Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
A WAG displacement is simulated with several combinations following a water flood. The non-wetting phase hysteresis
of hysteresis models and wetting phase options for water-wet model is based on the theory developed by Land4 and
systems. A comparison of the results from the WAG Carlson2. This method is often called the WAG Hysteresis
displacements shows a factor of two difference in the oil Model. The net effects of this model is that the relative
recovery in the water-wet system. permeability of the non-wetting gas phase and of the wetting
water phase, decrease with each cycle of the WAG injection
Recommendations are given of how to properly use the process.
options and choices in a hysteresis simulation. Laboratory
experiments are suggested in order to determine the correct Additional description and the mathematical details of these
hysteresis option for the reservoir of interest. hysteresis models are available in the original publications or
the ECLIPSE E100 and E300 Technical Descriptions.7, 8
Hysteresis in oil-wet systems will be discussed in a later
publication. The Killough, Carlson, and WAG Hysteresis models will be
discussed in the remainder of this paper.
Hysteresis Models
In a simple displacement where fluid saturations start at their Laboratory experiments show that the hysteresis model should
connate saturation and saturation always increases, the relative be applied to the non-wetting phase(s). This situation leaves a
permeability and capillary pressure values of the fluid are choice of which curves (imbibition or drainage) to be used for
found on the drainage or imbibition bounding curves. the relative permeabilities of the wetting phase (water in a
Hysteresis models were developed for the cases when the water-wet system). Or hysteresis could be also applied to the
initial fluid saturation is not an end point value or when a wetting phase. Thus, after choosing to use the Killough or
phase saturation increases and then decreases, or vise versa. Carlson hysteresis model, one then must decide how to handle
Killough1 and Carlson2 published mathematical techniques for the wetting phase.
calculating the scanning curves. Both techniques are
extensively used in the industry. This situation yields five options, called “type”, as specified in
Table 1.
Carlson’s method produces a scanning curve that is parallel to
the imbibition curve. One can simply shift the imbibition Table 1 Model and Wetting Phase Choices: Water-Wet
curve horizontally until it cuts the drainage curve at the Systems
saturation Shy, see Figure 1. The imbibition curve must be Type Model Curve used for the
steeper than the drainage curve at all values of kr. Failure to
wetting (water) phase
do this can result in a scanning curve that will cross to the
right of the drainage curve and may produce a negative value 0 Carlson drainage curve
of Sncrt. 1 Carlson imbibition curve
2 Killough drainage curve
Killough’s method does not have a simple geometric
3 Killough imbibition curve
interpretation. The equations show that Sncrt will always lie
between Sncrd and Sncri. 4 Killough hysteresis
Occasionally users use the same curve for both the primary The effects of the choice of these types on scanning curves
imbibition and primary drainage curve. One should be aware and oil recoveries from a WAG displacement will be seen in
that with Killough’s method, in this situation, the scanning following sections.
curves will not necessarily follow this coincidental curve,
except at its end points. Thus, the relative permeabilities may Two additional type possibilities exist: (1) no hysteresis and
not be what one would expect. (2) equilibrate with the drainage curve and simulated with the
imbibition curve. This last case we shall label “type” = -1.
Jargon8 (Marathon) has introduced a modification to
Killough’s method that overcomes the inconsistent scanning Three Phase Oil Relative Permeability Models
curves. To calculate the oil relative permeability in the presence of gas
and water phases requires the choice of a three-phase model.
In the case of a WAG (Water Alternating Gas) injection, The industry standard for this calculation are Stone’s5 first
Larsen and Skauge3 have developed a three phase hysteresis model, Stone’s6 second model, or the Cheshire7 model, which
model that calculates a relative permeability of a phase that is is the default in Eclipse7. Each of these models gives different
dependent on the history of the third phase in addition to its scanning curves as saturations change. Examples will follow
(the phase’s) saturation and history. This theory follows the in a later section.
observation that the gas relative permeability is higher when
displacing pure oil that when displacing a water-oil mixture
SPE 63147 COMPARISON OF RESERVOIR SIMULATION HYSTERESIS OPTIONS 3
Imbibition and Drainage Bounding Curves CASE 2: Initial water saturation = 0.2, oil saturation
To use the Killough, Carlson, or WAG Hysteresis models, one varying from 0.2 to 0.8, gas saturation varying from
initially needs properly defined imbibition and drainage 0.6 to 0.0. See Figure 7.
curves. That is, the end points of the curves must be
consistent and the orientation of the curves with respect to Given the initial saturation distribution and the three injection
each other must be correct. If the end points or the orientation fluids (water, oil, and gas), on some occasions several of the
of the curve are not correct, then the calculated scanning grid blocks initially experienced an imbibition process, and
curves most likely will be incorrect and the simulator will then drainage or vise versa. This created scanning curves that
assign incorrect relative permeabilities to the phases. An go up then down or loop. A detailed analysis of the
example of incorrect scanning curves will be shown in a displacements is required to fully understand the figures
following section. provided in later sections.
Figures 2 through 5 give examples of valid relative The combination of: (1) 2 cases; (2) water injection, oil
permeabilities and capillary pressures for a water-wet rock. injection, and gas injection; (3) the hysteresis options: no
The drainage and imbibition curves and the saturation change hysteresis, “type” = -1, 0, 1, 2, 3, and 4; and (4) the 5 plots for
direction where the curve is employed are noted on the each case (Pcow, Pcgo, krw, krg, kro) created a large number of
figures. The difference between the drainage and imbibition figures. In some cases the STONE 1 and STONE 2 three
curves is accentuated to aid the observations of the scanning phase models were used to see their effect on the scanning
curves. curves. In the discussions that follow, only a few key figures
will be presented.
Two key features of the bounding imbibition and drainage
curves will assure that the scanning curves are correctly Since one of the main uses of the hysteresis options is for
calculated. One, the relative orientation of the curves must WAG injection, the scanning curves are of interest during a
follow the curves provided. For example, in Figure 2, the WAG injection. To observe the scanning curves a four grid
drainage krowd curve must be above the imbibition krowi curve block reservoir was created with an initial connate water
(at a given saturation the drainage krowd value must be higher saturation (Sw = 0.2) and So = 0.8. Five WAG cycles of
than the imbibition krowi value.) Also one observes that the injection starting with water injection were simulated. All
krwi imbibition curve is above than the krwd drainage curve. “type” parameters were investigated and the WAG Hysteresis
Two, the end points of the curves must coincide. For model was also used. Also in some cases the STONE 1 and
example, in Figure 2, at the saturation Sw = 0.2 the imbibition STONE 2 three phase models were used to see their effect on
and drainage krow end points must meet. The same applies for the scanning curves. In all cases the initial oil saturation was
the krwi and krwd curves and at the Sw = 1.0 end point. A calculated by So = 1 - Sw - Sg.
guideline that will assure good scanning curves is that the
envelope formed by the imbibition and drainage curves should Reference 9 gives a detailed description of all of the results of
be closed (a closed loop). the analysis of hysteresis options in both a water-wet and an
oil-wet system and the Eclipse data sets to generate the
Description of the Simulation Models Used to Create scanning curves.
Scanning Curves
To create scanning curves for the various models a linear, Scanning Curves in a Water-Wet System
horizontal reservoir with 13 grid blocks was chosen, see Many cases were simulated in the study. The following are
Figure 6. Fluids were injected in the right hand side grid examples of some of the scanning curves created.
block and produced from the left-hand side grid block. When
the capillary pressure or phase relative permeability in a grid • Figure 8, Pcow, Case 1 reservoir with water injection –
block is plotted vs. the phase saturation in that block during Carlson hysteresis (“type” = 0). Note the Pc curves start
the displacement, the resulting curve reproduces one of the at various points along the drainage curve and form
bounding curves or shows a scanning curve. scanning curves (or the two bounding curves) as the water
saturation increases (imbibition).
For the three-phase water-wet system, it was necessary to • Figure 9, krw, Case 1 with water injection – Killough
simulate the injection of water, oil, and gas (separately) into hysteresis (“type” = 4). Note the krw curves are created
reservoirs with different initial conditions to display all of the similarly to the Pc curves of Figure 8.
scanning curves with the various hysteresis options. The two • Figure 10, krg, Case 2 with water injection – Carlson
initial reservoir conditions are as follows: hysteresis (“type” = 0). Note the krg curves are created
similarly to the Pc curves of Figure 8.
CASE 1: No initial gas saturation, water saturation • Figure 11, kro, Case 1 with oil injection – Killough
varying from 0.2 to 0.8, oil saturation varying from hysteresis (“type” = 4), Eclipse default three phase model.
0.8 to 0.2. See Figure 6. Viewing the oil relative permeability is more complicated
4 C. A. KOSSACK SPE 63147
since two sets of bounding curves are involved – krog and WAG Hysteresis Model Results
krow (imbibition and drainage). Two plots are given in Test of the WAG Hysteresis model demonstrate the
Figure 11. The left hand plot shows the kro scanning differences in the scanning curves generated with the WAG
curves plotted with the krog bounding curves vs. the Hysteresis model and the standard Killough or Carlson
saturation of oil. The right hand plot gives a mirror image hysteresis model.
of the same scanning curves (see line of symmetry
between the plots) and the bounding curves for krow vs. A Case 1 reservoir was subjected to gas injection with the
the inverted saturation of oil. For the scanning curves to WAG Hysteresis model activated using input values of the
be correct they must be inside of either the envelope Land’s parameter = 2.0 and the secondary drainage reduction
created by the krog bounding curves or the envelope factor7 α = 1.0. The krg reduction factor (for each cycle in the
created by the krow bounding curves. The case presented WAG) is given by:
contains valid scanning curves. α
S
Observing the difference in the scanning curves generated by wconnate
Killough or Carlson models is difficult with these plots. The S start
differences between the Killough and Carlson hysteresis w
models are most easily observed from a WAG displacement
and will be shown in a later section. An example of the resulting scanning curves is shown in
Figure 17. Here the krg values are lower than the initial
Invalid Scanning Curves Example drainage curve and independent of any input imbibition curve
Figure 12 show an example of incorrect scanning curves and as described Larsen and Skauge.3
what can happen if the bounding curves do not follow the
rules stated above. As one can observe, some calculated kro Since the WAG Hysteresis model is designed for Water-
values are much too high and outside both envelopes. These Alternating-Gas injections a 4 grid block linear, horizontal
scanning curves were created from a situation where the reservoir was created and initialized with Sw = 0.2 and So =
bounding curve end points did not coincide. The krg 0.8. Five WAG injection cycles (starting with water) were
imbibition table ended at a Sg value of 0.42 while the krg simulated with the WAG Hysteresis model and Killough
drainage table ended at a Sg of 0.0. Thus, the krg loop was not hysteresis (“type” = 4) for the oil phase.
closed. When one adds one additional point to the krg
imbibition table of krg = 0 at Sg = 0 the resulting scanning The interesting comparison is the krg scanning calculated with
curves are shown in Figure 13. These curves are correct. the Killough or Carlson model and with the WAG Hysteresis
model. Figure 18 shows the krg values in 3 of the four grid
Comparison of Scanning Curves for Different Three blocks (the black curve is from the grid block closest to the
Phase Models injector, green next, and red furthest) as the WAG injection is
The choice of the three-phase model makes a significant simulated. With the standard hysteresis model the krg does not
difference in the kro scanning curves. Figures 14, 15 and 16 decrease. Figure 19 shows the krg values when the WAG
show this difference. The reservoir is Case 1 with gas hysteresis model is used. Here the krg values are substantially
injection using Killough (“type” = 4) hysteresis. When gas is lower as the WAG cycles are simulated. In this process the
injected into this reservoir where initially Sg = 0, the oil gas mobility is much lower. Many additional results are
saturation in the grid blocks initially decreases. Some of the available in the full report.9
grid blocks away from the injector experience an increase in
the oil saturation as an oil bank passes by. Then the oil Results From a WAG Injection in a Linear Reservoir
saturation decreases to its residual value. This process creates To show the effect of the various hysteresis options, a WAG
the interesting loop shape in the scanning curves. injection was simulated into the 13 grid block linear,
horizontal reservoir with the initial Sw = 0.2 and So = 0.8 in all
Figure 14 gives the result from the Cheshire7 model (default grid blocks. Five WAG cycles were simulated starting with
Eclipse). Figures 15 and 16 gives the results when Stone I5 30 days of water injection followed by 30 days of gas
and Stone II6 models is used, respectively. A comparison injection. A total of 0.70 pore volumes were injected in the
shows that the Stone models calculate a much lower oil five cycles. Eight cases were simulated with the only change
relative permeability. Also some of the scanning curves for being the hysteresis option:
low oil saturations are outside of the bounding curve’s 1. Hysteresis OFF (drainage curves is used for all)
envelope for both Stone I and II. These kro values are not 2. Type = -1 (Equilibrate drainage, Sim. with imbibition)
correct. One must consider these results when choosing a 3. Type = 0 (Carlson, wetting phase uses drainage)
three-phase model. 4. Type = 1 (Carlson, wetting phase uses imbibition)
SPE 63147 COMPARISON OF RESERVOIR SIMULATION HYSTERESIS OPTIONS 5
5. Type = 2 (Killough, wetting phase uses drainage) and/or gas breakthrough times. A sensitivity study to the
6. Type = 3 (Killough, wetting phase uses imbibition) “type” parameter would show which model would
7. Type = 4 (Killough, all phases hysteresis) provide the best match of the history.
8. WAG Hysteresis (oil phase type = 4) 2. For prediction simulations of WAG displacements in the
absence of experimental results to define the hysteresis
The oil production rates, Figure 20, show a large variation model, sensitivity simulations should be run varying the
from effect of the hysteresis models. The cumulative oil “type” parameter so that the effect can be quantified.
produced, Figure 21, shows nearly a factor of 2 in the oil 3. To determine the proper hysteresis model for the reservoir
recovery. It is interesting to note that hysteresis “OFF” being evaluated, long core displacements using WAG
recovers the most oil in the 500-day simulation. The two injections should be run in the laboratory. Approximately
Carlson cases (types 0 and 1) recover the smallest amount of one pore volume should be injected in 5 or 6 cycles. The
oil. Type = -1 has a low initial oil rate and recovers more oil experiment should then be simulated and history matched
at the end of the simulation. Type = -1 is not a normal way to to determine the appropriate hysteresis model.
simulate since the relative permeabilities and capillary
pressures have discontinuous changes in the transition zone as Conclusions
the simulation starts and no hysteresis is used. 1. Properly oriented imbibition and drainage curves for
water-wet systems are provided as a guide for simulation
Figure 22 gives the gas production rate from the reservoir for engineers.
the eight cases. The Carlson hysteresis cases (type = 0 and 1) 2. A technique for showing scanning curves for the various
breakthrough gas first at about 125 days. The Killough hysteresis models is given. Example scanning curves are
hysteresis cases (type = 2, 3, and 4) breakthrough later at calculated and plotted for various hysteresis models and
about 200 days. With type = 2 the gas production stops at various three-phase models.
about 350 days and the breaks through again at 450 days. The 3. The scanning curves for the WAG Hysteresis model3 are
WAG Hysteresis case breaks through gas last (except for type shown.
= -1). This was expected since the WAG Hysteresis model 4. The results of a WAG displacement are compared for all
reduces the gas relative permeability with each cycle of the the water-wet hysteresis options. Substantial differences
WAG injection. in the results are observed and explained.
5. Recommendations to aid the simulation engineer’s choice
Figure 23 gives the water production rate from the reservoir. of hysteresis options are given.
There is a difference in breakthrough times of 250 days out of 6. A complete discussion of this topic and the Eclipse data
500 days between these cases. The key is what relative sets to generate the scanning curves are available from the
permeability curve is assigned to the water phase. Hysteresis author.9
types 1, 3, and 4 use the krw imbibition curve for the water
phase. In these cases breakthrough occurs first since the References
imbibition krw values are much higher that the drainage ones. 1. Killough, J. E.: ”Reservoir Simulation with History-
Hysteresis types 0 and 2 and the OFF case have a water Dependent Saturation Functions,” Trans. AIME 261,
breakthrough that is late since they use the drainage krw Page 37-48, 1976.
values, which are low. 2. Carlson, F. M.: “Simulation of Relative Permeability
Hysteresis to the Non-Wetting Phase,” SPE 10157, San
Recommendations: How to Choose the Correct Antonio, 1981.
Hysteresis Option 3. Larsen, J. A. and Skauge, A.: “Methodology for
The large variation in the results between the different cases in Numerical simulation with Cycle-dependent Relative
the WAG injection highlights the effect of the various Permeabilities,” submitted to SPEJ.
hysteresis models and options on displacements. As a 4. Land, C. E.: “Calculation of Imbibition Relative
simulation/reservoir engineer, one would hope to use the Permeability for Two- and Three-Phase Flow from Rock
model that most closely represents the rock-fluid interaction in Properties,” Soc. Pet. Eng. J., page 149-156, June 1968.
the field of interest. This information is not easy to obtain 5. Stone, H.: Trans AIME, 249, page 214-218, 1970.
since standard special core analysis does not provide this. The 6. Stone, H.: Can Pet. Tech., 12, page 53-61, 1973.
following recommendation will help engineers determine 7. ECLIPSE 100 Technical Description, Schlumberger
which “type” parameter should be used in simulations of their GeoQuest, 1999.
field: 8. ECLIPSE 300 Technical Description, Schlumberger
1. When history matching a WAG displacement the GeoQuest, 1999.
sensitivity of the results to the “type” of hysteresis model 9. Kossack, C. A.: “ECLIPSE Users How to Guide for
should be made when no information is available to Hysteresis: The Effect of Hysteresis Options in Eclipse”,
specify choice of the hysteresis model. As seen in the Schlumberger Holditch-Reservoir Technologies
WAG simulation in the linear reservoir, the choice of Publication, Denver, 1999.
“type” has a significant effect on oil production and water
6 C. A. KOSSACK SPE 63147
Acknowledgement
The author thanks Schlumberger Holditch-Reservoir
Technologies for permission to participate in preparation and
publication of the paper.
Figure 7. Case 2 Numerical Grid for Creating Scanning Figure 10. Reservoir Case 2 with Water Injection, Carlson
Curves Hysteresis (“type”=0) Scanning Curves for Gas Relative
Permeability
Figure 13. Corrected Scanning Curves by the Addition of One Figure 16. Reservoir Case 1 with Gas Injection, Killough
Point to Tables to Create a Loop with the Bounding Curves Hysteresis Model (“type”=4) kro Scanning Curves with Stone
II Three Phase Model
Figure 19. Four Grid Block Reservoir with 5 Cycle WAG Figure 22. Gas Production Rate from Eight Simulation Cases
Injection, WAG Hysteresis Model Scanning Curves for Gas Using Different Hysteresis Models – 13 Grid Block Linear
Relative Permeability Reservoir with WAG Injection
Figure 20. Oil Production Rate from Eight Simulation Cases Figure 23. Water Production Rate from Eight Simulation
Using Different Hysteresis Models – 13 Grid Block Linear Cases Using Different Hysteresis Models – 13 Grid Block
Reservoir with WAG Injection Linear Reservoir with WAG Injection