THE COMBINED INFLUENCE OF
GRAVITY, FLOW RATE AND SMALL SCALE HETEROGENEITY
ON CORE-SCALE MULTIPHASE FLOW OF CO2 AND BRINE
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF ENERGY
RESOURCES ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Chia-Wei Kuo
Dec 2012
iv
v
Abstract
The purpose of the research is to understand and predict the combined influences of
viscous, gravity and capillary forces in heterogeneous rocks over the range of conditions
relevant to storage of CO2 in deep underground geological formations. The study begins
by quantifying the separate and combined influences of flow rate, gravity, and sub-core
capillary heterogeneity on brine displacement using the 3-D simulator TOUGH2 (Kuo et
al. 2010). These studies demonstrate that the average saturation depends on the capillary
and gravity numbers in a predictable way. Based on the insight gained from numerical
simulation, this work develops an approximate semi-analytical solution for predicting the
average steady-state saturation during multiphase core flood experiments over a wide
range of capillary and gravity numbers as well as a wide range of heterogeneity.
Although computational technology has improved greatly, running high-resolution 3D
models including capillarity, gravity, and capillary heterogeneity still takes a significant
amount of computational effort. The new solution provided here is a quick and easy way
to estimate the flow regimes for horizontal core floods.
A two dimensional analysis of the governing equations for the multiphase flow system
at steady-state is used to develop the approximate semi-analytical solution. We have
developed a new criterion to identify the viscous-dominated regime at the core scale
where the average saturation is independent of flow rate. Variations of interfacial tension,
core permeability, length of the core, and the effects of buoyancy, capillary and viscous
forces are all accounted for in the semi-analytical solutions. We have also shown that
three dimensionless numbers (NB, Ngv, Rl) and two critical gravity numbers (Ngv,c1, Ngv,c2)
are required to properly capture the balance of viscous, gravity, and capillary forces.
There is good agreement between the average saturations calculated from the 3-D
simulations and the analytical model. This new model can be used to design and interpret
multiphase flow core-flood experiments, gain better understanding of multiphase flow
displacement efficiency over a wide range of conditions and for different fluid pairs, and
perhaps even provide a tool for studying the influence of sub-grid scale multiphase flow
phenomena on reservoir-scale simulations.
vi
One practical application for the new semi-analytical solution is to help design and
interpret core flood experiments, including assuring that relative permeability
measurements are made in the viscous dominated regime, evaluating potential flow rate
dependence, influence of core-dimension on a multiphase flow experiments, and
influence of fluid properties on the experiments.
New guidelines and suggestions for making relative permeability measurements are
presented. Results are based on a combination of high resolution of 3D simulations and
core-flooding experiments with X-ray CT scanning of saturation distributions. Effects of
flow rate, permeability, surface tension, core length, boundary conditions, sub-core scale
heterogeneity, and gravity over a range of fractional flows of CO2 are systematically
investigated. Synthetic “data sets” are generated using TOUGH2 and subsequently used
to calculate relative permeability curves. A comparison between the input relative
permeability curves and “calculated” relative permeability is used to assess the accuracy
of the “measured” values. Results show that for a modified capillary number
(Ncv=kLpc*A/H
2μCO2qt) smaller than 15, flows are viscous dominated. Under these
conditions, saturation depends only on the fractional flow and is independent of flow rate,
gravity, permeability, core length and interfacial tension. For modified capillary numbers
less than 15, accurate whole-core relative permeability measurements can be obtained
regardless of the orientation of the core and for a high degree of heterogeneity under a
range of relevant and practical conditions. Importantly, the transition from the viscous to
gravity/capillary dominated flow regimes occurs at much higher flow rates for
heterogeneous rocks. For modified capillary numbers larger than 15, saturation gradients
develop along the length of the core and accurate relative permeability measurements are
not obtained using traditional steady state methods. However, if capillary pressure
measurements at the end of the core are available, or can be estimated from
independently measured capillary pressure curves and the measured saturation at the inlet
and outlet of the core, accurate relative permeability measurements can be obtained even
when there is a small saturation gradient across the core.
vii
Acknowledgments
First and foremost, I owe my deepest gratitude to my advisor, Professor Sally M. Benson,
who not only served as my supervisor but also encouraged and challenged me throughout
my graduate program. Sally has been providing her wide knowledge and advice during
my Ph.D. study and gave me a lot of freedom to explore the research area what I am
interested in. I am truly thankful for her patience, inspiring guidance and the unlimited
time she has spent to correct my proposal, paper, and this dissertation. This dissertation
could not have been done without her encouragement and help and I could not express
my sincere appreciation to her.
I wish to thank all the members of my reading and examination committees:
Professor Tony Kovscek and Margot Gerritsen for their careful reading of the entire
dissertation and many constructive comments on this work as well as supervising my
qualification and Ph.D. dissertation. I am greatly indebted for their guidance and advice.
My special thanks to Professor Keith Loague for agreeing to chair my oral defense and
the examiner Professor Jef Caers who gave me insightful suggestions during my
qualification.
I am blessed with having kind and generous friends and colleagues during my Ph.D.
study. I like to thank to the former and current group members in the Benson’s group.
Special thanks to Ljuba Miljkovic who taught me at the very beginning for how to run the
TOUGH2 simulation step by step; Mike Krause who helped me a lot either for the
problems I accounted for and Boxiao Li who discuss the detail of the simulation; Jean-
Christophe Perrin, Sam Krevor, Ronny Pini, and Lin Zuo who gave me the general idea
of the experimental setting; Ariel Esposito, Christin Strandli, and Whitney Sargent who
shared the room together; Ethan Chabora, Karim Farhat, and Da Huo for their friendship
and useful suggestions. In summary, working with them was a valuable experience for
me.
viii
In addition, thank to Ni (Panithita Rochana), Prare (Monrawee Pancharoen),
Yangyang Liu, Xiaochen Wang, Yang Liu, Zhe Wang and all the other students of the
Department of Energy Resources Engineering for making my years at the department so
much easier and enjoyable. I also like to gratefully acknowledge the financial support of
the Global Climate and Energy Project (GCEP) at Stanford University.
Finally, I owe my deepest gratitude and respect to my family for their tremendous,
unconditional support during these years. I am deeply grateful to my dearest fiancé
Hsuan-Tung Chu, waiting for me all these years with his understanding, patience, and
love enabled me to complete this work. This dissertation is dedicated with love to my
family.
ix
Contents
ABSTRACT ................................................................................................................................................... V
ACKNOWLEDGMENTS ........................................................................................................................... VII
CONTENTS ................................................................................................................................................. IX
LIST OF TABLES ...................................................................................................................................... XII
LIST OF FIGURES .................................................................................................................................... XIV
CHAPTER 1 INTRODUCTION ..................................................................................................................... 1
1.1 BACKGROUND/STATEMENT OF THE PROBLEM ................................................................................................. 1 1.2 OUTLINE OF THE APPROACH......................................................................................................................... 5 1.3 DISSERTATION OUTLINE/ORGANIZATION ........................................................................................................ 6
CHAPTER 2 LITERATURE REVIEW .......................................................................................................... 8
2.1 MULTIPHASE FLOW IN THE VISCOUS AND CAPILLARY LIMITS ............................................................................... 8 2.2 MULTIPHASE FLOW WITH VISCOUS, CAPILLARY AND GRAVITY FORCES ................................................................. 9 2.3 CAPILLARY HETEROGENEITY ....................................................................................................................... 11 2.4 RELATIVE PERMEABILITY MEASUREMENTS .................................................................................................... 13 2.5 DISCUSSION ............................................................................................................................................ 16
CHAPTER 3 SIMULATION METHODOLOGY ........................................................................................ 18
3.1 TOUGH2 MP/ECO2N ........................................................................................................................... 20 3.2 CORE DESCRIPTIONS AND GRID SIZE ............................................................................................................ 24 3.3 INPUT PARAMETERS ................................................................................................................................. 25
3.3.1 Boundary Conditions ................................................................................................................... 25 3.3.2 Initial Conditions ......................................................................................................................... 27 3.3.3 Input Capillary Pressure .............................................................................................................. 27 3.3.4 Input Relative Permeability ......................................................................................................... 30 3.3.5 Input Injection Flow Rate ............................................................................................................ 31
3.4 OUTPUT PARAMETERS .............................................................................................................................. 33 3.4.1 Slice-Average Quantities (SCO2, PCO2, Pc) ...................................................................................... 33 3.4.2 Simulated CO2 Saturation Distributions ...................................................................................... 34
CHAPTER 4 3D NUMERICAL AND 2D ANALYTICAL STUDIES FOR THE HOMOGENEOUS
CORES .......................................................................................................................................................... 36
4.1 METHODOLOGY ....................................................................................................................................... 37 4.1.1 Simulation Study ......................................................................................................................... 37
4.2 THEORETICAL ANALYSIS ............................................................................................................................. 44 4.2.1 2D General Solution .................................................................................................................... 44 4.2.2 2D General Solution with Simulation Constraints ....................................................................... 46
4.3 RESULTS ................................................................................................................................................. 48
x
4.3.1 Comparison between Simulations and the Approximate Semi-Analytical Solution ................... 48 4.3.2 Model Validation ........................................................................................................................ 50
4.4 DISCUSSION ........................................................................................................................................... 53 4.5 CONCLUSIONS ........................................................................................................................................ 56
CHAPTER 5 3D NUMERICAL AND 2D ANALYTICAL STUDIES FOR THE HETEROGENEOUS
CORES ......................................................................................................................................................... 58
5.1 HETEROGENEOUS REPRESENTATIONS .......................................................................................................... 59 5.1.1 Random Permeability Distribution (3D) ..................................................................................... 59 5.1.2 Porosity-based Permeability Distribution (3D) ........................................................................... 60 5.1.3 Permeability Distribution Summary ........................................................................................... 62
5.2 SIMULATION STUDIES............................................................................................................................... 64 5.2.1 Heterogeneity Effects ................................................................................................................. 64 5.2.2 Interfacial Tension and Permeability Effects .............................................................................. 65
5.3 THEORETICAL ANALYSIS OF MULTIPHASE DISPLACEMENT EFFICIENCY IN HETEROGENEOUS CORES ........................... 68 5.3.1 2D General Solution.................................................................................................................... 69 5.3.2 2D General Solution for Heterogeneous Rocks Using Simulation Constraints ........................... 71 5.3.3 Approximate Semi-Analytical Solution ....................................................................................... 74
5.4 VERIFICATION OF THE ANALYTICAL MODEL ................................................................................................... 77 5.4.1 Different Interfacial Tension and Permeability .......................................................................... 77 5.4.2 Different Fractional Flows of CO2 (HC Model) ............................................................................ 78 5.4.3 Different Core Dimensions .......................................................................................................... 79 5.4.4 Different Heterogeneity ............................................................................................................. 80
5.5 PROCEDURES FOR USING THE ANALYTICAL SOLUTIONS ................................................................................... 82
CHAPTER 6 CO2 AND BRINE RELATIVE PERMEABILITY IN HETEROGENEOUS ROCKS .......... 85
6.1 SIMULATION OUTPUTS ............................................................................................................................. 86 6.2 CALCULATION OF RELATIVE PERMEABILITY ................................................................................................... 89
6.2.1 Relative Permeability Calculated when ΔPw=ΔPCO2 .................................................................... 89 6.2.2 Relative Permeability Calculated by True Pressure Drops (ΔPw=ΔPCO2–ΔPc) .............................. 94
6.3 SENSITIVITY STUDIES FOR DIFFERENT CORE PROPERTIES ................................................................................. 97 6.3.1 Effects of Heterogeneity ............................................................................................................. 97 6.3.2 Effects of Core Length (15.24-45.72 cm) .................................................................................... 99 6.3.3 Effects of Absolute Permeability (31.8-3180 md) ....................................................................... 99 6.3.4 Effects of Interfacial Tension (7.49-67.41 mN/m) .................................................................... 100 6.3.5 Effects of Gravity ...................................................................................................................... 102
6.4 DISCUSSION OF RELATIVE PERMEABILITY MEASUREMENTS ............................................................................ 103 6.4.1 Observations from the Numerical and Semi-Analytical Models ............................................... 103 6.4.2 General Rule of Thumb for Reliable Relative Permeability Measurements .............................. 104
CHAPTER 7 CONCLUSIONS AND FUTURE WORK ........................................................................... 108
7.1 SUMMARY AND CONCLUSIONS OF THE PRESENT WORK ................................................................................ 108 7.2 DIRECTIONS FOR FUTURE RESEARCH ......................................................................................................... 110
APPENDIX A 2D ANALYTICAL DERIVATION FOR THE HOMOGENEOUS MODEL .................. 113
A.1 DETAIL OF 2D GENERAL SOLUTION .......................................................................................................... 113 A.2 DERIVATION OF PARAMETERS A AND B ...................................................................................................... 116
APPENDIX B 2D ANALYTICAL DERIVATION FOR THE HETEROGENEOUS MODEL ............... 118
B.1 GENERAL SOLUTION .............................................................................................................................. 118
xi
B.2 ANALYTICAL SOLUTION WITH SIMULATION CONSTRAINTS .............................................................................. 122 B.3 HETEROGENEOUS FACTOR, ΣLNK/LN(KMEAN)................................................................................................... 125
APPENDIX C FLOW REGIMES ............................................................................................................... 126
C.1 CRITICAL GRAVITY NUMBERS ................................................................................................................... 126
APPENDIX D PARAMETER TABLES FOR HETEROGENEOUS CORES.......................................... 128
APPENDIX E VALIDATION OF TOUGH2 SIMULATIONS ................................................................ 129
E.1 RELATIVE CONVERGENCE TOLERANCE Ε1=10-5
............................................................................................. 129 E.2 GPRS .................................................................................................................................................. 130
NOMENCLATURE .................................................................................................................................... 131
GREEKS ................................................................................................................................................... 132 SUBSCRIPTS ............................................................................................................................................ 133
BIBLIOGRAPHY ....................................................................................................................................... 134
xii
List of Tables
TABLE 1. 1 COMPARISON OF THE DIFFERENCES AND SIMILARITIES BETWEEN THE THREE-
DIMENSIONAL NUMERICAL SIMULATIONS AND TWO-DIMENSIONAL THEORETICAL
DERIVATION. .............................................................................................................................................. 6
TABLE 2. 1 SUMMARY OF THE LIMITING CASES (ZHOU ET AL. 1994). ........................................ 10
TABLE 2. 2 SUMMARY OF THE LIMITING CASES (YORTSOS 1995). .............................................. 11
TABLE 3. 1 SIMULATION SUMMARY. .................................................................................................. 19
TABLE 3. 2 SUMMARY OF ECO2N. ........................................................................................................ 20
TABLE 3. 3 BOUNDARY CONDITION SUMMARY. ............................................................................. 27
TABLE 3. 4 INPUT PARAMETER VALUES FOR RELATIVE PERMEABILITY AND CAPILLARY
PRESSURE CURVES FIT. .......................................................................................................................... 29
TABLE 3. 5 THE EFFECT OF POROSITY, PERMEABILITY, AND CAPILLARY PRESSURE ON CO2
SATURATION DISTRIBUTION. ............................................................................................................... 30
TABLE 4. 1: SUMMARY OF SENSITIVITY STUDIES ........................................................................... 39
TABLE 4. 2: VALUES OF UNKNOWN VARIABLES USED TO MATCH THE BASE CASE. ............ 48
TABLE 4. 3: BEREA CORE PROPERTIES AND FLUID PROPERTIES USED IN THE
HOMOGENEOUS CORES FOR THE BASE CASE .................................................................................. 49
TABLE 4. 4: SUMMARY OF FLOW REGIONS FOR GENERAL CASES .............................................. 54
TABLE 5. 1 SUMMARY OF SIMULATION CASES WITH DIFFERENT DEGREES OF
HETEROGENEITY. .................................................................................................................................... 63
TABLE 5. 2: SUMMARY OF SENSITIVITY STUDIES FOR HIGH CONTRAST AND RANDOM 3
MODELS. ..................................................................................................................................................... 66
TABLE 5. 3: SUMMARY OF CONSTANT COEFFICIENTS IN EQ. 5.12. ............................................. 74
TABLE 6. 1 SYNTHETIC INPUT PARAMETERS FOR EVERY GRID IN THE SIMULATIONS FOR
THREE DIFFERENT MODELS. ................................................................................................................ 87
TABLE 6. 2 SUMMARY OF DIFFERENT FLOW RATES FOR HOMOGENEOUS CORES WITH 95%
FRACTIONAL FLOW OF CO2. CAPILLARY NUMBERS ARE CALCULATED BASED ON EQ. 4.1
xiii
AND EQ. 4.2, RESPECTIVELY. ................................................................................................................. 91
TABLE 6. 3 SUMMARY OF DIFFERENT FLOW RATES FOR HETEROGENEOUS CORES (HIGH
CONTRAST MODEL: ΣLNK/LNKMEAN=0.168) WITH 95% FRACTIONAL FLOW OF CO2. ................... 94
TABLE 6. 4 SUMMARY OF DIFFERENT FLOW RATES FOR DIFFERENT HETEROGENEOUS
CORES: HOMOGENEOUS, RANDOM 2, KOZENY-CARMAN, HIGH CONTRAST AND RANDOM 3
MODELS. ..................................................................................................................................................... 98
TABLE 6. 5 SUMMARY OF DIFFERENT LENGTHS OF HETEROGENEOUS CORES (HIGH
CONTRAST MODEL: ΣLNK/LNKMEAN=0.168). ........................................................................................... 99
TABLE 6. 6 SUMMARY OF THE FIRST CRITICAL CAPILLARY NUMBER NCV,C1 FOR DIFFERENT
HETEROGENEITY CORES. ..................................................................................................................... 104
TABLE D.1: SUMMARY OF SENSITIVITY STUDIES FOR HIGH CONTRAST MODELS (HC). ..... 128
TABLE D.2: SUMMARY OF SENSITIVITY STUDIES FOR RANDOM 3 MODELS. ......................... 128
TABLE D.3: SUMMARY OF SENSITIVITY STUDIES FOR DIFFERENT HETEROGENEOUS
MODELS. ................................................................................................................................................... 128
xiv
List of Figures
FIGURE 1. 1 DIFFERENT FLOW REGIMES DOMINATED BY DIFFERENT FORCES IN THE
STORAGE RESERVOIR. .............................................................................................................................. 3
FIGURE 1. 2 VARIOUS RELATIVE PERMEABILITY CURVES AT INSITU CONDITIONS FOR CO2
/BRINE SYSTEM (A) BASAL CAMBRIAN, (B) COOKING LAKE, (C) NISKU, (D) WABAMUN
(LOW PERMEABILITY), (E) ELLERSLIE, (F) VIKING, (G) WABAMUN (HIGH PERMEABILITY),
AND (H) FRIO SANDSTONE ...................................................................................................................... 4
FIGURE 3. 1 THE EXPERIMENTAL STEADY STATE THREE-DIMENSIONAL VIEWS OF CO2
SATURATION IN THE CORE FOR A GIVEN FRACTIONAL FLOW OF CO2 AT A GIVEN FLOW
RATE. THE FLUIDS WERE INJECTED FROM RIGHT TO LEFT (PERRIN ET AL, 2009). ............... 18
FIGURE 3. 2 SIMULATION CORE OF 25 25 31 GRID BLOCKS OF UNIFORM SIZE. ................. 19
FIGURE 3. 3 SPACE DISCRETIZATION IN THE INTEGRAL FINITE DIFFERENCE METHOD. ..... 21
FIGURE 3. 4 CO2 SATURATION ALONG THE BEREA SANDSTONE CORE FOR DIFFERENT
FRACTIONAL FLOWS OF CO2 AT A TOTAL INJECTION FLOW RATE 2.6 ML/MIN: (A)
EXPERIMENTAL RESULTS; (B) HIGH CONTRAST MODEL WITH BOUNDARY CONDITION
PC=0; (C) HIGH CONTRAST MODEL WITH BOUNDARY CONDITION DPC/DX=0. ......................... 26
FIGURE 3. 5 LABORATORY CAPILLARY PRESSURE DATA WITH A CURVE FIT USED IN
SIMULATIONS. .......................................................................................................................................... 28
FIGURE 3. 6 THE RANGE OF CAPILLARY PRESSURE CURVES IN THE SIMULATIONS. THE
VALUES OF INPUT PARAMETERS ARE A1=0.007734, B1=0.307601, Λ1=2.881, Λ2=2.255, SP=0 AND
Σ =22.47 DYNES/CM. ................................................................................................................................. 29
FIGURE 3.7 INPUT RELATIVE PERMEABILITY CURVES FOR CO2 AND BRINE WITH SWR =0.15,
NW =7, AND NCO2=3 AND THE EXPERIMENTAL DATA. ..................................................................... 31
FIGURE 3.8 CONCEPTUAL MODEL OF THE RESERVOIR USED TO ESTIMATE THE RANGE OF
RELEVANT FLOW VELOCITIES. ............................................................................................................ 32
FIGURE 3.9 THE VOLUMETRIC FLOW RATE AND ITS CORRESPONDING DISTANCE OF CO2
PLUME AT THE RESERVOIR (100 METER THICKNESS AND 1 MT CO2/YR INJECTION RATE). 32
FIGURE 3.10 CO2 SATURATION DISTRIBUTION AT STEADY-STATE FOR 95% FRACTIONAL
xv
FLOW OF CO2 AT A TOTAL INJECTION FLOW RATE 1.2 ML/MIN. .................................................. 33
FIGURE 3.11 SATURATION DISTRIBUTION FOR HOMOGENEOUS AND HETEROGENEOUS
CORES WITH GRAVITY, USING GRID WITH PIXELS (150, 150, 31) TO GRID BLOCKS (25, 25, 31).
....................................................................................................................................................................... 35
FIGURE 4. 1: AVERAGE CO2 SATURATION AS A FUNCTION OF (A) TRADITIONAL CAPILLARY
NUMBER CA, (B) ALTERNATIVE CAPILLARY NUMBER NCV, AND (C) ALTERNATIVE
GRAVITY NUMBER NGV. THE INTERFACIAL TENSION Σ AND THE PERMEABILITY OF THE
BASE CASE ARE 22.47 MN/M AND 430 MD, RESPECTIVELY. THE SENSITIVITY STUDIES OF
INTERFACIAL TENSION ARE ILLUSTRATED IN THE LEFT HAND SIDE OF THE FIGURES AND
PERMEABILITY ON THE RIGHT. ............................................................................................................ 41
FIGURE 4. 2: PRESSURE GRADIENTS FOR THREE DIFFERENT FORCES AS A FUNCTION OF
ALTERNATIVE GRAVITY NUMBER NGV. THE INTERFACIAL TENSION AND THE
PERMEABILITY OF THE 3Σ CASE ARE 67.41 MN/M AND 430 MD, RESPECTIVELY. .................... 42
FIGURE 4. 3: THE FUNCTION ΔJ HAS A LINEAR DEPENDENCY ON THE SQUARE ROOT OF
ASPECT RATIO, RL. .................................................................................................................................... 44
FIGURE 4. 4 (LHS) COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF
CAPILLARY NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS
FOR THE HOMOGENEOUS OR BASE CASE (Σ, K); (RHS) FRACTIONAL FLOW CURVE BASED
ON OUR INPUT RELATIVE PERMEABILITY CURVES (EQ.3.6). ........................................................ 48
FIGURE 4. 5 COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR THE
SENSITIVITY CASES OF INTERFACIAL TENSION. LHS: (3Σ, K), RHS: (Σ/3, K) .............................. 50
FIGURE 4. 6 COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR THE
SENSITIVITY CASES OF DIFFERENT DIMENSIONS OF THE CORE. ................................................ 51
FIGURE 4. 7 COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR THE
SENSITIVITY CASES OF DIFFERENT FRACTIONAL FLOWS OF CO2. ............................................. 52
FIGURE 4. 8 COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR SIX
DIFFERENT INPUT RELATIVE PERMEABILITY CURVES. THOSE SIMULATIONS SHARE THE
SAME CONSTANT PARAMETERS IN TABLE 4.3. ................................................................................ 53
FIGURE 4. 9 (LHS) THE FINER GRIDS HAVE GRID DIMENSIONS 0.884MM X 0.884MM X 5.08MM
(GRID NUMBERS: 53X53X31) WHILE THE COARSER GRIDS HAVE DIMENSIONS 1.874MM X
xvi
1.874MM X 5.08MM (GRID NUMBERS: 25X25X31); (RHS) CFL NUMBER (UXΔT/ΔX<1) FOR
DIFFERENT FLOW RATE. ........................................................................................................................ 56
FIGURE 5. 1: FOUR DIFFERENT RANDOM PERMEABILITY PROFILES AND THE
CORRESPONDING AVERAGE CO2 SATURATIONS IN THE VISCOUS-DOMINATED REGIME. .. 60
FIGURE 5. 2: AVERAGE PERMEABILITY AND AVERAGE CO2 SATURATION ALONG THE
FLOW DIRECTION FOR THREE DIFFERENT HETEROGENEOUS CORES (HOMOGENEOUS
MODEL, KOZENY-CARMAN MODEL, AND HIGH CONTRAST MODEL). ....................................... 62
FIGURE 5. 3: AVERAGE CO2 SATURATION AS A FUNCTION OF ALTERNATIVE CAPILLARY
NUMBER NCV. THE INTERFACIAL TENSIONS Σ FOR ALL CASES IS 22.47 MN/M AND THE
EFFECTIVE PERMEABILITY VARIES FROM 254 TO 570 MD. ........................................................... 65
FIGURE 5. 4: AVERAGE CO2 SATURATION AS A FUNCTION OF TRADITIONAL CAPILLARY
NUMBER CA, ALTERNATIVE CAPILLARY NUMBER NCV, AND ALTERNATIVE GRAVITY
NUMBER NGV FOR TWO HETEROGENEOUS CORES: HIGH CONTRAST MODEL (LEFT) AND
RANDOM 3 MODEL (RIGHT). THE INTERFACIAL TENSION Σ IS KEPT AS A CONSTANT, 22.47
MN/M. .......................................................................................................................................................... 67
FIGURE 5. 5: PRESSURE GRADIENTS AS A FUNCTION OF ALTERNATIVE CAPILLARY
NUMBER NCV FOR THREE PHYSICAL FORCES. .................................................................................. 68
FIGURE 5. 6: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR THE
SENSITIVITY CASES OF DIFFERENT DEGREES OF HETEROGENEOUS CORES. ......................... 74
FIGURE 5. 7: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR HIGH
CONTRAST MODEL AT A WIDE RANGE OF PERMEABILITY AND INTERFACIAL TENSION. .. 75
FIGURE 5. 8: THE DEPENDENCE OF Ε AND Ω IN TERMS OF THE MODIFIED BUCKLEY-
LEVERETT SATURATION (SBLHETE
) AND THE BOND NUMBER (NB) FOR HIGH CONTRAST AND
RANDOM 3 MODELS. ............................................................................................................................... 76
FIGURE 5. 9: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY
NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS FOR (A) HIGH
CONTRAST MODEL AND (B) RANDOM PERMEABILITY MODEL IN A WIDE RANGE OF
PERMEABILITY AND INTERFACIAL TENSION. ................................................................................. 78
FIGURE 5. 10: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF
CAPILLARY NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS
FOR HIGH CONTRAST MODEL AT DIFFERENT FRACTIONAL FLOWS OF CO2. .......................... 79
FIGURE 5. 11: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF
xvii
CAPILLARY NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS
FOR THE SENSITIVITY CASES OF DIFFERENT ASPECT RATIO. ..................................................... 80
FIGURE 5. 12: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF
CAPILLARY NUMBER NCV BETWEEN HOMOGENEOUS ANALYTICAL PREDICTIONS AND
SIMULATION RESULTS FOR THE SENSITIVITY CASES OF SMALL DEGREES OF
HETEROGENEOUS CORES. ...................................................................................................................... 81
FIGURE 5. 13: COMPARISON OF AVERAGE CO2 SATURATION AS A FUNCTION OF
CAPILLARY NUMBER NCV BETWEEN THEORETICAL VALUES AND SIMULATION RESULTS
FOR THE SENSITIVITY CASES OF DIFFERENT DEGREES OF HETEROGENEOUS CORES. ........ 82
FIGURE 6. 1 OVERVIEW OF SCIENTIFIC APPROACH. ........................................................................ 86
FIGURE 6. 2 3D POROSITY DISTRIBUTION OF BEREA SANDSTONE. ............................................. 87
FIGURE 6. 3 CO2 SATURATION DISTRIBUTION AT STEADY-STATE FOR 95% FRACTIONAL
FLOW OF CO2 AT A TOTAL INJECTION FLOW RATE 1.2 ML/MIN. .................................................. 88
FIGURE 6. 4 FLOW RATE EFFECT ON CO2 SATURATION ALONG THE HOMOGENEOUS CORE
AT A 95% FRACTIONAL FLOW OF CO2 WITH FLOW RATES RANGING FROM 0.1 ML/MIN TO 6
ML/MIN. ....................................................................................................................................................... 90
FIGURE 6. 5 RELATIVE PERMEABILITY CALCULATED BY THE SAME PRESSURE DROP
(ΔPW=ΔPCO2) FOR HOMOGENEOUS CORE WITH 430 MD PERMEABILITY AT DIFFERENT FLOW
RATES. ......................................................................................................................................................... 92
FIGURE 6. 6 THE EFFECT OF HETEROGENEITY ON CO2 SATURATION ALONG THE CORE AT A
FRACTIONAL FLOW OF 95% OVER A WIDE RANGE OF FLOW RATES. ........................................ 93
FIGURE 6. 7 RELATIVE PERMEABILITY CALCULATED BY THE SAME PRESSURE DROP
(ΔPW=ΔPCO2) FOR HETEROGENEOUS CORE WITH 318 MD PERMEABILITY AT DIFFERENT
FLOW RATES. SMALL PICTURE SHOWS THE SAME RELATIVE PERMEABILITY CURVES AT
LOG SCALE. ................................................................................................................................................ 94
FIGURE 6. 8 FLOW RATE EFFECT ON RELATIVE PERMEABILITY CALCULATED BY THE TRUE
PRESSURE DROPS (ΔPW ΔPCO2) FOR HOMOGENEOUS AND HETEROGENEOUS CORE AT
VARIOUS FLOW RATES. .......................................................................................................................... 95
FIGURE 6. 9 (LHS) RELATIVE PERMEABILITY CALCULATED USING THE SAME PRESSURE
DROP FOR BOTH FLUIDS (FIGURE 6.7); (RHS) RELATIVE PERMEABILITY CALCULATED BY
THE CORRECTED PRESSURE DROPS FOR HIGH CONTRAST MODELS AT VARIOUS FLOW
RATES. ......................................................................................................................................................... 96
FIGURE 6. 10: RELATIVE PERMEABILITY CALCULATED BY THE TRUE PRESSURE DROPS
xviii
FOR FIVE DIFFERENT HETEROGENEOUS CORES IN THE VISCOUS-DOMINATED REGIMES:
HOMOGENEOUS AND THE RANDOM 2 CORES (Q=0.5 ML/MIN), KOZENY-CARMAN MODELS
(Q=1.2 ML/MIN), HIGH CONTRAST MODELS (Q=2.6 ML/MIN), THE RANDOM 3 CORES (Q=6
ML/MIN) AND THE INPUT RELATIVE PERMEABILITY CURVES. ................................................... 98
FIGURE 6. 11 (LHS) BRINE DISPLACEMENT EFFICIENCIES FOR THREE DIFFERENT LENGTHS
OF HOMOGENEOUS CORE WITH CAPILLARY NUMBER RANGING FROM 10 TO 107; (RHS)
RELATIVE PERMEABILITY CALCULATED BY THE TRUE PRESSURE DROPS FOR
HOMOGENEOUS CORES AT 0.1 ML/MIN FLOW RATES. ................................................................... 99
FIGURE 6. 12 (LHS) BRINE DISPLACEMENT EFFICIENCIES FOR THREE DIFFERENT
PERMEABILITY VALUES WITH CAPILLARY NUMBER NCV RANGING FROM 10 TO 105; (RHS)
PERMEABILITY EFFECTS ON RELATIVE PERMEABILITY CALCULATED IN THE TRUE
PRESSURE DROPS FOR HETEROGENEOUS CORE (HIGH CONTRAST MODEL) AT 6 ML/MIN
FLOW RATES. .......................................................................................................................................... 100
FIGURE 6. 13 (LHS) AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY NUMBER
NCV FOR HOMOGENEOUS AND HIGH CONTRAST MODELS WITH THREE DIFFERENT VALUES
OF INTERFACIAL TENSIONS; (RHS) INTERFACIAL TENSION EFFECTS ON RELATIVE
PERMEABILITY CALCULATED IN THE TRUE PRESSURE DROPS FOR HETEROGENEOUS
CORE (HIGH CONTRAST MODEL) AT 6 ML/MIN FLOW RATES. ................................................... 101
FIGURE 6. 14 AVERAGE CO2 SATURATION AS A FUNCTION OF CAPILLARY NUMBER NCV FOR
HOMOGENEOUS, KOZENY-CARMAN (SMALL HETEROGENEITY) AND HIGH CONTRAST
(LARGE HETEROGENEITY) MODELS WITH AND WITHOUT GRAVITY (1G/0G). ...................... 102
FIGURE 6. 15 AVERAGE CO2 SATURATION AS A FUNCTION OF ALTERNATIVE CAPILLARY
NUMBER NCV, AND ALTERNATIVE GRAVITY NUMBER NGV FOR HOMOGENEOUS AND HIGH
CONTRAST MODELS. ............................................................................................................................. 105
FIGURE 6. 16 LABORATORY CAPILLARY PRESSURE DATA. ........................................................ 105
FIGURE 6. 17 HETEROGENEOUS PARAMETER Τ IN TERMS OF NORMALIZED STANDARD
DEVIATION. ............................................................................................................................................. 107
FIGURE C.1: PRESSURE GRADIENTS FOR VISCOUS, CAPILLARY AND GRAVITY FORCES FOR
DIFFERENT ASPECT RATIOS. .............................................................................................................. 126
FIGURE C.2: PRESSURE GRADIENTS FOR VISCOUS, CAPILLARY AND GRAVITY FORCES FOR
DIFFERENT INTERFACIAL TENSION AND PERMEABILITY VALUES. ........................................ 127
FIGURE C.3: PRESSURE GRADIENTS FOR VISCOUS, CAPILLARY AND GRAVITY FORCES FOR
DIFFERENT PERMEABILITY VALUES FOR HIGH CONTRAST MODEL. ...................................... 127
FIGURE E. 1 CO2 SATURATION AS A FUNCTION OF PORE VOLUME INJECTED (PVI) FOR
xix
DIFFERENT FLOW RATES. .................................................................................................................... 130
FIGURE E. 2 GPRS VS TOUGH2 RESULTS FOR SLICE X=16 (LHS) AND X=21 (RHS) .................. 130
1
Chapter 1
Introduction
CO2 sequestration in saline aquifers requires knowledge of multiphase flow of CO2 and
brine, specifically, the interaction between viscous, gravity, and capillary forces in
heterogeneous rocks at the reservoir scale. This is an important and challenging task, to
which insights from core-scale experiments and simulations can be applied. The
purposes of this work are to first, gain insight about these processes by examining core
scale experiments, second, quantify the influence and interplay of these forces using
numerical simulation, and finally, to develop a theoretical model to predict the average
saturation during multiphase displacements and identify different flow regimes based on
dimensional variables. As an application of these ideas, the results are used to design
core laboratory experiments for obtaining accurate relative permeability data for
heterogeneous and homogeneous cores.
1.1 Background/Statement of the Problem
Carbon dioxide capture and sequestration in deep geological formations is one of the
most important technologies for climate change mitigation (IPCC, 2005). Although
depleting or abandoned oil and gas reservoirs are available in some regions like Texas in
the US, the Middle East, Russia, and Alberta in Canada, they have lower capacity than is
needed to store worldwide CO2 emissions from large stationary sources. Furthermore,
these depleted reservoirs are not common all over the world. Moreover, their capacities
are available only when the reservoirs are depleted or if CO2 sequestration is combined
with CO2–EOR. Thus, storage of CO2 in depleted or abandoned oil and gas fields is
2
limited. Therefore, the large storage capacity of worldwide distributed saline aquifers
makes them good options to store CO2 captured from power generation stations and
industrial sources, which are the largest stationary sources emitting CO2 (IPCC, 2005).
Moreover, sequestration of CO2 in deep saline formations is immediately accessible
compared to the other options.
Carbon dioxide and water are immiscible under the conditions anticipated in
geological sequestration reservoirs (1,000 to 3,000 m depth). Therefore, the migration of
injected CO2 and brine displacement will be governed by the physics of multiphase flow.
There are many studies of multiphase flow of oil and water in petroleum reservoirs, but
because the viscosity, density, interfacial tension of CO2 is different that oil, conclusions
drawn from oil and water systems may not be applicable in CO2 and brine systems. For
example, since the density difference between CO2 and brine is typically much larger
than between oil and water, buoyancy forces will play a greater role. Similarly, the low
viscosity CO2 relative to oil, results in smaller pressure gradients and higher mobility to
CO2. Moreover, the flow velocities in a typical CO2 storage project will be much lower
than during oilfield recovery, especially far from the injection well and during the post-
injection period.
To date, actual experience with multiphase flow of CO2 in saline aquifers is limited to
the Sleipner Project (Kaarstad, 2004) and a number of smaller pilot tests (ex: Frio Project,
2006). Key issues related to the feasibility of deep saline aquifer storage include:
1. What will be the average saturation of CO2 in the plume?
2. How far will the injected CO2 migrate?
3. How much CO2 will be dissolved in the native brine and how quickly?
4. How much CO2 will be immobilized by capillary trapping?
5. What fraction, if any, of the CO2 will leak up to the subsurface and how fast?
Answering these questions relies on proper understanding of the multiphase flow systems
and the complex interplay of viscous, gravitational, and capillary forces as well as large
and small scale heterogeneity. Figure 1.1 illustrates the conceptual CO2 migration in
deep saline aquifers. Three physical forces dominate CO2 flow behavior in different flow
3
regimes.
Figure 1. 1 Different flow regimes dominated by different forces in the storage reservoir.
The relative permeability of CO2/brine systems is an essential element to determine CO2
injectivity and migration, and to assess the safety of potential CO2 sequestration sites. To
date, very few relative permeability measurements have been made for CO2/brine
systems (Bennion and Bachu, 2005, 2006; Benson et al, 2006; Krevor et al., 2012). This
work was initially motivated by the need to increase the number of relative permeability
measurements for CO2/brine systems under realistic reservoir conditions. However, in
the course of these investigations the important role of sub-core scale heterogeneity
became evident (Perrin and Benson, 2009). This observation motivated a systematic
evaluation of sub-core scale heterogeneity and flow rates on brine displacement
efficiency during horizontal core floods. Figure 1.2 shows some relative permeability
data measured on a range of rock types. The wide range of rock types results in a high
degree of variability. Also, unusually high residual brine saturations appear in most of
samples, the reason for which needs to be understood.
4
Figure 1. 2 Various relative permeability curves at insitu conditions for CO2 /brine
system (a) Basal Cambrian, (b) Cooking Lake, (c) Nisku, (d) Wabamun (low
permeability), (e) Ellerslie, (f) Viking, (g) Wabamun (high permeability), and (h) Frio
Sandstone
Several potential methods can be applied to study CO2/brine systems:
Core-Scale Laboratory Experiments: Experimental studies for both drainage and
imbibition displacements at representative reservoir pressure and temperatures are
needed to improve our ability to answer these questions
Petrophysical Studies: Pore-scale features such as size and shapes of the pore
spaces in the rock can be used to study multiphase flow properties using pore-
network models.
Numerical Simulation: Numerical simulations intended to replicate multiphase
flow experiments, in which CO2 is injected into a brine saturated rock, is one way
to test our understanding of the underlying physics. Numerical simulation also
can be used to develop scaling relationships and interpret field-scale behavior of
injected CO2.
Multiphase Flow Theory: Improving fundamental understanding of multi-phase
flow in CO2/brine systems to better predict both the experimental and numerical
results, and developing approaches for up-scaling laboratory measured relative
5
permeability curves for use in reservoir simulations will also be needed.
Predicting experimental or numerical results quantitatively can save a significant
amount of computational time and provide systematic insights into the physical
controls on multiphase flow under a wide range of conditions and for different
fluid pairs.
Field Scale Studies: it is important to study the CO2 distribution, migration, and
flow movement at the reservoir scale to test our knowledge gained from
numerical and theoretical investigation.
1.2 Outline of the Approach
In this work, numerical simulations and theoretical analysis are performed to study the
issues described in the previous section. First, we perform 3D high resolution
simulations to match the experimental data qualitatively and quantitatively. Second, we
do sensitivity studies including a wide range of rock types with different types of
heterogeneity to investigate the potential effects on the CO2/brine multiphase flow
system. Based on a series of numerical experiments, a two-dimensional theoretical
model incorporating gravity, capillary pressure, and sub-core heterogeneity is developed
to predict the three-dimensional results. Finally, the volume-averaged relative
permeability values calculated from the simulation outputs are investigated. Table 1.1
compares the similarity and the difference between these two different approaches. In
summary, this comprehensive study presents numerical and theoretical research efforts
that provide new understanding of CO2 migration in heterogeneous reservoirs.
6
Table 1. 1 Comparison of the differences and similarities between the three-dimensional
numerical simulations and two-dimensional theoretical derivation.
Miscible Diffusion Compressible Isotropic Gravity Capillarity Heterogeneity
3D
Numerical
Simulation
NO NO YES YES YES YES YES
2D
Theoretical
Derivation
NO NO NO YES YES YES YES
1.3 Dissertation Outline/Organization
The dissertation is organized as follows. Chapter 2 summarizes the relevant literature on
multiphase flow and the existing theoretical analysis pertinent to this work. In the first
part of this chapter, research focused on flow systems in which viscous or capillary forces
are dominant, the so-called “viscous and capillary limits” are reviewed. The second
section reviews studies considering the combined effects of capillarity and gravity. The
last section reviews a series of relative permeability studies focusing on several issues
including flow injection rate, gravity segregation, and sub-core scale heterogeneity.
Chapter 3 presents the methodology for numerical simulations performed in this
study. The short term goal of core-scale numerical simulations is to model the behavior
of CO2 at steady-state while the long term goal is be able to generalize small scale results
to large scale or at least gain insight from core-scale studies. Through these studies we
will ultimately develop a better understanding of interactions between physical forces as
well as the role of sub-core scale heterogeneity on multi-phase flow. A description of
the simulator, initial and boundary conditions, and required input parameters are
presented.
A numerical and theoretical investigation of the influence of flow rate, gravity, and
capillarity on brine displacement efficiency in homogeneous porous medium is presented
in Chapter 4. This chapter starts with the results of three-dimensional high resolution
numerical simulations. Sensitivity studies on core absolute permeability and interfacial
tension are presented to illustrate three regimes dominates by the viscous force, buoyancy
force, and capillary force, respectively. A two-dimensional semi-analytical solution that
7
includes gravity and capillarity is developed based on relevant dimensionless numbers.
Model validation is applied to a wide range of core dimensions, different fractional flows
of CO2 as well as different input relative permeability functions.
Based on the knowledge gained from the homogeneous studies, a further
investigation in heterogeneous porous media both numerically and analytically is
presented in Chapter 5. The first part of Chapter 5 shows results of 3D numerical
simulations with a wide range of degree of capillary heterogeneity simulated at core-scale
to understand the role of sub-core scale heterogeneities. Two types of heterogeneity are
studied: one uses a random log-normal distribution to generate 3-D permeability maps.
The other type studied is based on a real rock where the heterogeneous porosity
distribution measured using a CT scanner is used to generate the permeability map. The
second part of this chapter presents a 2D analytical solution incorporating the influence
of heterogeneity. A general form of the solution is first derived and the constraints from
3D high-resolution simulations are applied to obtain the final semi-analytical solution.
The validation of analytical model for heterogeneous cores is presented next. Finally, a
summary of how to use these homogeneous and heterogeneous analytical results are
provided.
Chapter 6 presents one of applications for the theoretical model developed in
Chapters 4 and 5, specifically, sensitivity studies on volume-averaged (up-scaled) relative
permeability that accounts for the role of sub-core scale heterogeneity on multi-phase
flow in CO2/brine systems. Correct procedures for calculating relative permeability are
discussed. After that, a general rule for calculating reliable relative permeability
measurements, even in heterogeneous cores, is provided.
Finally, conclusions and recommendations for future work are discussed in Chapter 7.
8
Chapter 2
Literature Review
Chapter 2 summarizes the relevant literature on multiphase flow and the existing
theoretical analyses pertinent to this work. In the first part of this chapter, research
focused on flow systems in which viscous or capillary forces are dominant, the so-called
“viscous and capillary limits” are reviewed. The second section reviews studies
considering the combined effects of capillarity and gravity. The third section reviews the
small-scale heterogeneity. The fourth section reviews a series of relative permeability
studies focusing on several issues including flow injection rate, gravity segregation, and
sub-core scale heterogeneity. Finally, a discussion based on the literature is provided that
identifies gaps in current understanding.
2.1 Multiphase Flow in the Viscous and Capillary
Limits
Recently, the influence of flowrate, gravity, and capillarity on immiscible CO2/brine two-
phase flow system at the core scale has been studied in laboratory experiments (Perrin et
al. 2009; Perrin and Benson 2010; Krevor et al. 2011) as well as using 3-D numerical
simulation (Krause et al. 2011; Kuo et al. 2010; Shi et al. 2010). Although theoretical
analysis of multiphase flow systems has identified flow regions controlled by different
types of flow behavior, most of them have focused on oil/water systems where the effect
of gravity has been neglected (Lenormand et al. 1988; Pickup and Stephen 2000; Jonoud
and Jackson 2008; Hussain et al. 2011). Without considering gravity, the flow regime is
often studied for two extreme limits: one is when the viscous force is sufficiently large
9
such that the capillary pressure is negligible. In this viscous limit, fractional flow is
uniform. The other limit is when the flow rate is very low and capillary forces dominate
the fractional flow distribution. In the capillary limit, capillary equilibrium is assumed.
Characterizing different flow regimes is important and very useful for upscaling.
Dimensionless numbers are often used to identify different regimes since they combine
the effects of flow rate, interfacial tension, the permeability of the core, and gravity
(Fulcher et al. 1985; Avraam and Payatakes 1995; Skauge et al. 1997; Skauge et al. 2000;
Virnovsky et al. 2004; Cinar et al. 2006).
However, the boundaries between viscous and capillary limits are ambiguous in the
literature. Jonoud and Jackson (2008) develop a new dimensionless group using three
dimensionless numbers (transverse and longitudinal Peclet numbers and end-point
mobility ratio: PeT, Pe
L, M
e) to characterize the balance of viscous and capillary forces
and use numerical simulations to determine empirically the threshold values of these two
limits. Their results show that the balance between viscous forces along the direction of
flow and capillary forces perpendicular to the flow control the boundaries.
2.2 Multiphase Flow with Viscous, Capillary and
Gravity Forces
For gas/oil or gas/water systems, the density difference between two the fluids is large,
and the effect of gravity becomes important (Rossen and Duijn 2004; Nordbotten et al.
2005; Kopp et al. 2009). Strong gravitational forces could lead to distinct gravity
segregation between the injected nonwetting phase and wetting phase. In addition, when
flow rates are low, gravity-capillary equilibrium is established.
Comprehensive 2D analyses of the combined effects of viscous, gravity and capillary
forces on the fluid flow in the transverse direction have been performed (Zhou et al.
1994; Yortsos 1995). Both of these works study incompressible flow in an anisotropic
system, and identify flow regions dominated by different forces in the vertical direction.
The corresponding simplified equations and conditions based on dimensionless groups.
Zhou et al. (1994) use modified gravity number NgvM/(1+M), modified capillary number
10
NcvM/(1+M), and the shape factor RL2 while Yortsos (1995) use gravity number NG,
transverse capillary number NCT and aspect ratios H/L, kav/kah as well as the RL2 as the
average scaling parameters. Definitions of these dimensionless parameters are provided
below.
*
2av c
cv CT L2
o t
k LpN = =N R
H μ u (2.1)
2av
gv G L
o t
Δρgk LN = =N R
Hμ u (2.2)
av
L
ah
kLR =
H k (2.3)
w
o
λM=
λ (2.4)
Zhou et al. (1994) establish approximate bounds for the transitions between regions by
examining a wide range of experimental data in the literature. However, most of the data
sources are focused on imbibition and miscible flow types. Table 2.1 lists five flow
regimes and their conditions.
Table 2. 1 Summary of the limiting cases (Zhou et al. 1994).
Flow Region Conditions
Capillary-dominated Crossflow Ncv>>Ngv and NcvM/(1+M)>>1
Gravity-dominated Crossflow Ngv>>Ncv and NgvM/(1+M)>>1
Capillary-gravity Equilibrium Ncv≈Ngv and NgvM/(1+M)>>1
Viscous-dominated Crossflow (Ncv+Ngv)M/(1+M)<<1
Vertical Crossflow Equilibrium (VCE) (Ncv+Ngv)M/(1+M)<<1 and RL2>>1
No-communication
(Viscous crossflow can be neglected) (Ncv+Ngv)M/(1+M)<<1 and RL
2<<1
Yortsos (1995) takes another approach to identify these flow regimes by using asymptotic
analysis. Table 2.2 summarizes five flow regimes and the corresponding conditions as
well as the qualitative comparison between three forces. Both the Zhou et al (1994)
11
model and the Yortsos (1995) model only characterize these different flow regions
without solving the simplified equations and they consider only transport in the
transverse direction.
Table 2. 2 Summary of the limiting cases (Yortsos 1995).
Flow Region Conditions Viscous Gravity Capillary
Viscous Fingering Ngv<<1, NCT<<H/L Strong Negligible Negligible
Viscous Fingering
with Dispersion Ngv~1, NCT~ H/L Strong Moderate Moderate
Gravity Tonguing Ngv~1/ RL2, NCT<<(kav/kah)L/H Moderate Strong Negligible
Capillary-gravity
Equil. Ngv~1/ RL
2, NCT~(kav/kah)L/H Moderate Strong Moderate
Capillary Equil.
Ngv~1/ RL
2, NCT>>(kav/kah)L/H Moderate Strong Strong
2.3 Capillary Heterogeneity
Capillary heterogeneity is known to be an important parameter affecting multiphase flow
of CO2 and brine, affecting properties such as saturation profile, capillary pressure and
relative permeability (Perrin and Benson 2010; Krevor et al. 2011; Pini et al. 2012;
Krause et al. 2011; Kuo et al. 2010; Shi et al. 2010). Moreover, capillary barriers within
the rock can affect flow behavior significantly. If the orientation of a capillary barrier is
perpendicular to the flow direction, non-wetting phase fluid in a water-wet system may
be trapped inside high permeability zones surrounded by low permeability ones
(Honarpour et al. 1995; Hamon and Roy 2000; Pickup and Stephen 2000). It is also
suggested that capillary heterogeneity provides a new trapping mechanism in carbon
sequestration (Saadatpoor et al 2010) in addition to structural trapping, residual trapping,
dissolution trapping and mineral trapping.
Rock heterogeneity is very common in geological formation and exists at every scale:
12
pore-scale (~m), grain scale (~mm), Core-scale (~ 0.1 to 10 cm), and field-scale (~ 10
cm and larger). These different scales of heterogeneity result in complexity when solving
multiphase flow problems. In this work, we are focused on understanding the influence
of spatial heterogeneity at core-scale on multiphase flow of CO2 and brine.
The analytical study of the capillary heterogeneity has been very limited due to the
complexity it introduces to the multiphase flow problem. Most studies of capillary
heterogeneity are limited to 1D or 2D analyses (e.g. Yortsos and Chang 1989, 1990;
Chang and Yortsos 1992; Chaouche et al. 1994; Hussain et al. 2011; Chen 2012). The
major conclusion from the 1D studies is that capillary heterogeneity leads to variable
saturation distributions. In general, for very low flow rates, the saturation distribution
follows the heterogeneity variation. For higher flow rates the situation is more complex:
viscous forces compete with capillary forces to control the saturation distribution at the
interfaces between different regions in the rock.
Yortsos and Chang (1990) studied the one-dimensional steady-state saturation
response to different heterogeneity both analytically and numerically. Permeability
variations are used to describe the heterogeneity of the core, the Leverett scaling
relationship is used to characterize capillary heterogeneity, and relative permeability is
assumed to be uniform. The 1D analytical model is accurate for large capillary numbers
Ncv (low enough flow rates). In addition, they also study the effect of permeability
heterogeneity on viscous dominated Buckley-Leverett type displacements (Chang and
Yortsos 1992) based on a similar 1D analysis. Three permeability distribution models are
studied: random step, sinusoidal, and correlated models. They conclude that the effect of
capillary heterogeneity on the saturation profile of a Buckley-Leverett displacement
(relatively high flow rates, small capillary number Ncv) is non-negligible once the
amplitude of permeability variation is high enough and the spatial correlation is small.
They also conclude that “further study of this problem and development of some practical
approaches would be desirable under-takings.”
Studying the influence of capillary heterogeneity effect in two-dimensions analytically
is even more difficult and complicated. Simple layered heterogeneous porous media
considering anisotropic permeability has been investigated in 2D (Zhou et al. 1994;
Yortsos 1995). However, its application is limited to miscible displacements and
13
immiscible displacements without gravity effects. In a recent study (Kuo and Benson,
2012), the combined effect of viscous, gravitational, and capillary forces at the core-scale
has been studied numerically and analytically. However, this study only considered
homogeneous porous media.
2.4 Relative Permeability Measurements
Saline aquifers have the largest potential capacity to store CO2 (IPCC 2005). However,
as compared to oil and gas reservoirs, where a century of experience exists regarding
multiphase displacement processes, our understanding of the fate and transport of CO2
and brine in saline aquifers is still limited. When CO2 migrates through a saline aquifer,
the interplay between viscous, capillary, buoyancy forces as well as structural
heterogeneities will determine how far and how fast the plume will move, how much CO2
will dissolve, and how much will be immobilized by residual trapping (Ide et al. 2007;
Kopp et al. 2009). An increasing number of studies based on numerical simulations both
at the reservoir (Bryant et al. 2006; Juanes et al. 2006; Flett et al. 2007; Han et al. 2011)
and core scale (Krause et al. 2011; Kuo et al. 2010; Shi 2010; Kuo and Benson, 2012) are
being widely applied to describe and quantify these processes: critical input parameters to
any simulation are the multiphase flow properties for the CO2/brine/rock system, such as
the capillary pressure and the relative permeability curves. The latter, specifically the
drainage relative permeability, is the main subject this literature review.
There are two main categories of laboratory techniques to measure relative
permeability curves: steady-state methods (Muskat 1937; Morse et al. 1947; Osoba et al.
1951; Abaci et al. 1992; Perrin and Benson 2010; Krevor et al. 2011) and unsteady-state
methods (Welge 1952; Jonhson et al. 1959; Bennion and Bachu 2006; Chalbaud et al.
2007). In the unsteady-state method, only a single phase is injected into the core to
displace in-situ fluids. Saturation equilibrium is not attained and thus it can significantly
reduce the time needed to measure the relative permeability curves. In the steady-state
method, two fluid phases are injected simultaneously at a fixed volumetric ratio and
constant rate until saturation and differential pressure along the core become constant.
14
Although the attainment of equilibrium for steady-state method might be time
consuming, the data can be interpreted directly with the multiphase flow extension of
Darcy’s law using the measured saturation and pressure drop (Abaci et al. 1992; Avraam
and Payatakes 1995). In this review, we focus on the steady-state relative permeability
measurement technique.
Currently there are limited laboratory data on CO2-brine relative permeability
(Bennion and Bachu 2005, 2006, 2007, 2008, 2010; Bachu and Bennion 2007; Perrin and
Benson 2010; Akbarabadi and Piri 2011, Krevor et al., 2012). However, the reliability of
such published relative permeability is directly affected by the quality of the measured
relative permeability curves, as recently highlighted in a review of published relative
permeability measurements (Müller, 2010). In particular, factors that could affect these
measurements are (a) the core heterogeneity that may be responsible for flow rate
dependency and incomplete fluid displacement; (b) capillary end effects that are not
properly accounted for; and (c) gravity segregation that may occur when relatively long
cores are used in a horizontal core-flooding system. In the following, the above
mentioned issues are discussed in more detail.
One of the biggest concerns about relative permeability measurements is the capillary
end effect (Leverett 1941). Whenever a capillary pressure gradient exists along the
porous medium, traditional approaches to calculate the relative permeability are
insufficient. One of the standard techniques used to minimize the capillary end effect is
injecting high flow rate since the capillary forces are small compared to viscous forces at
high flow rates. The influence of end effect on relative permeability becomes significant
at low rates or for low pressure gradients as saturation gradients increase with decreasing
flow rate (Leverett, 1941; Morse et al, 1947; Osoba et al., 1951; Caudle et al., 1951; Kyte
and Rapoport, 1958; Henderson et al., 1997). Based on that, early studies have argued
that relative permeability should be independent of flow rates and that if flow rate
dependency is observed, this should be attributed to the boundary effect (Osoba et al.
1951; Sandberg et al. 1958; Fulcher et al. 1985). However, there are a number of
research papers in the literature suggesting that relative permeability does depend on the
flow rate even when the end effect is carefully avoided (Henderson and Yuster 1948;
Caudle et al. 1951; Avraam and Payatakes 1995; Skauge et al. 2000; Virnovsky et al.
15
2004). Flow rate dependent relative permeability curves are attributed to inadequacy of
the multiphase extension of Darcy’s law for transient flow also upscaling (volume
averaging) of heterogeneous rocks with capillary heterogeneity (Ringrose et al., 1993;
Saad et al., 1995; Dale et al., 1997; Pickup and Stephen, 2000; Pickup et al., 2000;
Virnovsky et al., 2004; Lohne et al., 2006; Jonoud and Jackson, 2008).
Another method to minimize the capillary end effect is to increase the length of the
core. However, experiments carried out with long cores in a horizontal core-flooding set-
up may encounter the issue of gravity segregation (Skauge et al. 1997). Gravity
segregation needs to be considered when working with a fluid pair characterized by a
significant density difference, such as the oil/gas or supercritical CO2/brine system
(Chang et al. 1994; Rossen and Duijn 2004; Ide et al. 2007; Hesse et al. 2008) since the
large density difference can lead to gravity override based on the high Bond number and
hence causes both horizontal and vertical saturation gradients. Although using vertical
experiments to measure relative permeability can avoid gravity segregation (Cinar et al.
2006), the use of a vertical arrangement would make the use of X-ray CT scanning to
observe fluid saturation is quite challenging without purpose-designed equipment.
In addition, spatial variation of rock properties affects both the capillary pressure and
relative permeability-saturation relations (Honarpour et al. 1995; Hamon and Roy 2000;
Ataie-Ashtiani et al. 2002), but it also influences the spatial distribution of saturation
(Chaouche 1994; Perrin and Benson 2010; Krause et al. 2010; Shi et al. 2010). Various
degrees of heterogeneity affect CO2 trapping capacity (Oloruntobi and LaForce 2009)
and may cause flow rate dependency, high residual water saturation and low end-point
relative permeabilities observed from the CO2/brine core flood experiment (Bennion and
Bachu 2005; Perrin and Benson 2010; Akbarabadi and Piri 2011, Krevor et al., 2012). It
has been shown that including heterogeneity characteristics in numerical simulator grid
blocks can improve the accuracy of simulation prediction and enable reliable relative
permeability measurements (Honarpour et al. 1995; Virnovsky et al. 2004; Krause et al.
2011; Kuo et al. 2010).
16
2.5 Discussion
Based on this literature review and the introductory discussion, the following conclusions
can be drawn:
A comprehensive analysis including gravity, capillarity and local heterogeneity is
needed to understand the fundamental physics of multiphase flow of CO2 and
brine in reservoirs and is needed to potentially reduce the computational effort
when performing high-resolution large-scale simulations;
Although there is an increasing number of measurements of multi-phase flow of
CO2 and brine in reservoir rocks (Bennion and Bachu, 2005,2006; Perrin and
Benson 2010; Krevor et al. 2011; Pini et al. 2012), given the emerging importance
of this topic, studies are needed to develop a strong scientific foundation to
support sequestration in saline aquifers;
The large body of multiphase flow studies, particularly relative permeability in
oil/water and gas/liquid systems provides a good starting point for understanding
CO2/brine systems;
Since the fluid properties of CO2 are very different than oil, and because of the
fundamentally empirical nature of the relative permeability concept, studies are
needed to establish similarities and differences between multiphase flow
oil/water and CO2/brine systems;
Potential and unresolved influences of flow rate, capillary number and small scale
heterogeneity on relative permeability in CO2/brine systems need to be
investigated;
The end effect is an important factor that could lead experimental error. If we
want to investigate the flow rate dependence on relative permeability curves, the
end effect must be carefully understood and compensated for;
Based on recently studies of heterogeneity, the effect of heterogeneity may be the
reason for observed dependence of relative permeability on flow rate; and
Since the experiments to measure the influence of heterogeneity on relative
permeability are time consuming, numerical simulations can be used to simulate,
understand and interpret laboratory experiments of multiphase flow in typical
reservoir rocks.
17
Based on above conclusions, and the motivation to investigate the importance of
heterogeneity in CO2 flooding at multiple scales, the goal of this work is to provide a
comprehensive study of the fundamental physics both numerically and analytically, of the
combined influence of heterogeneity, viscous forces, gravity, and capillarity on
multiphase flow of CO2 and brine. Specifically, steady-state displacements in both
homogeneous and heterogeneous cores over a range of relevant conditions are studied.
For example, 3D coreflood simulations have studied the impact of sub-core heterogeneity
on CO2/brine displacements over a wide range of flowrates. Various degrees of
heterogeneity are generated based on the normal random distribution as well as for real
models of cores based on 3-D X-Ray tomography. A 2D analytical model considering
gravity and permeability heterogeneity is developed for predicting brine displacement
efficiency over a wide range of capillary numbers that a good agreement with the 3D
results. The analytical derivation is general and suitable for all flow rates, and is not
limited to high or low capillary and gravity number regimes.
18
Chapter 3
Simulation Methodology
Chapter 3 presents methodology for the numerical simulations performed in this study.
The goal of the simulations is to replicate CO2/brine core flood experiments that are
representative of typical reservoir condition occurring at sites of CO2 sequestration in a
saline aquifer. Once experimental data are simulated qualitatively, a series of sensitivity
studies are performed to investigate the effects of flow rate, gravity, permeability,
interfacial tension, and core length as well as sub-core heterogeneity on multiphase flow
of CO2 and brine. Through these studies we will ultimately develop a better
understanding of interactions between physical forces as well as the role of sub-core scale
heterogeneity on multi-phase flow of CO2 and brine. A typical CO2 saturation
distribution inside a Berea Sandstone rock sample is shown in Figure 3. 1. This
saturation distribution was measured while a mixture of CO2 (95%) and brine (5%) were
injected into the core and after the saturation distribution was no longer changing (Perrin
et al., 2009). One interesting and important feature of the experiment is that there is a
large portion of the core near the outlet end that is almost completely bypassed by the
CO2.
Figure 3. 1 The experimental steady state three-dimensional views of CO2 saturation in
the core for a given fractional flow of CO2 at a given flow rate. The fluids were injected
from right to left (Perrin et al, 2009).
19
The laboratory conditions, core properties as well as grid information listed in Table 3.1
are selected to replicate the laboratory experiments. The Berea sandstone core used in
the experiment is modeled by a three dimensional roughly cylindrical core (Figure 3. 2).
A total of 31 slices is used in the flow direction, including 29 rock slices, an “inlet” slice
at the upstream end of the core, and an “outlet” slice at the downstream end. All of the
simulations are carried out by co-injecting known quantities of CO2 and brine at a
constant flow rate into the inlet end of the core (Figure 3. 1). The downstream end of the
core is maintained at a constant pressure by a back-pressure pump.
TOUGH2-MP with the ECO2N module is used to conduct core-scale multiphase flow
simulations and will be discussed more detail in the next section. The goal of these
simulations is 3-fold: first, the experiments are simulated with the goal of qualitatively
replicating the major features of the experiments; second, simulations are used to conduct
sensitivity studies regarding the influence of viscous, gravity and capillary forces over a
wide parameter space of interest for CO2 sequestration; and finally, the simulations are
used to help develop and then to validate semi-analytical solutions for predicting the
average saturation in a core as a function of the gravity and capillary numbers.
Figure 3. 2 Simulation core of 25 25 31 grid blocks of uniform size.
Table 3. 1 Simulation summary.
Temperature and
Pressure
Fluid properties Berea Core Properties Simulation Grid Data
Tres=50˚C
Pres=12.4 MPa
μCO2=0.046 cp
μw=0.558 cp
ρCO2=608 kg/m3
ρw=1005 kg/m3
σ=22.4 mN/m
L=14.73 cm
H=4.69 cm
Φmean=0.202
kmean=430 mD
25x25x31 grids
Total 19375 grids
5.08 mm grid length
1.874 mm grid width
20
3.1 TOUGH2 MP/ECO2N
TOUGH2 is a general purpose numerical simulator for multi-component, multi-
dimensional, multi-phase fluids flowing in porous and fractured media (Pruess et al.,
1999). TOUGH2-MP, a massively parallel (MP) version of TOUGH2 (Zhang et al.,
2003), is used due to the intensive computational requirement in areas such as reservoir
engineering and CO2 geological sequestration. ECO2N, one of the fluid property
modules of the current version of TOUGH2-MP V2.0, can model thermodynamic and
thermophysical properties of CO2, H2O and NaCl accurately under the range of
conditions where CO2 storage in saline aquifers is likely (Pruess, 2005; Zhang et al.,
2007). Table 3. 2 lists the summary of ECO2N parameters implemented in all the
simulations performed in this work. There are three mass components (water, NaCl and
CO2) in the simulation. Flow systems are assumed to be isothermal, hence three mass
conservation equations are needed to be solved per grid block. Molecular diffusion is not
included in the simulations.
Table 3. 2 Summary of ECO2N.
Components
#1: Water #2: NaCl #3: CO2
Parameter Choices
NK=3 # of mass component
NEQ=3 # of balance equations per grid block (isothermal)
NPH=3 # of phases that can be present
NB=6 no diffusion
Primary Variables
two fluid phases (aqueous and gas)
P Pressure
T Temperature
SG Gas Phase Saturation
XNACL Salt Mass Fraction in Brine
A general description of the differential equations solved in TOUGH2 is provided in the
TOUGH2 User’s Guide (Pruess et al. 1999). The mass conservation equations in
TOUGH2 are discretized in space and time and solved by using the fully implicit finite-
21
difference method.
The mass conservation equations are discretized in space using the integral finite-
difference (IFD) method (Pruess, 1991) and discretized in time using first-order
backward finite difference:
κ
κ κn
nm nm n
mn
dM 1= A F +q
dt V (3.1)
κ,k+1 κ,k
κ,k+1 κ,k+1n n
nm nm n
mn
M -M 1= A F +q
Δt V (3.2)
The superscripts, κ, represent the mass component κ. The subscripts, nm, indicate a
suitable averaging at the interface between grid blocks n and m, (Anm), illustrated in
Figure 3. 3. In this work, all grid blocks have the same size of rectangular shape,
therefore the volume of each grid block is the same (Vn=Vm).
Figure 3. 3 Space discretization in the integral finite difference method.
Mn is the average value of M over Vn. The surface integrals are approximated as a sum
of the average flux over Anm. Fnm is the average value of the inward, normal component
of the flow term, F, over the surface Anm. qn denotes sinks and sources in grid block n.
Mass flux Fnm is a sum over individual phase fluxes Fβ,nm, given by Darcy’s law, times
mass fraction of component κ present in phase β,β
κx :
β
κ κ
nm β,nm
β
F = x F (3.3)
The discretized Darcy flux term is as follows:
22
rβ β β,n β,m
β,nm nm β,nm nm
β nmnm
k ρ P -PF =-k -ρ g
μ D
(3.4)
The discretized flux is expressed in terms of averages over the parameters for element Vn
and Vm. knm and krβ,nm are average absolute permeability and average relative
permeability to phase β over Anm, respectively. ρβ,nm is the interface density of phase β,
defined as the arithmetic average of densities between grid blocks n and m, (ρβ,n+ρβ,m)/2.
Dnm is the distance between the nodal point n and m, and gnm is the component of
gravitational acceleration in the direction from m to n. In this work, evaluation of
mobilities and permeability at interfaces are fully upstream weighted.
In addition, Eq. 3.2 shows that the flux and the sink/source terms are evaluated at the
next time step, tk+1
. This so-called fully implicit method can provide numerical stability
(unconditionally stable). Based on Eq. 3.2, the discrete mass conservation equations can
be written in a residual form (Pruess, 1991; Pruess et al., 1999):
κ k+1 κ k+1 κ k κ k+1 κ,k+1
n n n nm nm n n
mn
ΔtR x =M x -M x - A F x +V q =0
V
(3.5)
where κ
nR are the residuals of component κ for grid block n, the vector xk consists of
primary variables at the time tk, and Δt denotes the current time step size. These strongly
coupled nonlinear equations are solved simultaneously using the Newton-Raphson
iterative scheme:
k k+1 k kJ x x -x =-R x (3.6)
where ij i jJ = R / x is the elements of the Jacobian matrix. Equation (3.6) is solved
leading to:
κ,k+1
κ,k+1n
p i,p+1 i,p n i,p
i i
R- | x -x =R x
x
(3.7)
where xi,p represents the value of ith
primary variable at the pth
iteration step. All the
Jacobian matrix elements are evaluated by numerical differentiation. The Jacobian
matrix as well as κ,t+1
n i,pR x need to be recalculated at each iteration, and this may lead
23
to massive computational work for a large simulation. In the parallel code, equation (3.7)
is computed by all processors. Consequently, computational performance is significantly
improved compared to the TOUGH2 code.
Iteration is continued until the residuals κ,k+1
nR are smaller than the absolute
convergence value
κ,k+1
n 1 2|R | ε ε (3.8)
where ε1 is relative convergence tolerance while ε2 is the convergence criterion for the
accumulation terms κ
nM . These two parameters are TOUGH2 input parameters, and
defaults are ε2=1 and ε1=10-5
, respectively. However, it is hard to converge using such
small tolerance once the small-scale capillary heterogeneity is included. Therefore, after
comparing simulation with high and low tolerances, we use ε1=10-2
to increase
computational speed while preserving adequate computational rigor (validation of these
simulations can be found in Appendix E.1). A comparison between results from
TOUGH2 and another simulator, Stanford’s Genreal Purpose Reservoir Simulator
(GPRS), has been performed and the results are almost identical (Li, 2011) which
provides confidence in the accuracy of TOUGH2 (Appendix E.2). Note that even with
such high tolerance, it still requires significant computational effort to simulate highly
heterogeneous cores. Although convergence usually takes 3-4 iterations, the time step
size Δt is reduced if the convergence cannot be achieved after 8 iterations (default). The
time step size Δt we use here is such that the CFL number (uxΔt/Δx) is much smaller than
1, for example, 0.04-0.3. More details about the CFL numbers are provided in the
discussion section of Chapter 4. Having a low CFL avoids numercal errors caused by the
large time step size although it also requires significant computational power. To achieve
steady-state saturation distributions for typical core flooding experiments in
heterogeneous rocks requires about 24-48 hours on a cluster with 4-8 processors.
There are a number of reasons why TOUGH2 was selected for these calculations.
First, although we are ultimately interested in steady-state conditions, we are also
interested in the transient flow behavior that occurs prior to the establishment of the
steady state. Second, TOUGH2 is the most widely used simulator for studying
multiphase flow of CO2 and brine, and as such, has well documented performance over a
24
wide range of conditions relevant to this study. Finally, TOUGH2 has a robust equation
of state for CO2/Water/Salt mixtures, again a capability needed for simulating the core
flooding experiments which are conducted at reservoir pressures and temperatures.
3.2 Core Descriptions and Grid Size
In order to simulate the laboratory experiments, the Berea core must be replicated in the
simulation grid. The core is divided into NYNZNX grid elements in Y, Z, and X
direction respectively. X-ray CT scans of the core prior to injection provide a three
dimensional spatial porosity map. The simulation grid blocks are assigned unique
porosity values based on the experimental data. Correlations between permeability and
porosity are applied to every grid block to generate its corresponding permeability values.
Therefore, each grid now has been assigned its unique rock properties. From the
TOUGH2-MP simulation, given a fractional flow of CO2 and specific flow rate, the
corresponding pressure drop and CO2 saturation after the system reaches steady state can
be obtained.
The dimensions of simulated core are 4.69 cm in diameter and 14.73 cm long, and the
core is modeled by 25 25 31 grid blocks with dimensions 1.884mm 1.884mm
5.08mm. The grid size chosen for these experiments balances three factors. First the grid
size must be large enough for the continuum representation of the flow and transport
equations to be valid (i.e. Darcy’s law can be used). Microtomography of the Berea
sandstone indicates that a 1mm cube of the grid is large enough to create a representative
elementary volume for the continuum assumption to be valid (Silin et al. 2011; Pini et al.
2012). Second, the grid cells must be sufficiently small to avoid numerical dispersion
and preserve accurate spatial gradients in saturation and capillary pressure. Finally, since
these simulations are computationally intensive, the number of grid cells needs to be
limited to keep the simulations tractable. In light of these considerations, the dimensions
of the grid cells used in this paper are 1.8mm x 1.8mm x 5mm. A comparison showing
that the solutions obtained with this grid are the same as for a grid that is twice as fine is
provided in Chapter 4.
25
3.3 Input Parameters
Before performing numerical simulations, each grid block is assigned values for porosity,
permeability, capillary pressure and relative permeability. The permeability of each grid
element is assumed to be isotropic for the rock slices. In this section, detail of boundary
conditions, initial conditions as well as the input capillary pressure and relative
permeability curves will be discussed.
3.3.1 Boundary Conditions
The boundary conditions used in this study are selected to replicate the core flooding
experiments (Perrin and Benson 2010; Krevor et al. 2012; Pini et al. 2011). In the
experiments, CO2 and brine are mixed and co-injected through a tube and enter into a
diffuser plate to distribute evenly before entering into the upstream end of the core. To
avoid dry-out, carbon dioxide and water are pre-equilibrated at high pressure and
temperature (in this case, 50˚C and 12.4 MPa) prior to starting the experiment. The
amounts of CO2 and brine that enter each pixel are controlled by the relative mobility of
CO2 and brine (Eq. 3.1) such that the total is equal to the injection rate of each phase (Eq.
3.2):
r,β β r,β β,inlet β,1
β,i
β βi i
kk Δp kk p -pq =- =
μ Δx μ Δx
where β=w, CO2 (3.1)
i T
i
q =q
where qi=qw,i+qCO2,i (3.2)
To replicate the inlet boundary condition with the simulator, anisotropic permeability is
implemented in the inlet slice such that the total injected fluids is free to spread out over
the cross section area evenly and enter each cell in accordance with its mobility.
In the experiment, the downstream end of the core is maintained at a constant
pressure by a back-pressure pump. Under this situation, it’s not apparent which boundary
conditions will most closely replicate the experimental measurements. Therefore we test
two numerical boundary conditions to determine which one would most closely replicate
26
the saturation distributions observed at the outlet of the core (Figure 3. 4a). Both test
cases have imposed a time-independent Dirichlet boundary condition: the primary
thermodynamic variables (for example, P and T) remain unchanged in the outlet. One
boundary condition sets the capillary pressure to zero in the outlet slice of the core
(Figure 3. 4b):
c outletP | =0 (3.3)
The other boundary condition imposed the condition that there is no capillary pressure
gradient between the last rock slice and the outlet slice of the model (Figure 3. 4c):
c
outlet
dP| =0
dx
(3.4)
An example of the measured saturation distribution along the core for different fractional
flows of CO2 at a total injection rate of 2.6 ml/min flow rate is shown in Figure 3. 4a.
Similar saturation distributions have been measured for other rocks as described by
Krevor et al. (2012). A relatively uniform saturation profile is observed over the whole
core; in particular, there is no large saturation gradient at the outlet. If the boundary
condition with Pc=0 at the downstream end is imposed, a large saturation gradient occurs,
which is not observed in the experiments (Figure 3. 4b). The Dirichlet boundary
condition with the added constraint that dPc/dx=0 between the last slice in the core and
the outlet provides a much better match to the data (Figure 3. 4c). Consequently we use
this boundary condition in the rest of the simulations. Specifications for the boundary
conditions are listed in Table 3. 3.
(a) (b) (c)
Figure 3. 4 CO2 saturation along the Berea Sandstone core for different fractional flows
of CO2 at a total injection flow rate 2.6 ml/min: (a) Experimental results; (b) High
contrast model with boundary condition Pc=0; (c) High contrast model with boundary
condition dPc/dx=0.
outlet
27
Table 3. 3 Boundary condition summary.
Inlet slice Rock slices (29 slices) Outlet slice
Φmean,
kmean: Anisotropic
(kz=ky=100kx)
Pc=Pc,mean
Φi,
ki: Isotropic
Pc i ik
Φmean,
kmean: Isotropic
Dirichlet boundary condition
dPc/dx=0
3.3.2 Initial Conditions
All the TOUGH2 simulations are conducted at 50˚ C temperature and 12.4 MPa initial
pore pressure. Before injecting CO2 and brine into the core, all the pore-space in the
simulated core is saturated with brine as the initial condition. Water is pre-equilibrated
with dissolved CO2 and 10,000 ppm NaCl. Pre-equilibrating CO2 and water before
injection is important for avoiding mass transfer between phases. Specifically, if dry CO2
is injected into the core, water is vaporized into dry CO2 and results in lower water
saturation near the inlet. Lower saturation of water results in the higher capillary
pressure at the inlet. This “dry-out” phenomenon could lead to imbibition of water from
the downstream portions of the core to the inlet. To avoid these complexities, we pre-
equilibrate the fluids, as is done in the laboratory experiments. At the experimental
condition of 50˚C and 12.4 MPa, the mass fraction of dissolved CO2 in brine is about
4.8%.
When CO2 and brine are injected into the core, the brine is partially displaced by
CO2. A small amount of the CO2 dissolved in the brine may be transferred to the gas
phase, but this is only significant at very small fractional flows of gas (1%). Preliminary
simulations were performed iteratively to determine the maximum CO2 mass fraction
before exsolution occurs. At the experimental condition of 50˚C and 12.4 MPa, the mass
fraction of dissolved CO2 in brine is about 4.8%.
3.3.3 Input Capillary Pressure
Eq. 3.5 show the capillary pressure equation for each grid element i, Pc,i, and the
corresponding modified J-function (Silin et al. 2009):
28
i
c,i w w
i
φP (S )=σcosθ J(S )
k
(3.5a)
2 2
1
λ 1/λ w P
w 1 1 * *λ
P*
S -S1J(S )=A ( -1)+B (1-S ) , S =
1-SS
(3.5b)
where Sw is the average brine saturation; σ is the CO2-brine interfacial tension; θ is the
contact angle, a function of properties of the solid and two fluids. φi and ki are the
porosity and permeability values of grid element i. A1, B1, λ1, λ2 and Sp are five fitting
parameters to match the experimental capillary pressure curves.
For the rock used in the core flood experiment, the capillary pressure was measured
using mercury injection porosimetry (Figure 3. 5). Eq. 3.5 is used to curve fit the
laboratory-measured capillary pressure curve obtained from a sub-sample of the rock
core based on the measured interfacial tension (σmean), measured porosity and
permeability values (φmean and kmean). θ=0 is used in the simulation based on the
assumption of strongly water wet core. Once the five fitting parameters of the J-function
are adjusted to match the measured capillary pressure data, they remain constant for all of
the simulations. The capillary pressure curve for each element in the simulation is then
scaled by the relationship (Eq. 3.5). The fitting parameters in J-function are shown in
Table 3. 4. Figure 3. 5 shows a plot of Pc,i versus Sw. The match between the measured
and fitted curve is very good over the majority of the saturation range, and in particular
the saturation range observed in the experiments. Note that the entry capillary pressure is
zero in this work based on Eq. 3.5.
Figure 3. 5 Laboratory capillary pressure data with a curve fit used in simulations.
29
To determine the effect of heterogeneity on brine displacement efficiency, homogeneous
core and heterogeneous core are simulated at the same reservoir condition, with the same
initial and boundary conditions, same grid sizes and same fitting parameters for relative
permeabilities and capillary pressure. For the homogeneous core, φi = φmean and ki
=kmean, and therefore result in a uniform capillary pressure assigned to each grid cell. On
the other hand, the capillary pressure curve of each grid element is scaled by its porosity
and permeability values for the heterogeneous core. Hence, each grid element has its
own capillary pressure curve. The unique pair of porosity and permeability values results
in a range of capillary pressure curves. Figure 3. 6 illustrates the mean capillary pressure
curve (Pc,mean) and the bounds for the capillary pressure curves in the heterogeneous core.
These capillary pressure curves are a function of core heterogeneity. A higher degree of
heterogeneity will result in a wider range of saturations for a given capillary pressure.
Figure 3. 6 The range of capillary pressure curves in the simulations. The values of input parameters are
A1=0.007734, B1=0.307601, λ1=2.881, λ2=2.255, Sp=0 and σ =22.47 dynes/cm.
Table 3. 4 Input parameter values for relative permeability and capillary pressure curves
fit. Parameters of Relative Permeability
SCO2,r Swr nw nCO2
0 0.15 7 3
Parameters of Capillary Pressure Curve
A1 B1 λ1 λ2 Sp
0.007734 0.307601 2.881 2.255 0
30
The Influence of Capillary Pressure on Saturations
A unique capillary pressure curve assigned to each grid is a key parameter to enable
simulating experimental core flood saturations qualitatively. Table 3. 5 shows the results
of four simulations performed at the 50% fractional flow of CO2 with different input
parameters of the rock such as porosity, permeability and capillary pressure curves. The
results show that even when the permeability and porosity vary within the core, a uniform
capillary pressure curve across the core leads in all cases to a uniform saturation
distribution. The last column demonstrates that the heterogeneity observed in CO2
distribution is introduced only when assigning a unique capillary pressure curve to each
grid cell. Therefore, to replicate the kind of spatial variations in CO2 saturation observed
in the experiments, the capillary pressure characteristic curve must be different in each
grid element.
Table 3. 5 The effect of porosity, permeability, and capillary pressure on CO2 saturation
distribution.
Porosity Φ
Uniform Measured Measured Measured
Permeability k
Uniform Uniform
K-C Model
K-C Model
Capillary
Pressure Pc
Uniform
Uniform
Uniform
Various
Saturation
Distribution
Average
Saturation SCO2 24.38% 24.38% 24.26% 21.24%
3.3.4 Input Relative Permeability
The relative permeability curves used in the simulations for both homogeneous and
31
heterogeneous cores are power-law functions:
CO w2
2
2
n n
w CO ,r w wr
r,CO r,w
wr wr
1-S -S S -Sk = , k =
1-S 1-S
(3.6)
where Swr is the residual brine saturation, and nw
and nCO2 are the functional exponents
for the brine and CO2 curves respectively (Figure 3.7). These three parameters in the
relative permeability functions were chosen to fit the relative permeability data calculated
from experimental measurements conducted by Perrin and Benson (2010).
Figure 3.7 Input relative permeability curves for CO2 and brine with Swr =0.15, nw =7,
and nCO2=3 and the experimental data.
3.3.5 Input Injection Flow Rate
The range of injection flow rates is chosen from 0.001 ml/min to up to 24 ml/min (four
orders of magnitudes). In the laboratory, flow velocity (v) can be calculated based on
volumetric flow rates (qlab) and core cross section (Acore). Consider a homogeneous and
isotropic reservoir with a thickness of 100 meters and a fully penetrated injection well
where CO2 is injected uniformly at a constant rate of 1 Mt CO2/yr (Figure 3.8); We can
calculate the flow velocity as a function of distance from the well using a 1-D radial
geometry:
well lab
core
Q qv= =
2πrH A (3.7)
32
Figure 3.8 Conceptual model of the reservoir used to estimate the range of relevant flow
velocities.
For example, Figure 3.9 shows that laboratory flow rate 1.2 ml/min corresponds to CO2
plume at 4.2 meters away from the injection well; 0.1 ml/min corresponds to 100 meters
away and 0.01 ml/min corresponds to 1 kilometer away.
Therefore these flow rates span the range of conditions expected in the near-will
region (1.2 ml/min) to the leading edge of the plume (0.001 ml/min), which may be up to
5 km or more from the injection well. Based on this range of values, the interplay of
viscous, capillary and gravity forces in core flood experiments is investigated at different
gravity and capillary numbers representative of those expected for a typical sequestration
project.
Figure 3.9 The volumetric flow rate and its corresponding distance of CO2 plume at the
reservoir (100 meter thickness and 1 Mt CO2/yr injection rate).
33
3.4 Output Parameters
For a given flow rate, simulations of co-injection of CO2 and brine are run until the
pressure drop and core-averaged saturation stabilize. All of the simulations have been
confirmed to run long enough (more than at least 10 pore volumes injected) to reach
steady-state. Important output parameters include grid cell CO2 saturation, CO2 pressures
and capillary pressures.
3.4.1 Slice-Average Quantities (SCO2, PCO2, Pc)
Here we also evaluate slice average quantities along the length of the core such as the
slice average CO2 saturation (SCO2), slice-average pressure in the CO2 phase (PCO2), and
slice-average capillary pressure (Pc). Figure 3.10 shows a typical simulation result: with
the CO2 saturation distribution, pressure drop across the core, and the core-averaged CO2
saturation.
Figure 3.10 CO2 saturation distribution at steady-state for 95% fractional flow of CO2 at a
total injection flow rate 1.2 ml/min.
The pressure drops across the core are defined as the difference between the inlet and the
outlet values:
2 2 2CO CO ,inlet CO ,outletΔP =P -P (6.1a)
w w,inlet w,outletΔP =P -P (6.1b)
Since Pc= PCO2–Pw, the water pressure drop can be rewritten in terms of the two output
parameters ΔPCO2 and ΔPc:
34
2 2 2w CO ,inlet CO ,outlet c,inlet c,outlet CO cΔP = P -P - P -P =ΔP -ΔP (6.2)
3.4.2 Simulated CO2 Saturation Distributions
In addition to measuring the average saturation of the core, we can also measure the
saturation distribution. For the homogeneous cores with no buoyancy force, the CO2
saturation distribution is uniform and equal after steady state for every flow rate.
Saturation gradients only occur along the horizontal direction during the transient period
while the front is advancing through the core, and shortly thereafter when the fractional is
approaching the fractional flow at the inlet.
However, once gravity is included, the saturation distribution changes significantly.
Figure 3.11 compares the saturation distribution for the homogeneous and the
heterogeneous cores. It shows that not only does the average brine displacement
efficiency decrease when flow rates are lower, but the distribution also changes. For high
flow rates, the saturation gradient is very small and only in the vertical direction due to
the density difference between CO2 and brine. However, saturation gradients occur along
both vertical and horizontal directions when gravity effect is relevant.
When comparing the saturation distribution between the homogenous and
heterogeneous cores, it is evident that the combination of gravity and heterogeneities
influences the CO2 distribution.
35
Figure 3.11 Saturation distribution for homogeneous and heterogeneous cores with
gravity, using grid with pixels (150, 150, 31) to grid blocks (25, 25, 31).
In summary, simulations are performed to replicate CO2/brine core flood experiments
that are representative of the typical reservoir condition occurring at sites of CO2
sequestration in saline aquifers. The conditions for the laboratory core flood experiments
are chosen to replicate temperatures and pressures typical of reservoir conditions (Perrin
et al 2009; Perrin and Benson 2010). For example, for the simulations described here, the
pressure is 124 bars (1800 psi) and the temperature is 50˚C. For the experiments,
injection rates are chosen to correspond to the flow rates in the near well region in the
field. Pore velocities range from 10-5
m/s to 10-3
m/s, as would be expected in the near-
well region for a typical storage project.
36
Chapter 4
3D Numerical and 2D Analytical Studies
for the Homogeneous Cores
In this chapter, we are interested in the drainage flow behavior in the CO2/brine two-
phase flow in the horizontal core. The main objective of this paper is to develop a greater
understanding of steady-state displacement mechanisms relevant to CO2 sequestration at
the core scale under reservoir conditions and to understand the transitions between
different flow regimes. Potential applications of the results include: 1) establishing the
bounds over which relative permeability can be accurately measured in horizontal core-
flood experiments; and 2) developing model experimental systems for studying
multiphase flows for a wide range of fluid pairs, geometric configurations and rock
properties. Additionally, the results have relevance for understanding reservoir-scale
processes, particularly at the sub-grid scale, where intra-gridblock processes may have an
influence on flow and transport parameterizations. Here we provide a new analytical
solution to identify different flow regimes in homogeneous and isotropic systems based
on the similar dimensionless groups used in Zhou et al. (1994). A two dimensional
analysis of the governing equations accounting for viscous, gravity, and capillary forces
at steady-state is used to develop an approximate semi-analytical solution for predicting
non-wetting phase saturations during core-flood experiments. A systematic parametric
study of the flow regimes is performed numerically to help identify flow regions. For
example, a wide range of flow rates, permeability, interfacial tension, and different
lengths of the core are investigated. Finally, the semi-analytical solution is validated by
comparing the results under a wide range of parameters, including different fractional
37
flows and relative permeability curves.
4.1 Methodology
The overall methodology for this study contains two parts. First, we conduct 3D high
resolution simulations of core scale CO2/brine flow over a wide range of injection
flowrates. We simulate the injection of a constant ratio of CO2/brine at a given flowrate
into a horizontal and initially brine saturated core. The corresponding brine displacement
efficiency is assessed when the system reaches steady-state (defined as the time when the
saturation is no longer changing and the fractional flows of CO2 and brine are equal at the
inlet and outlet of the core). Sensitivity studies incorporating a wide range of
permeability, core lengths and interfacial tension values have also been studied to
generalize these results. Second, a theoretical analysis of the multiphase flow equations
is used to develop an approximate semi-analytical solution for predicting the average
saturation in the core as a function of gravity number, Bond number and several
dimensionless parameters. Finally, we compare the results of simulations and
approximate semi-analytical solutions to test the validity of the solution over a wide
range of parameters.
4.1.1 Simulation Study
Initially, we focus on 95% fractional flow of CO2 and 5% brine, which are injected
simultaneously into a simulated core at a wide range of flowrates. Later we consider a
wide range of fractional flows. In this chapter, we assume a homogeneous core with the
average petrophysical properties of a Berea Sandstone (mean porosity of 0.202 and mean
permeability of 430 md). The capillary pressure gradient between the last slice of the
core and the outlet slice is set to zero to minimize end effects. This boundary condition
has been found to most accurately replicate saturation distributions measured during core
flooding experiments. The input curves for the capillary pressure and relative
permeability used in this section have been illustrated in Figure 3. 4 and Figure 3. 6.
The details regarding the simulation have already explained in Chapter 3. The
influences of flow rate, interfacial tension, and core permeability on the CO2/brine flow
38
systems are studied: flow rate is changed by several orders, from 0.001 ml/min up to 60
ml/min which corresponds the CO2 velocity from near well region (~m) to the leading
front of the plume (~10 km) based on the analysis shown in Section 3.3.5; interfacial
tension and core permeability are varied by two orders of magnitudes to study a range of
values. Additionally, a range of core lengths and heights is also included.
Figure 4. 1 a through c show the core average CO2 saturations for the homogeneous
cores as a function of traditional capillary number Ca, alternative capillary number Ncv,
and gravity number Ngv for two sensitivity studies using the definitions shown in Eqs.
4.1-4.3 (Fulcher et al. 1985; Lake 1989; Zhou et al. 1994).
2t coCa= u μ σ (4.1)
2
*
c
cv 2
co t
kLpN =
H μ u (4.2)
2
gv
co t
ΔρgkLN =
Hμ u (4.3)
where ut is the total average Darcy flow velocity, μCO2 is CO2 viscosity, pc* is a
characteristic capillary pressure and is defined as the displacement capillary pressure
(more detail in Chapter 6), L the length of the core, H the height of the core, Δρ=ρw-ρg is
the density difference between CO2 and brine, and g is the acceleration of gravity. The
values of Ca, Ncv and Ngv are controlled by varying injection rates, interfacial tension,
and permeability where the other parameters are kept as constants. The wide range of
IFT values is purely hypothetical and used solely to explore the sensitivity to IFT values.
Higher flowrates, lower permeability, smaller density differences and lower capillary
pressure have higher Ca, or lower Ncv and Ngv. The high flowrate regime is
representative of the near well region. Lower flowrates, hence the lower Ca or higher Ncv
and Ngv, are representative of the leading edge of the plume and during fluid
redistribution in the post-injection period. We can also define Bond number NB as a ratio
of gravity to capillary numbers:
gv
B *
cv c
N ΔρgHN =
N p (4.4)
39
Since capillary pressure depends on the interfacial tension and permeability (Eq. 3.5),
different pairs of σ and k result in different pc*, and hence different Bond number NB.
Table 4.1 shows the summary of four sensitivity studies and their corresponding Bond
number.
Table 4. 1: Summary of Sensitivity Studies
σ, mN/m k, md pc*, Pa NB
Base Case (σ, k) 22.47 430 3000 0.061
Sensitivity 1 (3σ, k) 67.41 430 9000 0.022
Sensitivity 2 (σ/3, k) 7.49 430 1000 0.198
Sensitivity 3 (σ, 0.1k) 22.47 43 9487 0.019
Sensitivity 4 (σ, 10k) 22.47 4300 947 0.192
All the simulation results shown in Figure 4. 1 are for a fractional flow of 95% CO2. The
sensitivity studies for interfacial tension are illustrated in the left hand side of the figures
and permeability on the right. When comparing the average saturation in terms of
traditional capillary number Ca, alternative capillary number Ncv and gravity number Ngv,
the efficiency of brine displacement clearly falls into three separate regimes: a viscous-
dominated regime where the saturation is independent of Ca, Ngv and Ncv; a gravity-
dominated regime where the average saturation is strongly dependent on the Ngv and Ncv;
and a capillary-dominated regime characterized by lower average saturations and smaller
sensitivity to variations in Ngv and Ncv.
The transitions from the viscous- to gravity-dominated regimes and from the gravity-
to capillary-dominated regimes are dependent on the interfacial tension, permeability and
flow rate (Figure 4. 1a and Figure 4. 1b). Results show that the transition points occur
earlier when gravity has a stronger influence (large Bond number), for example, for the
lower interfacial tension and higher permeability cases. Plotting the same data in terms
of gravity number (Ngv) results in the same transitions for different interfacial tensions or
core permeabilities, or simply, different Bond numbers (Figure 4. 1c). Here we define
the first transition, from the viscous- to gravity dominated regime, as the critical number
1 and the second transition, from gravity- to capillary-dominated regime, as the critical
number 2. The dashed lines shown in Figure 4. 1c indicate the two critical numbers.
40
In the viscous-dominated regime, defined as Ngv ≤ Ngv,c1, the brine displacement
efficiency is independent of Bond number. Average saturations are also accurately
predicted based on the Buckley-Leverett theory which neglects capillarity and gravity.
Macroscopically, the capillary pressure and CO2 saturation are uniform along the vertical
and horizontal axes of the core. The saturation gradient or the capillary pressure gradient
are very small (<5%) and only in the vertical direction.
As the flow rate decreases, gravity starts to have a significant effect on the multiphase
flow behavior and the viscous and gravity forces become similar in magnitude. In the
gravity-dominated regime (Ngv,c1 < Ngv < Ngv,c2), the brine displacement efficiency is
highly flowrate dependent and insensitive to the Bond number (Figure 4. 1c). Gravity
override leads to saturation gradients in the vertical direction, with higher saturation at
the top of the core, as would be expected for buoyancy driven flows. As a result of this, a
saturation gradient is also established along the horizontal axis of the core. In this
regime, the saturation gradient is large and exists along both vertical and horizontal
directions. This decreasing trend of brine displacement efficiency changes gradually over
two orders of magnitude of flow rate.
Capillary forces dominate when flowrates are reduced further. In this capillary-
dominated regime (Ngv,c2 ≤ Ngv), the effect of gravity is small and the brine displacement
efficiency is strongly dependent on capillary forces, and hence the Bond number.
Capillary pressure gradients control the displacement of the fluids, tending towards
achieving capillary-gravitational equilibrium based on the capillary pressure curve of the
core. In this regime the saturation gradient is small and mainly in the horizontal
direction.
41
(a)
(b)
(c)
Figure 4. 1: Average CO2 saturation as a function of (a) traditional capillary number Ca,
(b) alternative capillary number Ncv, and (c) alternative gravity number Ngv. The
interfacial tension σ and the permeability of the base case are 22.47 mN/m and 430 md,
respectively. The sensitivity studies of interfacial tension are illustrated in the left hand
side of the figures and permeability on the right.
Figure 4. 1 illustrates the pressure gradients associated with the viscous, gravity and
capillary forces as a function of gravity numbers for the high interfacial tension case (3σ).
The values in Figure 4. 1 are obtained directly from the output of the high resolution 3-D
simulations using TOUGH2 described in Chapter 3. The viscous and capillary pressure
drops in the flow direction (ΔPv,x and ΔPc,x) are calculated based on the slice-average
pressure difference between inlet and outlet grid cells while the pressure drop due to
gravity (ΔPgrav) is defined as ΔρgH. Note the pressure obtained from TOUGH2 is the gas
42
phase pressure. It is clear that the viscous force is dominant before the first transition.
The transition between the viscous and gravity dominated flow regime occurs when the
viscous pressure gradient in the flow direction (ΔPv,x/L) and the pressure gradient of
gravity in the vertical direction (ΔPgrav/H) have the same magnitude. In the second
regime, gravity force eventually exceeds the viscous force while capillary force also
increases to have the similar magnitude. The second transition, from the gravity to the
capillary dominated regime, occurs when the viscous pressure drop and the capillary
pressure drop in the flow direction have the same magnitude (ΔPv,x≈ΔPc,x), which implies
the capillary-dominated regime occurs when the water pressure drop across the core has
almost vanished. The pressure gradients for more sensitivity cases can be referred to
Appendix C.
Figure 4. 2: Pressure gradients for three different forces as a function of alternative
gravity number Ngv. The interfacial tension and the permeability of the 3σ case are 67.41
mN/m and 430 md, respectively.
Based on Figure 4. 1c and Figure 4. 2, the transition between the viscous and gravity-
dominated regimes occurs when Ngv ≤ 54 while the capillary-dominated regime begins
when Ngv ≥ 220. The first critical gravity number can be derived from first principles,
that is, when the buoyancy pressure gradient equals the viscous pressure gradient.
Assuming uniform saturation, the viscous pressure drop across the core can be calculated
based on the multiphase extension of Darcy’s law:
43
2 2 2
2
2 2
r,CO CO t COv,x
CO v,x
CO r,CO
kk f u μΔpu =- Δp =
μ L kk
L (4.5)
uCO2 is Darcy flow velocity of CO2 and fCO2 is CO2 fractional flow. The pressure gradient
due to buoyancy forces is ΔPgrav= ΔρgH. Therefore, the first critical gravity number
(Ngv,c1) can be calculated by:
2 2
2 2 2 2 2 2
r,CO r,COgrav
gv
CO t CO r,CO CO t CO CO lv,x
Δρgkk kΔp H Δρg= = N 1
f u μ kk f u μ f RΔp L (4.6)
2
2
CO l
gv,c1
r,CO
f RN
k (4.7)
Rl =L/H is the shape factor or the so-called the aspect ratio. Using Eq. 4.7, the first
critical number is 53.9, which agrees well with the results from the simulations shown in
Figure 4. 2. Similar analysis can be used to derive the second critical number, Ngv,c2.
The capillary pressure of multiphase flow system is given by pc=pc*J. The ratio of the
viscous pressure drop to the capillary pressure drop is
2 2 2 2 2 2
22
2v,x CO t CO r,CO CO t CO CO
* * 2
c r,CO cvc r,CO c
Δp f u μ kk f u μ fL L= =
|Δp | ΔJ k ΔJNp ΔJ kk p H
L
(4.8)
2 2 2
2 2 2
2 2 2
v,x CO l CO l CO l
cv,c2 gv,c2 B
c r,CO cv r,CO r,CO
Δp f R f R f R= =1 N = or N =N
|Δp | k ΔJN k ΔJ k ΔJ
(4.9)
However, it is more difficult to predict Ngv,c2 since ΔJ depends on the average saturation
at the inlet and outlet ends of the core, which depends on a number of variables.
Therefore, an empirical method, based on an evaluation of the sensitivity studies, is used
to calculate the second critical number:
3/2
l
gv,c2
BL
RN =α where α=12.79737
S (4.10)
The second critical number depends on the aspect ratio and the saturation predicted by
the Buckley-Leverett solution (SBL). Comparing Eq. 4.9 and Eq. 4.10 implies that ΔJ is
proportional to Rl0.5
. If we treat ΔJ as a fitting parameter and analyze a series of
sensitivity studies on the aspect ratio, it indeed shows a linear dependency on Rl0.5
(Figure
4.3). Therefore, it provides confidence in Eq. 4.10. The concept of three different
44
regimes and the associated critical numbers will be used in the analytical solution in the
following sections.
Figure 4. 3: The function ΔJ has a linear dependency on the square root of aspect ratio,
Rl.
4.2 Theoretical Analysis
4.2.1 2D General Solution
We first simplify our 3-dimensional problem into 2 dimensions (x-z direction). The
properties of this 2D porous medium are homogeneous and isotropic. Porosity and
permeability are constant. Mass conservation equations and pressure equation for
incompressible flow are:
j j j,x j,z
j
S S u uφ u =φ =0
t t x z
(4.11)
tu 0 (4.12)
with the condition Sw+2COS =1. uj,x and uj,z are Darcy velocity of phase j in the x and z
direction, respectively. Darcy flow velocities for both phases are given by
j j rj
j j j j,x j,z j
j
p p ku =-λ k , +ρ g u , u where λ =
x z μ
(4.13)
where uj, λj, pj, ρj and μj are Darcy flow velocity, relative mobility, pressure, density, and
45
viscosity of phase j, respectively. Boundary conditions used in the simulation are given
as followings. At the top and the bottom boundaries,
uw,z=ug,z=0 (4.14)
In the flow direction x, total volumetric flow rate is sum of water and gas flow:
uw,x+ug,x=ut (4.15)
Using Darcy’s equation and the boundary conditions, the mass conservation of gas phase
now becomes:
g g gt c c
w
S Mkλ Mkλu p p1 1- + Δρg- =0
t φ kλ x 1+M φ z z 1+Mx
(4.16)
M= λw/λg is mobility ratio. To non-dimensionlise the equation, we define xD=x/L,
zD=z/H, tD= t/tchar= tut/φL, and pc(Sw) =pc*J (Sw). The characteristic time tchar is chosen as
the time it would take for one pore volume to flow through the core at a velocity of ut.
Substituting all the defined terms into Eq. 4.16 yields the dimensionless mass
conservation equation:
2
g rg rg rg
gv cv
D D D l D D D D
S Mk Mk Mk1 1 J J+ +N -N + =0
t x 1+M z 1+M R x x 1+M z z 1+M
(4.17)
where Ncv and Ngv are shown in Eq. 4.2 and 4.3. Assuming steady-state (Eq. 4.18) and
using separation of variables, we can obtain 2D time independent CO2 saturation SG (Eq.
4.19). The derivation is detailed in Appendix A for reference.
g
D
S=0
t
at steady state (4.18)
2
lD
cvB D B D
R- ax
N-N bz -N bz
1 2 3 4SG C e +C e +C e +C (4.19)
Unknown variables a and b are functions of saturation and mobility ratio M (see
Appendix A.2). C1, C2, C3 and C4 are functions of xD and zD. Based on this solution, the
time independent CO2 saturation SG depends only on its position xD and zD, M, as well as
46
the dimensionless numbers Rl, Ncv and NB.
4.2.2 2D General Solution with Simulation Constraints
Since it is difficult to integrate Eq. 4.19 to obtain the average core saturation, we assume
that the saturation at a particular point (x0,D, z0,D) is representative of the core average
saturation2COS :
2CO 0,D 0,DS SG x ,z (4.20)
Now C1, C2, C3 and C4 become constants and the two terms in the exponent ax0,D and
bz0,D are unknown variables which will be determined later by fitting the simulation
results. To eliminate some unknown parameters, we use several observations from the
simulation results. First, Figure 4. 1c shows that average CO2 saturations are independent
of gravity number Ngv when Ngv ≤ Ngv,c1:
2l
B 0,D2 gv,c1
R- N ax
CO N
gv gv,c1
gv
S 0 when N N e 0
N
(4.21)
Therefore, in the viscous-dominated regime, Ngv ≤ Ngv,cl, the core average saturation
becomes the Buckley-Leverett solution SBL:
B 0,D
2
-N bz
CO 3 4 BLS C e +C S . (4.22)
The average saturation in this regime is independent of Bond number, which implies
C3=0 and C4= SBL. Second, when Bond number equals to zero (g=0), saturations are
observed to be a constant SBL (Kuo et al. 2010):
2l
0,Dcv
2
R- axN
CO 1 2 BL BLS C +C e +S S (4.23)
Eq. 4.23 results in C2= -C1. Based on above constraints, Eq. 4.19 can be rewritten as the
following form:
2
l0,D
B 0,D cv
2
R- ax
-N bz N
CO 1 BLS C e -1 e +S (4.24)
Letting the term ax0,D ≡d1Ngv,cl/NB, we need to satisfy2
l 1-R de 0 based on Eq. 4.21. Once
47
we choose an appropriate value of d1 to satisfy this condition, we can replace the
unknown term ax0,D in terms of d1, NB, and the critical gravity number Ngv,c1. Similarly,
Figure 4. 1c also illustrates that saturations are not sensitive to Bond number when Ngv ≤
Ngv,c2, which leads to
2 B 0,DCO -N bz
gv gv,c2
B
S0 when N N e 0
N
(4.25)
Defining the exponent term bz0,D≡d2/Ngv,c2, choosing an appropriate value of d2 to satisfy
2 B-d N Ngv,c2e 0 , and the saturations are now determined by Bond number and gravity
number, shown in Eq. 4.26.
2l gv,c1B
2 1gv,c2 gv
2
R NN-d -d
N N
CO 1 BLS =C e -1 e +S
(4.26)
C1, d1, and d2 are fitting parameters which can be determined by simulation results or
experimental data. Eq. 4.26 shows that the core average saturation depends on two terms.
First, in the high flowrate regime or for Ngv ≤ Ngv,c1, the average saturation (and hence,
SBL) is determined solely from the fractional flow curve based on Buckley-Leverett
theory (1942) which neglect gravity and capillary pressure:
SBL = SBL(2COf ) (4.27)
where CO2 fractional flow, fCO2, is a function of mobility ratio, M:
2COf =1 1+M (4.28)
If we know the input relative permeability curves, the viscosity of the two fluids and the
CO2 fractional flow, we can determine the corresponding SBL.
Second, the core average saturation also depends on a correctional term which
combines the interactions between all three forces through the dimensionless numbers
and the aspect ratio. Therefore, Eq. 4.26 is the modified Buckley-Leverett solution; it not
only depends on the viscous, capillary and gravity forces but also depends on the size of
multiphase flow system. In the literature, people often solve mass conservation equation
numerically by simplifying the conservation equations into different flow regimes (Zhou
48
et al. 1994; Yortsos 1995) or by generalizing the traditional mass conservation equations
without providing the explicit form of solutions (Hassanizadeh and Gray 1993b). Here
we provide a tool which can actually predict the solution of the CO2 saturation explicitly.
4.3 Results
4.3.1 Comparison between Simulations and the Approximate Semi-
Analytical Solution
Figure 4. 4 shows that the average CO2 saturation of the base case as a function of
capillary number Ncv (LHS) and the CO2 fractional flow curve (RHS). To compare our
theoretical predictions with the simulation results shown earlier, in addition to SBL and
two critical numbers Ngv,c1 and Ngv,c2 described above, we need to determine variables C1,
d1 and d2. Table 4. 2 provides the values of dimensionless parameters and the fitting
parameter (C1, d1 and d2) for the base case with interfacial tension σ and core
permeability k (see Table 4.3 for parameter values). As shown, the semi-analytical
solution compares well with the simulation results and captures the major features and
transitions of the curve.
Figure 4. 4 (LHS) Comparison of average CO2 saturation as a function of capillary
number Ncv between theoretical values and simulation results for the homogeneous or
base case (σ, k); (RHS) fractional flow curve based on our input relative permeability
curves (Eq.3.6).
Table 4. 2: Values of unknown variables used to match the base case.
Ngv,c1 Ngv,c2 C1 d1 d2
54 220 934.18 0.3747 1
49
Table 4. 3: Berea core properties and fluid properties used in the homogeneous cores for
the base case
mean φ mean k,
md σ, mN/m L, cm H, cm Rl μbrine, cp μCO2, cp
0.202 430 22.47 14.73 4.69 3.1445 0.558 0.046
By analyzing the fitting parameters C1, d1, d2 for a wide range of input parameters, we
have been able to establish that the fitting parameters d1 and d2 depend on the aspect ratio
of the core and that C1 depends on the Bond number:
11
1 11
l l
dd = where d =2.0894
R R (4.29a)
l l
2 22
22
R Rd = where d =5.57607
d (4.29b)
11
1 11
B
CC = where C 56.78
N (4.29c)
Ideally, one would derive the constant coefficients such as α, d11, d22 and C11 from first
principles. However, it is not clear how to develop a closed-form solution for
determining these coefficients. Consequently, we use an empirically based approach to
identify these coefficients based on curve fitting to the simulation cases. The general
form of core average CO2 saturation now can be predicted based on Eq. 4.26 and 4.29 as
well as the critical numbers in Eq. 4.7 and 4.10.
Figure 4. 5 show the sensitivity studies for permeability (43 md and 4300 md) and
interfacial tension (7.49 mN/m and 67.41 mN/m) respectively. This wide range of
sensitivity studies covers a realistic range of relevant parameters such as permeability in
the field. The range of interfacial tension is much larger than expected and selected
solely to explore how variations in interfacial tension would affect the results. Those
figures compare the simulation results and the predicted values based on Eq. 4.26 and Eq.
4.29. As shown, we can replicate the simulation results quite well, especially the
transition from the viscous- to gravity-dominated regime. However, a slight mismatch in
the capillary-dominated regime occurs for the cases when the capillary force is strong
(0.1k and 3σ). When the capillary force is relatively small such as 10k (4300 md) and
50
σ/3 (7.49 mN/m), the semi-analytical solution matches best in the transition from
viscous- to gravity-dominated regime but deviates slightly in the gravity- and capillary-
dominated regime.
Figure 4. 5 Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of interfacial
tension. LHS: (3σ, k), RHS: (σ/3, k)
4.3.2 Model Validation
Here we test the validity of Eq. 4.26 by using different core dimensions (Figure 4. 6),
different fractional flows of CO2 (Figure 4. 7), and different input relative permeability
curves (Figure 4. 8).
Different Core Dimensions
We have carried out simulations with the aspect ratio Rl ranging from 1.6 to 16. The
solution provides accurate values for the average core saturation within this range (Figure
4. 6). Since the input relative permeability curves and the fractional flow of CO2
(fCO2=0.95) for these cases are the same as for the base case, the Buckley-Leverett
solution SBL is still 0.324. As shown, the semi-analytical solution predicts the average
saturation quite well for different core geometries. In addition, Eq.4.7 shows that the
51
larger aspect ratio results in a bigger Ngv,c1, which implies that the gravity-dominated
regime occurs at larger values of Ngv if the system has a larger aspect ratio.
Figure 4. 6 Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of different
dimensions of the core.
Different Fractional Flows of CO2
Figure 4. 7 shows the average CO2 saturations as a function of capillary numbers for CO2
fractional flows of 0.79, 0.51 and 0.34. Although the same input relative permeability
curves are used, different fractional flow of CO2 results in different Buckley-Leverett
solution SBL and hence different krg(SBL) values. Again, the semi-analytical model
predicts average saturations very well for the entire range of fractional flows.
52
Figure 4. 7 Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of different
fractional flows of CO2.
Different Relative Permeability Curves
The final cases to test our semi-analytical solution use different relative permeability
curves with 0.95 fractional flow of CO2. The other parameters such as those in Table 4.3
are the same. For these cases, the Buckley-Leverett solution SBL and krg(SBL) are
different due to various relative permeability curves. Figure 4. 8 shows the average CO2
saturations as a function of capillary numbers for six different input relative permeability
curves. The first four cases (the 1st and 2
nd rows of Figure 4. 8) use the same form of
relative permeability functions shown in Eq. 3.6. The base cases use parameters nw=7,
nCO2=3, and Swr=0.15 (IRP1) while the other three cases change one or two parameters at
a time. One has a lower residual brine saturation (IRP2), one has a higher exponent of
brine nw (IRP3), and the last one has a higher exponent of nCO2 as well as a lower residual
brine saturation (IRP5). The 2D model predicts the lower residual brine saturation
(IRP2) and the higher nw (IRP3) very well while the predictions for the higher nCO2 and
lower residual brine saturation (IRP5) have slight deviations in the gravity- and capillary-
dominated regime. However, the semi-analytical solution still captures the transition
from viscous- to gravity-dominated regime quite accurately. The last two cases (the 3rd
53
row of Figure 4. 8) use Corey’s equation with residual saturation 0.05 and 0.15,
respectively. The results show that even with different relative permeability functions,
the semi-analytical solutions still match the simulation results quite well.
Figure 4. 8 Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for six different input relative
permeability curves. Those simulations share the same constant parameters in Table 4.3.
4.4 Discussion
In previous sections, we have shown that simulation results and theoretical predictions
agree quite well. The average CO2 saturation is determined by three forces. The relative
magnitudes of gravity to capillary forces and viscous to gravity forces are characterized
by the Bond number NB and gravity number Ngv. The combined effect of these forces
results in Eq. 4.26. The semi-analytical model provides a simple tool to determine
54
approximate values for the critical numbers. Eq. 4.29 works well for all the cases tested
in this paper.
Table 4. 4 summarizes three different flow regimes and the corresponding core
average CO2 saturation expressions based on Eq. 4.26. At high flow rates or the so-called
viscous-dominated regime (Ngv ≤ Ngv,c1), the brine displacement efficiencies for
homogeneous cores are flowrate independent at the given fractional flow of CO2. The
value of SBL can be predicted from Buckley-Leverett theory neglecting capillarity and
gravity. When buoyancy forces begin to dominate multiphase flow (Ngv,c1 < Ngv < Ngv,c2),
average saturations are highly flow rate dependent. In this gravity-dominated regime,
the average saturation is mainly dependent on gravity number. When Ngv > Ngv,c2,
saturation becomes less sensitive to the gravity number. In the capillary-dominated
regime, the average saturation asymptotically approaches to a constant value.
Table 4. 4: Summary of flow regions for general cases
Flow region Conditions Steady State CO2 saturation
Viscous-dominated regime Ngv ≤ Ngv,c1
2CO BLS S
Gravity-dominated regime Ngv,c1 < Ngv < Ngv,c2
2l gv,cl
1gv
2
R N-d
N2 11
CO BL
gv,c2
d CS S - e
N
Capillary-dominated regime Ngv,c2 ≤ Ngv
2
2
1 l gv,cl2 11
CO BL
gv,c2 gv
d R Nd CS S - 1-
N N
Some degree of mismatch between the semi-analytical solution and TOUGH2
simulations are observed in both the gravity- and capillary-dominated regimes. The
differences between the 2D analytical solution and 3D simulation may be explained by a
number of reasons. First, we assume that all of the cases share the same fitting
parameters C11, d11, α and d22 once they are adjusted to match the simulation results of a
reference case. However, C1 is a function of (x0,D, z0,D) and hence a function of NB,
Ngv,cl, and Ngv,c2. Based on the derivation in Appendix A.2, the coefficients d11 and d22
are functions of several parameters such as mobility ratio M, the derivative of J-function,
and relative permeability coefficients aw and ag since the first critical number Ngv,cl
depends on fCO2/krg while the second critical number depends on SBL and hence fCO2 (see
55
Eqs. A-19). In addition, the fitting parameter α also changes slightly for different cases:
g w
11
a +ad =f
J'
(4.30a)
g w
22
a M-ad =f
J'
(4.30b)
2B rg CO Lα=f N ,k ,f ,R (4.30c)
Therefore, the assumption used to develop this solution requires that the combined effects
of other parameters are constant. We show that this assumption is reasonably good for
the cases studied here, but it is not strictly correct. The semi-analytical predictions for
the lower flowrate regimes can be improved by slightly adjusting these coefficients.
However, the differences are small and the correct functional form to adjust these
coefficients systematically is not evident. Further investigation could generalize
solutions by exploring the sensitivity to these parameters.
Second, the mismatch in the capillary-dominated regime (Figure 4. 4) could be
influenced by the outlet boundary condition we use for the simulation (no Pc gradient in
the flow direction at the outlet). Although this boundary is chosen to match the
experiments at moderate (~cm3/min) flow rates, we find that even for lower velocities (in
the gravity-dominated regime), the analytical predictions are close to the simulation
results. The potential artifact may occur only at very low flowrate, for example, when
Ngv > 105.
Finally, the gridding and the time step size could potentially contribute to the
mismatch observed. As discussed in Section 3.2.1, the size of the grid used in this paper
(1.8mm x 1.8mm x 5mm) is chosen to balance several considerations. The LHS of
Figure 4. 9, which compares results from the grid used for these studies to another with ½
the size in the y and z directions (0.88 mm) yields essentially identical results, with a
maximum deviation in the average saturation in the core of less than 2%. The good
agreement between the higher and lower grid resolution simulations suggests that grid
refinement is not the cause of the discrepancy between the analytical solution and
numerical results. In addition, to avoid numerical artifacts caused by the time-step size,
the initial time step is chosen to have a small CFL number (uxΔt/Δx<1) for every flow
56
rate, shown on the RHS of Figure 4. 9. For subsequent time steps, it is automatically
adjusted by TOUGH2 to higher or lower values during a simulation run dependent on the
convergence rate. After breakthrough, the time step size will increase up to the
maximum time step size Δtmax set up at the beginning.
For instance, for the case of the viscous-dominated regime such as 1.2 ml/min, the
initial time step is set as 8 seconds (CFL number=0.05) and it reduces to the smaller time
step size (4 seconds) or increases to the larger time step size (16, 32, or 64 seconds)
dependent on the convergence rate. The corresponding CFL numbers are 0.025, 0.1, 0.2,
and 0.4, which all smaller than 1.
Figure 4. 9 (LHS) The finer grids have grid dimensions 0.884mm x 0.884mm x 5.08mm
(grid numbers: 53x53x31) while the coarser grids have dimensions 1.874mm x 1.874mm
x 5.08mm (grid numbers: 25x25x31); (RHS) CFL number (uxΔt/Δx<1) for different flow
rate.
4.5 Conclusions
A new semi-analytical solution has been developed to predict the influence of gravity and
capillary numbers on the average saturation expected during multiphase flow
experiments. Although computational technology has improved greatly, running high
resolution 3D models including capillarity and gravity still takes a significant amount of
computational effort. The new solution provided here is a quick and easy way to
estimate the flow regimes for horizontal core floods. Practical applications include
helping to design core flood experiments, including assuring that relative permeability
measurements are made in the viscous dominated regime, evaluating potential flow rate
dependence, influence of core-dimension on a multiphase flow experiments, and
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
1 10 100 1000 10000 100000
Ave
rage
CO
2 Sa
tura
tio
n
Capillary Number, Ncv
Grid Effect (Homogeneous)
CoarseScale: 25x25x31
FineScale: 53x53x31 0
0.2
0.4
0.6
0.8
1
0.01 0.1 1 10 100
CFL
nu
mb
er
flow rate q (ml/min)
CFL number
57
influence of fluid properties on the experiments. In addition, having a semi-analytical
solution provides a useful tool for investigating multiphase fluid displacement efficiency
over a wide parameter space of practical interest. Other applications, not explored in this
paper, include investigation of upscaling strategies in the transition between the viscous,
gravity and capillary dominated regimes. The new semi-analytical solution can be used
to estimate the average saturation over a wide range of conditions in terms of several
important dimensionless numbers such as mobility ratio, relative permeability to gas
evaluated at SBL, aspect ratio Rl, Bond number NB and gravity number Ngv.
The solution has been compared to 3D high resolution simulations to study the effects
of flowrate, gravity, interfacial tension, core-length, and core permeability on two-phase
immiscible flow. The proposed 2D semi-analytical technique predicts the brine
displacement efficiency for 3D homogeneous CO2/brine two-phase flow simulations very
well when the Bond number ranges from 0.02-0.2 and aspect ratio ranges from 1.57-
15.72. As mentioned before, the sensitivity studies cover the wide range of permeability
and interfacial tension values and hence results in this range of Bond number, which is in
the parameter range of interest for CO2 sequestration in deep aquifers.
Theoretical predictions match the corresponding simulation results not only for the
base case but also for many sensitivity cases. It can also apply to various fractional flows
of CO2 as well as different input relative permeability curves. It is expected that the first
critical number can be defined accurately even for the larger aspect ratios outside the
range studied in this work, for example, Rl=100, which could be happened in the sub-grid
cell of field simulations with 100 meters in length and 1 meter in height. However,
further investigation is needed to confirm this conclusion. The limitations of the
analytical solution will be discussed in detail in Chapter 7.
58
Chapter 5
3D Numerical and 2D Analytical Studies
for the Heterogeneous Cores
In the Chapter 4, a 2D analytical solution was developed to predict the average saturation
in homogeneous cores during steady state core flood experiments over a wide range of
relevant conditions. The analytical predictions show excellent agreement with the
simulation results. Based on the knowledge gained from the homogeneous studies,
further investigations into the behavior of heterogeneous rocks are presented in this
chapter. Another important variable, capillary heterogeneity, is added into the 2D
analytical analysis to provide more realistic predictions for heterogeneous rocks.
In order to provide a direct comparison to the results for homogeneous rocks, the
same grid size, initial and boundary conditions, fitting parameters for capillary pressure
and relative permeability curves for the base case illustrated in Chapter 4 are used. For
example, the capillary pressure gradient between the last slice of the core and the outlet
slice is set to zero to minimize end effects. The input curves for the capillary pressure
and relative permeability implemented in all the simulations shown in Eq. 3.5 and Eq. 3.6
are used. Every simulation is performed at the reservoir condition 50˚C and 12.4 MPa.
We also focus on 95% fractional flow of CO2 and 5% brine injected simultaneously into
a simulated core at a wide range of flow rates. Additional details regarding the
simulations have been provided in Chapter 3. The only parameters changed are rock
properties of porous media such as the porosity, permeability and capillary pressure
curves of each grid cell in the simulation.
59
5.1 Heterogeneous Representations
The first part of Chapter 5 will present results of 3D numerical simulations with a wide
range of heterogeneity to understand the role of sub-core scale heterogeneities. To
generate different heterogeneity distributions, two methods are used. The first method
for assigning heterogeneity uses a constant porosity distribution (φ=0.202) and the
permeability distribution is generated based on a random log-normal distribution with a
standard deviation of σlnk. The other type of heterogeneity uses measured values of Berea
porosity distribution obtained from X-ray CT scanner to generate its corresponding
permeability distribution based on a porosity-permeability relation. Both methods are
implemented with isotropic permeability (kx=ky=kz) for every grid cell and the mean
permeability ranges from about 250 md to 570 md. In the heterogeneous cases, the
capillary pressure curve is calculated based on a modified form of the Leverett J-function
(Eq. 3.5) using the known porosity and permeability distributions. Each grid element
has a unique pair of porosity and permeability values; hence a unique capillary pressure
curve.
5.1.1 Random Permeability Distribution (3D)
Four different degrees of permeability heterogeneity are created based on a random log-
normal distribution. As mentioned above, porosity is uniform, which is a reasonable
assumption, since the variation of porosity is often small, unlike the variation in
permeability, which could be up to several orders of magnitude. Once the porosity and
permeability of a grid element is assigned, the corresponding capillary pressure curve is
generated based on Eq. 3.5. The relative permeability of each grid cell is identical (Eq.
3.6). With these inputs, the steady-state average CO2 saturation from a series of core-
scale simulations can be obtained. The standard deviations for the four different cases
and their permeability distributions are listed in Table 5.1.
Figure 5. 1 illustrates both the slice-average permeability and the corresponding
steady-state average saturation (at the 6 ml/min total injection rate) along the flow
direction for four random distributions. Case 1 (lnk=0.02) has the most uniform
permeability profile while the Case 4 (lnk=2.65) has the widest range of slice averaged
60
permeability values along the length of the core. It is observed that large variations over
small length scales result in a relatively large variation in the saturation profile, similar to
1D result in the literature (Yortsos and Chang 1989, 1990; Chang and Yortsos 1992). It
is also consistent with the experimental results showing that relatively homogeneous
cores result in the smoothly varying saturation profiles, while the saturation distributions
of very heterogeneous cores are variable (Krevor et al. 2012). Importantly, these graphs
also show that for randomly distributed heterogeneity, the larger the degree of
heterogeneity, the lower average saturation in the core.
Figure 5. 1: Four different random permeability profiles and the corresponding average
CO2 saturations in the viscous-dominated regime.
5.1.2 Porosity-based Permeability Distribution (3D)
The other type of heterogeneity distribution is generated using a porosity-based approach,
which is commonly used in the literature (Mavko and Nur 1997; Pape et al. 1999; Krause
et al. 2008). We have two permeability fields generated from porosity-permeability
61
models based on the measured porosity values of Berea Sandstone (Perrin et al. 2009).
One of three-dimensional permeability maps is generated from the Kozeny-Carman
equation (Kozeny, 1927; Carman, 1937):
3i
i 2i
φk =S
(1-φ )
(5.1)
S is a scaling factor that assures that the average of all permeability values is equal to the
average permeability of the core, 430 md. φi is a pixel value in the porosity map, and ki
is the corresponding value calculated using Eq. (5.1). To increase the contrast in
permeability and hence increase the degree of heterogeneity, an alternative empirical rock
property model is created as follows:
4
ii=exp 64φ -6k (5.2)
Figure 5. 2 illustrates the slice-averaged porosity profile and its two corresponding
permeability profiles based on Eq. 5.1 and 5.2. The porosity of Berea Sandstone varies
within a small range. Comparing Figure 5. 1 and Figure 5. 2, the permeability
distributions generated from the porosity values of a real rock change more smoothly
than the random distributions and have spatially correlated low permeability features
aligned sub-parallel to the axis of the core. The corresponding saturation profile is shown
in the bottom of Figure 5.2. For the Kozeny-Carman permeability model, the results are
nearly indistinguishable from the homogeneous case, which is consistent with the results
from the random distribution with a small value of lnk. For the model with a higher
degree of heterogeneity (High Contrast model), the slice-average saturation varies
significantly along the length of the core and the average saturation is lower. These
results are qualitatively consistent with the results from the random distribution.
62
Figure 5. 2: Average permeability and average CO2 saturation along the flow direction
for three different heterogeneous cores (Homogeneous model, Kozeny-Carman model,
and High Contrast model).
5.1.3 Permeability Distribution Summary
Table 5. 1 illustrates all the 3D permeability distributions used for this study: Random 1-
4 cases and the two porosity-based permeability fields (Eq. 5.1 and 5.2). Although the
isotropic permeability is implemented for all the grid element, there is some degree of
anisotropy observed for the Random 3 and Random 4 cores due to the shape of the grid
cells used in the simulation (rectangular shape). Therefore it results in some spatial
correlation of properties in the flow direction. However, the effect of anisotropy caused
by the shape of grids would not affect our results significantly; hence “Random”
distribution cores are used in this work to refer these quasi-random distribution models.
The standard deviation of permeability ranges from 0 to 2.65. Although the statistical
mean permeability is chosen to be 430 md for most of the cases, the absolute
permeability or the so-called effective permeability for different heterogeneous
63
representations are different. Effective permeability kmean is calculated based on the
single phase Darcy’s law: injecting solely brine into a brine-saturated core at a given flow
rate and calculating pressure drop across the core after steady-state. The effective
permeability is around 430 md for the cores with a small degree of heterogeneity. The
degree of heterogeneity is represented by the parameter σlnk/ln(kmean). A relatively
homogeneous medium has a small σlnk/ln(kmean) and its permeability variation approaches
to zero, while a highly heterogeneous porous medium can have large values of
σlnk/ln(kmean). In this study, σlnk/ln(kmean) ranges from about 0 to 0.5.
Table 5. 1 Summary of Simulation Cases with Different Degrees of Heterogeneity.
Permeability
Realizations
Porosity
φ
Effective
Permeability
kmean (md)
Standard
Deviation
σlnk
Heterogeneous
Factor
σlnk/ln(kmean)
Homogeneous
0.202 430 0 0
Random 1
0.202 432 0.0236 0.0039
Random 2
0.202 430 0.2540 0.0419
Berea
(Kozeny-
Carman)
various 430 0.2757 0.0455
Berea
(High
Contrast) various 318 0.9602 0.1680
Random 3
0.202 366 1.3679 0.2343
Random 4
0.202 254 2.6542 0.4747
64
5.2 Simulation Studies
Similar to the approach used to study homogeneous cores, we investigate the significance
of flow rate, interfacial tension, and core permeability on the CO2/brine flow systems for
moderately heterogeneous cores, or specifically, the High Contrast and Random 3
models. Flow rate is changed by several orders, from 0.005 ml/min up to 24 ml/min;
interfacial tension and core permeability are again varied by two orders of magnitudes to
study a range of values. Sensitivity studies for a wide range of permeability values are
provided later in this section. Different values of CO2 fractional flows are also studied.
5.2.1 Heterogeneity Effects
For the base case of the homogeneous study, we injected a fractional flow of 95% CO2
over a wide range of flow rates at reservoir conditions. To study the effects of
heterogeneity on brine displacement efficiency, we keep the conditions the same as for
the homogeneous study. Figure 5. 3 shows the core average CO2 saturations at 95%
fractional flow for the two cores over a wide range of flow rates, hence capillary
numbers. Although the two types of heterogeneity exhibit different behaviors, it is clear
that a higher degree of heterogeneity (higher σlnk/lnkmean) results in a lower average
saturation in the core. It also increases the flowrate dependency of brine displacement
efficiency, specifically the transition from the viscous dominated regime to gravity and
the capillary controlled regimes occurs at lower capillary numbers.
Note that the saturations for the Kozeny-Carman model are lower than for the
Random 2 model even though they have similar heterogeneity factor (σlnk/lnkmean~0.05).
This is mainly due to the presence of a capillary barrier close to the outlet end of core that
prevents CO2 from entering a portion of the core (not visible in Table 5.1 but shown later
in Figure 6.2).
65
Figure 5. 3: Average CO2 saturation as a function of alternative capillary number Ncv.
The interfacial tensions σ for all cases is 22.47 mN/m and the effective permeability
varies from 254 to 570 md.
5.2.2 Interfacial Tension and Permeability Effects
A series of sensitivity studies on interfacial tension and core permeability for the
heterogeneous cores are performed. Table 5.2 shows the summary of sensitivity studies
and their corresponding Bond numbers for the High Contrast model and Random 3 model
respectively. Based on capillary pressure function (Eq. 3.5), the effect of interfacial
tension changes from 3σ to σ/3 is similar to the effect of permeability changes from 0.1k
to 10k. Simulations have confirmed this conclusion and therefore only results for a wide
range of permeability values are shown in Figure 5. 4. However, the results of brine
displacement efficiency with three different interfacial tension values in terms of
alternative capillary number Ncv will be discussed later in Chapter 6.
66
Table 5. 2: Summary of Sensitivity Studies for High Contrast and Random 3 Models.
High Contrast Models σ, mN/m k, md pc*, Pa NB
Base Case (σ, k) 22.47 318 3489 0.0523
Sensitivity 1 (3σ, k) 67.41 318 10466 0.0174
Sensitivity 2 (σ/3, k) 7.49 318 1163 0.1568
Sensitivity 3 (σ, 0.1k) 22.47 31.8 11032 0.0165
Sensitivity 4 (σ, 10k) 22.47 3180 1103 0.1654
Random 3 Models σ, mN/m k, md pc*, Pa NB
Base Case (σ, k) 22.47 366 3252 0.056
Sensitivity 1 (σ/3, k) 7.49 366 1084 0.168
Sensitivity 2 (σ, 0.1k) 22.47 36.6 10283 0.018
Sensitivity 3 (σ, 10k) 22.47 3680 1028 0.177
The core-averaged CO2 saturations as a function of traditional capillary number Ca,
alternative capillary number Ncv, and gravity number Ngv using the definitions shown in
Eqs. 4.1-4.3 are plotted in Figure 5. 4. The left hand side provides the results for the
High Contrast model while the right hand side shows the results of the Random 3 model.
For ease of comparison, sensitivity studies on permeability for the homogeneous cores
are also illustrated in the same graph. Solid lines represent the homogeneous saturations
while dashed lines represent the heterogeneous ones.
As discussed extensively in Chapter 4, plotting the average saturations for
homogeneous cores in terms of gravity number can distinguish three flow regimes
clearly. For the heterogeneous cores, the efficiency of brine displacement also falls into
three separate regimes: a viscous-dominated regime where the saturation is independent
or nearly independent of Ca, Ngv and Ncv; a viscous-capillary transition regime where the
average saturation is strongly dependent on the dimensionless numbers; and a capillary-
dominated regime characterized by low saturations with a small dependence on the
dimensionless variables. However, unlike the case for homogeneous cores where the
gravity number can be used to normalize the results, for heterogeneous permeability
distributions, sensitivity studies are normalized better in terms of the transverse capillary
number Ncv (Eq. 4.2), shown in Figure 5. 4. This suggests that Ncv is a better
dimensionless number to distinguish the transitions when considering capillary
heterogeneity. The ratio of capillary to viscous forces is a critical measure for assessing
the influence of capillary trapping in heterogeneous systems (Rapoport 1955; Ringrose et
67
al 1996)
Results show the transitions from the viscous- to viscous-capillary transition regimes
and from the transition to the capillary-dominated regimes for both types of
heterogeneous cores occur earlier for higher degrees of heterogeneity. Capillary
heterogeneity not only reduces the average saturation in the viscous-dominated regime
but also increases the flowrate dependency, which implies that higher flow rates are
required to reach the viscous-dominated regime.
Figure 5. 4: Average CO2 saturation as a function of traditional capillary number Ca,
alternative capillary number Ncv, and alternative gravity number Ngv for two
heterogeneous cores: High Contrast model (left) and Random 3 model (right). The
interfacial tension σ is kept as a constant, 22.47 mN/m.
68
Figure 5.5 illustrates the pressure gradients for the three forces as a function of capillary
numbers for the High Contrast and the Random 3 models, respectively. These
demonstrate that the viscous force is always greater than the buoyancy force within the
flowrate range we are interested in. Viscous and capillary forces are much more relevant
than the gravity force once considering capillary heterogeneity. Unlike for the
homogeneous cores, where the first transition can be defined by the point at which
gravity and viscous forces are equal, for heterogeneous cores, viscous forces are always
greater than the gravity force. However, the second transition occurs when the viscous
pressure drop and the capillary pressure drop in the flow direction have the similar
magnitude, similar to the homogeneous results shown in Figure 4. 1.
Figure 5. 5: Pressure gradients as a function of alternative capillary number Ncv for three
physical forces.
5.3 Theoretical Analysis of Multiphase
Displacement Efficiency in Heterogeneous Cores
The second part of this chapter presents a 2D analytical solution incorporating the
influence of heterogeneity. The analysis is an extension of the methods developed in
Chapter 4 for homogeneous cores. A general form of the solution is first derived and the
constraints from 3D high-resolution simulations are applied to obtain the final semi-
analytical solution. The concept of derivation is explained generally here. The detailed
derivation is provided in Appendix B.
69
5.3.1 2D General Solution
To predict the simulation results observed in the previous section, we consider steady-
state, two-dimensional flow (x-z direction) of two immiscible phases. The properties of
this 2D porous medium are heterogeneous and each grid cell has isotropic properties.
Using standard terminology and notation, the mass conservation equation of gas phase is
described by the following:
g g gt c c
w
S Mkλ Mkλu p p1 1- + Δρg- =0
t φ kλ x 1+M φ z z 1+Mx
(5.3)
The approach used to develop the analytical solution is similar to the derivation in
Chapter 4, except now the permeability and hence the capillary pressure values vary
spatially:
c w w
φp S , x, z σcosθ J S
k x, z (5.4)
For the analysis in this chapter, we consider only permeability heterogeneity (and
consequent capillary heterogeneity) while porosity, interfacial tension, and contact angle
remain constant. However, the same approach could also be applied to these different
types of heterogeneities. To non-dimensionlize the equation, we define xD=x/L, zD=z/H,
and tD=tut/φL as well as introducing a dimensionless variable τ(xD, zD) to represent
heterogeneity:
D D
D D
mean
k x ,zτ x ,z
k (5.5)
Based on the definition of Eq. 5.4, we can obtain pc= [σcosθ(φ/kmean)0.5
]J/τ. In addition,
for simplification, we introduce variables defined by Chang and Yortsos (1992) as H=-
fwkrgJ’ and G=fwkrgJ. Substituting all the variables into Eq. 5.3, the dimensionless mass
conservation equation at steady-state becomes:
2
2cv cv
CO cv B rg w cv2 2
D D D D D D D Dl l
N Nτ τ SG SGf G +N N k f τ +G = Hτ +N Hτ
x x z z x x z zR R
(5.6)
70
where the capillary and gravity numbers have the same definitions in Chapter 4 except
kmean is the permeability:
22
*
mean c mean
cv gv2
co tco t
k Lp Δρgk LN = and N =
Hμ uH μ u (5.6b)
fCO2=1/(1+M) and fw=M/(1+M) are fractional flows of CO2 and water respectively.
Important variables are the alternative capillary number or the so-called inverse
macroscopic capillary number Ncv (Eq. 5.6b), gravity number Ngv (Eq. 5.6b), aspect ratio
Rl =L/H, and the heterogeneity function τ(xD, zD). Incorporating capillary heterogeneity
into the mass conservation equations results in a high degree of complexity. The right
hand side of Eq. 5.6 is known as the capillary dispersion term. Using a similar strategy to
that used for development of the semi-analytical solution for homogeneous cores and
assuming all the variables are continuous functions (for saturation this will only be true if
all grid cells have a zero capillary entry pressure); we can obtain the general form of two-
dimensional time-independent CO2 saturation SG for the heterogeneous cores in terms of
Bond number NB, capillary number Ncv, aspect ratio Rl, and the heterogeneity term τ (Eq.
5.7a). The unknown coefficients C1Hete
, C2Hete
, C3Hete
and C4 not only depend on location
but also on the heterogeneity of the rock (Eq. 5.7b).
2l
DcvB D B D
R1- axτ N-τN bz -τN bzHete Hete Hete
1 2 3 4SG= C e +C e +C e +C (5.7a)
D D D Dz x x z-ε lnτ z -ε lnτ x -ε lnτ x -ε lnτ zHete Hete Hete
1 D D 1 2 D D 2 3 D D 3C x ,z ,τ =C e e , C x ,z ,τ =C e , and C x ,z ,τ =C e (5.7b)
The definitions of two variables, a and b, in the exponent terms are the same as defined in
the homogeneous analysis (Eq. A-19). C1, C2, C3 and C4 are functions of coordinates xD
and zD. (lnτ)x and (lnτ)z are the heterogeneity gradients in the flow direction x and
vertical direction z, respectively. The variable ε is defined as Jb-1 where J is the
Leverett-J function. A detailed derivation of Eq. 5.7 is provided in Appendix B.1.
71
5.3.2 2D General Solution for Heterogeneous Rocks Using
Simulation Constraints
Since it is difficult to integrate Eq. 5.7 to calculate the average core saturation, we use the
same strategy in the Chapter 4 to assume that the saturation at a particular point (x0,D,
z0,D) is representative of the core average saturation2COS :
2CO 0,D 0,DS SG x ,z (5.8)
Now C1Hete
, C2Hete
, C3Hete
, C4 and τ become some single valued functions of x0,D and z0,D.
Similarly, to eliminate some unknown coefficients, we apply several observations from
the simulation results to the general solution. First, in the viscous-dominated regime
(Ncv<Ncv,c1), the average saturation is independent of capillary and Bond numbers:
2 2CO CO
cv B
S S 0 and 0
N N
(5.9)
Second, in the capillary-dominated regime (Ncv>Ncv,c2), the average saturation is
independent of the Bond number:
2CO
B
S0
N
(5.10)
Applying simulation constraints to the general equation and using the homogeneous
solution derived in the previous chapter as a model, we can eliminate several unknown
coefficients. A detailed derivation is provided in Appendix B.2.
As mentioned before, the parameter τ (Eq. 5.5) is related to the heterogeneity factor,
σlnk/ln(kmean). The mathematical representation of the parameter τ in terms of the
heterogeneity factor is provided in Eq. 5.11a. See Appendix B.3 for a detailed
description of the derivation.
1 0,D lnk lnk2 0,D 2 0,D
2 0,D mean mean
β x σ σ1+ β z ωβ z
β z ln k ln k
0,D 0,D 0,D o,D meanτ x ,z k x ,z k τ=e =e
(5.11a)
β1 and β2 are proportionality factors between σlnk/ln(kmean) and the spatial derivatives of
the permeability field, (lnτ)x and (lnτ)z, shown in Eq. 5.11b. ω is a fitting parameter
dependent on β1 and β2 as well as x0,D and z0,D (Eq. 5.11c).
72
lnk lnk
2 1z x
D mean D mean
σ σln τ ln τlnτ β and lnτ β
z ln k x ln k
(5.11b)
1 0,D
2 0,D
β xω 1+
β z (5.11c)
More investigation and discussion about τ and ω will be provided in the next section.
Applying all the simulation constraints to the general solution (Eq. 5.7), we can predict
the average CO2 saturation for heterogeneous cores in terms of several dimensionless
numbers:
2 Hete2 l cv,c1
Hete 1cv,c2 cv
2
d R N- dN NHete HeteBL
CO 1 BLHete
BL
SS C e - e +S
S
(5.12a)
lnk
2 0,D0,D meanz
σ-εβ z
-ε lnτ z ln kHete
BL 1 BL 1S h τ S where h τ e =e (5.12b)
2
3/2CO lHete Hete l
cv,c1 cv,c2 HeteHeteB B BLrg BL
f R R1 1N = and N = α
τN τN Sk S
(5.12c)
l lHete 11 11
1 1 2ε
B 22l l
R RC d1C = , d = and d =
N dτ R R (5.12d)
The average saturation of the heterogeneous core is controlled by the Bond number NB,
capillary number Ncv, the Buckley-Leverett solutions SBL and SBLHete
, and the aspect ratio
Rl once the two critical capillary numbers (Ncv,c1Hete
and Ncv,c2Hete
) are calculated (Eq.
5.12a). Moreover, the heterogeneous solution naturally reduces to the homogeneous
solution when τ=1.
Eq. 5.12b shows that the average CO2 saturation of the heterogeneous core in the
viscous-dominated regime is SBLHete
, which is a modified Buckley-Leverett solution that
takes into account heterogeneity. The capillary heterogeneity has a significant effect on
the average saturation of a BL displacement. The effect is important even for large flow
rates within the viscous-dominated regime if the degree of heterogeneity is large (Figure
5. 3). This is because the capillary heterogeneity terms GNcv(τ/xD)/Rl2 and
GNcv(τ/zD) in the mass conservation equation (Eq. 5.6) are not negligible even for
small values of Ncv once we have large variations of heterogeneity (large τ/xD and
τ/zD). h1 is a heterogeneous factor, dependent on the heterogeneity gradient in the
73
vertical direction (lnτ)z as well as the parameter ε, shown in Eq. 5.12b. The variable ε is
considered as a fitting parameter to match the simulation results. Since h1<1, the average
saturation for the heterogeneous cores is always lower than the Buckley-Leverett
solution, which is reasonable and consistent with the simulation results (Figure 5. 3).
The definitions of two critical capillary numbers, Ncv,c1Hete
and Ncv,c2Hete
, are shown in
Eq. 5.12c. They can be related to the critical gravity number for the homogeneous
studies based on the Eq. 4.4, Ncv=Ngv/NB. Eq. 5.13 relates the two critical numbers
between the homogeneous cores and the heterogeneous cores for easier comparison.
Homo Hete
gv,c1 BLHete Homo Hete
cv,c1 cv,c1 BL
B
N S1 1N = = N S
τ N τ (5.13a)
Homo HeteHetegv,c2 BLgv,c2Hete Homo Hete
cv,c2 cv,c2 BL
B B
N SN 1N = = = N S
N τN τ (5.13b)
For the homogeneous porous medium, τ=1, hence lnτ =0, we can obtain SBLHete
=SBL.
Therefore Ncv,c1/2Hete
=Ncv,c1/2Homo
. Note that the smaller critical numbers are obtained
with a larger degree of heterogeneity (large τ).
Coefficients C1Hete
, d1 and d2 are shown in Eq. 5.12d where d1 and d2 are the same
functions used in the homogeneous solution. C1Hete
can be related to the C1Homo
as
follows:
Hete Homo
1 1ε
1C = C
τ (5.14)
C11, α, d11, and d22 are constant parameters, which have already been determined
previously (Eq. 4.27) by curve matching the semi-analytical solution with the simulation
results for homogeneous cores. The values are provided in Table 5.3.
In summary, the average saturation for the heterogeneous cores can be predicted once
we can obtain the correct values of τ and ε. These parameters will be determined by
matching the semi-analytical solution with the numerical results.
74
Table 5. 3: Summary of constant coefficients in Eq. 5.12.
d11 d22 C11 α
2.0894 5.57607 56.78 12.79737
5.3.3 Approximate Semi-Analytical Solution
From the simulation results shown in Figure 5. 3, we can obtain the modified Buckley-
Leverett solutions SBLHete
from cores with different degrees of heterogeneity. Figure 5.6
plots SBLHete
as a function of normalized standard deviation factor σlnk/ln(kmean) for
random permeability cores (LHS) and all the permeability cores listed in Table 5.1
(RHS).
Figure 5. 6: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of different
degrees of heterogeneous cores.
A near-perfect linear correlation between SBLHete
and σlnk/ln(kmean) for cores with
randomly distributed heterogeneity is observed in Figure 5.6:
Hete lnk lnk
BL BL
mean mean
σ σS =0.324-0.1788 1-0.55185 S
ln k ln k
(5.15)
For cores with structured heterogeneities a reasonably good correlation is also observed.
Based on the linear relation shown earlier, the heterogeneous function h1 can be
approximated as follows:
75
lnk2 0,D
mean
σHete -εβ zln kBL lnk
1 2 0,D
BL mean
S σ=h =e 1-εβ z
S ln k (5.16)
Once we obtain the value of h1 based on Eq. 5.16, we can evaluate the heterogeneous
factor τ in terms of this known parameter:
lnk
2 0,D1
mean
σωω β z 1-hln k ετ=e =e
(5.17)
In conclusion, the only fitting parameters introduced in Eq. 5.12 for the heterogeneous
cores now become ω and ε while the other parameters have been already introduced in
the homogeneous studies. By adjusting these two coefficients, the semi-analytical
predictions agree with the simulation results quite well. Figure 5.7 shows good
agreement between the simulated results and the theoretical predictions for the average
CO2 saturations of base cases for the High Contrast model (LHS) and the Random3
model (RHS), respectively.
Figure 5. 7: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for High Contrast model at a wide
range of permeability and interfacial tension.
However, it is difficult to find a general functional form for ω and ε since they depend on
several factors. For example, they depend on the spatial gradients in heterogeneity for
both the x and z directions. For the random fields, the proportional factor β1 is equal to
β2. We can expect to have different values of ω for spatially correlated permeability
distributions. In addition, Figure 5.8 plots the dependence of ε and ω in terms of
potential factors such as the modified Buckley-Leverett saturation (SBLHete
), and the Bond
76
number (NB) based on the sensitivity studies of CO2 fractional flow, permeability and
interfacial tension. It shows that different heterogeneous cores have different
dependency. Further investigation is needed in the future to generalize solutions by
exploring the sensitivity to these parameters.
Figure 5. 8: The dependence of ε and ω in terms of the modified Buckley-Leverett
saturation (SBLHete
) and the Bond number (NB) for High Contrast and Random 3 models.
77
5.4 Verification of the Analytical Model
The validation of analytical model for heterogeneous cores based on Eq. 5.12 is presented
in this section. As mentioned earlier, it is difficult to obtain the correct functional form of
ε and ω; hence these two parameters are treated as fitting parameters. These two values
are different for different sensitivity cases and are provided in Appendix D with other
relevant dimensionless numbers.
5.4.1 Different Interfacial Tension and Permeability
Figure 5.9a shows the sensitivity studies of High Contrast model for permeability (31.8
md and 3180 md) and interfacial tension (7.49 mN/m and 67.41 mN/m) respectively.
Those figures compare the simulation results and the predicted values based on Eq. 5.12.
As shown, we can replicate the simulation results in general quite well, especially for the
high permeability and low interfacial tension cores. When the capillary force is relatively
small, the semi-analytical solution matches best in the transition from viscous- to
viscous/capillary-dominated regimes. Although a slight mismatch to the simulation
results occurs in the cases of High Contrast model when capillary force is strong (0.1k
and 3σ), the overall predictions are captured the transitions between regimes. The
prediction is even better for the random distribution core (Random 3) shown in Figure
5.9b for both permeability values (36.6 md and 3660 md).
The differences at the higher capillary numbers of High Contrast model could be due
to the lack of information related to the correlated permeability distribution observed in
the cores. Further studies to include the correlation length in the semi-analytical solution
could generalize these results to a wider range of heterogeneity distributions.
78
(a) Porosity-based permeability: High Contrast model
(b) Random permeability distribution: Random 3
Figure 5. 9: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for (a) High Contrast model and (b)
Random permeability model in a wide range of permeability and interfacial tension.
5.4.2 Different Fractional Flows of CO2 (HC Model)
Figure 5.10 shows the average CO2 saturations of the High Contrast model as a function
of capillary numbers for four different CO2 fractional flows, 0.95, 0.79, 0.51 and 0.34.
The heterogeneous factor τ stays constant for four cases since it is the same core.
Although the same input relative permeability curves are used, different fractional flow
of CO2 results in different Buckley-Leverett solution SBLHete
and hence different
krg(SBLHete
) values. Again, the semi-analytical model predicts average saturations very
well.
79
Figure 5. 10: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for High Contrast model at different
fractional flows of CO2.
5.4.3 Different Core Dimensions
One sensitivity study on the aspect ratio Rl is performed to test the semi-analytical model.
The solution works well for the limited cases studied (Figure 5. 11). Since the input
relative permeability curves and the fractional flow of CO2 (fCO2=0. 95) for this case are
the same as for the base case of the High Contrast model, we can expect the modified
Buckley-Leverett solution SBLHete
to remain the same.
80
Figure 5. 11: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of different
aspect ratio.
5.4.4 Different Heterogeneity
The final cases to test the semi-analytical solution use different degrees of heterogeneity,
listed in Table 5. 1, with 0.95 CO2 fractional flow. All of the cases have the same
simulation parameters, namely, relative permeability curves, capillary pressure curves,
same reservoir conditions, and same grid size, etc. The only thing changed is the
permeability distribution of porous medium which is characterized by the heterogeneity
factor, σlnk/ln(kmean). The other important input parameters such as those in Table 4. 3 are
the same.
For these cases, different heterogeneity results in different values of τ. Therefore the
modified Buckley-Leverett solution SBLHete
and krg(SBLHete
), the slope coefficient C1Hete
as
well as the two critical numbers Ncv,c1Hete
and Ncv,c2Hete
are expected to be different for
each case. Figure 5.12 and Figure 5.13 shows the average CO2 saturations as a function
of capillary numbers for six models. We use the semi-analytical solution for the
homogeneous cores (Eq. 4.26) to predict the small degree of heterogeneity cases
(σlnk/lnkmean < 0.0455) shown in Figure 5.12 while the heterogeneous semi-analytical
solution is used to predict the large degree of heterogeneity cases (σlnk/lnkmean > 0.0455)
shown in Figure 5.13. The values of ω and ε for each case are listed in Table D.3
(Appendix D).
Figures 5.12 and 5.13 show that the 2D semi-analytical model predicts the average
81
saturation for different heterogeneous cores very well. Specifically, the average
saturation in the viscous dominated region is accurately predicted and the transitions
between different flow regimes occur at the correct capillary number. The results show
that even for different types of permeability distributions, the semi-analytical solutions
still match the simulation results well.
Figure 5. 12: Comparison of average CO2 saturation as a function of capillary number Ncv
between homogeneous analytical predictions and simulation results for the sensitivity
cases of small degrees of heterogeneous cores.
82
Figure 5. 13: Comparison of average CO2 saturation as a function of capillary number Ncv
between theoretical values and simulation results for the sensitivity cases of different
degrees of heterogeneous cores.
5.5 Procedures for Using the Analytical Solutions
In this section, instructions for how to use the semi-analytical solutions for predicting
average saturations and flow regimes during homogeneous and heterogeneous core floods
are provided.
The following input parameters are required:
Porosity
Permeability
Capillary pressure curve (Eq. 3.5)
Relative permeability function (Eq. 3.6)
If your core is relatively homogeneous (σlnk/lnkmean < 0.05), the average saturation can be
predicted quite well based on the homogeneous analytical solution:
83
First determine SBL based on fractional flow curves
Calculate dimensionless numbers (Rl, Ngv, NB), two critical numbers, and
coefficients based on
2
2
CO l
gv,c1
r,CO
f RN
k (4.7)
3/2
l
gv,c2
BL
RN =α where α=12.79737
S (4.10)
l l11 11
1 1 2
B 22l l
R RC dC = , d = and d =
N dR R (4.29)
Obtain the average saturation for the homogeneous cores based on Eq. 4.26:
2l gv,c1B
2 1gv,c2 gv
2
R NN-d -d
N N
CO 1 BLS =C e -1 e +S
(4.26)
To predict the average saturation for moderately to highly heterogeneous cores:
First determine the heterogeneous factor σlnk/ln(kmean)
Calculate SBLHete
using Eq. 5.15
Hete lnk lnk
BL BL
mean mean
σ σS =0.324-0.1788 1-0.55185 S
ln k ln k
(5.15)
Once we obtain SBLHete
, we can evaluate
Hete
BL
1
BL
Sh
S (5.16)
Calculate τ based on Eq. 5.17a. For Bond numbers in the range of 0.017 to 0.17,
the average values of parameter and based on all the sensitivity studies in this
work can be estimated as 3.3 and 89, respectively. The percentage errors of
range from 2.6%-42% while the percentage errors of range from 21%-74%.
Further investigation of and on relevant dimensionless numbers such as
SBLHete
and NB are required to approximately predict the values of and .
84
lnk
2 0,D1
mean
σωω β z 1-hln k ετ=e =e
(5.17a)
Calculate dimensionless numbers (Rl, Ngv, NB) and two critical numbers based on
Eq. 5.12
2
3/2CO lHete Hete l
cv,c1 cv,c2 HeteHeteB B BLrg BL
f R R1 1N = and N = α
τN τN Sk S
(5.12c)
l lHete 11 11
1 1 2ε
B 22l l
R RC d1C = , d = and d =
N dτ R R (5.12d)
Finally obtain the average saturation for the heterogeneous cores based on Eq.
5.12a:
2 Hete2 l cv,c1
Hete 1cv,c2 cv
2
d R N- dN NHete HeteBL
CO 1 BLHete
BL
SS C e - e +S
S
(5.12a)
In summary, we provide an analytical solution to predict the brine displacement
efficiency for both homogeneous and heterogeneous cores based on the calculated
dimensionless group. For relatively homogeneous core (σlnk/lnkmean < 0.05), the average
saturation can be predicted quite well based on the homogeneous analytical solution. For
moderately to highly heterogeneous cores (σlnk/lnkmean > 0.05), the theoretical solution is
provided and can be estimated qualitatively.
85
Chapter 6
CO2 and Brine Relative Permeability in
Heterogeneous Rocks
Motivated by the multiphase flow literature regarding CO2/brine core flood experiments,
the issues described in Section 2.4 are addressed in this chapter. This chapter also
demonstrates one of the practical applications of the numerical studies described in
Chapter 3 and the analytical solutions provided in Chapters 4 and 5. The objective in this
chapter is to investigate systematically what parameters have a significant impact on
reliable drainage relative permeability measurements on rock cores. In particular, 3D
high resolution core-scale simulations are conducted to study the independent as well as
the combined effect of flow rate, capillary pressure, gravity, and rock heterogeneity, thus
allowing identification of the operational regimes under which reliable measurements of
relative permeability can be obtained using steady-state horizontal core flood
experiments. The concept of first critical capillary number, Ncv,c1, which defines the
viscous-dominated regimes introduced in Chapters 4 and 5 will be used later in this
chapter for the discussion.
In this so-called viscous-dominated regime where the average saturation of the core is
independent of capillary or gravity number, heterogeneity results in spatially varying and
lower average CO2 saturation (Figure 5. 3) as compared to that expected for a uniform
core. Consequently, the effective relative permeability for the whole core is different
than the intrinsic relative permeability of each individual voxel in the core. Saturations in
this “viscous dominated regime” vary spatially in response to the establishment of
gravity-capillary equilibrium in the core.
Therefore, for the practical interest, the concept of effective relative permeability for
86
the heterogeneous rocks is used in this chapter, and the viscous-dominated regime is
referred to the “capillary equilibrium viscous-dominated regime” or the “quasi viscous
dominated-regime” in this work.
The overall methodology for this study is illustrated in Figure 6. 1. First, we define a
set of properties for the core, including the spatial distribution of permeability (k),
porosity (φ), capillary pressure (Pc) and relative permeability curves (kr). These are then
used as input for simulations using TOUGH2 (Pruess et al., 1999) that mimic the core-
flooding procedures used for making steady-state relative permeability measurements.
Outputs from the simulation include the pressure drop across the core in both fluid
phases, as well as, the CO2 saturation distribution. These outputs are used as synthetic
“data sets” for calculating the relative permeability of the core. The influence of
flowrate, rock heterogeneity, core length, gravity, interfacial tension on the accuracy of
the calculated relative permeability curves is systematically studied by varying these
parameters over a wide range of values. Based on the comparison between the input and
calculated relative permeability curves we draw conclusions about the important sources
of error for these calculations as well as the conditions over which accurate
measurements can be obtained.
Figure 6. 1 Overview of scientific approach.
6.1 Simulation Outputs
To determine the effect of sub-core scale heterogeneity on CO2/brine multiphase flow,
three different degrees of heterogeneities including homogeneous, low contrast (Kozeny-
Carman model) and High Contrast models are simulated. In addition, to study the flow
rate effect, the injection rate is varied from 0.1 to 10 ml/min.
The core flooding experiment conducted by Perrin and Benson (2010) is simulated
for making steady-state relative permeability measurements. For the heterogeneous
Define simulation
input parameters: φ,
k, Pc, kr
Generate steady-state
synthetic data sets
(SCO2, ΔPCO2)
Calculate relative
permeability
(Steady-State Method)
87
cores, the three-dimensional porosity map (Perrin and Benson 2010) shown in Figure 6. 2
is used to generate the corresponding permeability map with the porosity-permeability
relationships as shown in Table 6.1. The Kozeny-Carman (KC) equation generates the
low contrast permeability map while the exponential function of porosity-permeability
relation generates a higher degree of heterogeneity, called High Contrast model. The
equations for the models are shown in Table 6.1 as well as their normalized standard
deviations in σlnk/ln(kmean). The permeability of each grid element is assumed to be
isotropic. These two heterogeneous models are compared to a homogeneous one to study
the effect of heterogeneity on the multiphase flow system.
Figure 6. 2 3D porosity distribution of Berea sandstone.
Table 6. 1 Synthetic input parameters for every grid in the simulations for three different
models.
σlnk/lnk Porosity Permeability (md)
Capillary
Pressure
(Pa)
Input
Relative
Permeability
Homogeneous
Model 0
Φi = Φmean
ki =kmean
Measured
Pc Curve Eq. 3.7
Kozeny-
Carman
Model
0.05
Φi
ki Φi3/(1-Φi)
2
Pc,i
i ik
(Eq. 3.6)
Eq. 3.7
High Contrast
Model 0.17
Φi
ki exp(Φi4)
Pc,i
i ik
(Eq. 3.6)
Eq. 3.7
For a given flow rate, simulations of co-injection of CO2 and brine are run until the
pressure drop and core-averaged saturation stabilize. All of the simulations have been
88
confirmed to run long enough (more than 10 pore volumes injected) to reach steady-state.
Important output parameters include grid-cell CO2 saturations, CO2 pressures and
capillary pressures. Here we also evaluate slice averaged quantities along the length of
the core such as the slice-average CO2 saturation (SCO2), slice-average pressure in the CO2
phase (PCO2), and slice-average capillary pressure (Pc). Figure 6. 3 shows a typical
simulation result, including the CO2 saturation distribution, pressure drop across the core,
and the core-averaged CO2 saturation.
Figure 6. 3 CO2 saturation distribution at steady-state for 95% fractional flow of CO2 at a
total injection flow rate 1.2 ml/min.
The pressure drops across the core are defined as the difference between the average inlet
and the outlet slice values:
2 2 2CO CO ,inlet CO ,outletΔP =P -P (6.1a)
w w,inlet w,outletΔP =P -P (6.1b)
Since Pc= PCO2–Pw, the water pressure drop can be rewritten in terms of the two output
parameters ΔPCO2 and ΔPc:
2 2 2w CO ,inlet CO ,outlet c,inlet c,outlet CO cΔP = P -P - P -P =ΔP -ΔP (6.2)
The pressure drops in each phase are used to calculate the corresponding relative
permeability values based on the simplified Darcy’s equation, shown in Eq. 6.3. As
shown in Eq. 6.2, when Pc is the same in the first and last slice of the core, the pressure
gradient drop across the core is the same in both phases.
89
6.2 Calculation of Relative Permeability
For horizontal, 1D, immiscible, two-phase flow in homogeneous and isotropic porous
media at core scale, Darcy’s law neglecting the gravity effect takes the form:
2 2
2
2
r,CO COr,w w
w CO
w CO
kk ΔPkk ΔPq = A , q = A
μ L μ L (6.3)
where qw and qCO2 are the volumetric flow rates of brine and CO2 respectively, μw and
μCO2 are the brine and CO2 viscosities, A is the core cross section, and L is the length of
the core. In this chapter, Eq. 6.3 is used to calculate the relative permeability as a
function of the average saturation in the core since the core average saturation and the
pressure drop across the core are known. Eq. 6.3 is valid once the saturation and the
pressure gradients along the core are constants. We can expect Eq. 6.3 is no longer
correct once there is a saturation gradient and hence a capillary pressure gradient across
the core. Simulations are repeated at a number of fractional flows to construct the full
relative permeability curve.
6.2.1 Relative Permeability Calculated when ΔPw=ΔPCO2
It is often assumed that the pressure drop in both phases is equal (Geffen et al. 1951;
Avraam and Payatakes 1995), which requires that the capillary pressure is constant along
the length of the core. Moreover, it is assumed that the measured pressure drop
accurately reflects the pressure drop in at least one of the phases. In this case, the relative
permeability can be calculated from:
2 2 2
2
2
CO r,CO COr,w
w CO
w CO
ΔP kk ΔPkkq = A , q = A
μ L μ L (6.4)
In this section, the relative permeability values are calculated based on the assumption
that ΔPw=ΔPCO2 and ΔPCO2 is evaluated by Eq. 6.1a. In the subsequent discussion, for
simplicity, results are presented for a range of flow rates and for two different degrees of
heterogeneity (σlnk/lnkmean=0 and 0.17). The results are later generalized in terms of
90
capillary number (Ncv) in Section 4.1.
Homogeneous Cores
For homogeneous cores, the effect of flow rate on brine displacement efficiency has
already shown in Figure 4. 1. Figure 6. 4 illustrates CO2 saturation as a function of the
distance from the inlet at a 95% fractional flow of CO2 over a large range of flow rates
(0.1 ml/min-6 ml/min). The saturation is uniform across the core at high flow rates
where no saturation gradient exists and hence there is no capillary pressure gradient along
the core. Decreasing the flow rate below this regime leads to a saturation gradient in the
flow direction.
As mentioned in Section 3.3.1, the flow rate dependence shown in Figure 6. 4 is not
due to the traditional capillary end effect since the outlet boundary condition does not
force the two fluids to have the same pressure. The observed saturation gradients exist at
low flow rates because gravity and capillary pressure are included in the simulation.
Gravity causes some small amount of flow in the vertical direction and consequently the
saturation of CO2 is higher near the top of the core. This in turn creates higher-than-
average factional flow of CO2 near the top of the core as the fluid moves away from the
inlet boundary. The net effect is to cause a saturation gradient along the length of the
core.
Figure 6. 4 Flow rate effect on CO2 saturation along the homogeneous core at a 95%
fractional flow of CO2 with flow rates ranging from 0.1 ml/min to 6 ml/min.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SCO
2
x
6 ml/min
3.6 ml/min
1.2 ml/min
0.6 ml/min
0.3 ml/min
0.2 ml/min
0.1 ml/min
91
Table 6. 2 Summary of different flow rates for homogeneous cores with 95% fractional
flow of CO2. Capillary numbers are calculated based on Eq. 4.1 and Eq. 4.2,
respectively.
Flow Rate
ml/min Ca
Ncv
(Ncv,c1~887)
Saturation
Gradient
∆Sg/Sg,inlet
Regime
6 1.1E-7 35.8 0% Viscous-dominated
2.6 4.7E-8 82.5 0% Viscous-dominated
1.2 2.2E-8 178.8 0.5% Viscous-dominated
0.5 9.0E-9 429 1% at around transition
0.1 1.8E-9 2145 15% Gravity-dominated
To study the flow rate effect on relative permeability, five different injection rates are
picked to obtain the corresponding relative permeability curve. Table 6. 2 summarizes
these cases, for instance, the corresponding capillary numbers, the saturation gradients
across the core, and the flow regimes.
Figure 6. 5 compares the input relative permeability curve with the relative
permeability calculated assuming the same pressure drop for two fluids using Eq. 6.4. As
shown, the calculated relative permeability is identical to the input values when the flow
rates are close or in the viscous-dominated regime, which corresponds to the negligible
saturation gradients observed in Figure 6.4 (Table 6. 2). On the other hand, a roughly
15% saturation gradient along the flow direction results in a significant deviation of
wetting phase relative permeability (0.1 ml/min). Eq. 6.4 is no longer valid for the
wetting phase since pressure drops for the two fluids are different once saturation
gradients occur. Using the pressure gradient in the CO2 phase overestimates the pressure
drop in the water phase leading to underestimation of the water-phase relative
permeability.
92
Figure 6. 5 Relative permeability calculated by the same pressure drop (ΔPw=ΔPCO2) for
homogeneous core with 430 md permeability at different flow rates.
Heterogeneous Core (High Contrast Model)
The flow rate effect on brine displacement efficiency for the High Contrast model has
already been shown in Figure 5. 5. The High Contrast model is used to study the effects
of flow rate as well as sub-core scale heterogeneity on relative permeability curves since
its permeability distribution is generated based on the measured porosity values of the
real rock and it has a moderate degree of heterogeneity. Figure 6. 6 compares the average
CO2 saturation along the length of the core between the homogeneous and the
heterogeneous cores at the same flow rates. The general trends observed in the
homogeneous core can apply to the heterogeneous one. First, the slice-averaged
saturation is relatively uniform in the high flowrate regime (q>1.2 ml/min). The source
of saturation variation along the core is due to the core heterogeneity. Second, a large
saturation gradient across the core starts to occur once the flow rate is below the viscous-
dominated regime. Comparing the homogeneous and the heterogeneous cores, it is clear
that the core heterogeneity will enhance the flow rate dependency, decrease the average
saturation, and increase the saturation gradient.
As mentioned before, although the constant Buckley-Leverett saturation or the
intrinsic relative permeability of heterogeneous cores can be obtained once we have
reached an extreme high flow rate (>100 ml/min), it is unrealistic to use such high flow
0
0.1
0.2
0.3
0.4
0.5 0.6 0.7 0.8 0.9 1
Kr
Sw
Homogeneous Core, 430md, Same Pressure Drop
kr (6)
kr (2.6)
kr (1.2)
kr (0.5)
kr(0.1)
kr_input
93
rates in the core flood experiments. Therefore, the effective relative permeability of the
heterogeneous cores is used to compare with the homogeneous results.
Figure 6. 6 The effect of heterogeneity on CO2 saturation along the core at a fractional
flow of 95% over a wide range of flow rates.
Relative permeability curves calculated using the same pressure drop for both fluids are
illustrated in Figure 6. 7. Table 6. 3 summarizes the results of these calculations. Similar
to the homogeneous results, when saturation gradients are small, the relative permeability
is independent of the flowrate (<3%), which demonstrates that the relatively uniform
slice-averaged saturation results in the rate-independent relative permeability values
(Figure 6. 7). In addition, once large saturation gradients develop (6%, 15%, and 35%),
the wetting phase relative permeability is underestimated significantly. For the same
flow rate, the heterogeneous core results in larger saturation gradients compared to the
homogeneous core. In general, the rate-independent drainage relative permeability can
be obtained even with the highly heterogeneous core once the flow rate is high enough to
eliminate saturation gradients from one end of the core to the other. It is not required that
saturation gradients are eliminated in the middle of the core – as these results from the
heterogeneity of the rock.
On the other hand, even in the viscous-dominated regime, the CO2 relative
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
CO
2 S
atu
rati
on
X
Homo: 6
Homo: 1.2
Homo: 0.5
Homo: 0.3
Homo: 0.1
Hete: 6
Hete: 2.6
Hete: 1.2
Hete: 0.5
Hete: 0.3
Hete: 0.1
94
permeability is higher than the input value. This occurs because the effective relative
permeability for the non-wetting phase of the heterogeneous core is higher than for a
homogeneous core. This is a well known phenomenon as described by Corey and
Rathjens (1956) and Honarpour et al. (1994).
Figure 6. 7 Relative permeability calculated by the same pressure drop (ΔPw=ΔPCO2) for
heterogeneous core with 318 md permeability at different flow rates. Small picture
shows the same relative permeability curves at log scale.
Table 6. 3 Summary of different flow rates for heterogeneous cores (High Contrast
model: σlnk/lnkmean=0.168) with 95% fractional flow of CO2.
Flow Rate
ml/min Ca
Ncv
(Ncv,c1~80)
Saturation
Gradient
∆Sg/Sg,inlet
Regime
6 1.1E-7 30.7 ~0% Viscous-dominated
2.6 4.7E-8 70.9 ~3% at around transition
1.2 2.2E-8 154 ~6% Gravity-dominated
0.5 9.0E-9 369 ~15% Gravity-dominated
0.1 1.8E-9 1840 ~35% Gravity-dominated
6.2.2 Relative Permeability Calculated by True Pressure Drops
(ΔPw=ΔPCO2–ΔPc)
As discussed in Sect. 6.1, both the average CO2 pressure and the average capillary
pressure along the flow direction at steady-state are known as simulation outputs, hence
95
the pressure drops of both the CO2 and water phases along the flow direction can be
determined accurately based on Eq. 6.1a and Eq. 6.2. The relative permeabilities now
can be calculated based on the following equations:
2 2 2
2
2
CO c r,CO COr,w
w CO
w CO
ΔP ΔP kk ΔPkkq = A , q = A
μ L μ L
(6.5)
Figure 6. 8 shows the drainage relative permeability data calculated based on the true
pressure drops in each phase over the same range of flowrates as shown in Figure 6. 5
and Figure 6. 7. Once the true pressure drop of the wetting phase is known and used in
the calculation, the wetting phase relative permeability collapses very nicely even with
15% saturation gradient along the core (0.1 ml/min for the homogeneous core and 0.5
ml/min for the High Contrast model). Injecting lower flow rates can still yield accurate
results as long as the true pressure drops of water and CO2 are taken into account.
Figure 6. 8 Flow rate effect on relative permeability calculated by the true pressure drops
(ΔPw ΔPCO2) for homogeneous and heterogeneous core at various flow rates.
Note that the flow rate effect is larger in the heterogeneous core, and therefore including
the capillary pressure drop when calculating relative permeability to water is not as
effective as for the homogeneous core. This method only improves the calculated
relative permeability to water over a smaller range of flow rates. Second, the effective
relative permeability to non-wetting phase is higher than for input relative permeability
curve for each grid cell even flow rate is in the viscous-dominated regime.
As shown above, to obtain more accurate relative permeability at lower flow rates,
capillary pressure gradient (ΔPc) needs to be included in the calculation (Eq. 6.5). This
96
concept may apply to core flood experiments and give us a more reliable relative
permeability. However, capillary pressure gradients in general are not measured in the
experiment. It is possible to estimate capillary pressure gradients based on the average
saturation values at the inlet and outlet slices of the core. Once saturations at the ends of
the core are measured (e.g. using X-Ray CT scanning), the corresponding capillary
pressure values can be estimated from independently measured capillary pressure curves:
c,inlet c inlet c,outlet c outlet P =P S , P =P S (6.6)
Therefore
c c inlet c outletΔP =P S P S (6.7)
Comparing to the relative permeability calculated using the same pressure drop for both
fluids (Figure 6.7 or LHS of Figure 6.9), it is observed that including this corrected
capillary pressure drop in the pressure drop of water, the accuracy of relative
permeability to water for the heterogeneous core at 1.2, 0.5, and 0.1 ml/min flowrates can
be increased (Figure 6.9). The deviations for these lower flowrate cases are smaller if the
corrected capillary pressure drop is considered. Once the saturations at the ends of the
core are known, the application of this methodology can improve the accuracy of the
measured relative permeability curves even at lower flowrates and in the presence of
saturation gradients.
Figure 6. 9 (LHS) Relative permeability calculated using the same pressure drop for both
fluids (Figure 6.7); (RHS) relative permeability calculated by the corrected pressure
drops for High Contrast models at various flow rates.
0
0.1
0.2
0.3
0.4
0.5 0.6 0.7 0.8 0.9 1
Kr
Sw
High Contrast Model, 318md, Corrected Pressure Drop
kr (6)
kr (2.6)
kr (1.2)_Correction
kr (0.5)_Correction
kr (0.1)_Correction
kr_input
97
6.3 Sensitivity Studies for Different Core
Properties
In Chapters 4 and 5, we investigated the sensitivity of the average saturation of the core
over a wide range of sub-core scale heterogeneity, core dimensions, interfacial tension
and absolute permeability. In the rest of this chapter, we provide support for the
conclusion that the effective relative permeability of a core can be measured accurately in
the viscous-dominated regime or even in the gravity-dominated regime, if the correct
pressure gradients are used in each phase. The results will be combined with information
derived from the 2D semi-analytical solutions to establish a “rules of thumb” for making
accurate effective relative permeability measurements even in heterogeneous cores using
horizontal core floods.
6.3.1 Effects of Heterogeneity
The core-averaged saturations with different degrees of heterogeneity have already been
illustrated in Figure 5. 3 with capillary numbers Ncv ranging from 10 to 105. It can be
easily concluded from the figure that once the capillary number is small enough to be in
the viscous-dominated regime, the brine displacement efficiency is independent of flow
rate, and is expected to have a nearly uniform slice-averaged saturation profile even with
large capillary heterogeneity (Figure 5. 1 and Figure 5. 2). Based on the previous results,
it is reasonable to hypothesize that reliable relative permeability can be obtained as long
as the saturation is relatively uniform.
To validate this conclusion, we use the homogeneous core and the four heterogeneous
models to obtain the relative permeability curves calculated based on the true pressure
drops in each phase (Figure 6. 10). The injection flow rates required for the cores to
reach the viscous dominated regime depend on the degree and nature of heterogeneity.
For example, the higher flowrate or the smaller Ncv is required to reach viscous-
dominated regime for the higher degree of heterogeneity (larger σlnk/lnkmean). Table 6.4
lists all the relevant details of these five cases.
The effective relative permeability depends on the degree of heterogeneity. Figure 6.
98
10 illustrates that the effective relative permeability to water krw is not as sensitive as the
krg to the small scale heterogeneity, for example, the water relative permeability krw is
almost identical when the heterogeneity factor σlnk/lnkmean < 0.17. The effective relative
permeability to the gas phase calculated based on Darcy’s law is larger for more
heterogeneous cores while the effective water relative permeability is smaller for the
largest degree of heterogeneity (Random3).
Table 6. 4 Summary of different flow rates for different heterogeneous cores:
Homogeneous, Random 2, Kozeny-Carman, High Contrast and Random 3 models.
q,
ml/min σlnk/lnkmean Ca Ncv Regime
Homo 0.5 0 9.0E-9 429 Viscous-dominated
Random2 0.5 0.0419 9.0E-9 429 Viscous-dominated
KC 1.2 0.0455 2.2E-8 180 Viscous-dominated
HC 2.6 0.1680 4.7E-8 70 Viscous-dominated
Random3 6 0.2343 1.1E-7 33 Viscous-dominated
Figure 6. 10: Relative permeability calculated by the true pressure drops for five different
heterogeneous cores in the viscous-dominated regimes: homogeneous and the Random 2
cores (q=0.5 ml/min), Kozeny-Carman models (q=1.2 ml/min), High Contrast models
(q=2.6 ml/min), the Random 3 cores (q=6 ml/min) and the input relative permeability
curves.
0.001
0.01
0.1
1
0.5 0.6 0.7 0.8 0.9 1
Kr
Sw
True Pressure Drop
σ/lnk=0
σ/lnk=0.04
σ/lnk=0.05
σ/lnk=0.17
σ/lnk=0.23
kr_input
99
6.3.2 Effects of Core Length (15.24-45.72 cm)
In order to assess whether or not the flow rate dependency observed in the previous
results depends on the length of the core, the average CO2 saturations as a function of
capillary numbers for the three different core lengths are shown in the LHS of Figure 6.
11. The aspect ratios Rl are 3.14, 6.29, and 9.43 respectively. The RHS illustrates the
corresponding relative permeability with 0.1 ml/min injection flow rate. Table 6.5
summarizes the information of these three cases. Simulation results show that even with
up to 15% saturation gradient, we can still obtain the intrinsic relative permeability for
different lengths of homogeneous cores.
Figure 6. 11 (LHS) Brine displacement efficiencies for three different lengths of
homogeneous core with capillary number ranging from 10 to 107; (RHS) Relative
permeability calculated by the true pressure drops for homogeneous cores at 0.1 ml/min
flow rates.
Table 6. 5 Summary of different lengths of heterogeneous cores (High Contrast model:
σlnk/lnkmean=0.168).
Homo Flow Rate
ml/min Ca Ncv
Saturation
Gradient
∆Sg/Sg,inlet
Regime
L 0.1 1.8E-9 2145 15% Gravity-dominated
2L 0.1 1.8E-9 4290 10% Gravity-dominated
3L 0.1 1.8E-9 6435 7% Gravity-dominated
6.3.3 Effects of Absolute Permeability (31.8-3180 md)
Figure 6. 12 (RHS) is an example of relative permeability calculated from the true
pressure drops at 6 ml/min flow rate for the highly heterogeneous cores with 31.8, 318,
and 3180 md permeabilities. The corresponding capillary numbers Ncv are around 9.72,
100
30.7, and 97.2 respectively. The brine displacements for 31.8 md and 318 md cases are
in the viscous-dominated regime while the displacement for 3180 md case slightly
deviates from the viscous-dominated regime (shown in the LHS of Figure 6. 12). The
heterogeneity not only increases the flow rate dependency but also reduces the sensitivity
of average saturation on permeability.
Nevertheless, the relative permeability sets for the three different permeability cores
are expected to be almost identical since their saturation gradients are small. Figure 6. 12
(RHS) shows that reliable relative permeability data can be obtained even for the
heterogeneous cores with a wide range of core permeability once the displacement is
within or near the viscous-dominated regime.
Figure 6. 12 (LHS) Brine displacement efficiencies for three different permeability values
with capillary number Ncv ranging from 10 to 105; (RHS) Permeability effects on relative
permeability calculated in the true pressure drops for heterogeneous core (High Contrast
Model) at 6 ml/min flow rates.
6.3.4 Effects of Interfacial Tension (7.49-67.41 mN/m)
Different pressure, temperature and water salinity will result in different interfacial
tension. The effect of interfacial tension on relative permeability of CO2/brine systems
has been reported in the literature (Fulcher et al. 1985; Bachu and Bennion, 2007).
Therefore, the sensitivity study of interfacial tension over the same range of capillary
number has also been simulated. As we mentioned earlier, the variation of interfacial
tension does not account for varying temperature and pressure conditions. The wide
0.001
0.01
0.1
1
0.5 0.6 0.7 0.8 0.9 1
Kr
Sw
True Pressure Drop
HC(6):10K
HC(6):K
HC(6):0.1K
kr_input
101
range of IFT values is purely hypothetical and used solely to explore the sensitivity of the
IFT values.
Figure 6. 13 (LHS) compares the average saturation of three different interfacial
tensions (7.49 mN/m, 22.47 mN/m, and 67.41 mN/m) between the homogeneous model
and the High Contrast model. It is observed that the range of interfacial tension effects
on the average saturation is smaller for the large degree of heterogeneity. If the core
sample is very heterogeneous, then it is always safer to use the smaller capillary numbers
Ncv or larger flow rates to enter the viscous-dominated regime.
Figure 6. 13 (RHS) is an example of relative permeability calculated from the true
pressure drops at 6 ml/min flow rate for the highly heterogeneous cores (High Contrast
Model) with 7.49, 22.47, and 67.41 mN/m IFT values. The corresponding capillary
numbers Ncv are around 10, 30.7, and 92 respectively. The brine displacements for 7.49
and 22.47 mN/m cases are in the viscous-dominated regime while the displacement for
67.41 mN/m case slightly deviates from the viscous-dominated regime. Similarly, the
relative permeability sets for the three different interfacial tension values are expected to
be almost identical since their saturation gradients are small. Reliable relative
permeability data can be obtained even for the heterogeneous cores with a wide range of
IFT values once the displacement is within or near the viscous-dominated regime.
Figure 6. 13 (LHS) Average CO2 saturation as a function of capillary number Ncv for
homogeneous and High Contrast models with three different values of interfacial
tensions; (RHS) Interfacial tension effects on relative permeability calculated in the true
pressure drops for heterogeneous core (High Contrast Model) at 6 ml/min flow rates.
0.001
0.01
0.1
1
0.5 0.6 0.7 0.8 0.9 1
Kr
Sw
True Pressure Drop
HC(6):σ/3
HC(6):σ
HC(6):3σ
kr_input
102
6.3.5 Effects of Gravity
Figure 6. 14 shows a sensitivity study on the effect of gravity for the homogeneous and
the two heterogeneous cores. The dashed lines represent simulations without considering
gravity (setting g=0) while the solid lines represent the simulations with gravity (setting
g=9.8 m/s2). It is shown that gravity override is eliminated even with horizontal
displacements when the capillary numbers Ncv are smaller than the critical values.
Furthermore it is shown that moderately heterogeneous cores are only minimally affected
by horizontal displacements. It is verified that the effect of gravity due to the density
difference between two fluids and the long core is small in the viscous-dominated regime,
as mentioned before. However, even without considering gravity in the simulation, flow
rate dependency is observed in the heterogeneous cores. In this condition, only capillary
force and viscous forces are competing with each other in our system. The transition is
much more abrupt than when gravity is included (Figure 6.14).
Figure 6. 14 Average CO2 saturation as a function of capillary number Ncv for
homogeneous, Kozeny-Carman (small heterogeneity) and high contrast (large
heterogeneity) models with and without gravity (1G/0G).
103
6.4 Discussion of Relative Permeability
Measurements
6.4.1 Observations from the Numerical and Semi-Analytical Models
Numerical simulations and semi-analytical studies of brine displacement efficiency in
homogeneous and heterogeneous cores have been presented in Chapter 4 and Chapter 5
respectively. The critical gravity numbers Ngv,c1 and Ngv,c2 define the transitions between
flow regimes, first from the viscous to gravity dominated regime and next, from the
gravity to capillary dominated regime. The different flow regimes are observed for both
the homogeneous and the heterogeneous models. Simulation results show that when the
capillary number Ncv is small enough (below the critical value), viscous forces dominate
in this regime. It is clearly shown in these chapters is that in the viscous-dominated
regime, the viscous force is much greater than gravity and capillary forces. Gravity
segregation can be neglected in this regime even with horizontal core flooding.
Consequently, the calculated effective relative permeability is independent of flow rate,
gravity and Bond number in this regime.
In the gravity-dominated regime, buoyancy of CO2 causes lower displacement
efficiency and results in a vertical saturation gradient, which leads to the deviations of
relative permeability values observed in Figure 6. 5 and Figure 6. 7. In this regime,
gravity not only causes the inaccuracy of relative permeability values but also results in
large flow rate dependency. Capillary heterogeneity will increase this flow rate
dependency.
For the cores with high permeability, low interfacial tension or a smaller degree of
heterogeneity, the two-phase flow displacement will encounter a stronger gravity effect
while the gravity effect is irrelevant for the cores with low permeability, high interfacial
tension or larger degree of heterogeneity. In addition, the highly heterogeneous cores
require the smaller Ncv to reach viscous-dominated regime and hence to obtain the
reliable relative permeability data (Table 6.6).
104
Table 6. 6 Summary of the first critical capillary number Ncv,c1 for different heterogeneity
cores.
Random 3 High Contrast Homogeneous
σlnk/lnkmean 0.234 0.168 0
Ncv,c1 (L) 168 222 887
Ncv,c1 (2L) - 419 1774
It is consistent with previous studies that indicate that once a saturation gradient develops
along the core, the relative permeability calculated based on Darcy’s law is no longer
valid (Avraam and Payatakes 1995). However, there are several methods to obtain the
reliable relative permeability. First, we can increase flow rates to minimize the saturation
gradients. Second, we can use true pressure drops for two fluids to get more reliable
relative permeability values even with 15% saturation gradient exists. Finally, if the
saturations at the inlet and outlet are known, we can increase the accuracy of relative
permeability to water by including the corresponding capillary pressure drop.
6.4.2 General Rule of Thumb for Reliable Relative Permeability
Measurements
In Chapter 4, it was shown that when the effect of gravity is important for the multiphase
flow system, we should use gravity number Ngv (Eq. 4.3) to non-dimensionalize the
saturation data, for example, for the homogeneous and mildly heterogeneous cores. On
the other hand, when the capillary heterogeneity is taken into account, the impact of
gravity is much smaller and capillary number Ncv (Eq. 4.2) is a better dimensionless
number to characterize our system (Chapter 5). The advantage of using appropriate
dimensionless numbers can be easily seen from Figure 6.16.
Since the critical capillary number depends on the rock heterogeneity, the higher the
degree of heterogeneity in the core, the smaller the capillary number Ncv required to get
the reliable relative permeability data.
105
Figure 6. 15 Average CO2 saturation as a function of alternative capillary number Ncv,
and alternative gravity number Ngv for homogeneous and High Contrast models.
In principle, we can calculate the first critical number based on Eq. 5.12c
2CO lHete
cv,c1 HeteB rg BL
f R1N =
τN k S
(5.12c)
where the Bond number is defined in Eq. 4.4
gv
B *
cv c
N ΔρgHN =
N p (4.4)
pc* is the characteristic capillary pressure of the medium, chosen as a so-called
displacement capillary pressure. The displacement capillary pressure is a capillary
pressure value at the brine saturation Sw equal to 1, and it is tangent to the major part of
the capillary pressure data. The pc* value for our core is about 3000 Pa, shown in Figure
6. 16.
Figure 6. 16 Laboratory capillary pressure data.
106
However, during the core-flood experiment, relative permeability evaluated at SBLHete
is
unavailable before we perform the relative permeability measurements. The parameter τ
is also dependent on σlnk/ln(kmean), which required detailed information on the rock.
Therefore, another practical method is proposed based on the simulation results for the
heterogeneous models. According to the simulation results (Figure 5.3, Figure 6.13 and
Figure 6.15), if the critical capillary number Ncv,c1 is chosen to be 15, relatively uniform
saturation profiles will be obtained for the most types of heterogeneity cores
(σlnk/ln(kmean)<0.5).
If the core is known to be a relatively homogeneous, a larger critical number can
result in the viscous-dominated regime, where we can expect to get a uniform saturation
and an intrinsic relative permeability data. The more information we know about the
core, the larger the Ncv (smaller flow rate) can be tolerated to obtain reliable relative
permeability data.
In summary, the accurate whole-core relative permeability measurements can be
achieved when
2
*
c
cv 2
Tco
kLp AN = 15
qH μ
(6.17)
which assures that the injection flowrate is in the viscous-dominated regime for a wide
range of heterogeneity with the aspect ratio Rl equals to 3.1445. Although additional
work is needed to establish the bounds for reliable relative permeability measurements
for a wide range of Rl, most of the laboratory cores have similar size of Rl, so it is
reasonable to use Eq. 6.17 to predict the first critical number qualitatively.
A hypothesis based on analytical and numerical results in Chapter 4, 5 and 6 is
provided as follows: reliable relative permeability measurements can be achieved once
we satisfy the conditions below. First, if the core is known as relatively homogeneous:
2
2
*CO lmean c
cv cv,c12
T B rg BLco
f Rk Lp A 1N = N
q N k SH μ
(6.18)
Second, if it is very heterogeneous:
107
2CO lHete
cv cv,c1 HeteB rg BL
f R1N N =
τN k S
(6.19)
For practical interest, we establish the relation of parameter τ in terms of the normalized
standard deviation σlnk/ln(kmean) based on all the sensitivity cases shown in this work
(Figure 6.17):
lnk
mean
στ 1+35
ln k
(6.20)
Eq. 6. 20 gives us an approximate estimation of τ, which can be used to estimate the first
critical number roughly based on Eq. 6.19.
Figure 6. 17 Heterogeneous parameter τ in terms of normalized standard deviation.
As mentioned, krg(SBLHete
) is unavailable before we perform the relative permeability
measurements, an iterative approach can be used to obtain a reliable input relative
permeability values by checking that Eqs. 6.18 and 6.19 are satisfied after the initial
measurements are made. However, Eq. 6.17 is an empirical approach and only suitable
for the aspect ratio Rl=3.1445. Finally, if there is no information about the core
properties, measurements should be made for a capillary number as small as possible (Eq.
6.17). Therefore, further investigation on the aspect ratio sensitivity studies for different
degrees of heterogeneity is required to generalize the results to a wide range of
conditions.
108
Chapter 7
Conclusions and Future Work
This dissertation addressed fundamental studies of multiphase flow of CO2 and brine in
heterogeneous porous media at the core-scale both numerically and analytically. The
combined influence of gravity, flow rate and small scale heterogeneity on core-scale
multiphase flow of CO2 and brine is an active and important research area due to the need
for storing CO2 emissions in deep saline aquifers. We now summarize this work and
make recommendations for future research directions.
7.1 Summary and Conclusions of the Present
Work
In this work, we developed and investigated new analytical techniques to study the
balance of three forces as well as the sub-core heterogeneity in multiphase flow system.
The following conclusions can be drawn from this study of the semi-analytical models:
1. A new semi-analytical solution has been developed to predict the influence of
gravity and capillary numbers on the average saturation expected during
multiphase flow experiments. The new solution provided here is a quick and easy
way to estimate the flow regimes for horizontal core floods. A general solution
and a specific solution at core-scale were provided for further development and
investigation.
2. Having a semi-analytical solution provides a useful tool for investigating
multiphase fluid displacement efficiency over a wide parameter space of practical
interest. The new semi-analytical solution can be used to estimate the average
109
saturation over a wide range of conditions in terms of several important
dimensionless numbers such as mobility ratio, average saturations in the viscous-
dominated regime (SBLHere
), relative permeability to gas evaluated at SBLHere
,
normalized standard deviation (σlnk/lnkmean), aspect ratio Rl, Bond number NB and
capillary number Ncv.
3. A summary of how to use analytical solution was provided. The semi-analytical
solution can be used to predict displacements in cores with small degrees of
heterogeneity.
4. For cores with a greater degree of heterogeneity, a modification of the solution is
provided that is capable of predicting average saturations even with a high degree
of heterogeneity.
5. Practical applications include helping to design core flood experiments, including
assuring that relative permeability measurements are made in the viscous
dominated regime, evaluating potential flow rate dependence, influence of core-
dimension on a multiphase flow experiments, influence of fluid properties on the
experiments, and influence of sub-core scale heterogeneity.
Potential applications of the analytical results include establishing the bounds over which
relative permeability can be accurately measured in horizontal core-flood experiments.
In this work, we investigated sensitivity studies on volume-averaged (up-scaled) relative
permeability that accounts for the role of sub-core scale heterogeneity on multi-phase
flow in CO2/brine systems. We now summarize this part of the work:
1. Increasing flow rate results in minimizing the saturation gradient caused by the
combined effects of capillary, viscous and gravity forces; hence, the relative
permeability approaches the maximum value asymptotically and stabilizes when
the uniform saturation is achieved.
2. Simulation results shown here indicate that the flow-rate dependent saturation
occur not only in the heterogeneous core but also in homogeneous cores. In
addition, we show that the heterogeneity will increase the flow-rate dependency,
but this flow-rate dependent behavior is mainly due to the complex interaction
between gravity, viscous, and capillary forces.
110
3. Correct procedures for calculating relative permeability are discussed.
4. Despite the complexity of heterogeneity, it is possible to obtain the accurate
relative permeability measurements for heterogeneous cores. The incomplete
fluid displacement is primarily due to the heterogeneity and unfavorable mobility
ratio, not gravity segregation. A flow rate-independent relative permeability can
be obtained even with the horizontal core orientation as long as the alternative
capillary number is small enough, in most of the cases, smaller than 15.
5. If the permeability of the core or the core heterogeneity is unknown, it is always
safer to choose a smaller alternative capillary number in order to get the rate-
independent relative permeability.
7.2 Directions for Future Research
Some possibilities and suggestions for improving the present study as well as for future
research directions are outlined below.
Accurate Critical Numbers: From the homogeneous studies, we can determine the
two critical numbers by comparing the pressure gradients associated with the
viscous, gravity and capillary forces as a function of gravity numbers. However,
it is not obvious how to determine the first critical number of the heterogeneous
cores using this methodology because the viscous pressure gradients are always
greater than the buoyancy pressure gradients (Figure 5.5). Based on Figure 6.13,
we can clearly identify the gravity-dominated regime by comparing simulations
with and without gravity. This approach can identify flow regimes more
accurately for the heterogeneous cores. Further investigation on different degrees
of heterogeneity is needed to give us insight into the two critical numbers.
Derive Fitting Parameters: It is difficult but also important to derive several fitting
parameters from fundamental principles to increase the robustness of the model.
Fourth Flow Regime: On the other hand, Figure 6.13 also shows a very interesting
result for the homogeneous cores. Even without including the gravity effect in the
111
simulation, a slight flow rate dependency occurs when Ncv > 4x105. This could be
evidence for a fourth flow regime or it might simply be due to the implemented
boundary condition. If this additional flow regime indeed exists, it is necessary to
include this in the analytical analysis. Further investigation is needed to resolve
this issue.
Correlation lengths: The heterogeneous porous medium created in this work is
either based on a one specific rock or using a random log-normal distribution. A
systematic study on cores with various correlation lengths can generalize and
extend the analytical solution to a wider range of rock types.
Anisotropic Porous Medium: It is important and realistic to study the influence of
anisotropic properties in both numerical simulations and the semi-analytical
solutions. The same strategy used to obtain the semi-analytical solution could be
applied provided constraints are provided by the simulations.
Different Aspect Ratios in Heterogeneous Porous Medium: Although
homogeneous semi-analytical solution predicts numerical results very well over a
wide range of aspect ratios, validation of heterogeneous semi-analytical solution
with different core dimensions is important and needed to test the 2D
heterogeneous model.
Upscaling: Investigation of upscaling strategies in the transition between the
viscous, gravity and capillary dominated regimes is very useful for increasing the
computational speed of high-resolution reservoir simulations. Additionally, the
results have relevance for understanding reservoir-scale processes, particularly at
the sub-grid scale, where intra-grid block processes may have an influence on
flow and transport parameterizations.
Compressibility: The effect of compressibility is included in the simulations.
Considering the compressibility in the analytical model may improve the
predictions.
Experimental Evaluation at Core Scale: Applying the semi-analytical solution to
the experimental systems for studying multiphase flows for a wide range of fluid
pairs, geometric configurations and rock properties is another way to test the
model.
112
Predict Pressure Drops: it will be very useful to generalize the semi-analytical
solution to the predict pressure gradients across the core. Once we can also
predict the pressure drops analytically, relative permeability can be calculated for
heterogeneous rocks based on the average saturation and the corresponding
pressure drop across the core.
The results shown in this work were mostly achieved with a specific core-shaped
geometry and a limited number of permeability distribution. The limitations inherent in
the results are discussed in the following. First, if there are two systems with the same
aspect ratio but the physical dimensions of the porous medium are meters instead of
centimeters, the dimensionless plots which show the saturations are expected to be still
valid. However, this needs to be confirmed with the large scale simulations.
Additionally, specific distributions of permeability and a specific measure of
heterogeneity are used in this study. The results from our work should be qualitatively
applied to a system which is encountered different heterogeneity than what examined
here. Also, to quantitatively predict the average saturation with different heterogeneity
distributions, a systematic investigation is needed to generalize results.
The most sensitive inputs of the results are the second critical number due to the
empirical definition obtained from the sensitivity studies. Although this definition works
reasonably well for most of the sensitivity studies, it is not as accurate as the first critical
number; therefore the errors will propagate from the homogeneous cores to the
heterogeneous cores.
The conclusions drawn from the relative permeability studies should apply to
unsteady-state relative permeability measurements, but further studies are needed to
confirm this. On the other hand, note that this work cannot be applied to the miscible
multiphase flow systems since all the simulations and all the theoretical work consider
the capillary pressure. In addition, the results shown here have only studied the physical
behavior of multiphase flow system, there is no chemical reactions considered in the
simulation or the theoretical analysis. Therefore, the results drawn from here may not
apply to carbonate systems, which have significant chemical reactions between the fluids
and rock.
113
Appendix A
2D Analytical Derivation for the
Homogeneous Model
A.1 Detail of 2D General Solution
2D mass conservation equations for incompressible flow with constant porosity and
permeability are:
j j j,x j,z
j
S S u uφ u =φ =0
t t x z
(4.11)
Pressure equation can be obtained with the condition Sw+2COS =1:
w,x g,x w,z g,z
t
u +u u +uu 0
x z
(4.12)
Therefore uw,x+ug,x=constant1 and uw,z+ug,z=constant2 where uj,x and uj,z are Darcy
velocity of phase j in the x and z direction, respectively. Darcy flow velocities for both
phases are given by
j j rj
j j j j,x j,z j
j
p p ku =-λ k , +ρ g u , u where λ =
x z μ
(4.13)
where uj, λj, pj, ρj and μj are Darcy flow velocity, relative mobility, pressure, density, and
viscosity of phase j, respectively. At the top and the bottom boundaries, uw,z=ug,z=0
hence results constan2=0. The total velocity in the z direction is vanished, therefore
uw,z=-ug,z. In the flow direction x, total volumetric flow rate is sum of water and gas
114
flow, uw,x+ug,x=ut. We can obtain the capillary pressure in the x and z directions in terms
of Darcy’s flow velocities:
w
w,x wg w g,xw,xc
g w g
g,x g
pu =-λ k
p -p uupx= -
p x x kλ kλu =-λ k
x
(A-1a)
w
w,z w w
g w g,zw,zc
w g
g w g
g,z g g
pu =-λ k +ρ g
p -pz uup= - + ρ -ρ g
p z z kλ kλu =-λ k +ρ g
z
(A-1b)
Apply two boundary conditions and rearrange we can obtain
g gt c c
g,x g,z
w
Mkλ Mkλq p pu = - and u = Δρg-
Akλ x 1+M z 1+M
(A-2)
Therefore we can solve velocity of gas phase ug
g t c c
g g,x g,z
w
Mkλ q p pˆ ˆ ˆ ˆu =u x+u z= - x+ Δρg- z
1+M Akλ x z
(A-3)
where M= λw/λg is mobility ratio
w rw w
w g
g rg g
λ k /μM= = , Δρ ρ -ρ
λ k /μ (A-4)
The mass conservation of gas phase now becomes:
g g gt c c
w
S Mkλ Mkλu p p1 1- + Δρg- =0
t φ kλ x 1+M φ z z 1+Mx
(4.16)
To non-dimensionlise the equation, we define xD=x/L, zD=z/H, tD=tut/φL, and pc(Sw)
=pc*J (Sw). Substituting all the defined terms into Eq. 4.16 yields the dimensionless mass
conservation equation for gas phase is
2
g rg rg rg
gv cv
D D D l D D D D
S Mk Mk Mk1 1 J J+ +N -N + =0
t x 1+M z 1+M R x x 1+M z z 1+M
(4.17)
where Ncv and Ngv are shown in Eq. 4.2 and Eq. 4.3, and Rl =L/H is the shape factor or
115
the so-called the aspect ratio. Since M, J, and krg all depend on saturations, we can
rewrite the dimensionless equation as
3 31 2
g g rg g rg g rg gcv
gv cv2
D g D g D D w D D w Dl
F FF F
S S Mk S Mk S Mk SNd 1 d dJ dJ+ +N + +N =0
t dS 1+M x dS 1+M z x dS 1+M x z dS 1+M zR
(A-5)
Or a more compact form
g g g g gcv
1 3 gv 2 cv 32
D D D D D D Dl
S S S S SN+F + F +N F +N F =0
t x x x z z zR
(A-6)
where F1, F2, and F3 are functions of saturations. At steady-state,
g
g D D
D
S=0 S SG x ,z
t
(A-7)
Assume Ncv and Ngv are independent of xD and zD, and then we can rewrite equations in
terms of xD and zD separately:
cv 3 cv 3
1 2 gv2
D 1 D D gv 2 Dl
N F N FSG SGF SG+ +F N SG+ =0
x F x z N F zR
(A-8)
Since xD and zD are independent, we assume that the dependence of the steady-state
solution SG on xD and zD is separated, that is:
D D D DSG x ,z =X(x )Z(z ) (A-9)
Substituting saturation back into the equation, and defining a=F1/F3, b=F2/F3, and Bond
number
gv
B *
cv c
N ΔρgHN =
N p (4.4)
Then Eq. (A-8) becomes
2 gvcv1
2
D D D B Dl
F NNF d 1 dX d 1 1 dZX+ + Z+ =0
X dx a dx Z dz N b dzR
(A-10)
116
To find a solution that satisfies Eq. (A-10), then
2l
Dcv
B D
cvR a1 D2 - xNDl
D 1 D
-N bz
D 2 D2 D
B D
N dΧΧ+ =const c z
dxR aΧ=A z e +c z
1 dΖΖ=B x e +c xΖ+ =const c x
N b dz
(A-11)
Substituting Eq. (A-11) back to Eq. (A-9) yields
2 2l l
D B D Dcv cv B D
R a R a- x -N bz - x
N N -N bz
1 D D 2 D D 3 D D 4 D DSG=C x , z e +C x , z e +C x , z e +C x , z (A-12)
where C1, C2, C3 and C4 are functions of xD and zD.
A.2 Derivation of Parameters a and b
From Eq. (A-5), F1, F2, and F3 are defined as
1 g 2
g g
d 1 1 dMF S = =-
dS 1+M dS1+M
(A-13a)
rg rg rg
2 g 2
g g g
Mk k dkd dM MF S = =
dS 1+M dS 1+M dS1+M
(A-13b)
rg rgw
3 g
w
Mk MkdJ SF S = =J'
dS 1+M 1+M (A-13c)
Assuming the derivative of relative permeability is proportional to itself,
rgrw
rw w rw rg g rg
w w
dkdkk '= =a k and k '= =-a k
dS dS (A-14)
Using different forms of relative permeability results in different aw and ag. In this paper,
aw and ag are as follows:
gw
w g
w wr wr
nna = and a =
S -S 1-S (A-15)
117
Since the mobility ratio M= λw/λg = μg krw/μw krg, the derivative of M with respect to Sw is
g -1 -2
rw rg rw rg rg w g
w w
μdMM'= = k 'k -k k k ' = a +a M
dS μ (A-16)
Substituting Eq. (A-14) and (A-16) back into (A-13) yields
1 g w g2 2
g g
d 1 1 dM MF S = =- = a +a
dS 1+M dS1+M 1+M
(A-17a)
rg g w rg
2 g
g
Mk Ma -a MkdF S = =
dS 1+M 1+M 1+M
(A-17b)
Therefore, we can derive the two variables a and b based on the definitions in Appendix
A.1:
w g1
3 rg
a +aF 1a= =
F J'k 1+M (A-18a)
g w2
3
a M-aF 1b= =
F J' 1+M (A-18b)
Since fCO2=1/(1+M), then a and b becomes
2
2
w g CO g w
CO
rg
a +a f a M-aa and b f
J' k J'
(A-19)
Therefore a and b are proportional to fCO2/krg and fCO2, respectively.
118
Appendix B
2D Analytical Derivation for the
Heterogeneous Model
B.1 General Solution
Substituting xD=x/L, zD=z/H, tD=tut/φL, τ(xD, zD)=sqrt(k/kmean) and pc(Sw) =pc*J (Sw)/τ
into Eq.(5.3), the dimensionless mass conservation equation at steady-state becomes
2
2 2 2cv
CO gv w rg w rg cv w rg2
D D D D D Dl
N J τ J τf +N f k τ = f k τ +N f k τ
x z x x z zR
(B-1)
Assuming all the variables are continuously differentiable. Differentiate the capillary
terms
2 2
D D D D D D
J/τ SG J' J τ J/τ SG J' J τ and
x x τ x z τ zτ τz
(B-2)
where J’=dJ/dSw. Substituting Eq. (B-2) into (B-1) and using the definition of Ms=fwkrg,
H=-fwkrgJ’=-MsJ’ and G=fwkrgJ= MsJ, then we can rewrite the steady-state mass balance
as follows, which is already shown in Eq. (5.6):
2
2cv cv
CO cv B rg w cv2 2
D D D D D D D Dl l
N Nτ τ SG SGf G +N N k f τ +G = Hτ +N Hτ
x x z z x x z zR R
(5.6)
The gradients of heterogeneity are defined as follows:
119
x z
D D
τ ττ and τ
x z
(B-3)
Substituting back into Eq.(5.6), and the left hand side becomes
2
2cv
CO x cv B S z2
D Dl
NSG d SG dLHS= f + Gτ +N N M τ Gτ
x dSG dSGR z
(B-4)
Apply the definition of F1, F2, and F3 used in Kuo and Benson (2012) and also provided
in Appendix A.2, then
2cv
1 x cv B 2 z2
D Dl
NSG d SG dLHS= F + G τ +N N F τ G τ
x dSG dSGR z
(B-5)
Differentiate the function G
S S 2 2
d dG= JM M J' JF H JF
dSG dSG (B-6)
Combine LHS and RHS of Eq.(5.6) into x-dependent and z-dependent terms:
2
2cv l
1 2 x cv B 2 2 z2
D cv D D D D Dl
N RSG SG SG SGF + H JF τ - Hτ +N N F τ H JF τ - Hτ 0
x N x x zR z z
(B-7)
Define
2
l
1 1 2 x
cv
RG = F + H JF τ
N (B-8)
2
2 B 2 2 zG =N F τ H JF τ (B-9)
We can rewrite Eq. (B-7) into a simpler form:
cv
1 cv 22
D 1 D D D 2 D Dl
N SG 1 SG SG 1 SGG - Hτ +N G - Hτ 0
x G x x G zR z z
(B-10)
Since
120
1 D D D 1 D D 1 D
2 D D D 2 D D 2 D
1 SG Hτ SG 1 SGHτ Hτ
G x x x G x x G x
1 SG Hτ SG 1 SGHτ Hτ
G z G z G zz z z
(B-11)
Assuming F1, F2, H, τx and τz are independent of xD and zD respectively, then
1 2 x 2 x
2 2 2
D 1 D D D1 1 1
2
2 B 2 2 z z B 2 2 z
2 2 2 2 2
D 2 D D D D2 2 2 2 2
G F τ F τ J'1 1 J SG
x G x x xG G G
G N F F τ 2ττ N F F τ J'1 1 τ J SG
z G z z zG G G G Gz
(B-12)
Eq. (B-12) is neglected since they are secondary effects, then
cv
1 cv 22
D 1 D D 2 Dl
N Hτ SG Hτ SGG SG- +N G SG- 0
x G x G zR z
(B-13)
We assume that the dependence of the steady-state solution SG on xD and zD is separated,
that is:
D D D DSG x ,z =X(x )Z(z ) (B-14)
cv cv 21
2
D 1 D D 2 Dl
N N GG Hτ dX Hτ dZX- + Z- 0
X x G dx Z G dzR z
(B-15)
To satisfy Eq. (B-15), which implies
1D
2D
Gx
1 D HτD 1 D1
Gz
Hτ2 D D 2 D
2
HτX- X'=const c z
Χ=A z e +c zG
HτZ- Z'=const c x Ζ=B x e +c x
G
(B-16)
where A(zD) and c1(zD), B(xD) and c2(xD) are parameters dependent on zD and xD
respectively. Substituting Eq. (B-16) into Eq. (B-14) yields the steady-state saturation as
2 1 2
D D D
G G Gz x z
Hτ Hτ Hτ1 D D 2 D D 3 D D 4 D DSG= C x , z e C x , z e C x , z e C x , z
(B-17)
121
C1, C2, C3 and C4 are functions of xD and zD. Since H=-F3, a=F1/ F3, and b= F2/ F3, and
assume ε=Jb-1, then we can obtain
2 2
l l1 2 x x 2
cv cv1 l x
3 cv
R RF + H+JF τ a+ -1+Jb τ
N NG R τa- -ε
Hτ -F τ -τ N τ τ (B-18)
2 2
B 2 2 z B z2 z
B
3
N F τ + H+JF τ N bτ + -1+Jb τG τ= =-N bτ-ε
Hτ -F τ -τ τ (B-19)
Use the previous definitions (Eq. B-3) and define the two heterogeneity terms
x z
x z
D D D D
τ τ1 τ ln τ 1 τ ln τlnτ and lnτ
τ τ x x τ τ z z
(B-20)
Therefore, we can rewrite the solution Eq. (B-17) as
2l
DD D D Dcvz x B D x z B D
R a- x
-ε lnτ z -ε lnτ x -ε lnτ x -ε lnτ zN τ-τN bz -τN bz
1 2 3 4SG= C e e e C e e C e e C
(B-21)
or
2l
DcvB D B D
R a- xN τ-τN bz -τN bzHete Hete Hete
1 2 3 4SG= C e C e C e C (B-22)
with coefficients
D Dz x-ε lnτ z -ε lnτ xHete
1 1 D DC C x , z e e (B-23a)
Dx-ε lnτ xHete
2 2 D DC C x , z e (B-23b)
Dz-ε lnτ zHete
3 3 D DC C x , z e (B-23c)
C1, C2, C3 and C4 are functions of xD and zD. Based on this solution, the time-
independent CO2 saturation SG depends on its position xD and zD, M, as well as the
dimensionless numbers Rl, Ncv, NB and τ.
122
B.2 Analytical Solution with Simulation
Constraints
To eliminate some of unknown coefficients, we apply the constraints from simulations
(Eq. 5.9 and 5.10) to the general equation. First, the average CO2 saturations (Eq. 5.8)
are independent of capillary number Ncv when Ncv ≤ Ncv,c1.
2l
0,D2 cv
R1ax
CO N
cv
S0 e 0
N
(B-24)
Therefore, in the viscous-dominated regime, Ngv ≤ Ngv,cl, the core average saturation
becomes
0,D B 0,DB D z
2
-ε lnτ z bN zN bzHete
CO 3 4 3 4S C e +C C e e +C
. (B-25)
Second, the average saturation in this regime is also independent of Bond number, which
leads to
2 0,D B 0,DzCO -ε lnτ z bN z3 4
0,D 3
B B B
S C C- bz C e e + 0
N N N
. (B-26)
Solving for
B 0,D B 0,DbN z bN z
3 3,0 4 4,0C =const e C e and C =const=C
. (B-27)
C3,0 and C4,0 are constants. Therefore, the average saturation in the viscous-dominated
regime (Eq. B-25) is simplified to
0,Dz
2
-ε lnτ z
CO 3,0 4,0S C e +C (B-28)
For the homogeneous core, τ=1, (lnτ)x=(lnτ)z=0, the average saturation in the viscous-
dominated regime is Buckley-Leverett solution, which results in C3,0+C4,0=SBL. For
mathematical convenience, assume C4,0=0, then C3,0=SBL. We can obtain the average
CO2 saturation for the heterogeneous core in the viscous-dominated regime:
0,Dz-ε lnτ zHete
BL 1 BL 1S h τ S where h τ e (B-29)
123
Since h1(τ)<1, the average saturation for the heterogeneous cores is lower than the
Buckley-Leverett solution SBL, which is consistent with the simulation results. The
deviation of saturation from Buckley-Leverett solution is dependent on the degree of
heterogeneity, (lnτ)z, mainly the gradient of heterogeneity in the vertical direction. Eq.
B-29 also implies that strong heterogeneity results in lower CO2 saturation. Now we can
rewrite Eq. (B-21) as
2l
0,DcvB D
2
aR1x
N-τN bzHete Hete Hete
CO 1 2 BLS C e C e +S
(B-30)
For the homogeneous cores, Eq. (B-30) can be reduced to
2l
0,DB 0,D cv
2
aRx
bN z N
CO 1 2 BLS C e +C e +S
(B-31)
When Bond number equals to zero (g=0), the average saturations are observed to be a
constant SBL (Kuo et al. 2010), which results in C2= -C1. Putting this constraint back to
the coefficients defined in Eq. (B-23) yields:
0,D 0,D 0,Dx z x-ε lnτ x -ε lnτ z -ε lnτ xHete Hete
1 1 2 1C C e e , C =-C e (B-32)
Define
0,Dx-ε lnτ x
2h τ =e (B-33)
Substituting h1, h2, and Eq. (B-32) back to Eq. (B-30), then we can obtain
2l
0,DB 0,D cv
2
R1ax
N bz N Hete
CO 1 2 1 BLS C h τ h τ e -1 e +S
(B-34)
Eq. (B-34) can be reduced to the homogeneous results when τ=1:
2
l0,D
B 0,D cv
2
Rax
N bz N
CO 1 BLS C e -1 e +S
(B-35a)
where
2CO11 BL
1 0,D 11 0,D
B rg BL B 22l
fC S1 1C = , ax = d and bz =
N k S N αdR
(B-35b)
and hence give us the expressions for the coefficients C1, ax0,D and bz0,D.where C11, d11, α
and d22 are constant parameters determined by curve matching the semi-analytical
124
solution with the simulation results for homogeneous cores. Substituting Eq. (B-35b)
into Eq. (B-34), the final equation for evaluating the average CO2 saturation can be
written as follows:
2l
0,DB 0,D cv
2
R1ax
N bz N Hete11
CO 2 1 BL
B 1
C 1S h h e - e +S
N h
(B-36a)
2CO0,D
11 HeteBrg BLl
fax 1 1 1= d
τ τ Nk SR
(B-36b)
Hete
BL
0,D
22
Sτbz =τ
αd (B-36c)
A more compact form can be rearranged:
2l
0,DB 0,D cv
2
R1- ax
-τN bz τ NHete HeteBL
CO 1 BLHete
BL
SS =C e - e +S
S
(B-37a)
where
0,Dz-ε lnτ zHete
BL 1 BL 1S h τ S where h τ e (B-37b)
0,Dx-ε lnτ x11 2 1Hete
1 2
B
C h τ h τC where h τ e
N (B-37c)
or finally in terms of critical numbers:
2 Hete2 l cv,c1
Hete 1cv,c2 cv
2
d R N- dN NHete HeteBL
CO 1 BLHete
BL
SS C e - e +S
S
(B-38a)
2
Homo Hete
gv,c1 BL CO lHete
cv,c1 HeteB B rg BL
N S f R1 1N = =
τ N τN k S
(B-38b)
Homo HeteHete 3/2gv,c2 BLgv,c2Hete l
cv,c2 Hete
B B B BL
N SN αR1N = = =
N τN τN S (B-38c)
125
B.3 Heterogeneous Factor, σlnk/ln(kmean)
In this section, we want to rewrite all the heterogeneous terms such as τ, h1(τ) and h2(τ) in
terms of the known factor, σlnk/ln(kmean). Assuming the heterogeneity changed in the x
and z direction, (lnτ)x and (lnτ)z are proportional to the σlnk/ln(kmean) respectively with
proportional factors β1 and β2:
lnk lnk
2 1z x
D mean D mean
σ σln τ ln τlnτ β and lnτ β
z ln k x ln k
(B-39)
Specifically, for random fields, the change rate of heterogeneity in both x and z direction
should be the same, therefore the proportional factor β1 is equal to β2. Since for the
homogeneous core, σlnk =0, τ=1, and (lnτ)x=(lnτ)z=0, we can solve lnτ and τ in terms of
the function σlnk/ln(kmean):
lnk lnk
2 D 1 D
mean mean
σ σln τ β z β x
ln k ln k (B-40a)
lnk lnk 1 D lnk
2 D 1 D 2 Dmean mean 2 D mean
σ σ β x σβ z β x 1+ β z
ln k ln k β z ln kτ=e =e
(B-40b)
Both Eq. (B-39) and Eq. (B-40) satisfy the homogeneous requirement. Substituting Eq.
(B-38) into the definitions of h1(τ) and h2(τ) yields
lnk
2 0,D0,D meanz
σ-εβ z
-ε lnτ z ln k
1h τ e =e (B-41a)
lnk
1 0,D0,D meanx
σ-εβ x
-ε lnτ x ln k
2h τ e =e (B-41b)
Therefore we can rewrite C1Hete
in terms of τ:
11 2 1Hete -ε11
1
B B
C h τ h τ CC τ
N N (B-42)
126
Appendix C
Flow Regimes
C.1 Critical Gravity Numbers
Figure C.1 and Figure C.2 illustrate the two critical numbers (dashed lines) and the
pressure gradients associated with the viscous, capillary and gravity forces for different
lengths of the cores as well as different interfacial tension and permeability values.
Figure C.1: Pressure gradients for viscous, capillary and gravity forces for different
aspect ratios.
127
Figure C.2: Pressure gradients for viscous, capillary and gravity forces for different
interfacial tension and permeability values.
Figure C.3: Pressure gradients for viscous, capillary and gravity forces for different
permeability values for High Contrast model.
128
Appendix D
Parameter Tables for Heterogeneous
Cores
Table D.1: Summary of Sensitivity Studies for High Contrast Models (HC).
Rl fCO2 kmean(mD) σ(mN/m) σlnk/ln(kmean) ω ε
Base(k, σ) 3.1445 0.95 318 22.47 0.1667 132 2.9
10k 3.1445 0.95 3180 22.47 0.1191 108 3.145
0.1k 3.1445 0.95 31.8 22.47 0.2776 145 3.153
σ/3 3.1445 0.95 318 7.49 0.1667 108 3.147
3σ 3.1445 0.95 318 67.41 0.1667 154 3.147
0.79f 3.1445 0.79 318 22.47 0.1667 78 3.2
0.51f 3.1445 0.51 318 22.47 0.1667 67 3.4
0.34f 3.1445 0.34 318 22.47 0.1667 65 3.5
2L 6.2890 0.34 318 22.47 0.1679 170 3.135
Table D.2: Summary of Sensitivity Studies for Random 3 Models.
Rl fCO2 kmean(mD) σ(mN/m) σlnk/ln(kmean) ω ε
Base(k, σ) 3.1445 0.95 366 22.47 0.2317 54 3.2
10k 3.1445 0.95 3660 22.47 0.1667 46 4.7
0.1k 3.1445 0.95 36.6 22.47 0.3740 62 2.6
σ/3 3.1445 0.95 366 7.49 0.2317 48 4.3
Table D.3: Summary of Sensitivity Studies for Different Heterogeneous Models.
Rl kmean(mD) σlnk/ln(kmean) Ncv,c1 Ncv,c2 ω ε
Homo 3.1445 430 0 886 3621 - -
Random2 3.1445 430 0.0419 850 3387 420 50
K-C 3.1445 430 0.0455 604 2283 183 10
H-C 3.1445 31.8 0.1667 173 618 132 2.9
Random3 3.1445 318 0.2317 170 464 54 3.2
Random4 3.1445 318 0.4793 131 213 31.5 2.7
129
Appendix E
Validation of TOUGH2 Simulations
E.1 Relative Convergence Tolerance ε1=10-5
A much smaller relative convergence tolerance ε1=10-5
is used to compare with the one
used in the simulations, ε1=10-2
. Figure E.1 shows the typical CO2 saturations as a
function of pore volume injected (PVI) for homogeneous cores with different flow rates.
It is observed that even with the large relative convergence tolerance, the results of
saturation are almost identical to the results with ε1=10-5
. Once capillary heterogeneity is
introduced into the simulations, it is extremely difficult to reach the steady-state
condition by using such smaller relative convergence tolerance to solve three-
dimensional mass conservation equations with both gravity and capillarity. Therefore,
the high convergence tolerance is acceptable due to the accuracy of the results and the
feasibility runoff simulating highly heterogeneous cores.
130
Figure E. 1 CO2 saturation as a function of pore volume injected (PVI) for different flow
rates.
E.2 GPRS
In order to validate the TOUGH2 simulation results, an alternative reservoir simulation
software, GPRS (General Purpose Research Simulator), is used for comparison to the
TOUGH2 results (Cao 2002; Jiang 2007; Li 2011). Figure E. 2 compares the grid
saturations of certain slices between GPRS and TOUGH2. The GPRS results are almost
identical to the TOUGH2 results (Li 2011).
Figure E. 2 GPRS vs TOUGH2 results for slice x=16 (LHS) and x=21 (RHS)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03
CO
2 S
atu
rati
on
PVI
Homogeneous: 0.1 ml/min
ε1=0.01
ε1=1E-50.00
0.02
0.04
0.06
0.08
0.10
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02
CO
2 S
atu
rati
on
PVI
Homogeneous: 0.01 ml/min
ε1=0.01
ε1=1E-5
0.00
0.02
0.04
0.06
0.08
0.10
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
CO
2 S
atu
rati
on
PVI
Homogeneous: 0.001 ml/min
ε1=0.01
ε1=1E-5
0.00
0.02
0.04
0.06
0.08
0.10
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02
CO
2 S
atu
rati
on
PVI
Homogeneous: 0.01 ml/min
ε1=0.01
ε1=1E-5
0.00
0.02
0.04
0.06
0.08
0.10
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
CO
2 S
atu
rati
on
PVI
Homogeneous: 0.001 ml/min
ε1=0.01
ε1=1E-5
131
Nomenclature
A cross section area of the core
A1 J-function fitting parameter
B1 J-function fitting parameter
C11 fitting parameter of analytical solutions for homogeneous cores
Ca traditional capillary number, μgut/σ
H height of the core [m]
J Leverett’s J function
L length of the core [m]
M mobility ratio, λw/ λg
NB Bond number, ΔρgH/pc*
Ncv alternative capillary number, kLpc*/H
2μg ut
Ngv gravity number, ΔρgkL/Hμg ut
Rl aspect ratio, L/H
S average saturation or scaling factor of Kozeny-Carman equation
SG steady-state CO2 saturation
Sp J-function fitting parameter
S* normalized brine saturation
a/b functions depending on water or CO2 saturation, fluid properties, and the
exponents of relative permeability functions, Eq. A-19
d11,22 fitting parameters of analytical solutions for homogeneous cores
f fractional flow, 1/1+M
g acceleration [m/s2]
h1 heterogeneous factor, Eq. 5.12b
k average permeability [mD]
132
kr relative permeability [mD]
n functional exponent of relative permeability
p pressure [Pa]
pc* characteristic capillary pressure [Pa]
t time
u Darcy velocity [N/m]
x x-coordinate
z z-coordinate
GREEKS
α proportional factor of second critical gravity number, 12.79737
β1,2 proportionality factors, which measure the degrees of heterogeneity in the x- and
z- directions, respectively, Eq. 5.11b
ε fitting parameter of analytical solutions for heterogeneous cores
λ1,2 J-function fitting parameter
λ relative mobility, kr/μ
μ viscosity [cp]
φ porosity
ΔP pressure difference between the average inlet and the outlet slice values
Δρ density difference between CO2 and brine [kg/m3]
σ CO2-brine interfacial tension [N/m] or standard deviation
θ contact angle, 0˚
τ heterogeneity parameter
ω fitting parameter of analytical solutions for heterogeneous cores
133
SUBSCRIPTS
c capillary
c1 critical number of transition 1
c2 critical number of transition 2
g gas (displacing fluid)
j phase
t total
w water (displaced fluid)
wr residual
BL Buckley-Leverret
CO2 CO2 (displacing fluid)
D dimensionless
134
Bibliography
1. Abaci, S., Edwards, J.S., Whittaker, B.N., 1992, Relative permeability measurements
for two phase flow in unconsolidated sands. Mine Water Environ. 11, 16.
2. Akbarabadi, M. and Piri, M., 2011, Geologic sequestration of supercritical carbon
dioxide: An experimental study of capillary trapping and relative permeability,
SCA2011-04.
3. Ataie-Ashtiani, B., Hassanizadeh, S.M., and Celia, M.A., 2002, Effects of
heterogeneities on capillary pressure-saturation-relative permeability relationships,
Journal of Contaminant Hydrology 56, 175-192.
4. Avraam, D. G., and A. C. Payatakes. 1995, Generalized relative permeability
coefficients during steady-state two-phase flow in porous media, and correlation with
the flow mechanisms. Transport in Porous Media 20: 135-168.
5. Bachu, S. and Bennion, B., 2007, Effects of in-situ conditions on relative
permeability characteristics of CO2–brine systems. Environ. Geol. doi
10.1007/s00254-007-0946-9.
6. Bennion, B. and S. Bachu, 2005, Relative permeability characteristics for
supercritical CO2 displacing water in a variety of potential sequestration zones in the
Western Canada Sedimentary Basin; Paper SPE 95547.
7. Bennion, B. and Bachu, S., 2006a, The impact of interfacial tension and pore size
distribution/capillary pressure character on CO2 relative permeability at reservoir
conditions in CO2-Brine systems; Paper SPE 99325.
8. Bennion, B. and Bachu, S., 2006c, Dependence on temperature, pressure, and salinity
of the IFT and relative permeability displacement characteristics of CO2 injected in
deep saline aquifers; Paper SPE 102138.
9. Bennion, B. and Bachu, S., 2007, Permeability and relative permeability
measurements at reservoir conditions for CO2-Water systems in ultra low
permeability confining caprocks; Paper SPE 106995.
135
10. Bennion, B. and Bachu, S., 2008, Drainage and imbibition relative permeability
relationships for supercritical CO2/Brine and H2S/Brine systems in intergranular
sandstone, carbonate, shale, and anhydrite rocks; Paper SPE 99326.
11. Bennion, B. and Bachu, S., 2010, Drainage and imbibition CO2/Brine relative
permeability curves at reservoir conditions for high-permeability carbonate rocks;
Paper SPE 134026.
12. Benson, S.M., and Surles, T., 2006, Carbon dioxide capture and storage: an overview
with emphasis on capture and storage in deep geological formations, The Proceedings
Special Issue, Institute of Electrical and Electronics Engineers (IEEE), Vol. 94, No
10, October 2006, DOI 10.1109/PROC.2006.883716.
13. Bryant, S.L., Lakshiminarasimhan, S., Pope, G.A., 2006. Buoyancy-dominated
multiphase flow and its impact on geological sequestration of CO2, SPE 99938,
presented at SPE/DOE Symposium on Improved Oil Recovery, Tulsa, April 22–26.
14. Buckley, SE. and Leverett, MC, 1942, Mechanism of fluid displacement in sands
,Trans. AIME 146,107.
15. Cao, H., Development of Techniques for General Purpose Simulators, PhD thesis,
Stanford University, (2002).
16. Caudle, B. H., Slobod, R. L., and Brownscombe, E. R., 1951, Further developments
in the laboratory determination of relative permeability, Trans. AIME 192:145–50.
17. Chalbaud, C., Lombard, J.-M., Martin, F., Robin, M., Bertin, H., Egermann, P., 2007,
Two Phase Flow Properties of Brine-CO2 Systems in a Carbonate Core: Influence of
Wettability on P c and k r, SPE 111420, Reservoir Characterization and Simulation
Conference SPE/EAGE Abu Dhabi, UAE, October 28–31.
18. Chang, Y.-B., Lim, M.T., Pope, G.A., and Sepehrnoori, K., 1994, CO2 flow patterns
under multiphase flow: heterogeneous field-scale conditions, SPE Reservoir
Engineering, 9. 208-16.
19. Chang, Y.C. and Mohanty, K.K., 1997, Scale-up of two-phase flow in heterogeneous
porous media. J Pet Sci Eng 18, pp. 21–34.
20. Chaouche, M., Rakotomalala, N., Salin, D., Xu, B. and Yortsos, Y. C.: 1994,
Capillary effects in drainage in heterogeneous porous media: continuum modelling,
experiments and pore network simulations.Chem. Eng. Sci., 49, 2447-2466.
136
21. Cinar, Y., Jessen, K., Berenblyum, R., Juanes, R. and Orr, F.M. Jr., 2006, An
experimental and numerical investigation of crossflow effects in two-phase
displacements, SPE Journal 11 (2): 216-226. SPE-90568-PA. DOI: 10.2118/90568-
PA.
22. Flett, M., Gurton, R. and Weir, G., 2007, Heterogeneous saline formations for carbon
dioxide disposal: Impact of varying heterogeneity on containment and trapping.
Journal of Petroleum Science and Engineering, 57 , pp. 106–118.
23. Fulcher, R. A., Ertekin, T., and Stahl, C. D., 1985, Effect of capillary number and its
constituents on two-phase relative permeability curves, J. Petrol. Tech., (Feb.), 249-
260.
24. Geffen T. M., Owens, W. W., Parrish, D. R. and Morse, R. D., 1951, Experimental
investigation of factors affecting laboratory relative permeability measurements,
Trans. AIME 192:99.
25. Hamon G. and C. Roy, 2000, Influence of heterogeneity, wettability and coreflood
design on relative permeability curves, SCA paper.
26. Han, Weon Shik, Kim, Kue-Young, Esser, Richard P., Park, Eungyu · McPherson,
Brian J., 2011, Sensitivity Study of Simulation Parameters Controlling CO2 Trapping
Mechanisms in Saline Formations, Transp Porous Med.
27. Hassanizadeh, S. M. & Gray, W. G. (1993b). Toward an improved description of the
physics of two-phase flow, Advances in Water Resources 16(1): 53–67.
28. Henderson, J. H., and S. T. Yuster. 1948, Studies in relative permeability. World Oil
V127, no.12, p.139-144.
29. Henderson, GD., Danesh, A, Tehrani, DH, and Peden, JM, 1997, The effect of
velocity and interfacial tension on relative permeability of gas condensate fluids in
the wellbore region, Journal of Petroleum Science and Engineering.
30. Hesse, MA, Orr, Jr FM, Tchelepi, HA, 2008, Gravity currents with residual trapping.
Journal of Fluid Mechanics 611: 35-60.
31. Honarpour, M.M., Cullick, A.S., Saad, Naji, and Humphreys, N.V., 1995, Effect of
Rock Heterogeneity on Relative Permeability: Implications for Scale-up, SPE 29311.
32. Hussain, F., Cinar,Y., and Bedrikovetsky, P., 2011, A Semi-Analytical Model for
Two Phase Immiscible Flow in Porous Media Honouring Capillary Pressure,
137
Transport in Porous Media.
33. Ide, S. T., Jessen, K., Orr, F. M., Storage of CO2 in saline aquifers: effects of gravity,
viscous, and capillary forces on amount and timing of trapping, J. Greenhouse Gas
Control 1 (2007): 481-491.
34. IPCC. IPCC special report on carbon dioxide capture and storage. In: Metz B,
Davidson O, de Coninck HC, Loos M, Meyer LA, editors. Prepared by working
group III of the intergovernmental panel on climate change. Cambridge, United
Kingdom and New York, NY, USA: Cambridge University Press; 2005.
35. Jiang, Y., Techniques for Modeling Complex Reservoirs and Advanced Wells, PhD
thesis, Stanford University, (2007).
36. Johnson, E.F., Gassler, D.P. and Naumann, V.O., 1959, Calculation of relative
permeability from displacement experiments, Trans. AIME, 216, p. 370.
37. Jonoud S., Jackson M., 2008, New criteria for the validity of steady-state upscaling.
Transport Porous Med. 71(1), 53–73.
38. Juanes, R., Spiteri, E.J., Orr, F.M. and Blunt, M.J., 2006, Impact of relative
permeability hysteresis on geological CO2 storage, . Water Resour Res, 42, p.
W12418 13p.
39. Kopp, A., Class, H., Helmig, R., 2009, Investigations on CO2 storage capacity in
saline aquifers—Part 1: dimensional analysis of flow processes and reservoir
characteristics. Int. J. Greenhouse Gas Control 3(3), 263–276.
40. Krause, M., Perrin, J.-C., and Benson, S. M., 2009, Modeling Permeability
Distributions in a Sandstone Core for History Matching Coreflood Experiments,
SPE126340.
41. Krause, M., Perrin, J.-C., and Benson, S. M., 2011, Modeling Permeability
Distributions in a Sandstone Core for History Matching Coreflood Experiments, SPE
Journal, volume 16, number 4, SPE126340 –PA.
42. Krevor, S., Pini, R., Li, B., and Benson, S., 2011, Capillary heterogeneity trapping of
CO2 in a sandstone rock at reservoir conditions, Geophysical Research Letters 38,
L15401.
43. Krevor, S., Pini, R., Zuo, L., and Benson, S., 2012, Relative permeability and
trapping of CO2 and water in sandstone rocks at reservoir conditions, Water
138
Resources Research, Vol. 48, W02532, 16 PP.
44. Kuo, C-W, Perrin, J-C., and Benson, S., 2010, “Effect of gravity, flow Rate, and
small scale heterogeneity on multiphase flow of CO2 and brine”, SPE132607, SPE
Western Regional Meeting, 27-29 May 2010, Anaheim, California, USA8.
45. Kuo, C-W and Benson, S., 2012, “Analytical Study of Effects of Flow Rate, Gravity,
and Subcore Scale Heterogeneities on CO2/Brine Multiphase Flow System in
Horizontal Corefloods”, SPE153954, SPE Western Regional Meeting, 21-23 March
2012, Bakersfield, California, USA.
46. Kyte, J. R. and Rapoport, L. A., 1958, "Linear Waterflood Behavior and End Effects
in Water-Wet Porous Medium", Trans. , AIME 213 , 423.
47. Lake, L. W., 1989, Enhanced Oil Recovery. Prentice Hall: Old Tappan, New Jersey.
48. Lenormand, R., Touboul, E., Zarcone, C., 1988, Numerical models and experiments
on immiscible displacements in porous media, J. Fluid Mech., 189 (1988), pp. 165–
187.
49. Leverett, M.C., 1941, Capillary behaviour in porous solids. J. Trans. AIME 142, 152–
169.
50. Li, B., Including Fine-Scale Capillary Heterogeneity in Modeling the Flow of CO2
and Brine in Reservoir Cores, M.S. thesis, Stanford University, (2011).
51. Mavko, G. and Nur, A., 1997, The effect of a percolation threshold in the Kozeny–
Carman relation, Geophysics 62, 1480–1482.
52. Morse, R. A., P. K. Terwilliger, and S. T. Yuster. 1947, Relative permeability
measurements on small core samples, Oil and Gas J.46, No. 16, 109.
53. Müller, Nadja, 2010, “Supercritical CO2-brine relative permeability experiments in
Reservoir Rocks—Literature Review and Recommendations “, Transport in Porous
Media DOI: 10.1007/s11242-010-9689-2.
54. Muskat, M., 1937, the Flow of Homogenous Fluids through Porous Media, McGraw-
Hill Book Company Inc, New York, London.
55. Nordbotten, J.M., Celia, M.A., Bachu, S., 2005, Injection and storage of CO2 in deep
saline aquifers: analytical solution for CO2 plume evolution during injection,
Transport Porous Media, 58, pp. 339–360.
56. Osaba, J. S., J. G. Richardson, J. K. Kerver, J. A. Hafford, and P. M. Blair. 1951,
139
Laboratory measurements of relative permeability. Trans. AIME 192:99.
57. Oloruntobi, O.S. and LaForce, T., 2009, Effect of aquifer heterogeneity on CO2
sequestration, EUROPEC/EAGE Conference and Exhibition, 8-11 June 2009,
Amsterdam, The Netherlands, SPE 121776.
58. Pape, H., Clauser, C., Iffland, J., 1999, Permeability prediction based on fractal pore-
space geometry, Geophysics, 64 (5), pp. 1447–1460
59. Perrin, J.-C., Krause, M., Kuo, C.-W., Miljkovic, L., Charob, E., Benson, S.M., 2009.
Core-scale experimental study of relative permeability properties of CO2 and brine in
reservoir rocks. Energy Procedia 1, 3515–3522.
60. Perrin, J-C., and Benson, S., 2010, An experimental study on the influence of sub-
core scale heterogeneities on CO2 distribution in reservoir rocks, Transport in Porous
Media.
61. Pickup, G.E., and Stephen, K.D., 2000, An assessment of steady-state scale-up for
small-scale geological models, Petroleum Geosci, 6 (2000), pp. 203–210.
62. Pini, R., S. C. M. Krevor, and S. Benson , 2012, Capillary pressure and heterogeneity
for the CO2/water system in sandstone rocks at reservoir conditions, Adv. Water
Resour., 38, 48–59.
63. Pruess, K. 2005, “ECO2N: A TOUGH2 Fluid property module for mixtures of water,
NaCl, and CO2,” Lawrence Berkeley National Laboratory Report LBNL-57952,
Berkeley, CA.
64. Pruess, K., Oldenburg, C. and Moridis, G, 1999, “TOUGH2 User’s Guide, V2.,”
Lawrence Berkeley National Laboratory Report LBNL-43134, Berkeley, CA,
Advances In Water Resources, 29,397-407.
65. Rossen W. R., van Duijn C. J., 2004, Gravity segregation in steady-state horizontal
flow in homogeneous reservoirs. J Pet Sci Eng 43, 99–111.
66. Sandberg C. R., and L. S. Gournay. 1958, The effect of fluid-flow rate and viscosity
on laboratory determinations of oil-water relative permeabilities, Trans. AIME 213:36
67. Shi, J.-Q., Xue, Z. and Durucan, S., 2010, Supercritical CO2 core flooding and
imbibition in Tako sandstone–Influence of sub-core scale heterogeneity, Int. J.
Greenhouse Gas Control.
68. Silin, D., Patzek, T., Benson, S., 2009, A model of buoyancy-driven two-phase
140
countercurrent fluid flow. Transport in Porous Media, 76, pp. 449–469.
69. Silin D, Tomutsa L, Benson S, Patzek T., 2011, Microtomography and pore-scale
modeling of two-phase fluid distribution. Transp Porous Media;86(2):495–515.
70. Skauge, A., G. Haskjold, T. Thorsen and M. Aarra, 1997, Accuracy of gas-oil relative
permeability from two-phase flow experiments, SCA-9707 paper.
71. Skauge, A., Hydro N., and Poulsen S., 2000, Rate effects on centrifuge drainage
relative permeability, SPE 63145 paper.
72. Virnovsky G.A., Friss H.A. and Lohne A., 2004, A steady-state upscaling approach
for immiscible two-phase flow. Transp. Porous Media 54: 167–192.
73. Welge, HJ., 1952, A simplified method for computing oil recovery by gas or water
drive, Trans. , AIME 195 , 91.
74. Zhang, K., YS Wu, C Ding, K Pruess. 2003, TOUGH2_MP: A parallel version of
TOUGH2. TOUGH symposium 2003.
75. Zhang, K., Doughty, C., Wu, Y.S., and Pruess, K., 2007, Efficient parallel simulation
of CO2 geologic sequestration in saline aquifers, SPE Reservoir Simulation
Symposium, Houston, TX.
76. Zhou, D., Fayers F. and Orr, F.M.J., 1994, Scaling of multiphase flow in simple
heterogeneous porous Media (SPE 27833). SPE Reservoir Eng. 12(3): 173–178.