful plot for obtaining a quick estimate of the properties of
CO
2
at a given oil reservoirs temperature and at the hydrostatic pressure due to the depth of the reservoir.


Figure 1


Laboratory Models of Sweep Efficiencies
It is possible to make laboratory model studies of displacement efficiency, using sand-pack or glass-bead-pack
models of, for example, a quarter of a five-spot well pattern. Experiments with such models must use either the
actual fluids for the field case, or else substitute fluids which have low vapor pressures and hence will not exert high
pressures on the top and bottom faces of the model. With actual field fluids, it is usually impossible to brace these
two faces of the model sufficiently to prevent bulging under pressure and the creation of a very rapid flow path next
to the surfaces. When substitute fluids are used, the mobility ratios of pairs of fluids must be matched. Their density
differences and the flooding velocity should be such that the viscous/gravity force ratio is as near as possible to the
field value at a similar spot in the well pattern. For miscible floods, it is also highly desirable that the ratio between
the rates of transverse dispersion and convective transport be made the same in the model as in the field for a
corresponding location in the pattern, since this ratio governs the extent of viscous fingering.
It is often impossible to match all of the dimensionless ratios of forces and rates to those in the field. in particular, in
these models it is very difficult to match the capillary/gravity force ratio to that in the chosen field prototype.
Nevertheless, it is possible to calculate the various ratios for the field case, and if any of them are far from a value of
1.0, it is not necessary to match them exactly to get behavior similar to that in the field.
Two studies at the University of Houston, performed under the direction of E.L. Claridge, employed just such a
model and experimental plan. The findings were subsequently presented to the SPE as "Optimum WAG Ratio vs
Rock Wettability in CO
2
Flooding" (Jackson, Andrews and Claridge, 1985) and "CO
2
Foam Flooding Performance
vs. Rock Wettability" (Lescure and Claridge, 1986). The experiments were actually performed with a light
hydrocarbon in place of CO
2
, but with the dimensionless ratios adjusted for the conditions of C0
2
/crude floods in
west Texas carbonate reservoirs. They apply equally well to hydrocarbon miscible flooding when the properties of
the reservoir, the crude oil and the solvent displacement conditions have the same ratios as those used in these
studies. It is worth emphasizing again that the physics of the process of miscible flooding does not depend on the
chemical nature of the oil or rock or solvent; it is instead their physical properties which affect the process behavior
(as long as there is no chemical reaction). If the viscosities, densities, etc., of the fluids and the porous medium are
such that the most important dimensionless ratios of forces and rates can be made the same in the model as in the oil
reservoir, then the model behavior can be a faithful replica of the actual field process. This can give insight into the
combined effects of controllable process design variables (for example, WAG ratio and slug size, or WAG ratio for
optimum performance in a reservoir of a given wetting state, whether oil-wet, water-wet, or somewhere in between).
The study on WAG ratio indicated that in a given, fairly homogeneous, layer of a low permeability reservoir (such
as the west Texas type of reservoir which served as the prototype), WAG operation is a detriment in a water-wet
system, but a help in a relatively oil-wet system. The study on the use of foam indicated that use of a properly
designed foam could improve oil recovery considerably as compared to results from using either straight slug
injection or WAG injection. Since a quarter-five-spot was used, the fluid mechanical effects we can expect of the
varying viscous/gravity ratio characteristic of flow in such patterns were present. Viscous fingering and gravity
layover occurred to the extent that they should. In tertiary floods, waterflood-trapped oil was remobilized and
pushed out as an oil bank (not during slug injection, up to a limited slug size, but during the subsequent water drive).
When oil broke through at the exit of the model in tertiary floods, solvent was always present with it; both solvent
and oil flowed out together during production of the extra oil due to solvent injection. This behavior has been
observed in field CO
2
floods, and in hydrocarbon miscible floods also.
An article by Claridge, Lescure and Wang (1988) shows that computer simulation is able to match the process
behavior described in both of the model studies mentioned above. At several SPE symposia (1988-90),
representatives from major oil companies have reported that the field behavior observed thus far has been very
similar to their computer simulation predictions. Hence, it appears that computer simulation with good input data
can give rather good performance predictions for miscible flooding. This is, of course, encouraging both to technical
staff and to company executives-particularly the latter, since they have had to make large economic decisions based
on the predictions of their technical staffs, who have, for the most part, relied on computer simulation.
It is, however, important to emphasize that good input data are vital to making good predictions: hence, the
particular need for the data derived from reliable laboratory tests.

Cross-Sectional Models of Vertical Sweep Efficiency
As a matter of historical record and as a means of demonstrating the serious curtailment of vertical sweep efficiency
by heterogeneity among layers, the early methods of Stiles (1949), of Dykstra and Parsons (1950) and of Dietz
(1953) are worthy of note. Because they are two-dimensional vertical slice models and make a variety of other ideal
(commonly unreal) assumptions, they are not suitable methods for calculating oil recovery curves for a real case.
Stiless Method
The simplest and first layer-model approach was that of Stiles (1949). The method is also described in Craig (1971).
A more complete description is given in Craft and Hawkins (1959). The basic assumption in the calculations is that
the division of fluid injection between layers is on the basis of each layers fraction of the total k
h
. The |
h
of each
layer determines the pore space in each layer, so that breakthrough in a given layer is obtained when the product of
rate and time equals the pore space in each layer (may be modified to read "movable saturation times pore space").
The writer has converted Stiles original equations into a set of dimensionless recursion equations for calculating the
dimensionless injection N
i
, oil production N
p
, and water cut f
w
(the calculations were originally intended for water-
flooding applications). These are given below In the original method, the porosity was assumed constant regardless
of permeability variation, but in the equations which follow it is assumed that the porosity varies:

for j = 1 to n. (A.95)


for j=1 to n (A.96)

for j = 1 to n (A.97)
We should also be able to obtain the oil recovery curve from the more fundamental relationship:
Np = (Np)bt+ *dNi (A.98)

Translated into discrete steps of N
i
this is:
(Np)j = (Ni)1 + * ANi for j = 2 to n (A.99)
However, this does not give the same results as Equation A.96, and the result is a very poor material balance. The
reason is the inconsistency in Stiles method of not using the mobility ratio when calculating Ni or Np, but using it
when calculating the water cut fw .
Dykstra-Parsons Method
The reservoir model here is the same as for Stiles method. However, the injection into layers is based on a changing
resistance as the front moves through each layer. It is assumed that the saturation changes at the front from
maximum oil saturation ahead to minimum oil saturation behind. Thus, the relative permeabilities at the end-point
saturations are assumed to apply, as well as the fluid viscosities, as expressed in the mobility ratio relation: M =
(k
rw
/k
ro
)/(
o
/
w
). The recursion equations for this case are:



for j=1 to n (A.100)


for j = 1 to n (A.101)

(A.102)
Equation A.98 still applies here. The material balance is better in this case, but is still not exact because the value of
w is calculated at intervals rather than continuously.
C. E. Johnson (1956) gives charts which are reproduced in Craigs monograph (1971) as Figures 8.1 to 8.4 in
Chapter 8. It is easy to calculate an oil recovery factor for a given value of the Dykstra-Parsons and for an
endpoint mobility ratio M, for four values of water/oil ratio (1, 5, 25 and 100), using the charts. However, the charts
do not enable the calculation of an oil recovery curve versus throughput, as the above method does.
Besides the failure to account for crossflow between layers, these two methods give an approximation to vertical
sweep efficiency which applies only to a cross-section, which might be construed to apply approximately to a line
drive, but not to more common patterns such as the five-spot and nine-spot patterns. As has been observed earlier,
the flow in well patterns is more nearly approximated by radial flow rather than linear flow.
Dietzs Method
The purpose of the Stiles and Dykstra-Parsons methods is to calculate the effect of layers of different permeability
on vertical sweep efficiency at any given throughput. Dietzs method serves a different purpose: calculating the
effect of gravity tonguing on vertical sweep efficiency, within a layer of uniform permeability.
The basic reference on this subject is Dietz (1953). In this section, I have converted Dietzs dimensional equations to
dimensionless form (not previously available in the literature). Note that Dietz used the symbol "a" for mobility
ratio, and that at that time the mobility ratio was widely defined as the reciprocal of mobility ratio M as presently
defined.
Using M = 1/a and N
i
= q t/(h LW AS
w
), with W = width = unity, and making x and y dimensionless by dividing
by L and h respectively, we obtain from equation (21) in Dietzs article the following equation:
(A.103)
Dietz made the judgement, based on limited calculations, that the second term on the right in the above equation
would be unimportant in comparison to the first term for cases of commercial interest (at high flooding rates, where
the ratio of viscous to gravity force N
v
/
g
would be large and its reciprocal small). Casting out the second term on the
right, which is a second order partial differential term, it is possible to integrate directly, because the first term on the
right does not contain either of the variables on the left. We obtain, as the equation for the shape of a gravity tongue
in a horizontal reservoir,
(x/L) = M N
i
/[1 + (M - 1) (y/h)]
2
(A.104)
The similar equation for a tilted reservoir may be obtained in the same way from Dietzs equation (19). As Dietz
noted, this equation does not contain the gravity force; it is present only in the ratio Nv/g in the second term, which
was discarded.
In Equation A.104, (x/L) is the horizontal dimensionless coordinate of a point on the interface between maximum oil
saturation and maximum water saturation (having the dimensionless vertical coordinate (y/h)), after injecting N1
movable pore volumes of water, at a mobility ratio M = (k
rw
/
w
)/(k
ro
/
o
), using the endpoint relative permeabilities
krw and kro. As written, the equation could describe a gravity tongue of an advancing higher-density fluid if (y/h) is
taken positive upward, or of an advancing lower-density fluid if (y/h) is taken as positive downward.
In Dietzs derivation, there is no flow upward or downward in forming the gravity tongue. The tongue forms
because, with an injected fluid of higher density than the fluid in place, the pressure difference between the injection
well and the production well is greater at the bottom of the formation than at the top, with proportional pressure
differences in between. The injected fluid thus moves more rapidly across the bottom of the formation than it does
across the top. If the mobility ratio is greater than one, the fluid accelerates as it moves, because it occupies an
increasingly high proportion of the distance between injector and producer. Hence, the tongue shape is not a slanted
straight line from top to bottom, as it would be if it were determined by the pressure difference alone.
Note that Equation A.104 is not valid for a mobility ratio of one, for then the equation would state that the frontal
location (x/L) is independent of height (y/h). This is inconsistent with the idea of more rapid movement across the
bottom due to a greater driving force
Equation A.104 cannot be valid when M is less than one either, for then we would obtain a tongue which is farther
advanced at the top than at the bottom.
Clearly, the integrated equation has nothing to do with fluid densities out in the formation. Rather, it has to do only
with the relative rate of advance of the fluid as a consequence of the mobility ratio, when it enters preferentially, for
whatever reason, either at the top or bottom of a homogeneous formation. Only the second term in the partial
differential equation concerns the gravitational force on the injected fluid out in the formation, and that term was
discarded in order to obtain a solution. Nevertheless, model studies have found that the integrated equation properly
describes the observed shape of a gravity tongue for mobility ratios greater than one, and for rates such that
significant vertical fluid motion due to fluid density differences does not occur in the formation during the time of
displacement (i.e., at relatively high rates). These are important limitations of Dietzs derivation and of his
integrated equation.
Continuing further to examine the consequences of the integrated equation (for M>1), we can see that the ratio
[(x/L)/Ni] expresses the abscissa of a given (y/h) value relative to the mean frontal position. This mean frontal
position is the integral average value of (x/L) over the total height h. This may be shown, for example, if we set Ni =
1.0 and proceed to integrate (x/L) * d(y/h) over the range from 0 to 1:
(A.105)
Since the average position of (x/L) represents the average extent of invasion of the system by the displacing fluid, it
should be equal to the movable pore volumes (or fraction of a movable pore volume) injected. Consequently, the
shape of the curve of the interface for a given value of M is of constant form relative to the average value of (x/L).
As Ni increases, the tongue shape simply stretches out in length as the average length of the invaded zone increases.
This is similar to the stretching of the curve of water saturation versus dimensionless distance in the Buckley-
Leverett solution of the simultaneous fractional flow problem in a linear system.
Equation A.104 leads directly to an equation for production after breakthrough, on the basis of the advance of the
point of the tongue (y/h = 0) to the end of the system (x/L = 1), at breakthrough, or past the end of the system after
breakthrough. The throughput required for breakthrough is obtained by setting y/h = 0 and x/L = 1, and it is seen
that Ni = 1/M.
Sweepout is obtained by setting y/h = 1.0, x/L = 1.0, at which point it is seen from Equation A.104 that Ni = M. For
intermediate throughputs, we can get the oil production, Np, as follows:
At breakthrough, Np = Ni = 1/M, and after breakthrough is given by 1/M plus the integral of the oil fractional flow
out of the production well times dNi. For x/L = 1, the oil fractional flow is equal to
Np = 1/M + (A.106)
where
1 - y/h = fraction of well height occupied by oil;
y/h = fraction of well height occupied by water;
M = mobility ratio
but
y/h = for x/L = 1 (A.107)
Substituting this for y/h in the preceding equation, we get:

N
p
= 1/M + (A.108)
Equation A.108 is the same as that obtained by Koval (1963) for viscous fingering of miscible fluids in a linear
system, except that Koval replaces M by an "effective mobility ratio" K (M
e
is used instead of K here to avoid
confusion with other uses of the letter K). Equation A.108 is also essentially the same as that obtained by van Meurs
and van der Poel (1958) for viscous fingering of two immiscible fluids, except that they allow for an apparent
stagnant fraction of the cross-section such that one minus the stagnant fraction is labeled b, so that the right hand
side is:
(A.109)
Obviously, if there were no stagnant fraction, then b = 1 and Equation A.109 becomes the same as Equation A.108.
We remarked above that the second term on the right hand side of Equation A.103 was discarded. It would seem that
we would need this term if we are really to see any gravity and flow-rate effect (i.e., effect of the viscous-to-gravity
force ratio N
v
/
g
). However, it is not obvious how it can be solved analytically along with the first term. Dietzs
method of approximation was to first assume only the first term and obtain a relationship between x/L and y/h, and
then use this expression to perform the double partial differentiation in the second term. This gives for the second
order partial differential:
(A.110)
Substituting Equation A.110 into Equation A.103 and integrating gives:

(A.111)
However, adding this approximate second order term has upset the material balance expressed by
(A.112)
The integral of the second term is .
At low values of Ni , this term is several orders of magnitude greater than the first term. This method of
approximating the second order term is therefore not satisfactory.
Another approach is to re-normalize the integral of (x/L)d(y/h), using Equation A.112 for x/L so that it equals Ni.
The result is curves of (y/h) versus (x/L) which, instead of being monotonically concave upward are like an inverted
integral sign. The bulge increases toward vertical as the value of N
v
/
g
increases. This reflects the tendency of a high
value of N
v
/
g
to make the front more vertical in shape, no matter what the mobility ratio. This is delineated more
clearly by the work of Hawthorne (1960). He shows that the tilt of the interface is a direct function of the viscous-
gravity force ratio.
Craig, Sanderlin, Moore and Geffen (1957) give results of an experimental study in a vertical-slice sand-pack model
in which the breakthrough sweep efficiency was plotted against the viscous/gravity force ratio, based on the
viscosity of the displaced fluid, which was oil. The resulting curves show a spread from the lowest mobility ratio,
with high sweep efficiency at low N
v
/
g
, to a mobility ratio of 1.0 reaching a high value of sweep efficiency only at
values of N
v
/
g
approaching 100. The curves for still higher mobility ratios never did get to high values of sweep
efficiency at breakthrough at N
v
/
g
values approaching 100.
If the value of N
v
/
g
were based on the viscosity of the invading water phase instead of the oil phase, all of the points
for mobility ratios up to 1.0 fell nearly upon each other. For mobility ratios above 1.0, the curves still fell below this
single curve. Furthermore, the curve for M = 1 so obtained could be fitted by equations derived by Gardner, Downie
and Kendall (1962). Their Equation 3, with M = 1 and the transverse flow parameter (ky/kx)*(L/h)
2
= 10, gave a
good fit to the Craig et al. curve for M = 1 plotted against N
v
/
g
basis water viscosity. Their Equation 2, with M = 1
but without the transverse flow parameter, did not give quite as good a fit, lying somewhat below the combined
curve of the Craig et al. data for M = 1. The observed curves for M > 1 lie below the curve for M = 1, as mentioned
above. This is because viscous fingering occurs for M > 1, and early breakthrough of the fingers occurs even at high
values of the viscous/gravity force ratio.
This is shown by van der Poel (1959). He gives a series of pictures of displacements in a vertical slice model, where
a gravity tongue of light solvent is seen at the top of the model at low injection rates (low v/g ratio); but as the rate
was increased, first a single finger and then a second finger appeared below the gravity tongue. Pozzi and Blackwell
(1963) show, by experimental work at low to medium values of the transverse flow or mixing group (compared to
the value of 10 which gave a good fit to the Craig et al. data), that viscous fingering then predominates over gravity
tonguing even at low values of the N
v
/
g
number, and then breakthrough sweep efficiency is higher than that due to
the gravity-dominated curve of Craig et al. That is, viscous fingering under such circumstances gives higher sweep
efficiency than does gravity tonguing, though both are low. Viscous fingering also sets an upper limit to
breakthrough sweep efficiency that depends on mobility ratio, for M> 1.
The data discussed are partly from immiscible displacements, and partly from miscible displacements. The
assumption in using such data together is that the capillary force is not very strong relative to gravity and viscous
forces in the immiscible displacements. Capillary force is a dispersive or mixing force in immiscible displacements,
and can negate by capillary rise much of the tendency toward a gravity tongue. This was not true for the cases
discussed here. However, the largely segregated flow due to gravity tongues or viscous fingers leads to Equation
A.106 as the recovery equation, compared to the Buckley-Leverett fractional flow solution for non-segregated flow
that would apply for strong capillary force relative to viscous or gravity forces. There are often reservoir situations
where largely segregated flow of immiscible phases occurs, either as a gravity tongue or as viscous fingers, and it is
important to know when this applies and hence what the governing equations are (i.e., those given above).
Vertical Sweep Efficiency from Dietzs Equations
The vertical sweep efficiency within a layer is directly measured by the ratio of N to Ni as obtained from the above
equations. It is only meaningful at and beyond breakthrough. At breakthrough, the vertical sweep efficiency is 1/M.
After breakthrough, it is equal to Np as obtained from Equation A.109 for a given M and throughput Ni , divided by
Ni. Since a linear flow system is assumed, this vertical sweep efficiency applies to linear flow, and only to a
homogeneous formation.

Immiscible Flood Calculations
This Section introduces the relationships used to describe immiscible displacements, beginning with a discussion of
fractional flow.
Derivation of the Buckley-Leverett Equation
Taking water as the component to be described by the continuity equation (the equation for the oil phase is
redundant since there are only two components), we have as the continuity equation:
A
|
[(oS
w
/ot)]x + q[(of
w
/ox)] t= 0 (A.37)
where
A = constant cross-sectional area of system perpendicular to x,
|= porosity (constant)
o= partial differential sign
S
w
= water saturation (a dependent variable)
t = a given value of time
x = partial derivative preceding this sign is for a fixed value of x
q = total flow rate through system (constant in both space and time)
f
w
= volume fraction of total flow which is water phase
x = a given value of length in the x-coordinate
t = partial derivative preceding this sign is for a fixed value of t
The coefficients of the partial derivatives may be removed by a change in the independent variables. We let T =
qt/A|L and X = A|x/AL = x/L, where L is the total length of the system in the coordinate x.
Equation A.37 then becomes:
(oS
w
/oT)X+(of
w
/oX)T=0 (A.38)
Since f
w
varies with Sw as well as with X and T, the chain rule allows us to write:
(of
w
/oT)S
w
= (of
w
/oX)T * (oX/oT)S
w
+(of
w
/oS
w
)T* (oS
w
/oT)X (A.39)
However, at a given value of Sw, fw is fixed and so the partial derivative on the left side of Equation A.39 is zero.
Substituting Equation A.38 in Equation A.39 and dividing by the derivative common to both terms, we obtain:
(oX/oT)S
w
= (of
w
/oS
w
)T (A.40)
Since f
w
depends only on S
w
, we may write the term on the right side of the equation as a total derivative, (df
w
/dS
w
).
Since the partial derivative of X with respect to T is at a particular value of S
w
,we must evaluate the derivative
(df
w
/dS
w
) at the same value of S
w
. For a given value of S
w
, however, the derivative (d
w
/dS
w
) is a constant, so the left
side of Equation A.40 is equal to this constant. We may therefore separate the variables in the partial derivative and
immediately integrate the equation:
X = (df
w
/dS
w
) * T for any given value of S
w
(A.41)
This is the general solution to the problem.
Note that the change of independent variables resulted in a variable X, which is the fraction of the pore volume of
the system traversed by a water saturation S
w
when a given fraction of a pore volume T of a certain fluid has been
injected. For a specific system, we must be able to determine the derivative (df
w
/dS
w
) for all possible values of S
w

The Fractional Flow Equation
To derive an equation for f
w
, we write Darcys Law for each phase:
q =k
rw
* k * A * grad u/
w
(A.42)
q
0
= k
ro
* k * A * grad u /
0
(A.43)
where grad u is the gradient of the potential u in the x direction (the potential u = p+Agh). By definition, q = q
w
+
q
0
; and we define the fractional flow of water as:
f
w
= q
w
/q (A.44)
Substituting for q
w
and q in the last equality, we get:
f
w
= 1/[1 + (k
ro
/k
rw
)(
w
/
o
)] (A.45)
Both k
rw
and k
ro
are unique functions of the water saturation Sw; these curves therefore reflect the properties of the
porous medium and not the properties of the fluids. The viscosities, on the other hand, are properties of the fluids,
and not of the rock. The fractional flow depends on both the ratio of the relative permeabilities and the ratio of the
viscosities. For a given displacement, however, the fluid viscosities are assumed to be constant, while the relative
permeabilities vary as the water saturation varies
A more general treatment of f
w
expresses the potential u separately for each phase as the sum of the pressure and the
head Agh for that phase, with the difference in phase pressures equal to the capillary pressure pc and the difference
in phase densities equal to A. This treatment gives the following expression for f
w
:
f
w
= (1+C+G)/[1+(k
ro
/k
rw
)(
w
/
o
)] (A.46)
where
C = grad p
c
/[q * o/kro * k * A] (A.47)
G = A* g * grad h/[q *
o
/k
ro
* k * A] (A.48)
Note that p
c
is a function of S
w
. The gradient of height, h, may be expressed as the sine of the dip angle of the
formation (which may be assumed constant). However, k
ro
is present in both C and G, so this generalized expression
for fw is much more complex than the simpler expression of Equation A.45 in which we neglected the capillary
pressure difference and tilt of the system. We will use the simpler Equation A.45 to illustrate further the behavior of
a waterflood, but it should be remembered that it is possible to take both of these effects into account; the general
solution still holds. From Equation A.45, then, we may write the derivative f
w
(= df
w
/dS
w
) as follows:
f
w
=-f
w
2
* (
w
/
o
) * [d(k
ro
/k
rw
)/dS
w
] (A.49)
It is convenient to express k ro and k as normalized exponential functions of water saturation S. We may write, for
example,
S=(S
w
-S
wc
)(1-S
wc
-S
orw
) (A.50)
k
ro
= [(k
ro
)S
wc
] * (1-S)
a
(A.51)
k
rw
= [(k
rw
)(1 - S
orw
)] * S
b
(A.52)
The constants a and b are empirical constants which enable fitting experimental data with Equations A.51 and A.52.
With these two equations, we may obtain the derivative (df
w
/dS
w
) analytically for any permitted value of S
w

(between S
wc
and 1-S
orw
, corresponding to S = 0 and S = 1).
The Welge Diagram
The 1w curve generally assumes an "S" shape between the limits of S
wc
and 1 - S
orw
. Over this same range, the slope
df
w
/dS
w
is initially zero. It rises to a maximum, and then decreases again to zero (or very near to zero). Therefore,
since the dimensionless velocity X/T of a given saturation is equal to df
w
/dS
w
, pairs of S
w
values will have the same
velocity. This suggests that if the system were initially at a low water saturation say S
wc
then both a slightly higher
water saturation and a much higher water saturation would move at the same velocity. However, a water saturation
near the middle of the saturation range would move even more rapidly than these extreme saturations, and would
reach the end of the system sooner.
This dilemma was resolved in discussions with the mathematician Von Neumann, who pointed out that an analogous
situation arises in the movements of bodies through a fluid at high speeds: a shock front occurs with a pressure
which moves more rapidly than all other values of pressure. In the case of the saturation front, there is a saturation
which overtakes all lower saturations. It may be found by drawing a tangent, which represents the shock front, from
this saturation on the f
w
- S
w
curve to the initial condition in the system.
For a waterflood starting at S
wc
,the tangent is drawn from the curve to the point f
w
=0, S
w
=S
wc
. For an oil flood
starting at 1 - S
orw
the tangent is drawn from the curve to the point f
w
= 1, S
w
= 1 - S
orw
. If the initial condition is any
intermediate value of f
w
on the f
w
curve, the tangent is drawn from the f
w
curve to that initial condition. This
assumes, of course, that it is possible to draw such a tangent.
If the injected fluid is water, and the initial condition is such that no tangent can be drawn from there to a higher
water saturation, then there is no shock front and the water saturation just gradually increases toward the limit
1 - S
orw
at the exit of the system. In such a case, however, it may still be possible to determine when the injected
water (not the connate water) will reach the exit of the system.
A tangent from the f w curve to f
w
= 0, S
w
= 0 gives the dimensionless velocity, if the tangent point on the f
w
curve
is at a higher f
w
than the initial condition of the system; otherwise this dimensionless velocity is the same as that of
the water at the initial state. The dimensionless velocity is the reciprocal fluid volume, expressed as a fraction or
multiple of a pore volume, required for the condition having that velocity to move from the entrance to the exit of
the system.
Welge also showed that when a waterflood shock front reaches the exit of the system, the average water saturation
in the system is found by extending the waterflood tangent to the top axis of the fw diagram (i.e., at f
w
= 1, S
w
=
S
awg
) Subtracting the initial water saturation from this average water saturation gives the saturation fraction of oil
which has been displaced. He showed that the average water saturation may be found similarly for any tangent to the
fw curve lying at a higher fw than the water breakthrough value by extending that tangent to the top axis. Thus,
choosing a series of points on the fw curve, and extending the tangents through those points to the top of the
diagram, gives a series of average water saturation values and the corresponding displaced oil. The water throughput
in pore volumes at each point is equal to the reciprocal of the slope of the tangent. Hence, it is possible by Welges
method to obtain and plot a curve of oil recovery versus water throughput.
It is also possible to plot water saturation versus the fractional length X of the system using the general solution: X =
(df
w
/dS
w
) * T where T is the fraction or multiple of a pore volume of fluid which has been injected. The derivative
(df
w
/dS
w
) is evaluated for a series of values of S
w
for which the product (df
w
/dS
w
) * T is less than or equal to 1.0, and
the value of fw is equal to or greater than the shock-front value (at the tangent point for the tangent drawn to the
initial condition).
Note that it is just as possible to calculate oil displacement of water as it is to calculate water displacement of oil.
Since laboratory core floods-in which capillary pressure end effects and gravity layover do not occur to a significant
extent-are fairly well represented by the linear system assumed in these derivations, the relationships given above
may be used to calculate what is going to happen during the core floods, once the relative permeability curves have
been determined.
For instance, it is possible to calculate how many pore volumes of water must be put through the core to arrive
within a given degree of closeness to the ultimate waterflood residual oil saturation, or to arrive at a given water cut.
Similarly, we can calculate the amount of throughput of oil phase required to obtain an average water saturation in
the core within a given degree of approach to connate water saturation.
Actual effluent from cores tends to be in the form of discrete drops when approaching the end point saturations, so
that for a long time there will be none of the displaced phase exiting the core; then a drop will appear that represents
the gradual accumulation at the effluent end of tiny amounts of the displaced phase. This discrete production
behavior is due to the small size of the system and would not happen on a large, reservoir scale. It is thus hard to
decide, on the basis of the effluent stream from a core, when a given waterflood or oilflood has gone on long enough
to achieve a desired state of water or oil saturation. It is better to use the Welge calculation to determine the
throughput required and to be satisfied when this is completed, rather than to watch the effluent stream and try to
decide on the basis of this observation. Of course, this requires that good quality relative permeability relationships
be established first. Since such good quality relative permeability data are needed for computer simulation purposes,
it is doubly important that these data be obtained. Some examples of this type of calculation, applied to laboratory
floods, are given below.
Example: Waterflood Calculations from Fractional Flow
This example is described in terms of a problem presented to a student for solution, with the resulting work shown in
response to the instructions.

S
wc
=0.2 S
orw
= 0.4
(k
ro
@S
wc
) = 0.8, ([k
w
]1-S
orw
)= 0.20

w
=0.7cp
o
=2.8cp
a = b = 2 (exponents in the Corey-type relative permeability equations)
Calculations are to be made analytically, using sufficient digits to obtain at least 4-digit accuracy for final calculated
quantities (round off only after all calculations are finished).
Detailed tasks are as follow:
Task 1
Starting the waterflood from the initial condition S
w
= S
wc
, find the values of S
wt
and f
wt
(subscript t indicates at
tangent point). Locate this point on a plot of fw versus Sw and draw the tangent. Call this plot Figure 1 .


Figure 1


The solution is as follows:
f
w
= 1/{l + [(0.6 - S
w
)/(S
w
- 0.2)]2},
(df
w
/dS
w
= f
w
2
* 0.8 * (0.6 - S
w
)(S
w
- 0.2)3 = f
wt

(the prime indicates a derivative).
The derivative of the fractional flow curve at the tangent point, which is the slope of the curve at that point, is set
equal to the slope of the tangent:
(df
wt
/dS
wt
) = f
wt
2
* 0.8 * (0.6 - S
wt
)(S
wt
- 0.2)
3
= (f
wt
- 0)(S
wt
- 0.2)
Substituting for f
wt
from the first equation and solving for S
wt
, we get
S
wt
=0.4828427125
f
wt
= 0.8535533906
(df
wt
/dS
wt
) = 3.017766953
See Figure 1 for the fractional flow curve and the tangent.
Task 2
Find the oil recovery O and water input W
i
=T at water breakthrough. Convert these values to N
p
and N
i
values by
dividing by the movable saturation range (0.4 in this case). Ten calculate values of throughput Wi and of oil
production Op for values of S
w
at which the water fractional flow (water cut) is 0.90, 0.95 and 0.98. These cover the
range of water cut used in the oil field as economic limits for continuation of waterflooding. Convert these values of
W
i
and O
p
to values of N
i
and N
p
. Plot the curve of oil recovery N
p
versus throughput N
i
in movable pore volumes as
Figure 2 .


Figure 2


The solution of Task 2 is as follows:
The average water saturation in the system, S
wav
, is equal to S
wt
plus (1 -f
wt
)/f
wt
= 0.5313708499. Then W
i
(equal to
Op at breakthrough) is equal to this average water saturation minus the water initially present, S
wc
: 0.5313708499 -
0.2 = 0.3313708499. As a check on this, it should also be equal to the reciprocal of ft: 1/3.017766953 =
0.3313708499. Then (N
i
)bt = (N
p
)bt = 0.3313708499/0.4 = 0.8284271248.
For f
w
= 0.90, S
w
= 0.5, f
w
= 2.4:
S
wav
=0.5 + (1 - 0.9)/2.4 = 0.541666667
O
p
= 0.3416666667
W
i
= 1/2.4 = 0.4166666667
N
p
= 0.8541666667
N
i
= 1.041666667.
For f
w
= 0.95, S
w
= 0.5253578013, f
w
= 1.564724736:
S
wav
= 0.5573123034
O
p
= 0.3573123034
W
i
= 0.6390900438,
N
p
= 0.8932807586, N
i
= 1.597725109
For f
w
= 0.98, S
w
= 0.55, f
w
= 0.896:
S
wav
= 0.5723214286
O
p
= 0.3723214286
W
i
= 1.1116071429
N
p
= 0.9308035714
N
i
= 2.790178571
See Figure 2 .
Task 3
For a water input W
i
= 0.3 (N
i
= 0.75) calculate points on the curve of S
w
versus X (= x/L) and plot it as Figure 3 .


Figure 3


The solution is as follows:
(x/L) = f
w
* T = f
w
* 0.3
A series of values of x/L vs S
w
is as follows ( Table 1 ):


Table 1


Polymer Flooding
Displacements with polymer solutions involve an immiscible displacement of oil by connate water, after which the
connate water is miscibly displaced by a polymer solution; at the same time there is continued immiscible
displacement of oil. These displacements may be calculated by a procedure similar to that for waterflooding but with
certain additions:
1. Polymer is adsorbed on the rock surface from the polymer solution, and the adsorption is almost completely
irreversible. It occurs quickly relative to the rate of flow, and thus takes place at the polymer solution front
as it moves through the system. Adsorption depletes polymer from the polymer solution, causing a certain
volume of water to appear at the front-this water is polymer solution that has lost its polymer content. This
volume or bank of water is usually assumed to have the same viscosity as the connate water or the flood
water which was injected ahead of the polymer solution. By the time polymer solution which has not lost
its polymer con-tent reaches the exit, a pore volume fraction b of polymer solution which has lost its
polymer content will have already exited the system. Compared to injected fluid, therefore, the appearance
of polymer solution at the exit is delayed by the pore volume fraction b.
2. Polymer does not invade all of the pore space that can be invaded by water. This is attributed to the large
size of the polymer molecules. Because of this, there is an excluded fraction b of the pore volume that is
not invaded or contacted by polymer, and the polymer solution therefore appears to have only a lesser pore
volume to flow through. This causes the polymer solution to appear sooner at the exit, by a throughput b in
pore volume fraction units. Thus, b and b are counter-acting effects, and the net delay of the polymer
solution front in arriving at the exit is, in pore volume fraction units, (b-b).
3. The polymer solution including the portion that has lost its polymer content by adsorption displaces most or
all of the connate water, so that the excluded pore volume for polymer solution is occupied mostly by water
that has lost its polymer by adsorption. Since the viscosity is the same, however, a quantity of water equal
to S
wi
(or S
wc
if the system is initially at S
wc
) plus (b-b) will exit the system before the polymer solution
will arrive. Distances along the Sw axis of the fractional flow diagram represent fractions of a pore volume.
The initial water is represented as the distance from S
w
= 0 to S
w
= S
wi
, and the additional quantity (b-b)
must be rep-resented by a distance to the left of S
w
= 0. Thus, a breakthrough tangent for polymer solution
must start at the point f
w
= 0, S
w
= -(b-b).
4. Since water flows ahead of the polymer solution, it must flow in accordance with a fractional flow curve
constructed from the relative permeabilities for oil and water and from the oil and water viscosities.
However, within the region containing polymer solution, flow must be in accordance with a fractional flow
curve constructed from the relative permeabilities and viscosities of oil and of polymer solution. It may
generally be assumed (though this is not necessarily true) that the relative permeability curve is the same
for polymer solution as for water. Thus, the polymer/oil fractional flow curve may be constructed using the
polymer solution viscosity instead of the water viscosity. The tangent mentioned in (3) must therefore be
tangent to the polymer fractional flow curve, not the water/oil curve. The fractional flow equation for
polymer will contain a new viscosity ratio (
p
/
o
), where p is the polymer solution viscosity at the flow
rate an&th
0
us the shear rate at which the displacement is conducted. This tangent is the breakthrough
tangent for polymer solution. It will cross the fractional flow curve for oil and water at a crossing point (f
wx
,
S
wx
) Breakthrough for ordinary water is represented by the line drawn from this point to the initial
condition in the system (f
w
= f
wi
,S
w
= S
wi
)
Conditions at the exit of the system may be described thus: for a throughput equal to the reciprocal of the slope of
the line joining initial conditions to the point f
wx
, S
wx
, fluids are produced at a fractional flow equal to f
wi
, in an
amount equal to that throughput. Then, the conditions at the exit jump from initial conditions to the conditions (f
wx
,
S
wx
), and stay at that condition with fluids exiting at constant water fraction (f
wx
) for an additional throughput equal
to the reciprocal of the slope of the polymer solution breakthrough tangent minus the water breakthrough
throughput. Then conditions at the exit jump to the point (f
pt
,S
pt
), which is the tangent point on the polymer/oil
fractional flow curve. From that point on, conditions at the exit slowly move along the polymer/oil fractional flow
curve toward the final point f
p
= 1.0, S
p
= 1 - S
orw

Occasionally, conditions may be such that the tangent point for the waterflood lies below the polymer tangent
crossing point f
wx
,S
wx
. In this case, the water breakthrough is at the waterflood tangent point on the water/oil
fractional flow curve. Conditions at the exit then move up the water/oil curve until they reach f
wx
,S
wx
and stay there
until the throughput (reciprocal of the slope) is equal to that the polymer tangent. Then conditions jump to the
tangent point on the polymer/oil curve, and follow that curve toward the upper limit of f
w
= 1.0.
DATA: Use the same relative permeability curves as before, with the polymer solution ten times as viscous as the
water, and with (b-b) = 0.10.
Figure 1 shows the fractional flow curve for this polymer displacement example.
Assuming that polymer solution is injected when the initial condition is (f
w
= 0, S
w
= S
wc
), we may calculate S
pt
and
f
pt
as follows:
f
pt
= 1/{1 + 10 * [(0.6 - S
pt
)/(S
pt
- 0.2)]
2
}
f
pt
= f
pt
2
* 8.0 * (0.6 - S
pt
)/(S
pt
- 0.2)
3
= (f
pt
- 0)/(S
pt
+ 0.1)
Solving for S
pt

S
pt
= 0.589223833, f
pt
= 0.9923929991, f
pt
= 1.439870404,
S
ptav
= 0.594506948, (O
p
)
pbt
= 0.394506948,
(W
i
)
pbt
= 0.6945069481
(N
p
)
pbt
= 0.98626737, (N
i
)
pbt
= 1.73626737
Equation of tangent line: f
p
= 1.439870404 * (S
p
+ 0.1)
We may solve the equation of the tangent line simultaneously with the equation of the water/oil fractional flow
curve to obtain S
wx
and thus f
wx
.:
1.439870404 * (S
wx
+ 0.1) = 1/{1 + [(0.6 - S
wx
)/(S
wx
- 0.2)]2)
Solving for S
wx
:
S
wx
= 0.474550149, f
wx
= 0.8272777535
(f
wx
- 0)/(S
wx
- 0.2) = 3.013211818 = slope of chord from initial conditions (f
w
= 0, S
w
= 0.2) to the point f
wx
, S
wx
The
reciprocal is equal to the throughput at water breakthrough, (W
i
)
wbt
= 0.3318717901,
(N
i
)
wbt
= (W
i
)
wbt
/0.4 = 0.8296794753
S
wxav
= S
wx
+ (1 - f
wx
) * (W
i
)
wbt
= 0.5318717901
(O
p
)
wbt
= 0.5318717901 - 0.2 = 0.3318717901 [= (W
i
)
wbt
, check]
(N
p
)
wbt
= (O
p
)
wbt
/0.4= 0.8296794753
The additional oil recovery during the period of simultaneous oil and water flow at the conditions (f
wx
, S
wx
) is equal
to the throughput difference in pore volumes, [(W
i
)
pbt
- (W
i
)
wbt
], times the fraction of oil flowing, (1 - f
wx
) This gives
the additional oil recovery as a pore volume fraction. We may convert this to movable pore volumes and add it to
the previous recovery to obtain (N
p
)
pbt
:
(0.694506948 - 0.3318717901) * (1.0 - 0.8272777535) = 0.0626351591
0.0626351591 + 0.3318717901 = 0.3945069492 = (O
p
)
pbt

(N
p
)
pbt
= 0.3945069492/0.4 = 0.9862673731 (checks value above)
A plot of the polymer flood production (N
p
) versus input (N
i
), each in movable pore volumes, for the polymer
flood, has been added to Figure 2 , with proper labels. We notice that when the polymer front breaks through at the
exit, the fractional flow jumps from f
wx
to f
pt
, and f
pt
is higher than the normal range of the economic limit of water
cut. Hence, the polymer flood terminates at that point.
For an input T of 0.3 (N
i
= 0.75), the saturation profile has been calculated for the polymer flood. This has been
added to Figure 3 , with proper labels to distinguish the ordinary waterflood profile from that of the polymer flood.
Hot Water Flooding
This is a miscible displacement of cold water by hot water, preceded by and accompanied by immiscible
displacement of oil. Under pressure, it is possible to heat water to temperatures well above the boiling point at
atmospheric pressure. If reservoir oil is heated by injecting hot water, the viscosity of the oil drops much more than
does the water viscosity. This changes the fractional flow curve to a more favorable condition than that of the
original curve. However, there is not much need for this unless the waterflood at reservoir temperature has
difficulties. This is illustrated by the comparison of the polymer flood with an ordinary waterflood. There is an
improvement in oil recovery by using polymer, but it is not large.
In this section, we change the situation by assuming that the viscosity of the crude oil is 100 times greater at the
same reservoir temperature of 100 F, where the viscosity of water is 0.7 cp. We assume that the water is heated at
450 psig and injected at this pressure at the wellhead, and we ignore heat losses passing down the well bore, so that
the hot water entering the formation is assumed to have the same temperature as saturated steam condensate at the
well head. This temperature is very close to 460 F, which is 360 F hotter than the reservoir.
In the calculations which follow, we see what this accomplishes. The hot water entering the reservoir must heat the
rock, and also the oil that is left behind the hot water front. This heat exchange occurs at the hot water front, and a
bank of water at reservoir temperature is created from injected hot water which has lost its excess heat (i.e., the heat
above that of the reservoirs temperature) to the rock and the remaining oil.
For purposes of these calculations, it is a close-enough approximation to assume that the remaining oil saturation is
the same as the waterflood residual oil saturation. This bank of water at reservoir temperature is similar to the bank
of water produced from solution by adsorption of polymer, and so we shall use the symbol b to designate the pore
volume fraction of cold water that is produced when the hot water front reaches the exit. We can calculate the value
of b by a heat balance, given the densities of the materials present and the average of their specific heats over the
temperature range. The heat given up by the pore volume fraction of water (b) is equal to the heat gained by the
parts of the system that are heated by this water, namely the rock and the residual oil saturation:
b*|*C
w
*
w
* (T
H
-T
C
)= [(1-|)*C
r
*
r
+|*S
orw
*C
o
*
o
] * (T
H
-T
C
) (A.53)
Note that the temperature difference cancels out of this equation, so that:
b=[(1-|)*C
r
*
r
+|*S
orw
*C
o
* p
o
]/( | * C
w
*
w
)
It is therefore necessary to provide data concerning densities and specific heats, as well as the rock porosity. These
are given below.
Note that the temperature difference cancels out of this equation, so that:
b = [(1 - |) * C
r
*
r
+ | * S
orw
* C
o
*
o
]/( | * C
w
*
w
)
It is therefore necessary to provide data concerning densities and specific heats, as well as the rock porosity. These
are given below.
DATA: The viscosity of the crude oil is 280 cp at 100 F and 0.64 cp at 460 F. The water viscosity is 0.7 cp at 100
F and 0.12 cp at 460K The specific heat of water (C
w
) over the range from 100 to 460 F is 1.0372 BTU/lb/degF
while that of the crude oil (C
o
) is 0.5139 and that of the rock (C
r
) is 0.15. The porosity ( ) of the reservoir rock is
0.2. The average density of the water (r
w
) is 56 lb/cu.ft. and that of the oil (r
o
) is 54 lb/cu ft, while that of the rock (r)
is 162 lb/cu ft.
Use of these data in Equation A.53 gives b = 1.86457. This is quite a large number compared to the size of b for
polymer flooding. It means that 1.86+ pore volumes of reservoir-temperature water derived from the injected hot
water, as well as the connate water initially present, must exit the system before the hot water front will reach the
exit.
The fractional flow curves for cold and hot conditions are calculated and plotted as Figure 4 , leaving space to plot
the point at minus 1.86457 saturation fraction on the horizontal axis, to the left of zero water saturation (which
ranges from 0.0 to 1.0). The tangent to the hot water/hot oil fractional flow curve starts from this point.


Figure 4


Next, we calculate the cold water/cold oil breakthrough tangent point (for the tangent starting at 0.2 on the S
w
axis),
and the hot water/hot oil tangent point
Solving,
S
wct
= 0.2398014874, f
wct
= 0.5497518566, f
wct
= 13.91234461
W
ict
= 0.0723990045 = O
pct
, N
ict
=0.l8O99751l4=N
pct

f
wht
= f
wht
2
* 0.6(0.6 - S
wht
)(S
wht
- 0.2) = (f
wht
- 0)/(S
wht
+ 1.8645708)
Solving again,
S
wht
= 0.5661653397, f
wht
= 0.9949031572
f
wht
= 0.4093034849, (W
i
)ht = 1/f
wht
= 2.443174898
S
whav
= S
wht
+ (1 - f
wht
) * (W
i
)ht = 0.5786178178, (O
p
)ht = 0.3786178178
(N
p
)ht = 0.9465445446, (N
i
)ht = 6.107937245
We write the equation of the hot water tangent line and solve it simultaneously with the equation of the cold
water/cold oil fractional flow curve to obtain the crossing point coordinates. This point lies above the cold water
tangent point. Therefore there will be a cold water breakthrough, described by the cold water tangent, then
conditions at the exit will move up the cold water fractional flow curve until the crossing point is reached.
Conditions remain stationary at this point until the hot water front arrives (at the required throughput) and then
conditions jump to the tangent point on the hot water fractional flow curve. At this point, the fractional flow of water
(water cut) exceeds the economic limit, so the process is finished. The curves of Np versus Ni for both a cold
waterflood and for the hot water flood are calculated and these recovery curves are plotted as Figure 5 .


Figure 5


For 0.07 pore volume of injected water, the water saturation profiles (vs X = x/L) are calculated for both the cold
waterflood and the hot waterflood just discussed and the results plotted as Figure 6 .


Figure 6


If saturated steam at the same pressure is injected instead of hot water, the cooled water bank pore volume fraction b
is considerably reduced due to the contribution of latent heat. An added advantage of steam over hot water and
polymer is a reduction in the residual oil to a value much lower than the waterflood residual oil saturation.



Suppose that the Lorenz diagram for rock samples from a given reservoir is found to be very closely
approximated by a quarter circle with its center at the lower right hand corner of the diagram. In this case,
the Lorenz coefficient is about:
1. 0.32
2. 0.37
3. 0.42
4. 0.47
5. 0.52
6. 0.57



Figure 1


Click Figure 1 for the solution.

For the quarter-circle Lorenz diagram of Exercise 1, suppose that we want to find the fractions of |
h
(on the
horizontal axis) corresponding to 1/3 fractions of kh (on the vertical axis). The fraction of |
h
corresponding to
the middle third of k
h
(from 1/3 to 2/3 on the vertical scale) is equal to about
1. 0.28
2. 0.26
3. 0.24
4. 0.22
5. 0.20
6. 0.18



Figure 1


Click Figure 1 for the solution.

For a formation described by the Lorenz diagram of Exercise 1 and the |
h
fractions of Exercise 2, the total |
h
is 10.0
feet and the total kh is 16,500 millidarcy/feet. The relationship between permeability and porosity was found to be:
L
n
k = 35|- 1, so we can write also that: L
n
k - 35 (|
h
/k
h
) k + 1 = 0. From this form of the equation we can solve for
the three values of permeability corresponding to each 1/3 of the k
h
. The smallest of the three values of permeability
thus determined is (in millidarcys)
1. 180
2. 150
3. 120
4. 90
5. 60



Figure 1


Click Figure 1 for the solution.

If the three values of permeability determined by the method of Exercise 3 are plotted on log-probability scales with
the permeabilities on the vertical log scale and the mid-points of the |
h
intervals on the horizontal cumulative %
probability scale, and a straight line is drawn through the three points, one can read the permeability at the 50%
point of |
h
as 180md and at the 84.1% point, as 48md. The value of the Dykstra-Parsons v is about:
1. 0.78
2. 0.83
3. 0.73
4. 0.68
5. 0.88
6. 0.63



Figure 1


Click Figure 1 for the solution.

The modified (by M.S. Graboski and T.W. Daubert, in SPE Reprint Book 15, 8690) Soave-Redlich-Kwong
equation of state is:
P = RT/(V-b) - (a*a)/[V(V+b)]
where a = 0.42747R
2
T
c
2
/P
c
and b = 0.08664RT
c
/P
c
, while o =[1+S(1-T
r
1/2
)]
2
, T
r
= T/T
c
, R = 10.73 psi/ft
3
/(lb/mol -
R), P
c
= critical pressure, T
c
= critical temperature, S = 0.48508 + 1.55171e - 0.15613e
2
, where e = Pitzers
acentric factor
For normal butane, P
c
= 550.7 psia, T
c
= 765.2 R, and e = 0.201. The molar volume V in ft
3
/lb mol at P = 1,377
psia, T = 925.9 R (434.3 F) is about:
1. 1.33
2. 0.98
3. 4.56
4. 1.75
5. 2.11
Then:
1,377 = [(10.73 * 925.9)/(V - 1.29175)] - [(52,328.6 * 0.84811)/(V * (V + 1.29175))]
Solving for V by trial-and-error,
v = 4.559395



Figure 1


Click Figure 1 for the solution.

The Peng-Robinson equation of state (SPE Reprint Book 15, p. 41-45) is:
P = RT/(V - b) - (a * a)/[V(V + b) + b(V - b)]
where
a = 0.45724R
2
T
c
2
/P
c

b = 0.0778RT
c
/P
c

a = [1 + k(1 - T
r
1/2
)]
2

k = 0.37464 + 1.54226e - 0.26992e
2

w = Pitzers acentric factor.
Note how similar this is to the modified Soave-Redlich-Kwong equation.
Calculate for normal butane (critical data and acentric factor the same as given in Exercise 5) the pressure at which
the molar volume is 1.57 ft
3
/lb mol, when the temperature is 100 F (560 R). This pressure, in pounds per square
inch absolute, is about:
1. 250
2. 500
3. 1,000
4. 2,000
5. 4,000



Figure 1


Click Figure 1 for the solution.
The compressibility factor Z = P
V
/R
T
can be derived from the Peng-Robinson equation of state in the form:
Z
3
- (1 - B)Z
2
+ (A - 3B
2
- 2B)Z - (AB - B
2
- B
3
) = 0
This is a cubic equation, solvable most easily by trial and error when the constants A and B are given: A =
aoP/(R
2
T
2
) = 0.45724oP
r
/T
r
2
and B = bP/RT = 0.0778Pr/Tr, where P
r
= P/P
c
and T
r
= T/T
c
. A cubic equation may
have more than one real root. In case the fluid in question exists in both liquid and vapor phase at the conditions
specified, the largest root is the z factor for the vapor phase and the smallest root for the liquid phase. At 200 F
(660 R) and 200 psia, normal butane is approximately on its vapor pressure curve. For these conditions the vapor
phase Z value is about:
1. 0.44
2. 0.54
3. 0.64
4. 0.74
5. 0.84
6. 0.94



Figure 1


Click Figure 1 for the solution.

For the conditions of Exercise 3 the smallest root of the Z factor equation is about 0.0527368. The specific gravity
of the liquid normal butane at 200 F and 200 psia (relative to the density of water at standard conditions = 62.4
lb/ft
3
), which can be calculated from this value of Z, using V = ZRT/P ft
3
/lb mol and the molecular weight which is
58.124 lb/lb mol, is about:
1. 0.65
2. 0.60
3. 0.55
4. 0.50
5. 0.45
6. 0.40



Figure 1


Click Figure 1 for the solution.

A reservoir crude oil analysis, in mole fractions, is as follows: C
1
= 0.28, C
2
= 0.03, C
3
= 0.04, iC
4
= 0.02, nC
4
=
0.03, iC
5
= 0.02, nC
5
= 0.04, C
6
= 0.05, C
7+
= 0.49. You want to estimate the mole fractions of each carbon number
fraction from C
7
through C
10
, and the mol fraction of C
11+
, for the purpose of calculating the properties of the lumped
gasoline component (iC
5
through C
10
). You may use the equation C
n
= C
6
*e
-a(n-6)
, where Cn is the mole fraction of a
given carbon number n and C6 is the mole fraction of C
6
. You can solve for the constant a by integrating C
n
d
n
from
n = 6 to n = infinity and setting the result equal to the sum of C
6
and C
7+
(mole fractions). Hint: the integral of b*e
-
ax
dx from zero to infinity is b/a. The mole fraction of C
11+
is about:
1. 0.27
2. 0.33
3. 0.30
4. 0.39
5. 0.36



Figure 1


Click Figure 1 for the solution.

In Figure 1 , the equation of the two-phase envelope is given for a certain temperature and pressure and a given
hydrocarbon mixture. The diagram, which is normally shown as a triangular diagram, has been converted into a
rectangular plot so that the equations of lines and curves can be derived more easily. The equation of the tie lines
and the equation of the two-phase envelope have been solved simultaneously to give the equation for x as shown on
the diagram. Two values of x are obtained, depending on the sign of the second term. One of these values of x is on
the upper part of the two-phase envelope, and the other is on the lower part, and the corresponding values of y can
be obtained from the equation of the tie lines. A value of m can be obtained for any mixture lying within the two-
phase envelope by substituting the values of y and x for that point in the equation of the tie line, and then the values
of y and x for the two ends of the tie line passing through that point can be calculated by substituting that value of m
into the equation for x. The lever-arm rule can then be used to determine the mole fraction of vapor phase and thus
the mole fraction of liquid phase for any such mixture.
Given a mixture which is 0.6 mole fraction of C
1
and 0.3 mole fraction of C
7+
, and using the equilibrium
relationships mentioned, calculate the mole fraction of vapor phase.
1. 0.27
2. 0.315
3. 0.365
4. 0.403



Figure 1


Click Figure 1 for the solution.
Calculate, from the equations determining the equilibrium of the figure in Exercise 10, what minimum mole fraction
of C
2
-6 must be in a mixture of C
7+
and C
2
-6 for this mixture to be first-contact miscible with pure C
1
.
1. 0.9
2. 0.85
3. 0.80
4. 0.75
5. 0.70
6. 0.65



Figure 1


Click Figure 1 for the solution.

For the equilibrium relations of Exercise 10, if a crude oil lying to the left of the critical tie line is being displaced by
a rich gas containing 0.84 mole fraction C
1
and 0.16 mol fraction C
2
-6, then a certain liquid composition must be
attained by multiple contact to achieve the miscible state. The mole fraction of C
1
in that liquid composition, which
lies on the binodal curve, is:
1. 0.7
2. 0.65
3. 0.60
4. 0.55



Figure 1


Click Figure 1 for the solution.

For the equilibrium relations of Figure 1 , a crude oil lying to the right of the critical tie line, with a composition of
0.4 mole fraction C
1
and 0.3 mole fraction C
2
-6, is being displaced by a vaporizing gas drive with pure C
1
. The
minimum enrichment of a gas phase composition lying on the binodal curve which will just achieve miscibility with
the initial crude oil composition contains the following mole fraction of C
2
-6:
1. 0.1332
2. 0.1376
3. 0.1294
4. 0.1428
5. 0.1125



Figure 1


Click Figure 1 for the solution.

It is often necessary, for computer simulation purposes, to characterize the components of a crude oil in the C
7+
fraction (commonly shown as the last component in PVT analyses of crude oils) as being composed of several
components designated as "pseudocomponents." The following properties are required: density, viscosity, molecular
weight, and either equilibrium K values or data with which these can be calculated by means of an equation of state.
With regard to density, a good value representing the compositional influence not only of paraffins (for which most
data are available) but also of cycloparaffins and aromatic compounds (for which there are few data) can be obtained
from a curve given by Roland in SPE Reprint Book 15 (p. 104), for density at 60 F, which is fitted by the equation:
(g/cc) = 0.59 + 0.0016(MW) - 0.000002(MW)
2

For a pseudo-component of molecular weight MW = 282.5, the C20 normal paraffin of that molecular weight has a
density of 0.792. The density calculated by Rolands equation is higher by the following amount:
1. 0.18
2. 0.15
3. 0.12
4. 0.09
5. 0.06
6. 0.03



Figure 1


Click Figure 1 for the solution.

In testing a cleaned core sample of a given formation rock to determine the longitudinal dispersion coefficient, a
typical procedure is (while injecting a given fluid at a known flow rate) to start at a given time injecting the same
fluid with a non-adsorbed tracer compound dissolved in it, and to continuously analyze the effluent stream for its
tracer concentration. The curve of this concentration, divided by the inlet fluid tracer concentration, may be plotted
against time. If this curve does not deviate too strongly from a symmetrical "s" shape, then the time when the
normalized concentration reaches 50% times the flow rate may be taken as the pore volume.
In a given case, the core diameter was 2.0 inches and its length was 6 inches. The time to reach the 50% point on the
effluent normalized concentration curve was 4 hours and 56 seconds. The porosity of the core had been
independently measured and found to be 0.1300. If this is correct, then the flow rate in cubic centimeters per minute
is:
1. 0.1667
2. 0.3333
3. 0.6667
4. 1.333
5. 2.667
6. 5.333



Figure 1


Click Figure 1 for the solution.

The text gives formulas derived from a paper by Brigham, as follow:
U = (N
i
- 1)/N
i
1/2

u
L
/K = [3.625/(U90% - U10%)]
2

where U is a dimensionless number, N
i
is the pore volumes of throughput at which a given value of the normalized
tracer concentration is observed, u is the linear velocity of the fluid through the pore area (e.g., cm/sec), L is the
length of the system (e.g., in this case 6 * 2.54 = 15.24 cm), and K is the longitudinal dispersion coefficient. In the
case described in Exercise 15, the 10% value of normalized concentration was observed at 2 hours 48 minutes and
40 seconds, while the 90% value was observed at 5 hrs 44 minutes and 10 seconds. A plot of U10%, U50%, and
U90% on a linear scale against the normalized concentration on a cumulative probability scale gives a straight line,
so the second equation given above is valid. From the data given, the value of K in cm2/sec can be calculated; it is
about:
1. 0.0001
2. 0.0002
3. 0.0006
4. 0.0010
5. 0.0016
6. 0.0032



Figure 1


Click Figure 1 for the solution.

In a given core flood, water is injected when the system is at a minimum water saturation of 0.15. The porous
medium has relative permeabilities for water and oil as follow: k
rw
= 0.1(S
w
- 0.15)1.5 and kro = 7.5(0.6 - S
w
)
3
, and
the viscosities of the oil and water are 1.4 cp and 0.7 cp, respectively. The fraction of the length of the system which
the saturation jump at the waterflood front has traversed when 0.29 pore volume fraction of water has been injected
is approximately:
1. 0.8
2. 0.7
3. 0.6
4. 0.5
5. 0.9



Figure 1


Click Figure 1 for the solution.

For the data of Exercise 17, the oil recovery as a fraction of pore space at 99% water cut (f
w
= 0.99) is about:
1. 0.36
2. 0.38
3. 0.40
4. 0.42
5. 0.44



Figure 1


Click Figure 1 for the solution.

You wish to waterflood a core that has the relative permeability characteristics of Exercise 17, starting at connate
water saturation, and with oil and water of the viscosities given in Exercise 17. The waterflood is to continue to the
point where the average water saturation is within 0.001 of (1 - Sorw), for the purpose of carrying out an EOR
process starting at approximately residual oil saturation. The pore volumes of water which must be injected to arrive
at the desired average water saturation (disregarding a capillary end-effect) is about:
1. 1,200
2. 1,000
3. 800
4. 600
5. 400
6. 200
7. 100
8. 50
9. 25
10. 10



Figure 1


Click Figure 1 for the solution.

Deppes method for calculating flow in well patterns consists of treating the flow as radial outflow from the
injection well and radial inflow to a production well. He assumed equal outer radii for the flow circles around the
injector and producer, and made the sum of the areas of the two circles equal the area of the pattern. If a symmetry
element of a well pattern is used, then there may be just a part of a complete circle for each well, but the same rule
applies: the sum of the areas of these part circles must equal the area of the symmetry element. Suppose that we are
dealing with the following element of a 2:1 staggered or indirect line drive:
Figure 1
Note that there is an injection well in the upper right corner and a production well in the lower left corner.


Figure 1


These are each 1/4 wells if the pattern is repeated over a large total area, which is assumed in this case.
The radius of each 1/4 circle according to Deppes method is, in feet, approximately:
1. 225
2. 250
3. 325
4. 400
5. 450
6. 517



Figure 1


Click Figure 1 for the solution.

In a field where there is only one mobile fluid (oil) present, and the borders of the field are far enough away from
wells that the effects of a boundary can be neglected, the pressure at any given point i in the flow field is given by:
m = M
P
i
= P
0
- [1/(4p)] * S{(q
m
/q) * Ln[(X
i
- X
m
)2 + (Y
i
- Y
m
)
2
]}
m = 1
in which X = x/L, Y = y/L, where L is a baseline length, such as the distance between wells, and x and y are the
coordinates of the point i and of the wells m; qm is the rate at well m (positive for injection, negative for
production), q is the total rate, and the dimensionless pressures are P
0
= k * k
ro
* h * (p
0
- p
w
)/(q * o), and P
i
= k * k
ro

* h * (p
i
- p
w
)/(q *
o
),
o
is the viscosity of the oil, k
ro
is the relative permeability of the porous rock to oil, k is the
total permeability of the rock, and h is the formation thickness. When k is expressed in darcys, the thickness h must
be in cm, the rate q in cm
3
/sec, the viscosity in cp, and the pressures in atmospheres.
In a given field, there were four producing wells at the corners of a square of side = 40,538 cm. The formation
thickness h was 500 cm, the permeability 0.2 darcy, and the only mobile fluid present was oil at an initial pressure of
200 atmospheres, a viscosity of 2.5 cp, and a relative permeability of 0.7. The total flow rate was 10,000 cm3/sec,
with a bottomhole pressure of 100 atmospheres in all four wells. Well No. 1 had dimensionless coordinates X
1
= 0 =
Y
1
, Well No. 2 had X
2
= 0, Y
2
= 1, Well No. 3 had X
3
= 1, Y
3
= 0, and Well No. 4 had X
4
= 1, Y
4
= 1. For the wells
numbered in this order, the fractional rates were q
1
/q = 0.35, q
2
/q = 0.14, q
3
/q = 0.28, and q
4
/q = 0.23.
Based on these data, the pressure, in atmospheres, in the center of the square is about:
1. 180
2. 170
3. 160
4. 150
5. 140
6. 130
7. 120
8. 110



Figure 1


Click Figure 1 for the solution.


In a line drive where there is more than one line of producers, the fractional length x/L and the value of Ni must be
referred to the total distance from the line of injectors to a given line of producers. The producing water cut in a
given well where the gravity tongue has already come by is given by the fractional height y/h times the mobility
ratio M, divided by [1 + (M - 1) * (y/h)]. The fractional distances x/L to the tip of the gravity tongue (given by
Dietzs Equation) is (x/L) = M * N
i
/[1 + (M - 1) * (y/h).
When the water cut in the first row of producers is 0.5 and M = 4, the value of y/h and hence of N
i
relative to the
movable pore volume between the injectors and this first line of producers can be calculated from the above
equation (since x/L = 1). Relative to the second line of producers, the fraction of a movable pore volume injected is
half as much, and L is doubled.
For this new L, the dimensionless distance x/L to the tip of the water tongue can be calculated; it is about:
1. 0.65
2. 0.7
3. 0.75
4. 0.8
5. 0.85
6. 0.9



Figure 1


Click Figure 1 for the solution.


Self Assessment
1. The main mechanism(s) in a miscible process are:


(A) Extraction


(B) Solubilization


(C) Vaporization


(D) Dissolution


(E) A and D


(F) A, B and D


(G) All of the above

2. At temperatures in the 100-130 F and with crude oils of 35-100 API, the miscibility
pressure of lean natural gas with crude oil is about:


(A) 2000-3000 psia


(B) 3000-4000 psia


(C) 4000-5000 psia


(D) 5000-6000 psia


(E) 6000-8000 psia

3. The principal difference(s) for miscible flooding with carbon dioxide include:


(A) Significant solubility with reservoir brine or injected flood water


(B) Superiority of reducing the viscosity of crude oil


(C) Its density being much less than that of crude oil at reservoir conditions


(D) A and B


(E) A and C


(F) All of the above

4. According to the Gibbs'' Phase Rule, the number of degrees of freedom, F is equal to: (
C = the number of components, and P = the number of phases present in the system)


(A) F = C + P


(B) F = P - C


(C) F = C + P - 2


(D) F = C - P + 2


(E) F = C + P + 1


(F) F = C - P - 1


(G) F = C + P - 1


(H) F = C - P + 1

5. The tangent to the curve at the pseudo-critical point is called:


(A) Plait point


(B) Tie-line


(C) Critical tie-line


(D) P-X diagram


(E) Bubble point

6. Nitrogen or flue gas is considered as:


(A) High-pressure miscible displacing agent


(B) High-temperature miscible displacing agent


(C) Low-pressure miscible displacing agent


(D) Low-temperature miscible displacing agent

7. Carbon dioxide is ______ solvent for crude oil ______ either propane or butane.


(A) as good ; as


(B) a better ; than


(C) a poorer ; than


(D) much superior ; than

8. The finger thicknesses are determined by:


(A) Formation heterogeneity


(B) Mobility ratio


(C) Velocity


(D) A and B


(E) A and C


(F) All of the above

9. Longitudinal dispersion coefficient K
L
is directly proportional to:


(A) Formation electrical resistivity factor


(B) Superficial velocity


(C) Grain size


(D) A and B


(E) B and C


(F) A and C

10. First equations pertaining to the sweep efficiencies of miscible floods for regular and
irregular well patterns are derived by:


(A) Claridge


(B) Muskat


(C) Deppe


(D) Thompson

11. The assumption(s) in oil recovery projects with miscible flooding include:


(A) Only one fluid is flowing


(B) Constant density


(C) Constant thickness, porosity and permeability


(D) Viscous fingering


(E) A, B and C


(F) A, B and D


(G) All of the above

12. Which of the following statements is FALSE?


(A) Linear displacement efficiency is impaired due to viscous fingering


(B) Areal sweep efficiency is impaired due to viscous fingering

(C) If the gravity effect is to prevent viscous fingering, there is a maximum rate of
displacement

(D) The degree of impairment in displacement efficiency becomes worse as the
mobility ratio decreases

13. The Dykstra-Parsons coefficient of permeability variation is defined as:


(A) v = k50% - k25%


(B) v = (k25% - k50%) / k25%


(C) v = (k50% - k84.1%) / k50%


(D) v = (k84.1% - k50%) / k84.1%

14. Which of the following assumptions is (are) TRUE for Koval''s method of fractional
flow calculations?


(A) Effective mobility ratio is 1.0


(B) Relative permeability of solvent is equal to its saturation fraction


(C) Segragation of fluids due to vertical flow in opposite directions is ignored.


(D) A and C


(E) B and C


(F) All of the above

15. Estimating oil recovery in miscible flooding processes with an extended version of
Deepe''s method takes into account:


(A) Viscous fingering


(B) Breakthrough sweep efficiency


(C) Layered systems


(D) A and B


(E) A and C


(F) All of the above

16. Black oil simulators are modified to handle viscous fingering by using _____________
method.


(A) Pseudo-relative permeability curves


(B) The mixing parameter


(C) Pseudo-layer models


(D) Transverse dispersion

17. About two-thirds of U.S. EOR production is from:


(A) Gas injection projects


(B) CO
2
flooding


(C) Thermal recovery methods


(D) Waterflooding

18. The multiple contact miscibility pressure can be determined experimentally by using:


(A) Podbielniak column approach


(B) Slim tube test


(C) Volume blending test


(D) Rising bubble test


(E) B and C


(F) B and D


(G) All of the above

19. Which of the following is NOT required to be known with Stone''s gravity segregation
model?


(A) Width and height of the formation


(B) Individual phase velocities


(C) Densities of gas and water


(D) Vertical permeability


(E) Horizontal permeability


(F) Ratio of phase flows

20. Which of the following statements is FALSE?


(A) The equation of state can be expanded algebraically to a cubic equation in Z

(B) For a single component, the largest root of the equation of state is the Z value for
liquid phase

(C) In combining constants a and b in the equation of state, arbitrary interaction
constants are needed

21. Which of the following is (are) needed to compute the concentration of a tracer in a
flowing stream?


(A) The average flow velocity


(B) Dispersion coefficient


(C) Distance in the direction of flow


(D) Porosity


(E) A and B


(F) A, B and C


(G) All of the above

22. Which of the following is NOT a characteristic of polymer flooding?


(A) Polymer does not invade all of the pore spaces that can be invaded by water


(B) Polymer is adsorbed on the rock surface from the polymer solution


(C) The adsorption is almost completely reversible


(D) A breakthrough tangent for polymer solution must start at the point, f
w
= 0.

23. Effective mobility concept is used to account for:


(A) Irregular well pattern configurations


(B) Off-centered injectors/producers


(C) Heterogeneity in the system


(D) Gravity segregation


(E) Viscous fingering

24. To determine streamlines of fluid flow, it is necessary to assume:


(A) Single fluid flow


(B) Constant permeability, porosity and thickness


(C) Constant density


(D) Two dimensional flow


(E) A and C


(F) A, B and C


(G) All of the above

25. Which of the following statements is TRUE about determining vertical sweep
efficiency?

(A) Dykstra-Parsons method includes the effect of gravity tonguing within a layer of
uniform permeability


(B) For mobility ratios one or less than one, gravity tonguing calculations fail


(C) Dietz's method is the simplest approach


(D) Stile's method includes mobility ratio in the calculations

Submit Your Answers
Note: Your answers CANNOT be changed after they have been submitted. Check your answers
thoroughly BEFORE you submit them.

Você também pode gostar