experimental, numerical, and soft computing-based …

EXPERIMENTAL, NUMERICAL, AND SOFT

COMPUTING-BASED ANALYSIS OF THE VAPEX

PROCESS IN HEAVY OIL SYSTEMS

A Thesis

Submitted to the Faculty of Graduate Studies and Research

In Partial Fulfilment of the Requirements

For the Degree of

Doctor of Philosophy

in

Petroleum Systems Engineering

University of Regina

By

Mehdi Mohammadpoor

Regina, Saskatchewan

July, 2014

Copyright 2014: M. Mohammadpoor

UNIVERSITY OF REGINA

FACULTY OF GRADUATE STUDIES AND RESEARCH

SUPERVISORY AND EXAMINING COMMITTEE

Mehdi Mohammadpoor, candidate for the degree of Doctor of Philosophy in Petroleum Systems Engineering, has presented a thesis titled, Experimental, Numerical, and Soft Computing-Based Analysis of the Vapex Process in Heavy Oil Systems, in an oral examination held on July 11, 2014. The following committee members have found the thesis acceptable in form and content, and that the candidate demonstrated satisfactory knowledge of the subject material. External Examiner: *Dr. Japan Trivedi, University of Alberta

Supervisor: Dr. Farshid Torabi, Petroleum Systems Engineering

Committee Member: Dr. Fanhua Zeng, Petroleum Systems Engineering

Committee Member: Dr. Paitoon Tontiwachwuthikul, P Systems Engineering

Committee Member: **Dr. Ezeddin Shirif, Petroleum Systems Engineering

Committee Member: Dr. Nader Mobed, Department of Physics

Chair of Defense: Dr. Andrei Volodin, Department of Mathematics & Statistics *Via tele-conference **Not present at defense

I

ABSTRACT

There are significant heavy oil and bitumen resources in Canada. Considering increasing

energy demands, these abundant resources are a potential energy source. Regardless,

looking for an economically viable and environmentally friendly heavy oil recovery

technique is essential for exploiting not just these resources, but all future heavy oil

resources.

The problems with highly viscous heavy oil reservoirs—excessive heat loss to the

surrounding formations, low permeability carbonate reservoirs, and the large amount of

CO2 emitted during these thermal processes—introduce economic and environmental

drawbacks for thermal methods. In fact, solvent-based heavy oil recovery methods have

recently gained attention due to the potential environmental and economic advantages

over the thermal processes.

In this research, an extensive experimental investigation was carried out to evaluate the

effect of solvent type and drainage height, as the key parameters of VAPEX in heavy oil

recovery. To accomplish this goal, two large, visual rectangular, sand-packed VAPEX

models with 24.5 cm and 47.5 cm heights were employed to run the experiments using

Plover Lake heavy oil (5650mPa.s) with a low permeability (6~9 D) sand pack. Propane,

methane, CO2, butane, propane/CO2 mixture, and propane/methane mixture were

considered as respective solvents for the experiments. Various parameters were

monitored and recorded during the course of experiments.

Moreover, separate experiments were carried out at the end of each VAPEX experiment

to measure the asphaltene precipitation at different locations of the VAPEX models. To

II

observe the drainage height effect in more detail, a comprehensive image analysis was

completed during the solvent chamber evolution.

As a result, it was determined that drainage height has a significant impact on production

rate and heavy oil recovery. The results prove the complexity of the effect of drainage

height and the up-scaling issues with the VAPEX process. Furthermore, in terms of

solvents, propane showed the best recovery performance due to its favourable low vapour

pressure and high solubility. Ultimately, promising recovery performance after

introducing CO2 and methane as the carrier gases was observed.

Separate experiments were conducted to obtain adequate PVT data for the heavy oil and

solvents used in this study. A numerical simulation study was carried out to match

experimental results and investigate the effect of well spacing, permeability, and

diffusivity on the VAPEX process.

Finally, the data gathered from the experiments were combined with available data in the

literature and a soft computing approach was utilized to develop a model that predicts the

recovery performance of the VAPEX process. Several experimental studies together with

various analytical models have been proposed to simulate and describe the performance

of the VAPEX process. However, due to the complexity of the mechanisms associated

with the solvent injection process (i.e., diffusion and gravity drainage processes), such

models are incapable of accurately predicting the production rate during the VAPEX

process. In this research, artificial neural networks (ANN) technique was utilized to

tackle the limitations that analytical methods encounter where there is uncertainty, and

imprecision.

III

ACKNOWLEDGMENTS

I would like to express my profound gratitude and appreciation to my supervisor, Dr.

Farshid Torabi, for his guidance, encouragement, suggestions, and support throughout the

course of this project. I learnt a lot from his great personality as well as his scientific

knowledge, creativity, and experience. I feel privileged to have had Dr. Torabi as my

supervisor during these years of study.

I am also grateful to Dr. Mobed, Dr. Tontiwachwuthikul, Dr. Zeng, and Dr. Shirif for

serving as members of my examination committee, and for their constructive suggestions.

In addition, I would like to acknowledge the financial support from the University of

Regina Faculty of Graduate Studies and Research (FGSR) and Natural Sciences and

Engineering Research Council (NSERC) Canada, and also thank Dr. Shirif for providing

lab space.

I am also very grateful to my friend, Mr. Abbasali Dehghan Tazerjani, for his help to

prepare and program the image analysis software. I would like to thank Mr. Ali Abedini

for his help and technical discussions.

Last, but not least, a heartfelt thank you to Asal, my wonderful wife, and to my family for

their patience and unrelenting support during this study.

IV

DEDICATION

This dissertation is dedicated to my beloved wife, Asal, my dearest parents, Elyas and

Kobra, and my dear brother and sisters.

V

TABLE OF CONTENTS

ABSTRACT ......................................................................................................................... I

ACKNOWLEDGMENTS ................................................................................................ III

DEDICATION .................................................................................................................. IV

LIST OF TABLES ........................................................................................................... XII

LIST OF FIGURES ....................................................................................................... XVI

NOMENCLATURE ................................................................................................... XXVI

CHAPTER 1: INTRODUCTION ....................................................................................... 1

1.1 Background ......................................................................................................... 1

1.2 Vapour extraction (VAPEX)............................................................................... 5

1.3 Objectives ........................................................................................................... 8

1.4 Organization of the thesis ................................................................................... 9

CHAPTER 2: LITERATURE REVIEW .......................................................................... 11

2.1 Heavy oil recovery methods ............................................................................. 11

2.1.1 Waterflooding ............................................................................................... 11

2.1.2 Cold heavy oil production (CHOPS) ............................................................ 12

2.1.3 Gas EOR methods ......................................................................................... 14

2.1.4 Thermal EOR processes ................................................................................ 14

2.1.5 Chemical EOR processes .............................................................................. 16

VI

2.1.6 Emerging EOR technologies......................................................................... 17

2.2 Vapour extraction (VAPEX)............................................................................. 19

2.2.1 Solvent requirement ...................................................................................... 22

2.3 VAPEX mechanism .......................................................................................... 24

2.3.1 Molecular diffusion ....................................................................................... 27

2.3.2 Physical dispersion........................................................................................ 33

2.4 Asphaltene precipitation ................................................................................... 35

2.4.1 Asphaltene precipitation in VAPEX ............................................................. 39

2.5 Economic and environmental advantages ......................................................... 42

CHAPTER 3: EXPERIMENTAL SETUP, MATERIALS, AND PROCEDURE ........... 45

3.1 Experimental setup............................................................................................ 45

3.1.1 Solvent injection unit .................................................................................... 45

3.1.2 Physical models ............................................................................................ 49

3.1.3 Solvent and liquid production unit ................................................................ 57

3.1.4 Data acquisition unit ..................................................................................... 60

3.2 Materials ........................................................................................................... 65

3.2.1 Sand............................................................................................................... 65

3.2.2 Heavy oil ....................................................................................................... 65

3.2.3 Injection solvents and back pressure gas ...................................................... 67

3.3 Experimental procedure .................................................................................... 70

VII

3.3.1 Preparation .................................................................................................... 70

3.3.1.1 Sand packing........................................................................................ 70

3.3.1.2 Porosity measurement.......................................................................... 71

3.3.1.3 Oil saturation ....................................................................................... 73

3.3.1.4 Permeability measurement .................................................................. 76

3.3.2 VAPEX experiments ..................................................................................... 77

3.3.3 Residual oil saturation and asphaltene content measurement ....................... 79

3.3.3.1 Residual oil saturation measurement ................................................... 81

3.3.3.2 Asphaltene content measurement ........................................................ 81

3.3.4 Cleaning ........................................................................................................ 84

CHAPTER 4: EXPERIMENTAL RESULTS AND DISCUSSION ................................ 85

4.1 VAPEX performance ........................................................................................ 88

4.1.1 Effect of drainage height ............................................................................... 88

4.1.1.1 Recovery factor and produced oil rate ................................................. 88

4.1.1.1.1 Propane injection ............................................................................. 88

4.1.1.1.2 Methane injection ............................................................................ 91

4.1.1.1.3 CO2 injection ................................................................................... 94

4.1.1.1.4 Butane injection............................................................................... 94

4.1.1.1.5 Propane/CO2 injection ..................................................................... 99

4.1.1.1.6 Propane/methane injection ............................................................ 102

4.1.1.2 Solvent utilization factor (SUF) ........................................................ 105

4.1.1.2.1 Propane injection ........................................................................... 105

VIII

4.1.1.2.2 Methane injection .......................................................................... 105

4.1.1.2.3 CO2 injection ................................................................................. 108

4.1.1.2.4 Butane injection............................................................................. 108

4.1.1.2.5 Propane/ CO2 injection .................................................................. 108

4.1.1.2.6 Propane/methane injection ............................................................ 112

4.1.1.3 Viscosity, density, molecular weight, and hydrocarbon components for

the produced oil................................................................................................... 112

4.1.1.3.1 Propane injection ........................................................................... 112

4.1.1.3.2 Methane injection .......................................................................... 117

4.1.1.3.3 CO2 injection ................................................................................. 121

4.1.1.3.4 Butaneinjection.............................................................................. 121

4.1.1.3.5 Propane/CO2 injection ................................................................... 128

4.1.1.3.6 Propane/methaneinjection ............................................................. 128

4.1.2 Effect of solvent type .................................................................................. 135

4.1.2.1 Small model ....................................................................................... 135

4.1.2.1.1 Recovery factor and produced oil rate .......................................... 135

4.1.2.1.2 Solvent utilization factor (SUF) .................................................... 138

4.1.2.1.3 Viscosity, density, molecular weight, and hydrocarbon components

for the produced oil ......................................................................................... 138

4.1.2.2 Large model ....................................................................................... 142

4.1.2.2.1 Recovery factor and produced oil rate .......................................... 142

4.1.2.2.2 Solvent utilization factor (SUF) .................................................... 145

IX

4.1.2.2.3 Viscosity, density, molecular weight, and hydrocarbon components

for the produced oil ......................................................................................... 145

4.2 Residual oil saturation..................................................................................... 149

4.3 Asphaltene precipitation ................................................................................. 152

4.3.1 Effect of drainage height ............................................................................. 155

4.3.1.1 Propane injection ............................................................................... 155

4.3.1.2 Methane injection .............................................................................. 157

4.3.1.3 CO2 injection ..................................................................................... 157

4.3.1.4 Butane injection ................................................................................. 157

4.3.1.5 Propane/CO2 injection ....................................................................... 161

4.3.1.6 Propane/methane injection ................................................................ 161

4.3.2 Effect of solvent type .................................................................................. 164

4.3.2.1 Small model ....................................................................................... 164

4.3.2.2 Large model ....................................................................................... 166

4.4 Image analysis (IA) ......................................................................................... 169

4.5 Effect of injection-production wells connection ............................................. 179

4.5.1 Small model ................................................................................................ 179

4.5.2 Large model ................................................................................................ 187

4.6 Scale-up: ......................................................................................................... 194

4.7 Dimensionless VAPEX number, Ns calculation: ............................................ 204

CHAPTER 5: PVT STUDIES AND NUMERICAL SIMULATION ........................... 206

X

5.1 Viscosity and density measurement ................................................................ 206

5.2 Vapour pressure .............................................................................................. 206

5.3 Solubility measurement .................................................................................. 210

5.4 Solvent volume fraction in heavy oil for VAPEX tests .................................. 213

5.5 Numerical simulation ...................................................................................... 216

5.5.1 Model construction ..................................................................................... 216

5.5.2 Injection and production wells’ constraints ................................................ 221

5.5.3 History matching ......................................................................................... 221

5.5.4 Effect of well configurations ...................................................................... 230

5.5.5 Effect of permeability ................................................................................. 232

5.5.6 Effect of grid thickness ............................................................................... 232

5.5.7 Effect of time step ....................................................................................... 232

CHAPTER 6: SOFT COMPUTING APPROACH ........................................................ 237

6.1 Data handling procedures ............................................................................... 239

6.1.1 Data acquisition .......................................................................................... 239

6.1.2 Data normalization ...................................................................................... 240

6.2 Neural network development .......................................................................... 243

6.3 Sensitivity analysis.......................................................................................... 252

6.4 Comparison of results ..................................................................................... 256

CHAPTER 7: CONCLUSIONS AND RECOMMENDATIONS .................................. 264

XI

7.1 Conclusions ..................................................................................................... 264

7.2 Recommendations ........................................................................................... 268

REFERENCES ............................................................................................................... 269

Appendix A ..................................................................................................................... 291

Appendix B ..................................................................................................................... 301

Appendix C ..................................................................................................................... 310

XII

LIST OF TABLES

Table 2-1: Standard methods for asphaltene precipitation measurement (after Speight,

2004) ................................................................................................................................. 38

Table 3-1: DFM Specifications ......................................................................................... 48

Table 3-2: Physical models dimensions ............................................................................ 50

Table 3-3: List of experimental equipment ....................................................................... 64

Table 3-4: Compositional analysis result of the injection heavy oil with viscosity of 5650

mPa.s at 21 °C ................................................................................................................... 68

Table 4-1: Operating conditions of the VAPEX experiments .......................................... 87

Table 4-2: Compositional analysis result of the produced heavy oil after propane injection

in small model ................................................................................................................. 114

Table 4-3: Compositional analysis result of the produced heavy oil after propane injection

in large model ................................................................................................................. 115

Table 4-4: Compositional analysis result of the produced heavy oil after methane

injection in small model .................................................................................................. 118

Table 4-5: Compositional analysis result of the produced heavy oil after methane

injection in large model .................................................................................................. 119

Table 4-6: Compositional analysis result of the produced heavy oil after CO2 injection in

small model ..................................................................................................................... 122

Table 4-7: Compositional analysis result of the produced heavy oil after CO2 injection in

large model...................................................................................................................... 123

XIII

Table 4-8: Compositional analysis result of the produced heavy oil after butane injection

in small model ................................................................................................................. 125

Table 4-9: Compositional analysis result of the produced heavy oil after butane injection

in large model ................................................................................................................. 126

Table 4-10: Compositional analysis result of the produced heavy oil after propane/ CO2

injection in small model .................................................................................................. 129

Table 4-11: Compositional analysis result of the produced heavy oil after propane/ CO2

injection in large model .................................................................................................. 130

Table 4-12: Compositional analysis result of the produced heavy oil after propane/

methane injection in small model ................................................................................... 132

Table 4-13: Compositional analysis result of the produced heavy oil after

propane/methane injection in large model ...................................................................... 133

Table 4-14: Produced oil properties for the small model ............................................... 140

Table 4-15: Produced oil properties for the large model ................................................ 147

Table 4-16: Produced oil properties ................................................................................ 185

Table 4-17: Produced oil properties ................................................................................ 192

Table 5-1: Vapor pressure of solvents used in this study at 21 °C ................................. 208

Table 5-2: Properties of small simulation model ............................................................ 217

Table 5-3: Properties of large simulation model ............................................................. 218

Table 6-1: Data range for various input and output parameters used in this study ........ 242

Table 6-2: Summary of the results for some selected training and testing trials ............ 249

Table 6-3: Error analysis for various techniques to predict drainage rate ...................... 263

Table A-1: Production method versus heavy oil resource (1) (after Clark, 2007) ......... 291

XIV




Table A-5: Technology versus production method (1) (after Clark, 2007) .................... 295





Table A-10: Technology versus production method (6) (after Clark, 2007) .................. 300

Table B-1: The experimental data on VAPEX experiments conducted by different

researchers....................................................................................................................... 301


researchers (Cont'd) ........................................................................................................ 302











XV





XVI

LIST OF FIGURES

Figure ‎1-1: Canadian heavy oil deposits (from Canadian Association of Petroleum

Producers) ........................................................................................................................... 2

Figure ‎1-2: (a) Total Canadian crude oil and production (Canada’s National Energy

Board), (b) Oil production from EOR methods in US (reprinted after ASPO-USA) ......... 3

Figure ‎1-3: (a) VAPEX in typical layout of heavy oil reservoir, (b) Concept of VAPEX

(after Upreti et al., 2007)..................................................................................................... 7

Figure ‎2-1: Mechanisms of VAPEX process .................................................................... 25

Figure ‎2-3: Yield of asphaltene precipitation for various hydrocarbon solvents (after

Speight, 2007) ................................................................................................................... 37

Figure ‎2-4: Effect of asphaltene content on heavy oil viscosity (after Luo and Gu, 2005)

........................................................................................................................................... 41

Figure‎3-1: Digital flow meter (DFM) ............................................................................... 47

Figure ‎3-2: Plexiglas slabs, (a) Large model slab (b) Small model slab .......................... 51

Figure ‎3-3: Gaskets, (a) Large model gasket, (b) Small model gasket ............................. 52

Figure ‎3-4: Steel cover protectors, (a) Large model, (b) Small model ............................. 53

Figure ‎3-5: Physical models assembled on a steel frame mounted on steel stand ............ 54

Figure ‎3-6: The schematic of the large physical model and its sand pack cavity ............. 55

Figure: ‎3-7: The schematic of the small physical model and its sand pack cavity ........... 56

Figure ‎3-8: High pressure back-pressure regulator (BPR) ............................................... 58

Figure ‎3-9:Two-phase separators ...................................................................................... 59

XVII

Figure ‎3-10: Wet test meters (WTM) ............................................................................... 61

Figure ‎3-11: Schematic diagram of the experimental set-up ............................................ 62

Figure ‎3-12: Experimental setup ....................................................................................... 63

Figure ‎3-13: Screen analysis for Ottawa sand #530 ......................................................... 66

Figure ‎3-14: Hydrocarbon composition of injected oil..................................................... 69

Figure ‎3-15: Sand-packed VAPEX models ...................................................................... 72

Figure ‎3-16: The schematic of the oil saturation set-up ................................................... 74

Figure ‎3-17: Oil saturated sand packs ............................................................................... 75

Figure ‎3-18: Sample locations, (a) Small model, (b) Large model .................................. 80

Figure ‎3-19: Schematic of the set up used to separate the oil from the sand .................... 82

Figure ‎3-20: Schematic of the set up used to measure the asphaltene content of the oil

samples .............................................................................................................................. 83

Figure ‎4-1: The recovery factor after propane injection in VAPEX models .................... 89

Figure ‎4-2: The produced oil rate after propane injection in VAPEX models ................. 90

Figure ‎4-3: The recovery factor after methane injection in VAPEX models ................... 92

Figure ‎4-4: The produced oil rate after methane injection in VAPEX models ................ 93

Figure ‎4-5: Recovery factor after CO2 injection in the VAPEX models ......................... 95

Figure ‎4-6: Produced oil rate after CO2 injection in the VAPEX models ........................ 96

Figure ‎4-7: Recovery factor after butane injection in the VAPEX models ...................... 97

Figure ‎4-8: Produced oil rate after butane injection in the VAPEX models .................... 98

Figure ‎4-9: Recovery factor after propane/CO2mixture injection in the VAPEX models

......................................................................................................................................... 100

Figure ‎4-10: Produced oil rate after first propane/CO2 injection in VAPEX models ..... 101

XVIII

Figure ‎4-11: Recovery factor after propane/methane mixture injection in the VAPEX

models ............................................................................................................................. 103

Figure ‎4-12: Produced oil rate after propane/methane mixture injection in the VAPEX

models ............................................................................................................................. 104

Figure ‎4-13: Solvent utilization factor (SUF) after propane injection in VAPEX models

......................................................................................................................................... 106

Figure ‎4-14: Solvent utilization factor (SUF) after methane injection in VAPEX models

......................................................................................................................................... 107

Figure ‎4-15: Solvent utilization factor (SUF) after CO2 injection in VAPEX models ... 109

Figure ‎4-16: Solvent utilization factor (SUF) after butane injection in VAPEX models 110

Figure ‎4-17: Solvent utilization factor (SUF) after propane/CO2 injection in VAPEX

models ............................................................................................................................. 111

Figure ‎4-18: Solvent utilization factor (SUF) after propane/methane injection in VAPEX

models ............................................................................................................................. 113

Figure ‎4-19: Compositional analysis of the produced oil after propane injection .......... 116

Figure ‎4-20: Compositional analysis of the produced oil after methane injection ......... 120

Figure ‎4-21: Compositional analysis of the produced oil after CO2 injection ................ 124

Figure ‎4-22: Compositional analysis of the produced oil after butane injection ............ 127

Figure ‎4-23: Compositional analysis of the produced oil after propane/CO2 injection .. 131

Figure ‎4-24: Compositional analysis of the produced oil after propane/methane injection

......................................................................................................................................... 134

Figure ‎4-25: Effect of the solvent type on recovery factor in small model .................... 136

Figure ‎4-26: Effect of the solvent type on produced oil rate in small model ................. 137

XIX

Figure ‎4-27: Effect of solvent type on solvent utilization factor (SUF) for small model139

Figure ‎4-28: Effect of solvent type on hydrocarbon components in small model .......... 141

Figure ‎4-29: Effect of the solvent type on recovery factor in large model ..................... 143

Figure ‎4-30: Effect of the solvent type on produced oil rate in large model .................. 144

Figure ‎4-31: Effect of solvent type on the solvent utilization factor (SUF) for large model

......................................................................................................................................... 146

Figure ‎4-32: Effect of solvent type on hydrocarbon components in large model .......... 148

Figure ‎4-33: Residual oil saturation profile for various solvents in the small model ..... 150

Figure ‎4-34: Residual oil saturation profile for various solvents in the large model ..... 151

Figure ‎4-35: Asphaltene precipitate after conducting the asphaltene measurement tests

......................................................................................................................................... 153

Figure ‎4-36: Schematic of the locations of each heavy oil samples in the physical models

......................................................................................................................................... 154

Figure ‎4-37: Effect of drainage height on asphaltene precipitation at different locations in

the small and large models after propane injection ........................................................ 156


the small and large models after methane injection ........................................................ 158


the small and large models after CO2 injection .............................................................. 159


the small and large models after butane injection ........................................................... 160


small and large models after propane/CO2 injection ...................................................... 162

XX


small and large models after propane/methane injection ................................................ 163

Figure ‎4-43: Effect of solvent type on asphaltene precipitation at different locations in the

small model ..................................................................................................................... 165

Figure ‎4-44: Effect of solvent type on asphaltene precipitation at different locations in the

large model...................................................................................................................... 167

Figure ‎4-45: (a) Asphaltene precipitation close to the injection point, (b) Asphaltene

streaks on the sand pack at the end of experiments ........................................................ 168

Figure ‎4-46: The interface of the coded software for IA ................................................ 170

Figure ‎4-47: The procedure for conducting IA in the small model: (a) The coordinates of

the image are specified, (b) The interface curve is defined, and (c) The oil and solvent

zones are schematically reprinted by the software ......................................................... 171

Figure ‎4-48: The procedure for conducting IA in the large model: (a) The coordinates of

the image are specified, (b) The interface curve is defined, and (c) The oil and solvent

zones are schematically reprinted by the software ......................................................... 172

Figure ‎4-49: Solvent chamber evolution in small model after propane injection........... 174

Figure ‎4-50: Solvent chamber evolution in large model after propane injection ........... 175

Figure ‎4-51: Sweep efficiency of various solvents in the small model .......................... 177

Figure ‎4-52: Sweep efficiency of various solvents in the large model ........................... 178

Figure ‎4-53: Effect of connection establishment between the injection and production

wells on the recovery factor in the small model ............................................................. 181


wells on the produced oil rate in the small model .......................................................... 182

XXI


wells on the asphaltene precipitation in the small model ............................................... 183

Figure ‎4-56: Solvent chamber evolution in small model after propane injection (first

injection scenario) ........................................................................................................... 186


wells on the recovery factor in the large model .............................................................. 188


wells on the produced oil rate in the large model ........................................................... 189


wells on the asphaltene precipitation in the large model ................................................ 191

Figure ‎4-60: Solvent chamber evolution in large model after propane injection (first

injection scenario) ........................................................................................................... 193

Figure ‎4-61: The results obtained for up-scaling the stabilized drainage rate based on the

proposed exponent by Butler (1994), (n=0.5). The dotted line is the drainage rate

prediction based on Butler’s model; the data points for different solvents are the

experimental results obtained in this study. .................................................................... 199


proposed exponent by Yazdani (2007), (n=1.1). The dotted line is the drainage rate

predicted based on Yazdani’s model; the data points for different solvents are the



proposed exponent by Yazdani (2007), (n=1.3). The dotted line is the drainage rate

XXII

predicted based on Yazdani’s model; the data points for different solvents are the


Figure ‎4-64: The results obtained for up-scaling the stabilized drainage rate. The dotted

line is the drainage rate predicted based on n=1.2; the data points for different solvents

are the experimental results obtained in this study. ........................................................ 202

Figure ‎4-65: Linear regression for the results obtained for different solvents in the small

and large physical models ............................................................................................... 203

Figure ‎4-66: Effect of drainage height and solvent type on dimensionless VAPEX

number, Ns ....................................................................................................................... 205

Figure ‎5-1: Densities and viscosities of the heavy oil used in this study at various

temperatures and atmospheric pressure .......................................................................... 207

Figure ‎5-2: Two-phase envelopes for propane/CO2 and propane/methane mixtures ..... 209

Figure ‎5-3: Schematic of the experimental set-up used for solubility measurement tests

......................................................................................................................................... 211

Figure ‎5-4: Solubility of (a) propane, (b) methane, (c) CO2, and (d) butane at 21°C .... 212

Figure ‎5-5: Solvent volume fraction in the produced oil from the small model for various

solvents at 21°C .............................................................................................................. 214

Figure ‎5-6: Solvent volume fraction in the produced oil from the large model for various

solvents at 21°C .............................................................................................................. 215

Figure ‎5-7: (a) 2D view of the simulated model with the injection and production wells

for the small physical model, (b) 3D view of the simulated model with the injection and

production wells for the small physical model ............................................................... 219

XXIII

Figure ‎5-8: (a) 2D view of the simulated model with the injection and production wells

for the large physical model, (b) 3D view of the simulated model with the injection and

production wells for the large physical model ................................................................ 220

Figure ‎5-9: Experimental and simulation results for the recovery factor after injecting

propane in the small model ............................................................................................. 224


propane in the large model .............................................................................................. 225


butane in the small model ............................................................................................... 226


(a)CO2 and (b) methane in the small model.................................................................... 227


(a) propane/CO2 and (b) propane/methane mixtures in the small model ........................ 228

Figure ‎5-14: Chamber evolution after 26 h in (a) simulated small model, (b) laboratory

model............................................................................................................................... 229

Figure ‎5-15: Effect of well configuration on the recovery factor. For the first well

configuration, the injection well is located at the top of the model and 24 cm above the

production well; for the second well configuration, the injection well is located 16 cm

above the production well; for the third well configuration, the injection well is 4 cm

above the production well; for the forth well configuration, the injection well is 24 cm

above the production well; and for the fifth well configuration, the injection well is at the

right top corner of the model; the production well is at the left bottom corner. ............. 231

Figure ‎5-16: Effect of permeability on the recovery factor after injecting propane ....... 234

XXIV

Figure ‎5-17: Effect of grid thickness on the recovery factor after injecting propane ..... 235

Figure ‎5-18: Effect of time step change on recovery factor after injecting propane ...... 236

Figure ‎6-1: Schematic of an artificial neural network .................................................... 238

Figure ‎6-2: Data distribution for training and testing sets; stabilized drainage rate vs. (a)

height (cm), (b) injection pressure (kPa), (c) porosity (%), (d) permeability (D), and (e)

viscosity (mPa.s) ............................................................................................................. 241

Figure ‎6-3: An example of network training procedure; plot of: (a) predicted outputs by

network for training data sets, (b) predicted outputs by network for validation data sets,

(c) predicted outputs by network for testing data sets, and (d) predicted outputs by

network for the whole group of data sets chosen for training procedure ........................ 247

Figure ‎6-4: Schematic of the developed BP network; there are 20 neurons on the first

hidden layer and 15 neurons on the second hidden layer. The transfer functions used for

hidden layers were log sigmoid functions, and linear transfer function was used for output

layer................................................................................................................................. 250

Figure ‎6-5: Output of the developed network vs. the actual data after simulating the

model with training data sets .......................................................................................... 251

Figure ‎6-6: Output of the developed network vs. the actual data after simulating the

model with testing data sets ............................................................................................ 253

Figure ‎6-7: Relevancy (r) factor for various parameters to the production rate ............. 255

Figure ‎6-8: Plot of predicted stabilized drainage rate by eq. 6.15 versus actual data sets

for testing ........................................................................................................................ 257


for testing ........................................................................................................................ 258

XXV


for testing ........................................................................................................................ 259


for testing ........................................................................................................................ 260


for testing ........................................................................................................................ 261


for testing ........................................................................................................................ 262

XXVI

NOMENCLATURE

Symbols Definitions

A Specific pore surface area, L2

Cmax Maximum solvent concentration

Cmin Minimum solvent concentration

Cp Coefficient of variance

Cs Solvent concentration

D Dispersion coefficient, L2t-1

Deff Effective diffusivity, L2t-1

Do Molecular diffusivity, L2t-1

Dp Particle size, L

Ds Diffusivity of solvent in bitumen, L2t-1

F Formation electrical diffusivity, L2t-1

H Drainage or model height, L

M Molecular weight of solvent, [g/mol]

Ns VAPEX number

P Pressure, ML-1

t-2

Pf Final Pressure, ML-1

t-2

Q Stabilized drainage rate per unit length of the horizontal well, L2t-1

R Universal gas constant, ML2t-2

N-1

T-1

T Temperature, T

XXVII

V Molal volume of solvent, L3N

-1M

-1

VA Molar volume of solute, L3N

-1

VB Molar volume of solvent, L3N

-1

Z Compressibility factor

d Diameter, L

g Gravitational acceleration, L2t-2

k Permeability, L2

m Mass, M

m1 Original mass of the heavy oil sample, M

m2 Mass of the dried particulate, M

n Moles of solvent, N

t Time, t

x Effective molecular weight of the solvent with respect to the diffusion

process

Greek symbols

γ Skewness

λ Mass transfer enhancement coefficient

μ Viscosity, ML-1

t-1

μmix Viscosity of mixture at solvent concentration, ML-1

t-1

τ Tortuosity

σ Surface tension, Mt-2

Porosity

Δρ Density difference between solvent and bitumen, ML-3

ΔSO Change in oil saturation

XXVIII

Abbreviations

ANN Artificial neural networks

APAD Average percent arithmetic deviation

BPR Back pressure regulator

CAT Computer assisted tomography

CERI Canadian Energy Research Institute

CHOPS Cold heavy oil production

CSS Cyclic steam stimulation

DFM Digital flowmeters

EOR Enhanced oil recovery

GOR Gas oil ratio

IA Image Analysis

MEOR Microbial enhanced oil recovery

MSE Mean square error

MW Molecular weight

NMR Nuclear magnetic resonance

PSD Particle Size Distribution

SAGD Steam assisted gravity drainage

SOR Solvent oil ratio

SRC Saskatchewan Research Council

SUF Solvent utilization factor

XXIX

VAPEX Vapour extraction

WTM Wet test meter

1

1. CHAPTER 1: INTRODUCTION

1.1 Background

Canada has significant crude oil resources, 50% of which are heavy oil and bitumen. The

application of numerous heavy oil recovery techniques has led to the recovery of small

portions of this oil. However, in many cases, more than 90% of the oil remains in place.

Improved profitability, technological advances, huge reserve size, low geological risk,

and low capital investment have drawn attention to heavy oil production from many

companies (Chugh et al., 2000). Figure 1-1 shows the distribution of Canadian heavy oil

reserves.

Increasing the capillary number and/or lowering the mobility ratio are the basic principles

of enhanced oil recovery (EOR) methods. EOR processes are mainly divided into four

categories: thermal, gas, chemical, and other. In addition, oil production from EOR

projects continues to supply an increasing percentage of the world’s oil. About 3% of the

worldwide production now comes from EOR processes, and this portion is increasing

each year. Figure 1-2(a) shows the amount of Canadian oil to be produced based on the

approved projects by major oil companies, while Figure 1-2(b) shows the oil production

in the U.S. using various EOR techniques. Based on these figures, the importance of

choosing the most feasible recovery technique is increasingly important to petroleum

engineers.

2

Figure ‎1-1: Canadian heavy oil deposits (from Canadian Association of Petroleum Producers)

3

(a)

Date (year)

1975 1980 1985 1990 1995 2000 2005 2010 2015

Enh

ance

d pr

oduc

tion

(bb

l/da

y)

0

200x103

400x103

600x103

800x103

1x106

Chemical

Gas injection

Thermal

Total EOR

(b)

Figure ‎1-2: (a) Total Canadian crude oil and production (Canada’s National Energy Board), (b) Oil

production from EOR methods in US (reprinted after ASPO-USA)

4

While the choice of injectants has widened considerably, petroleum engineers still must

choose an injection fluid and/or feasible recovery process to maximize the recovered oil

from the reservoir. Not surprising, screening criteria have evolved through the years to

help petroleum engineers make appropriate decisions. However, in recent years,

computer technology has improved the application of screening criteria through the use

of artificial intelligence techniques; yet, the reliability of such programs depends on input

data quantity and accuracy.

The continuing rise in demand, the decline in conventional domestic production, and the

belated development of alternatives to petroleum combine to increase the importance of

seeking new resources and methods for enhanced oil recovery. Enhanced oil recovery

could offset some of this dependence; though, the amount, cost, and timing of the EOR

contribution are highly uncertain.

Furthermore, selecting and implementing an EOR method requires several steps. Initially,

reservoir properties and formation fluid characteristics are used as a preliminary technical

screening guide for any possible EOR method. After the selection of candidate methods,

basic static tests are carried out. Then, more practical methods will be chosen and

subjected to flow studies in porous media where a semi-realistic environment is

introduced. Next, pilot projects demonstrate the viability of the selected method. Finally,

assuming success at the lower screening levels, a field-wide EOR project is implemented.

Of course, economic studies are conducted throughout all screening levels (Goodlet et al.,

1986).

5

1.2 Vapour extraction (VAPEX)

The vapour extraction (VAPEX) or vapour assisted petroleum extraction process is the

solvent analog of the steam-assisted gravity drainage process (SAGD), which reduces oil

viscosity by diluting the in-situ heavy oil or bitumen with the help of injected vapourized

solvents. The idea of injecting the solvent vapours to enhance oil recovery was first

proposed by Allen (1974) (Allen 1976;Chatzis and James, 2007).

Later, Butler and Mokrys (1989) introduced a brief discussion on the process during a

study of Athabasca and Suncor Coker feed bitumen. This process was later named

VAPEX in 1990 (Butler and Mokrys, 1990). In this process, the horizontal production

well is located near the bottom of the pay zone and the injection well is completed right

above the production well. A schematic of the process is shown in Figure 1-3.

First, the solvent is injected through the injector to form an initial vertical solvent vapour

chamber between the injector and the producer. The vapour chamber spreads, and the oil-

solvent interface becomes stabilized by gravity. The drainage is controlled by molecular

diffusion of solvent vapour into the bitumen (Butler and Mokrys, 1990). There is a phase

change during VAPEX when the solvent diffuses into the oil at the solvent-oil interface.

During this phase change, changes to temperature, pressure, and concentration occur at

the contact interface. This causes asphaltene deposition. The asphaltene precipitation has

a contentious effect on the VAPEX process. While the asphaltene precipitation may

improve the quality of heavy oil by reducing its viscosity, it may also alter the wettability

of rock and consequently plug pores.

6

Since the VAPEX process is a non-thermal process, compared to SAGD, it might be

considered as an energy efficient and environmentally friendly process. This is because it

does not require steam generation and water recycling. Therefore, it is significantly fuel

efficient. Also, because this is a non-thermal process, there will be no CO2 emissions,

and, in the case of the CO2-based VAPEX process, it will help sequester CO2

underground.

7

Figure ‎1-3: (a) VAPEX in typical layout of heavy oil reservoir, (b) Concept of VAPEX (after Upreti et al.,

2007)

(a)

8

1.3 Objectives

The main issue for implementing the VAPEX technique is a lack of knowledge about

upscale performance estimates. Other researchers have proven that current analytical

models under predict the VAPEX produced oil rate. Such assumptions, and the limited

operating conditions under which these analytical models are developed, make them

unsuitable to predict recovery performance (in most cases). Conversely, because of the

complex effect of diffusion and dispersion mechanisms on the process, specifically in

porous media like sand-packed models, there are still issues in upscaling the laboratory

results to the field scale.

This research focused on providing an extensive study of VAPEX process performance

by considering the injection of different solvents in large-scale physical models and by

combining the experimental results, numerical analysis, and soft computing tools to

model the production of heavy oil through the VAPEX process. Ultimately, the goal of

this study is to develop a tool that can accurately predict the production rate and

performance of the VAPEX process when applied to a field scale. More specifically, the

following tasks will be carried out:

1. Build and design an experimental set-up with two large visual physical models to

conduct VAPEX experiments.

2. Conduct several tests using different solvents in two different models with

different sizes to produce a wide range of data.

3. Investigate the effect of model size and solvent type on recovery performance.

4. Monitor asphaltene precipitation at different physical model locations using

different solvents.

9

5. Match the experimental results by numerically simulating the process and

comparing the performance of the simulated model with the experimental results,

thereby highlighting simulation issues.

6. Gather, categorize, and pre-process the obtained data to develop a soft-

computing-based model.

7. Train, validate, and test a soft-computing-based model to predict the recovery

performance after implementing the VAPEX technique.

1.4 Organization of the thesis

Chapter 1 provides background on heavy oil resources, heavy oil recovery methods, and

the VAPEX process. It also includes an introduction to the research objectives and

structure of the dissertation.

Chapter 2 includes an extensive literature review on the heavy oil recovery techniques

and VAPEX mechanisms. Moreover, it contains a complete literature review on the

feasibility of different experimental studies on VAPEX, which includes solvent selection,

diffusion and dispersion during VAPEX, asphaltene precipitation and environmental as

well as economic considerations for VAPEX.

Chapter 3 provides a detailed explanation of the experimental set-up and the equipment

and materials used. In addition, the experimental procedure is discussed fully.

Chapter 4 describes the experimental results. These results are analysed and discussed in

detail. Next, the effect of model size and solvent type is investigated. Furthermore, the

asphaltene precipitation experiments are shown and the results are explained for each

solvent.

10

Chapter 5 presents the results for the PVT experiments and measurements. Moreover,

these results were incorporated into a compositional simulator (i.e., CMG’s STARS

package (Computer Modelling Group Ltd., Inc.) to simulate the VAPEX process and

match the experiments.

Chapter 6 describes the soft computing approach utilized in this study to predict the

recovery performance after implementing VAPEX technique. The experimental results

alongside a wide range of data gathered from the literature were employed, and an

artificial neural network (ANN) was utilized to develop a model to predict the oil

drainage rate after conducting VAPEX. Furthermore, the validity of the developed model

was tested by comparing the results with available prediction techniques.

Finally, Chapter 7 summarizes the experimental, numerical, and soft computing results

for the VAPEX tests that have been conducted. This chapter includes the highlighted

results as conclusions. In the second section of this chapter, the recommendations for

future work are explained in detail.

11

2. CHAPTER 2: LITERATURE REVIEW

2.1 Heavy oil recovery methods

For the first EOR method to implement in a specific reservoir, the candidate reservoir and

the recovery mechanism of the EOR method should be studied in detail. In terms of

Canadian reservoirs, the first methods implemented are waterflooding, cold production,

or steam flooding. While chemical flooding and other emerging new technologies are

mostly coupled with the above mentioned methods, implementing these technologies is

highly dependent on economic profitability.

2.1.1 Waterflooding

For nearly 50 years, heavy oil waterfloods have operated in Saskatchewan and Alberta. If

a waterflood is located in an area near heavy oil cold production, it is often classified as

heavy oil waterflooding. When defining heavy oil, the emphasis is mostly on oil gravity;

however, there is another important controlling parameter: oil viscosity. In fact, problems

obtaining consistent heavy oil viscosity measurements and confusion about whether

available viscosity values were collected using dead oil, live oil, or something in between

means that researchers often pay less attention to viscosity terms (Miller, 1995, Miller,

2005 and Miller, 2006).

Furthermore, Forth et al. (1996) conducted a review of Golden Lake heavy oil field that

determined the areal sweep was very poor and the viscosity variation affected the

waterflooding performance. On the other hand, Smith (1992) studied the causes of the

successful waterfloods in Wainwright and the Wildmere areas around Lloydminster.

12

Smith found that induced fracture networks allow the formation to simultaneously filter

the input water, which is quite dirty and plugs the formation and has a negative impact on

filtrate disposal. Moreover, Adams (1982) noted that injected water channeling was so

severe that converted mature injectors sometimes became low water cut producers shortly

after conversion. Then, Turta et al. presented several ‘toe to heel’ waterflooding papers

(Turta et al., 2002, Turta et al., 2003 and Zhao and Turta, 2004) that show that no

permeability restrictions are present in the vertical direction. However, this assumption

limited the applicability of their proposed method. Additionally, Stephen et al. (1995)

studied the effect of well spacing reduction from 20 acres to 10 acres by infill drilling,

but they did not observe any significant recovery. Finally, Mai and Kantzas (2007)

performed a set of ambient temperature laboratory core floods and found that capillary

forces, which are often neglected due to the high oil viscosity, are in fact important even

in heavy oil systems.

2.1.2 Cold heavy oil production (CHOPS)

Cold production refers to the use of operating techniques and specialized pumping

equipment to aggressively produce heavy oil reservoirs without applying heat. Production

remained stable up to 1991, with a yearly output averaging 3.6 million m3; after cold

production became more common, the production tripled to 11 million m3/year in 2003

(Nakutnyy and Renouf, 2009). Sand production—a function of (1) the absence of clays

and cementation materials, (2) oil viscosity, (3) the producing water cut and GOR, and

(4) pressure drawdown rate—is the basis of cold production (Chugh et al., 2000).

Several researchers have investigated the appropriate candidates for cold production

(Dusseault and Geilikman, 1995; Chugh et al., 2000; Dusseault et al., 2000). According

13

to the studies conducted by these researchers, the appropriate reservoir for cold

production should have the following specifications:

High oil viscosity (2,000 ~ 30,000 cp) because the higher the oil viscosity, the greater

the drag force on a sand particle (Chugh et al., 2000).The IFT between gas and oil

should increase with a decrease in oil API. Dusseault and El-Sayed (2000) said that in

more viscous oils (μ > 15,000 cp), despite several attempts. CHOP has not yet been

economically successful.

Unconsolidated formations with less cement bands, which are better candidates for

cold production (Chugh, 2000 & Dusseault, 1995). According to Dusseault, most of

these reservoirs have porosity of 29% to 31%.

Low initial water production and preferably no bottom water (< 40% water cut)

(Chugh et al., 2000).

High initial reservoir pressure because the better the initial drawdowns, the more the

well will cleanup (i.e., will produce sand with oil) (Chugh et al., 2000). As such, the

reservoirs should be buried at depths of 300m to 600m (Dusseault and Geilikman,

1995).

Reservoir thickness of 8 to 15 m; however, the thinnest reported is ~4 m and the

thickest is ~30 m. The sand lithology varies from quartz arenites (>95% SiO2) to

arkoses or litharenites with ~15% feldspar grains, ~20% siliceous volcanic shards,

and ~5-8% lithic fragments (Dusseault et al., 2000).

While these specifications lead to better efficiency, cold production can be improved by

forming high permeability channels (wormholes) in the formation. In fact, Tremblay et

al. (1997; 1998) as well as Tremblay and Forshner (1998) studied wormhole growth

14

under solution gas drive. They visualized wormhole growth during oil flow through a

horizontal sand pack. The wormhole developed in the higher porosity region, which

indicates that the wormhole likely followed the weaker (higher porosity) sand.

2.1.3 Gas EOR methods

In these methods, the injectant can be dry gas, enriched gas (hydrocarbon miscible), CO2,

nitrogen or flue gas, or combinations of these injectants. These methods increase

capillary number. They are also called solvent flooding, miscible-gas flooding, or simply

gas flooding methods. N2 and flue gas are the cheapest possible injectants. In the

literature, successful projects use these cheap gases (Taber, 1988; Taber, 1990; Moritis,

1994; Babadagli et al., 2008; and Sahin et al., 2008). Moreover, for a miscible flood, the

main factor for screening criteria is average pressure, and this parameter is dependent on

depth (Taber, 1988 and Moritis, 1994). Ultimately, the advantages of carbon dioxide

flooding in comparison to N2 and flue gas are that CO2 is very soluble in oils at reservoir

pressure; it reduces the oil viscosity before miscibility is achieved between CO2 and

crude oil, and CO2 will stay dissolved in crude oil (Martin and Taber, 1992).

2.1.4 Thermal EOR processes

Thermal methods lower mobility ratio by decreasing oil viscosity. Since the effect of

temperature is especially pronounced for viscous crudes, these processes are normally

applied to heavy crudes. Thermal methods are divided into in-situ combustion, cyclic

steam stimulation (CSS), hot waterflooding, steam-assisted gravity drainage, and steam

flooding (Prats, 1982 and White and Moss, 1983).

Furthermore, in-situ combustion has been extensively field tested (Farouq, 1972; Farouq

and Meldau, 1979 and Chu, 1982). According to many, in-situ combustion is feasible

15

under a wide variety of field conditions, and, next to waterflooding, it could become the

most widely used recovery method. In fact, a well-designed fireflood could be expected

to recover 50 percent of the oil in place, and could make a profit, especially if

simultaneous or intermittent water injection with air is employed (Farouq and Meldau,

1979). Actually, Yannimaras introduced a new method to screen crude oils for

applicability of the air-injection/ in-situ combustion process (Yannimaras and Tiffin,

1995).

Steam flooding is usually used in reservoirs containing high viscosity crude oils that are

difficult to mobilize by methods other than thermal recovery. Good steamflooding

projects require thick, shallow deposits with high oil saturations and good permeabilities.

Advancements in steam injection applications have made it possible to apply the new

technology in previously unsuitable reservoirs. In addition, the introduction of steam

assisted gravity drainage (SAGD) has transformed the huge quantities of tar sand oil in

Alberta to proven oil reserves, moving Canada to second place in terms of oil reserves

worldwide behind only Saudi Arabia (Shin and Polikar, 2005). SAGD has different

applications that have been mentioned in the literature (Mendoza et al., 1999; Mendoza

and Herrera, 2001; Sedaee and Rashidi, 2006 and Bagci, 2006). SAGD is carried out at

very small pressure gradients, which helps stabilize the process, avoiding the high-

pressure gradients that can potentially lead to channeling and isolation of parts of the

reservoir. Furthermore, Shale barriers can present a challenge to SAGD (Alvarado and

Manrique, 2010). Kumar et al. (1995) simulated a cyclically steamed well in Cymric

field, San Joaquin Valley, California. Their results show that fluid flow from the well to

the reservoir is primarily through the hydraulic fracture induced by the injected steam.

16

CSS is also known as steam soak, or huff and puff. Wong et al. (2003) presented a field

review of the Pikes Peak steam project, showing key performance indicators of CSS and

steam drive in non-bottom water. He concluded that CSS has been conducted

successfully with economic steam/oil ratios (SORs) in areas with up to 4 m of bottom

water by injecting significantly larger steam slugs in what is termed a “drive, block, and

drain process” (Wong et al., 2003). Moreover, Williams et al. (2001) studied the effects

of discontinuous shales on multizone steamflood performance in the Kern River field.

Kern River is a shallow, heavy oil field. According to the literature, this field has been on

steam flooding since the mid-1960s (Bursell and Pittman, 1975; Belvins and Billingley,

1975; Oglesby et al., 1982; Restine, 1983 and Restine et al., 1987). Williams et al. (2001)

concluded that discontinuous shales allow significant oil drainage from upper to lower

sands, as well as fluid migration across zones, and small pattern-element or single-sand

models cannot adequately explain observed field behaviour in this type of reservoir.

2.1.5 Chemical EOR processes

Chemical processes are not widely used, especially when compared to thermal and gas

injection methods. In fact, replacing the trapped oil is approximately 10 times more

difficult than replacing continuous oil (Chatzis and Morrow, 1983); therefore, the chosen

chemical must be very efficient.

The chemical processes are divided into polymer, alkaline, and surfactant/polymer

methods. Regardless of category, chemical flooding methods require low to moderate oil

viscosities and moderate to high permeability; the latter is for favourable water injection

(Maerker and Gale, 1990 and Baviere et al., 1995). However, alkaline flooding consists

17

of injecting solutions of sodium hydroxide, sodium carbonate, and sodium silicate or

potassium hydroxide into the reservoir (Mayer et al., 1983 and Shutang et al., 1995).

Cooke et al. (1974) describe a new method for alkaline flooding. In the process, they

submitted that the alkaline water must be saline rather than fresh water. The use of saline

water causes the sand to become oil-wet in the presence of the alkaline water. High

salinity also leads to the formation of a water-in-oil type of emulsion that does not form

in the other processes. Furthermore, Dong (2008) investigated the effective viscosity of

water-in-oil emulsions in porous media experimentally using four different qualities of

water-in-oil emulsions flowing through sand packs of different permeabilities at different

injection flow rates. He found that the effective viscosity of an emulsion in a sand pack

decreased with increasing flow rate and that the relative variation was minimal.

Ultimately, the emulsions exhibited a higher effective viscosity in a higher permeability

sand pack than in a lower permeability sand pack.

2.1.6 Emerging EOR technologies

New methods for downhole dielectric heating using electromagnetic radiation,

microwaves and radiofrequency are now applicable in enhancing heavy oil recovery

(Emmons et al., 1986 and Islam et al., 1991). To that end, researchers have conducted

successful laboratory experiments using different heating techniques. The heating of

formation fluids and porous media can improve oleic phase mobility relative to the

aqueous and gas phases that enhance oil recovery (Fanchi, 1990 and Islam, 1999). In fact,

electromagnetic frequency can enhance the heavy oil recovery by more than 50%

(Ovalles et al., 2001 and Ovalles et al., 2002).

18

Beckman (1926) first proposed the concept of using microorganisms to enhance oil

recovery (MEOR). Then, Zobel studied this subject more closely in 1950 (Zobel, 1946

and 1947). Since then, different MEOR technologies have been developed to enhance the

oil recovery. For example, some microbial methods aid in paraffin removal while others

are designed to modify heavy oil. Likewise, other methods use microorganisms to

produce chemicals, such as surfactants, polymers, or solvents that are useful in oil

recovery processes, either in above-ground facilities or in situ. However, unfortunately,

MEOR has not gained credibility in the oil industry due to technical and economic

constraints (Maudgalya, 2007).

Another method, which is now becoming more common, is VAPEX. VAPEX is an

energy-efficient method of recovering high viscosity heavy oil and bitumen from

reservoirs. The process uses a solvent in the miscible displacement of bitumen or heavy

crude oil. VAPEX improves energy efficiency and reduces emissions and operating costs.

However, production rates with this process are lower than with traditional steam

processes. In the conventional VAPEX process, a mixture of vapourized solvent (propane

and/or butane) and a commercially available non-condensable gas (methane, natural gas)

is injected into the reservoir to reduce oil viscosity. While the VAPEX process became

less attractive with the increase of gas price, injecting CO2 will decrease solvent cost.

Moreover, CO2 is more soluble in heavy oils than methane. On the other hand, this can be

environmentally important because, nowadays, CO2 sequestration itself is an important

environmental issue.

Yazdani et al. (2005) did a scale-up for the VAPEX method and studied the effects of

drainage height and grain size on production rates in the VAPEX process. They found

19

that minor changes in heavy oil composition do not significantly affect the observed

drainage rates. They also observed that scaled-up, stabilized oil-drainage rates are much

higher than the predictions published in the literature. Thus, the VAPEX process may be

more widely applicable than previously thought.

In Appendix A, a summary of applicability of some EOR techniques found in the

literature is provided. Tables A-1 to A-4 estimate which production method applies to

each heavy oil resource. Tables A-5 to A-10estimate the potential impact of specific

technologies on various subsurfaces deeper than 50m, which constitute 90% of Canada’s

heavy oil resource production methods and 100% of the US’s and Venezuela’s resource

production methods. The potential impacts have been rated “high”, “medium”, “low”,

and “unknown” (Clark, 2007).

2.2 Vapour extraction (VAPEX)

The VAPEX process is the solvent analog of SAGD, which reduces oil viscosity by

diluting the in-situ bitumen with vapourized solvents. The idea of injecting solvent

vapours to enhance oil recovery was first proposed in 1974 by Allen (Allen, 1974, Allen,

1976 and James and Chatzis, 2007), in which, the Cyclic Steam Stimulation (CSS)

process was varied by alternating steam and solvent. The solvents used in his experiments

were butane and propane. Because of the low oil recovery, the idea was not field tested.

Later, Allen (1976) improved the idea by injecting a mixture of two gases: one gas as the

carrier gas and the other one as the solvent. Then, Butler and Mokrys (1989) introduced a

brief discussion on the process during a study on Athabasca and Suncor Coker feed

bitumen. They named it VAPEX in 1990 (Butler and Mokrys,1989 and Butler and

Mokrys, 1990). In this process, the horizontal production well is located near the bottom

20

of the pay zone and the injection well is completed right above the production well. Even

though this process is significantly slower than SAGD, using vapour to reduce oil

viscosity and increase operating temperature will make VAPEX economically

advantageous.

The low cost of the injected solvent, which can be recovered and recycled, the

applicability of this method in thin and low-porosity-good permeability reservoirs are the

key advantages of the VAPEX process (Yazdani, 2007). Indeed, the energy requirements

for VAPEX are less than thermal methods. Besides, the thermal recovery methods cannot

be implemented in the reservoirs with bottom aquifer (Das, 1998, James et al., 2007,

Rahnema et al., 2008 and Pourabdollah, 2013).

In VAPEX, diluted oil becomes less viscous along the boundary of the vapour chamber

and drains via gravity toward the production well, which is directly located below the

injection well. Of note, long horizontal wells are required to obtain reasonably high

production rates because gravity drainage is a slow recovery process (Jiang and Butler,

1996). Nevertheless, the vapour chamber forms around the injection well in the swept

zone by pore spaces filling with solvent vapour. The mixing of solvent and bitumen

occurs mainly by molecular diffusion and convective dispersion mechanisms that are

combined during the solvent and bitumen mixing process (Das and Butler, 1998).

However, in the mixing process, convective dispersion is more important than molecular

diffusion (Nghiem et al., 2001). Regardless, the solvent dispersion coefficient is an

important parameter governing the efficiency of bitumen dilution during the VAPEX

process. In short, to estimate the oil recovery after implementing VAPEX, an accurate

dispersion coefficient estimate is crucial. However, while there is not any proven

21

methodology to predict dispersion coefficient, Karmaker and Maini (2003) proposed a

new technique to extract the net dispersion coefficient using a 2-D magnetic resonance

imaging tool.

Yazdani and Maini (2005) did a scale-up for the VAPEX method and studied the effects

of drainage height and grain size on production rates in the VAPEX process. In their

research, it was found that minor changes in heavy oil composition do not significantly

affect the observed drainage rates. It was also observed that scaled-up, stabilized oil-

drainage rates are much higher than the predictions published in the literature. Thus, the

VAPEX process may be more widely applicable than previously thought.

Even though the most suitable solvents for the process are propane and ethane, a mixture

of butane, propane and ethane may suffice depending on reservoir pressure and

temperature (Karmaker and Maini, 2003). Regardless of solvent selection, the optimum

injection point is near the dew point where the vapour phase has maximum solubility and

there is maximum diffusivity in the liquid phase (Talbi and Maini, 2003).

Increases in the price of gas have hurt the feasibility of the conventional VAPEX method.

Currently, CO2, as a carrier gas, is a good alternative because of its low cost and higher

solubility than methane, meaning it will dilute better. Besides, from an environmental

point of view, CO2 sequestration through CO2 injection in heavy oil reservoirs can be one

of the most promising technologies for mitigating atmospheric CO2 concentration (Manik

et al., 2003 and Mohammadpoor et al., 2012). However, CO2 mixture-heavy oil systems

form multiple liquid phases that reduce the gravity drainage effectiveness by introducing

complex relative permeability effects (Talbi and Maini, 2003).

22

2.2.1 Solvent requirement

Several factors affect appropriate injection solvent selection criteria. These factors

include equilibrium pressure, molecular weight, density difference, solubility, diffusivity,

reservoir temperature, and pressure (Upreti et al., 2007). Specifically, when a low

molecular weight vapourized solvent is injected into the reservoir near its dew point, the

solubility of the vapourized solvent reaches its maximum near its dew point (Upreti et al.,

2007 and Das, 1995). This causes additional viscosity reduction by deasphalting.

Additionally, a higher density difference between a vapourized solvent and heavy oil

results in superior gravity drainage (Das and Butler, 1998).

As an injected solvent, propane is common in VAPEX studies. After all, Das and Butler

found propane and butane to be the most effective solvents for VAPEX (Das and Butler,

1994). They found that propane diffuses faster and produces higher production rates.

Moreover, in further investigations, Butler and Jiang (2000) found that a 50:50 mixture of

butane and propane has approximately the same performance of pure propane and is

better than a pure butane injection. Likewise, Kok et al. (2009) and Yildirim (2003)

utilized propane and butane solvents on light, medium, and heavy oil in a Hele-Shaw cell

at three different injection rates. In the experiments with heavy oil, butane had the highest

injection rate, even better than propane. However, with the other two rates, both solvents

showed almost identical performance. Ultimately, the results revealed that propane

provides better results than butane in almost all injection rates for light and medium oils.

Azin et al. (2005) found that in the reservoirs with relatively low viscosity, higher

injection rates create higher oil production. Alternatively, in the reservoirs with high

23

initial viscosity, the higher injection rate results in a system pressure increase that may

cause an early solvent breakthrough.

Another viable and economically sound option is implementing non-condensable and

inert gases as carrier gases. For instance, methane and CO2 are suitable carrier gases for

solvent injection during the VAPEX process (Talbi and Maini, 2003).That is, they found

that the carbon dioxide and propane mixture showed better results than the methane and

propane mixture at 600 psig. Thus, at higher pressures, where carrier gas concentration is

increased, oil production decreased using either of the above-mentioned mixtures.

Furthermore, Torabi et al. (2012) found that incorporatingCO2 into the solvent for a

VAPEX process is a viable option. For them, the non-condensable gas portion of the

solvent should be less than 60% of the total mixture. Through several simulation runs

with different solvent compositions, they found that replacing a portion of the methane in

the solvent with CO2 resulted in equal or greater recovery factors in most simulations.

Frauenfeld and Lillico (1999) patented the use of a mixture of hydrocarbon solvents to

increase the effectiveness of solvent-assisted heavy oil recovery processes. They utilized

mixtures of propane, ethane, and butane with paired-injector and producer-well systems

or single-well cyclic systems.

After monitoring the effects of non-condensable and inert gases on the production history

of the VAPEX process, Chatzis et al. (2006) found that the accumulation of non-

condensable gas near the boundary reduced the advance rate of the VAPEX chamber.

They also found that the position of the vapour front is proportional to the square root of

time.

24

From an environmental perspective, many promote the mixing of the solvent with a non-

condensable gas such as CO2. It is also economically profitable, as doing so reduces

expensive solvent inventory. Therefore, CO2 sequestration through CO2 injection in

heavy oil reservoirs is one of the most promising technologies for mitigating atmospheric

CO2 concentration (Manik et al., 2003 and Mohammadpoor et al., 2012). However, the

lower rate of solvent mass transfer into the heavy oil and less heavy oil dilution may be

possible disadvantages of mixing the diffusing solvent with non-condensable gas (James,

2009).

2.3 VAPEX mechanism

In the first step of the VAPEX, the solvent is injected through the injector to form an

initially vertical solvent vapour chamber between the injector and the producer. The

vapour chamber then spreads, and gravity stabilizes the oil-solvent interface. Here,

molecular diffusion of solvent vapour into the bitumen controls drainage (Butler and

Mokrys, 1990). Thus, in order to maximize solvent vapour contact with the reservoir, the

injection and production wells should be drilled horizontally. Figure 2-1 shows the

VAPEX mechanism schematic.

There are two types of gravity drainage flow during the VAPEX process: boundary

drainage and transition film drainage (Roopa and Dawe, 2007). In fact, Roopa and Dawe

(2007) found that the rate of film drainage that occurs in the three-phase flow processes

within the vapour chamber depends on the effects of temperature on viscosity, diffusion

coefficients, mass transfer, interfacial tension, and wettability.

25

Figure ‎2-1: Mechanisms of VAPEX process

Pay zone

Solvent vapour chamber

Horizontal injection well

Horizontal production well

Solvent/heavy oil interface

Draining of diluted oil Draining of diluted oil

26

Also worth mentioning, Yang and Gu (2005) found that capillary and gravity forces

control gravity drainage in porous media. They measured the interfacial tension between

the Lloydminster heavy oil and four solvents (methane, ethane, propane, and carbon

dioxide) at different pressures below their vapour pressures by applying the axisymmetric

drop shape analysis (ADSA) technique for the pendant drop case. They found that the

interfacial tension between heavy oil and a solvent is reduced linearly with pressure

(Yang and Gu, 2005). Furthermore, Cuthiell et al. (2006) investigated the effect of

capillary force, concluding that capillary mixing could significantly influence

diffusion/dispersion. In short, layering effects will increase mixing and drainage speed.

However, reservoir layer heterogeneity will typically be much greater than that in a

prepared sand pack, and this may significantly enhance VAPEX drainage rates. In their

study, Cuthiell et al. (2006) found that dispersive mixing was consistent with molecular

diffusion only in that there was little or no enhancement due to convective dispersion.

Absence of capillary pressure allows the vapour to achieve maximum vertical

propagation with the lowest amount of sideways leaching (Ayub, 2009). Because

capillary pressure tends to delay the gas production without affecting the overall

recovery, it produces a significant amount of asphaltene precipitation near the injection

well (Ayub and Tuhinuzzaman, 2007).

Along the same lines, Rostami et al. (2007) observed the dual effect of capillary force in

the VAPEX process. Capillary forces hinder solvent breakthrough, and the establishment

of well communication and chamber extensions is different from conventional cases. In

fact, cumulative oil production is increased because of solvent-oil relative permeability

alteration that occurs due to surface tension reduction (Rostami et al., 2007).

27

2.3.1 Molecular diffusion

Diffusion plays an important role in the VAPEX process. Indeed, solvent gas diffusion is

the main molecular reaction that accounts for gas absorption and, consequently, a

reduction in mixture viscosity (Upreti et al., 2007). Of note, during the molecular

diffusion, the gas first moves towards the oil–gas interface; then, the gas penetrates the

interface before penetrated gas diffuses in the oil body in the last stage (Pourabdollah et

al., 2013). Because of the importance of molecular diffusion, in order to determine the

amount of injection gas, the amount of heavy oil reserves that will undergo the viscosity

reduction, the time required for the viscosity reduction to take place, and the rate of oil

production, an accurate knowledge of gas solvent-heavy oil system diffusion is

necessary.

In that vein, two main methods experimentally determine the diffusion coefficients. The

first method is the direct method, in which different liquid samples undergo

compositional analysis at different times (Schmidtet al., 1982).

The second method is the indirect method, which includes change in volume, pressure,

solute volatilization rate, position of the gas liquid interface, and nuclear magnetic

resonance (Upreti et al., 2007). Actually, Renner (1988) proposed an in-situ method for

measuring molecular diffusion coefficients of CO2, methane, ethane, and propane. His

proposed correlation was a function of liquid viscosity, molecular gas weight, molar

volume of gas, as well as gas pressure and temperature. Likewise, Riazi (1996) proposed

a semi-analytical model for the estimation of mass transfer rates caused by diffusion

between a non-equilibrium gas and a liquid in a constant volume cell with a constant

temperature. In his method of measuring diffusion coefficients, no compositional

28

measurements are necessary. Furthermore, Grogan et al. (1988) found that CO2

diffusivity in both pure hydrocarbons and crude oil at reservoir conditions depends

primarily on solvent viscosity. Measurements are based on the direct observation of

interface motion caused by CO2 diffusion through oil or oil shielded by water. Then, they

determined the diffusion coefficients by fitting the mathematical models to the observed

interface motions.

Oballa et al. (1989) found that the diffusivity coefficient is strongly dependent on

concentration. In short, overall diffusivity reaches a maximum at an intermediate

concentration.

Another indirect method to calculate the diffusion coefficient is nuclear magnetic

resonance (NMR) (Afsahi and Kantzas, 2005; Salama and Kantzas, 2005; Wen and

Kantzas, 2005). Afsahi and Kantzas (2005) found that diffusivity of heptane into Cold

Lake bitumen in the presence of sand is approximately 10-6

to 10-7

cm2/s, which is within

the same order of magnitude of the solvent diffusivity into pure heavy oil and bitumen.

They also observed that diffusivity decreases as diffused solvent concentration into

bitumen increases over time before nearly stabilizing after a few hours of diffusion.

However, they concluded that although the diffusivity is a function of concentration, it is

constant in short time intervals.

In addition, Wen and Kantzas (2005) successfully implemented the nuclear magnetic

resonance (NMR) method to study solvent-heavy oil/bitumen mixture properties. They

used low-field NMR and x-ray computer-assisted tomography (CAT) scanning for

solvent diffusion measurements with heavy oil or bitumen systems. They found that

29

diffusion coefficients calculated from NMR data provided results that were reasonable

and similar to those obtained via CAT scan.

Hayduk and Cheng (1971) introduced a relationship between the diffusivity equation and

viscosity. Based on the experimental data, they concluded that a unique diffusivity-

solvent-viscosity relationship—one independent of temperature and solvent

composition—exists for each different diffusing substance. After plotting data on a log-

log paper, they found a linear relationship between viscosity and diffusivity. In fact, the

slope of the line appeared to depend on the diffusivity itself: the lower the diffusivity, the

higher the slope. They proposed the following relationship:

BAD ……………………………………………………………………...……… (2.1)

Constants A and B apply to each diffusing substance.

However, the above-mentioned equation does not consider the effect of porous media. In

the presence of a porous medium, “apparent diffusivity” accounts for the effect of porous

media on diffusivity. In 1963, Perkins and Johnson suggested porous media (both

unconsolidated packs and consolidated rocks) can, as networks of flow chambers, have

random size and flow conductivity that are connected by smaller-sized openings. Then,

they proposed the following equation to calculate the diffusion coefficient in such a

porous medium:

FD

D

o

1 ……………………………………………………………………….…….. (2.2)

30

where D is the apparent diffusivity, is the diffusivity, while F is the formation

electrical resistivity factor, and is the fractional porosity. Fatt (1958) also reported the

same diffusivity and formation factor relationship.

In addition, Grane et al. (1961) found that at sufficiently low flow rates, transverse and

longitudinal dispersion are equal and are determined by the coefficient of fluid molecular

diffusion and porous medium formation factor. However, at high flow rates in

consolidated media, transverse and longitudinal dispersion exist independent of fluid

properties and are proportional to flow velocity.

Researchers offer diverse correlations for calculating diffusivity coefficient. However,

each of these correlations is valid in its own range of assumptions.

Wilke and Change (1955) carried out experiments involving iodine and toluene diffusion

in a variety of hydrocarbon solvents; they proposed the following equation:

…………………………………………………………...… (2.3)

where D is in cm2/sec and the association parameter x is introduced to define the effective

molecular weight of the solvent with respect to the diffusion process. M is molecular

solvent weight, T is temperature in , while µ is viscosity in centipoises, and V is molar

volume of a solute at normal boiling point in cc/g.mole.

Moreover, Hiss and Cussler (1973) calculated the diffusion coefficients of n-hexane and

naphthalene in a series of hydrocarbon liquids with viscosities from to 5 kg.m−1

sec−1

(0.5 to 5000 cp) at 25°C. They used a Savart-plate-wave front-shearing

oD

6.0

2/18 )(

104.7V

TxMD

K

4105

31

interferometer that allows direct determinations at effectively infinite dilution. They

proposed the following equation to calculate the diffusivity:

3/2Da ……………...……………………….…………………………………….. (2.4)

Next, Hayduk et al. (1973) conducted experiments using the steady state capillary cell

method; they proposed the following equation to calculate diffusivity:

……………………………………………………………… (2.5)

where D is m2/s and µ is Pa.s.

Later, Hayduk and Minhas (1982) proposed correlations for solute diffusivity in aqueous

solutions:

…………………………………………......…. (2.6)

……………………………………………………………………..… (2.7)

where D is in cm2/sec, V is molar volume of solute at normal boiling point in cc/ mole, T

is temperature in , µ is viscosity in centipoises. For non-aqueous solutions:

………………………………………………...……. (2.8)

where D is in cm2/sec, VA and VB are molar volume of solute and solvent respectively at

normal boiling point in cc/ mole, T is temperature in , while µ is viscosity in

545.09100591.0 D

52.119.08 )292.0(1025.1 TVD

12.158.9

V

K

105.0

125.0

92.0

29.1

42.0

27.081055.1

A

B

A

B T

V

VD

K

32

centipoises, and σ is surface tension at the normal boiling point temperature in dyne/cm.

For paraffin solutions:

………………………………………………………..….. (2.9)

………………………………………………………………..…… (2.10)

where D is in cm2/sec, V is molar volume of solute at normal boiling point in cc/ mole, T

is temperature in , while µ is viscosity in centipoises.

Reid et al. (1987) proposed another correlation for calculating diffusivity (Yazdani,

2007):

…………………………………………………………………………. (2.11)

Das and Butler (1996) proposed the following equations. For propane as the solvent:

………………………………………………………….…… (2.12)

Moreover, for butane as the solvent:

…………………………………………………………….… (2.13)

where D is in m2/sec, and is viscosity in Pa.s.

Moreover, Upreti and Mehrotra (2002) found that the gas diffusivity increases with

temperature and pressure. They used a non-intrusive experimental method to calculate the

71.047.18103.13 VTD

791.02.10

V

K

d

RTD

3

46.0910306.1 D

46.0910131.4 D

33

diffusivity of CO2, methane, ethane, and nitrogen as a function of gas concentration in

bitumen. They provided the following correlation for the average diffusivity:

………………………………………………………… (2.14)

where D is in m2/sec, T is temperature in , and d0 and d1 are correlated coefficients.

2.3.2 Physical dispersion

The term dispersion refers to the additional mixing caused by concentration gradients or

uneven fluid flow when the fluids are flowing through the porous medium. Of note,

dispersions in the longitudinal and transverse fluid flow are not equal; hence, there are

two different types of dispersion: longitudinal and transverse (Perkins and Johnston,

1963).

Dispersion, or effective diffusion, is fluid mixing due to diffusion and convective motion.

During VAPEX, gravity drainage causes convective motion and an additional mixing that

results in dispersion. Heavy oil viscosity is reduced when the vapourized solvent diffuses

into the heavy oil. In addition, the diluted oil drains due to gravity.

Several factors in addition to molecular diffusion will improve the heavy oil recovery

during VAPEX. These factors include convection, increased interfacial area, and the

resulting continuous renewal of surface area exposed to the solvent in the porous media

caused by the oil drainage, increased gas solubility, and capillary phenomena at the

solvent-oil interface (Upreti et al., 2007).

)15.273(ln 10 TddD

C

34

Taylor (1953) showed that the distribution of concentration after introducing a soluble

substance into a fluid (flowing slowly through a small-bore tube) spreads under the

combined effect of molecular diffusion and velocity variation.

Likewise, Kapadia et al. (2006) developed a mathematical model to calculate the

dispersion coefficient of butane along with its solubility in Cold Lake bitumen. They

found that gas dispersion, as well as heavy oil and bitumen viscosity, were dependent on

composition.

More recently, El-Haj et al. (2009) conducted VAPEX experiments using three different

types of glass beads and Athabasca bitumen. They developed a mathematical model to

calculate the optimum interfacial mass fraction, and the dispersion coefficient of butane

in the medium. They found that the dispersion coefficients were three orders of

magnitude higher than the molecular diffusion. Other researchers have reported this

enhanced mass transfer as well (Dunn, 1989; Das and Butler, 1995; Boustani and Maini,

2001; Odenton et al., 2001 and Yazdani and Maini, 2005).

Furthermore, Ahmadloo et al. (2011) proposed a new correlation for effective diffusivity.

They performed VAPEX experiments with butane as the solvent and concluded that

capillary forces play an important role in the process. They represented this effect in their

correlation by introducing the term A, which stands for specific pore surface area. They

also included the drainage height as an important parameter:

………………………...……………… (2.15) 7956.06096.0555.171045.1 AhDeff

35

where Deff is in cm2/sec, h is drainage height in cm, A is specific pore surface area in m

2/g,

and is viscosity in mPa·s.

Finally, Abukhalifeh et al. (2013) studied the effect of drainage height on the

concentration-dependent dispersion coefficient of propane in heavy oil during the

VAPEX process. They found that the propane dispersion coefficient, the amount of

dissolved propane, and the oil production rate all increase with an increase in model

height.

2.4 Asphaltene precipitation

Asphaltenes are components of heavy oil with high molecular weights ranging from

1,000 to 2,000,000 g/mole. They are dark brown or black friable solids that do not have

any definite melting point. Asphaltenes have complex molecules, so much so that the

exact structure of their molecules is still unknown. In addition, they contain aromatic

rings and oxygen, nitrogen, and sulphur, as well as heavy metals such as vanadium,

nickel and iron. Canada’s heavy oil asphaltene components on average are: 45.25%

carbon, 52.45% hydrogen, 0.74% nitrogen, 0.68% oxygen, and 0.87% sulphur (Fredrich,

2005).

The liquids used to obtain asphaltenes from petroleum are non-polar solvents with low

boiling points. These include petroleum naphtha, petroleum ether, n-pentane, n-heptane,

and iso-butane (Speight, 2004). Furthermore, asphaltene constituents are insoluble in

methane, ethane, and propane.

The amount of asphaltene precipitation after using any of the mentioned separating

liquids depends on the solvent used, temperature, solvent concentration, and the time

36

during which the crude oil and the solvent are mixed together. Figure 2-2 shows the

asphaltene precipitation in weight% of the bitumen using different hydrocarbon solvents

(Speight, 2007).

As mentioned, the volume of solvent added to crude oil also affects asphaltene

precipitation. There are different standard methods for asphaltene precipitation. These

standard methods when using n-pentane and n-heptane as the solvent are listed in Table

2-1. During asphaltene separation, if insufficient amounts of liquid hydrocarbon are used,

resins appear within the asphaltene fraction. This isolates asphaltene from crude oil and

consequently creates an incorrect report about crude oil asphaltene content.

To more accurately calculate asphaltene levels, Speight (2004) listed the following

factors to measure precipitation amount and type:

1. Adding more than 30 mL of hydrocarbon per g of feedstock.

2. Using n-pentane or n-heptane.

3. Providing 8 to 10 hours of contact time.

4. Doing a precipitation sequence to eliminate absorbed resin.

The experimental studies in this research were conducted by injecting various solvents in

two low permeability large scale visual models to monitor the asphaltene precipitation

with higher accuracy. The specific well configuration used in this study and the large

dimensions of VAPEX physical help to monitor the VAPEX process and asphaltene

precipitation in more details.

37

Figure ‎2-2: Yield of asphaltene precipitation for various hydrocarbon solvents (after Speight, 2007)

38

Table ‎2-1: Standard methods for asphaltene precipitation measurement (after Speight, 2004)

Method Precipitant

Volume precipitant per g of

sample (mL)

ASTM D-893 n-pentane 10



ASTM D-3279 n-heptane 100

ASTM D-4124 n-heptane 100

IP 143 n-heptane 30

Syncrude method n-pentane 20

39

2.4.1 Asphaltene precipitation in VAPEX

A phase change occurs during the VAPEX process when the solvent diffuses into the oil

at the solvent oil interface. During this phase change, there will be change in temperature,

pressure, and concentration at the contact interface. This will result in asphaltene

deposition. The asphaltene precipitation has a contentious effect on the VAPEX process.

While the asphaltene precipitation may improve the quality of heavy oil by reducing the

viscosity, at the same time, it may alter the wettability of rock and consequently plug

some of the pores.

Mokrys and Butler (1993) conducted a deasphalting experiment in a pressure cylinder

packed with l mm glass beads. Cold Lake bitumen and Lloydminster heavy oil were

deasphalted. They used propane as a precipitant. They found that the viscosity of Cold

Lake bitumen was decreased by a factor of 300 and the viscosity of Lloydminster oil was

decreased by a factor of 50.

Nghiem et al. (2000) carried out phase behaviour calculations and a compositional

simulation of asphaltene precipitation for the VAPEX process with Lindbergh heavy oil.

They found that asphaltene precipitation occurs in the oil phase region that is adjacent to

the vapour chamber and spreads with the growth of the chamber. They observed that the

streaks of asphaltene precipitation in their simulation occur in the same way other

researchers reported. However, they observed wider streaks at the upper part of the

solvent chamber.

Luo and Gu (2005) performed experiments to determine the effect of asphaltene content

on heavy oil viscosity. They found that if the heavy oil asphaltene content is reduced

40

from 14.5% to zero, the sample heavy oil viscosity is reduced by 13.7 times. Their results

appear in Figure 2-3.

Moreover, Pourabdollah et al. (2011) carried out VAPEX experiments on sand packs

with Iranian bitumen to investigate the effect of vapour dew point and permeability on

the movement of asphaltene streaks. Their results demonstrated that when the solvent

pressure was less than its dew point, precipitated streaks irregularly remained on the

surface of glass beads. Furthermore, they observed that when the solvent pressure was at

the dew point, the precipitated streaks moved faster and their movement was in the

direction of live oil to the production well. Ultimately, they did not observe any

precipitation in a high permeable porous medium (Pourabdollah et al., 2010).

Pourabdollah et al. (2011) conducted VAPEX experiments in a modified sand pack

model where nanoclay particles were added to glass beads. They used montmorillonite as

nanoclay alongside glass beads and used propane as the solvent for VAPEX experiments.

The nanoclays acted as adsorbents in heavy oil to adsorb the asphaltene and decrease

bitumen viscosity. They observed that, in the nano-assisted sand pack, there was higher

asphaltene precipitation and a larger contact area that resulted in higher production rates.

41

Figure ‎2-3: Effect of asphaltene content on heavy oil viscosity (after Luo and Gu, 2005)

42

2.5 Economic and environmental advantages

The VAPEX process is a non-thermal process. This makes it more energy and

environmental efficient. Since no steam generation and water recycling are required, this

process is significantly fuel efficient. From an environmental point of view, and being

non thermal, there will be no CO2 emissions. In addition, in cases of CO2-based VAPEX

processes, it will also help sequester CO2.

In 1998, the Petroleum Recovery Institute (PRI) conducted a project of 16 participants

with nine research-performing organizations to evaluate the full project engineering and

commercial scale economics for the VAPEX process. They calculated the supply cost

economics for VAPEX oil production from the Athabasca oil sands, Cold Lake oil sands,

and Southeast Alberta heavy oil. Supply cost means the threshold price required for

satisfactory economics. Their results showed that VAPEX has economic and

environmental benefits. However, they emphasized that specific field conditions should

be studied in detail and that a pilot test should be carried out before implementing

VAPEX (Luhning et al., 2003).

Among all these advantages, the possibility of solvent loss and compression is a

drawback of the VAPEX method. On the other hand, the key drawback for this method is

lack of knowledge about the process specifically in real field-operating conditions. This

method is not well tested in the field and there is not much information available about

legitimate field design for facilities.

Luhning et al. (2003) performed an economic analysis based on the results of the history

matching experimental results. Their economic analysis was based on matching the

43

simulation results with the real field-operation costs as well as for estimations of real

field costs. The analysis was based on supply costs reported by Canadian Energy

Research Institute (CERI). However, they estimated such costs at an even higher rate

than the reported amounts for facility and VAPEX solvent costs. Although they

considered their estimation margins safe, the results showed economic profits and

attractiveness.

The environmental advantages of VAPEX process can be listed as follows (Luhning et

al., 2003):

1. CO2 sequestration: Because of solvent recycling, there will be pressure depletion

in the reservoir after oil production using VAPEX; this happens in most heavy oil

recovery methods. Moreover, CO2 can be a very good candidate to pressurize the

reservoir and maintain reservoir pressure as greenhouse gas sequestration takes

place. In fact, there was an amazing comparison regarding the amount of CO2 that

can be sequestered during a VAPEX process. They found that during an

Athabasca VAPEX-depleted reservoir, the amount of CO2 that can be

permanently sequestered is 0.5% to 2% of the total CO2 produced by all the cars

in Alberta.

2. Lower transportation costs: Due to the in-situ upgrading that takes place during

the VAPEX process, there will be less asphaltene in the produced oil. This means

less energy is needed for the transportation of heavy oil in pipelines as well as less

maintenance costs for the pipelines and facilities. This will consequently decrease

the future emissions of and lower facility bottlenecking in refineries.

44

3. Conservation of natural gas, water, and injection solvent: Since VAPEX is a non-

thermal process, steam is not generated so, there will not be any transformation of

water and clean natural gas to steam and greenhouse gas emissions. Alternatively,

because of the nature of the VAPEX process, solvent gas recovery will reduce the

costs and environmental effects.

4. Less surface and overlying reservoir disturbance: The number of surface facilities

required for the VAPEX process is significantly fewer than the facilities required

for thermal methods such as SAGD. This means less surface disturbance,

especially in areas where there are limitations on the land surface available for

production operations. However, since VAPEX is a low-pressure non-thermal

gravity drainage method, there will not be much pressure and temperature

disturbance to the overlying layers of the recovery zone. In the end, the residual

temperature and pressure of the reservoir after implementing the VAPEX method

is significantly unchanged.

45

3. CHAPTER 3: EXPERIMENTAL SETUP, MATERIALS,

AND PROCEDURE

The implementation of EOR techniques in any oil reservoir often carries enormous costs.

As such, a preliminary study is necessary to avoid the loss of natural resources and

unnecessary expenses. Therefore, running an experimental setup—a smaller

representative of the actual reservoir—will give a better understanding of the

mechanisms affecting a specific EOR technique. This will give the researchers the

opportunity to understand the key parameters affecting the process, to investigate the

uncertainties, and to design a more cost-effective pilot field test. To achieve this goal, a

comprehensive experimental study was designed and carried out in order to investigate

the applicability of injecting different solvents during the VAPEX process. To better

represent and simulate the actual conditions, two large VAPEX models were designed

and successfully used for these experiments.

3.1 Experimental setup

The experimental set up consists of four major units: a solvent injection unit, the VAPEX

physical models, a solvent and liquid production unit, and a data acquisition system. In

this section, each unit is explained in more detail.

3.1.1 Solvent injection unit

In this study, VAPEX experiments were conducted under constant pressure. In this case,

the solvent injection unit was composed of gas cylinders (propane, CO2, methane and

butane), gas pressure regulators, digital pressure gauges, solvent injection valves, and

46

digital flow meters calibrated specifically for each gas. The solvent was injected through

the pressure regulators to monitor the cylinder injection pressure through the digital flow

meters. Another pressure gauge records pressure at the point of injection to the model. In

addition, the flow rate and total solvent volume were recorded accurately with the digital

flow meters. Pure CO2, methane, propane, and butane gas (99%) cylinders were

purchased from Praxair and were used for the VAPEX tests. The injected gas was passed

through the AALBORG digital flow meters (DFM) (Figure 3-1) before entering the

packed models. Furthermore, four different DFMs were purchased from AALBORG, and

each one was calibrated specifically for each pure gas. The details about each DFM are

shown in Table 3-1. All these DFMs recorded the flow rate and total volume of injected

gas during the experiments, and they were all connected to the data acquisition unit for

continuous data recording.

47

Figure‎3-1: Digital flow meter (DFM)

48

Table ‎3-1: DFM specifications

No. Calibrated Gas Model Number Max. Pressure

(kPa)

Flow Rate

(mL/min)

1 Propane (C3H8) DFM26 6800 0-1000

2 Methane (CH4) DFM26 6800 0-1000

3 Carbon Dioxide (CO2) DFM26 6800 0-1000

4 Butane (C4H10) DFM27 680 0-2000

49

3.1.2 Physical models

The major components of the experimental set up were the VAPEX physical models.

Two 2-D rectangular VAPEX models with different sizes were used to carry out the

experiments. The dimensions of these models are shown in Table 3-2. It should be noted

that these physical models were used for another study by other researchers (Ahmadloo et

al., 2011). These visual slab models are made of Plexiglas plates with a stainless steel

frame. Figure 3-2 shows the Plexiglas slabs for the large and small models. To seal the

model pressure, gaskets were put between the slabs and the steel frame (Figure 3-3). In

addition, another steel protection cover was bolted to the Plexiglas plates to increase the

pressure tolerance of these models (Figure 3-4). These models are designed for a

maximum pressure of 1 MPa. The visual slabs limit the maximum operating pressure.

However, their transparency was necessary for visual observation of the solvent injection

process, specifically in terms of gas chamber evolution. The physical models were

assembled on a steel frame mounted on a steel stand with rotation capability for better

packing and cleaning purposes (Figure 3-5). Four injection/production ports were

designed on the top and bottom of the large model; there are two injection/production

points for the small model on the top and bottom sides. Figures 3-6 and 3-7 show the

schematic and dimensions of the physical models as well as the cavity space of each

model.

50

Table ‎3-2: Dimensions of physical models

Physical model Height (cm) Length (cm) Thickness (cm) Volume (cm3)

Small 24.5 20 5 2450

Large 47.5 38 5 9025

51

Figure ‎3-2: Plexiglas slabs, (a) Large model slab (b) Small model slab

52

Figure ‎3-3: Gaskets, (a) Large model gasket, (b) Small model gasket

53

Figure ‎3-4: Steel cover protectors, (a) Large model, (b) Small model

54

Figure ‎3-5: Physical models assembled on a steel frame mounted on steel stand

55

Figure ‎3-6: The schematic of the large physical model and its sand pack cavity

47

.5 c

m

38 cm

Injection

Production

Sand pack cavity

Physical Model

Steel cover protector

Plexiglas Slabs

56

Figure: ‎3-7: The schematic of the small physical model and its sand pack cavity

24.5

cm

20 cm

Injection

Production

Sand pack cavity

Physical Model

Steel cover protector

Plexiglas Slabs

57

3.1.3 Solvent and liquid production unit

The fluid production unit included production control valves, digital pressure gauges,

back-pressure regulators (BPR), nitrogen gas cylinders, separators, wet test meters

(WTM), and oil sample collectors. Digital pressure gauges were mounted at the

production points to monitor the outlet pressure. The BPRs (Figure 3-8) were used to

maintain the pre-specified pressure in each VAPEX model during the experiments. Two

more pressure gauges were mounted on each BPR to monitor the pressure of the gas line

connected to each BPR from the nitrogen gas cylinders. The produced oil and gas were

collected in two separators below each physical model. These separators were visual and

calibrated to record the volume of oil produced during the course of experiments. The

separators were made of Plexiglas with two stainless steel flanges on the top and bottom

to seal the vessel. The separators were bolted on a steel stand. These separators are shown

in Figure 3-9. In addition, there were two connection points on top of each separator and

one connection point on the bottom. The produced oil was collected from the BPR

through one of the top connection points. The other connection point is connected to

another valve toward the WTM in order to collect the free gas. This is the same for each

separator. The dead oil is collected from the bottom. The rate and total volume of the

produced gas are accurately measured with two WTMs.

58

Figure ‎3-8: High pressure back-pressure regulator (BPR)

Oil Inlet

Oil Outlet Gas Dome

59

Figure ‎3-9:Two-phase separators

60

The WTMs used for these experiments are shown in Figure 3-10. These WTMs were

Ritter TG05s. They were drum-type gas meters with a maximum pressure of 1 bar and

maximum flow rate of 60 L/h. They had refined stainless steel casings and polypropylene

measuring drums.

Furthermore, each WTM included a pulse generator, which was connected directly to the

data acquisition system through Rigomo software to record the flow rate and total volume

of the produced gas. Because of the large volume of the physical models, and as a result

the large amount of produced gas, conventional bubblers did not seem to be accurate due

the necessity to continuously refill the water cylinder. Ultimately, the oil sample

collectors were simply glass jars that collected the dead oil through the valves connected

to the bottom connection points of each separator.

3.1.4 Data acquisition unit

During the course of experiments, different parameters were recorded. This unit was

composed of a computer as well as special ports, converters, and pulse generators. The

rate and total volume of injected gas were recorded with DFMs. The data was stored

using DFM controller software offered by AALBORG. The rate and total volume of

produced gas were recorded with the WTM pulse generators. Then, the data were sent to

the computer and recorded with Ritter’s Rigomo software.

Figure 3-11 shows the experimental set up schematic, while Figure 3-12 shows the

experimental set up design, and Table 3-3 summarizes all the parts and equipment used

for building and designing the experimental setup.

61

Figure ‎3-10: Wet test meters (WTM)

62

Figure ‎3-11: Schematic diagram of the experimental set-up

63

Figure ‎3-12: Experimental setup

64

Table ‎3-3: List of experimental equipment

No. Device/ Model Number Manufacturer Quantity

1 Physical Models UOR Workshop 2

2 Separators UOR Workshop 2

3 Syringe Pump/ 1000D Teledyne isco 1

4 Vacuum Pump Fisher Scientific 1

5 Transfer Cylinder 3

6 Digital Flow Meter/ DFM26,

DFM27 Aalborg 4

7 Wet Test Meter/ TG05 Ritter 2

8 Digital Pressure Gauge/ Ashcroft 4

9 Digital Pressure Gauge/ Heise 1

10 High Pressure BPR Core Laboratories 2

11 Pressure Regulators Swagelok 5

12 Gas Cylinders Praxair 5

13 Vibrator/ ABU-38 Deca Vibrators Industries 1

14 Electric Balance/ Mettler TOLEDO 1

15 Heater/ Stirrer Fisher Scientific 1

16 Computer Dell 1

17 Light Source Underwriters Laboratories

Inc. 1

18 Digital Camera Canon 1

19 High Pressure Gas Sampler Swagelok 1

20 Two Way Valve Swagelok 10

21 Three Way Valve Swagelok 6

22 Ball Valve Swagelok 4

23 Check Valve Swagelok 4

24 1/8 “ Steel Tubing Swagelok 60 ft

25 1/4 “ Steel Tubing Swagelok 4 ft

26 1/8 “ Plastic Tubing Swagelok 30 ft

27 Buchner Flask PyrexPlus 1

28 Buchner Funnel Coorstek 2

29 Filter Paper/ No.2, No. 5 Whatman N/A

65

3.2 Materials

3.2.1 Sand

Ottawa sand #530 (Bell and Mackenzie Co. ltd., Canada) was used to pack the VAPEX

physical models. This is a white sand with a rounded grain shape and is 99.88% silicon

dioxide (SiO2). The specific gravity of the sand used for this study was 2.65 (γH2O=1.0).

Figure 3-13 shows the screen analysis for Ottawa sand #530. In each experiment,

approximately 4.3 kg of sand was used to pack the small model, while, for the large

model, approximately 16.5 kg of sand was used for packing.

3.2.2 Heavy oil

A heavy oil sample representing Saskatchewan heavy oil with a viscosity of 14271 mPa·s

at 21°C was used in this study. In order to reach the pre-specified dead oil viscosity for

the VAPEX experiments, kerosene was added to the oil sample. In order to get a

homogeneous oil sample, a mixture of kerosene and heavy oil was stirred for an hour,

and then the mixture was placed in an air bath that was heated for several hours.

Meanwhile, the mixture was stirred with a mixer. It was then cooled to 21°C before

viscosity was measured. This process was repeated if additional kerosene was needed to

get to the pre-specified viscosity. Then, the compositional analysis of the heavy oil

sample was obtained by using the simulated distillation method. The Saskatchewan

Research Council (SRC) lab undertook this analysis.

66

U. S. Sieve

0 25 50 75 100 125 150 175 200 225 250 275 300 325

wt

% R

etai

ned

0

10

20

30

40

Figure ‎3-13: Screen analysis for Ottawa sand #530

67

The results are shown in Table 3-4 and in Figure 3-14. It is obvious that there are no

hydrocarbon components under C9, and that the weight percent of C50+ is 10.54%. In

each experiment, approximately 1000 mL of heavy oil was used to saturate the small

model; for the large model, approximately 4000 mL of heavy oil was used for saturation.

3.2.3 Injection solvents and back pressure gas

Propane, butane, methane, nitrogen, and carbon dioxide gas cylinders were purchased

from Praxair Canada with the stated purity of 99.50%, 99.50%, 99.97%, 99.99%, and

99.99%. Propane, butane, methane, and carbon dioxide were injected as pure gases and as

mixture gases to be the solvent in the VAPEX experiments. The nitrogen gas was used

for the backpressure line to maintain the desired pressure using the BPRs for each test. It

was also used before starting the VAPEX experiments to conduct pressure leak tests.

68

Table ‎3-4: Compositional analysis result of the injection heavy oil with viscosity of 5650 mPa.s at 21°C

Carbon Number Mol.% Carbon Number Mol.%

C1 0.0 C31 1.20 C2 0.0 C32 1.16 C3 0.0 C33 0.80 C4 0.0 C34 0.76 C5 0.0 C35 0.97 C6 0.0 C36 1.02 C7 0.00 C37 0.61 C8 0.00 C38 0.57 C9 3.38 C39 0.95 C10 11.17 C40 0.96 C11 12.95 C41 0.53 C12 5.76 C42 0.58 C13 3.22 C43 0.80 C14 3.02 C44 0.75 C15 3.60 C45 0.50 C16 3.19 C46 0.49 C17 3.47 C47 0.51 C18 3.31 C48 0.50 C19 2.93 C49 0.39 C20 2.59 C50 0.38 C21 2.75 C51 0.42 C22 1.68 C52 0.41 C23 2.11 C53 0.38 C24 1.83 C54 0.33 C25 1.75 C55 0.31 C26 1.56 C56 0.31 C27 1.61 C57 0.29 C28 1.61 C58 0.31 C29 1.32 C59 0.30 C30 1.25 C60+ 6.44

502:WeightMolecular , mPa.s, 5650=oil 3

oil kg/m 971.53=

69

Figure ‎3-14: Hydrocarbon composition of injected oil

Carbon Number

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

C37

C38

C39

C40

+

Mo

l. %

0

2

4

6

8

10

12

14

70

3.3 Experimental procedure

Each of the VAPEX experiments was performed in three major steps. The first step was

preparation. During preparation, the model was packed with sand, pressure leaks were

tested, and the model was then vacuumed and saturated with oil. The next step was

running the experiments, which included the continuous solvent injection, monitoring the

process, and recording the data. The last step was unpacking and cleaning the model.

These steps are explained in detail in the following sub-sections.

3.3.1 Preparation

3.3.1.1 Sand packing

As mentioned earlier, the physical models were bolted on a movable stand with rotation

capability. For the packing, the VAPEX models were set into horizontal position while

one of the slabs on each model was bolted. The cavities of the VAPEX models were

packed with dry Ottawa sand. Then, the gaskets, second Plexiglas slabs, and steel

protection covers were bolted in sequence, and the models were set back to the vertical

position. At this point, additional sand was added with a funnel through the top injection

ports to pack the empty spaces. In order to achieve more homogeneous packing, wet

packing and simultaneous shaking were conducted; water was used for wet packing due

to solution glazing behaviours of other solvents such as acetone on Plexiglas slabs. Next,

a syringe pump was used to inject the water in the models. The models were saturated

with water through the top injection points, and they were vibrated to get uniform

packing. For vibrating the models, an ABU38 pneumatic ball vibrator from Deca

Vibrator Industries Inc. was used. The vibrating was continued for 24 hours. Then,

pressurized air was injected for 24 hours to dry the sand and prepare the sand packs for

71

porosity measurement. After the air injection, the models were vibrated again for several

hours. At the same time, the models were rotated manually back and forth at 45° angles

to add sand to any void space at the top portion of the models. At this point, the

connections and required fittings, valves, and piping were connected to the top and

bottom ports of the physical models. Then, nitrogen was injected into the models at the

maximum allowable operating pressure of the VAPEX models to conduct the pressure

test and look for any possible leakage. In the last step, the physical models were

evacuated with a Fisher Scientific vacuum pump. For the large model, the evacuation was

conducted for 7 hours in 1-hour intervals, and, for the small model, it was done for 3

hours in 1-hour intervals. Pressure gauges were mounted to monitor the vacuum process.

Figure 3-15 shows the sand-packed VAPEX models.

3.3.1.2 Porosity measurement

After evacuating the VAPEX models, the solvent injection and production points and any

other connection points at the inlet and outlet ports were sealed tightly. The imbibition

method (Dong et al., 2006) was used to measure the porosity. By measuring the volume

of the water imbibed in the sand pack, the pore volume of the sand pack was measured.

The ratio of the pore volume to the total bulk volume was determined to be the sand

pack’s porosity.

72

Figure ‎3-15: Sand-packed VAPEX models

73

3.3.1.3 Oil saturation

In this study, initial water saturation was not considered. So, before oil injection, the sand

pack was dried with pressurized air. To get uniform oil saturation in the VAPEX models,

the oil was injected to the VAPEX models through the bottom connection points.

Therefore, oil was injected through two valves for the small model and four valves for the

large model. For this purpose, a high-pressure transfer cell was employed and connected

to a Teledyne 1000D syringe pump. To push the piston upward, tap water was injected to

the lower portion of the transfer cell. Doing so displaced the oil into the VAPEX models.

Because of the pressure constraints of the physical models, the injection rate was very

low, which made the oil saturation process very slow. It took about 2 to 3 days to saturate

the small model and about 6 to 7 days to saturate the large model.

The oil saturation set-up schematic is shown in Figure 3-16. During the saturation period,

the pressures of the physical models were monitored carefully to avoid over pressuring.

Figure 3-17 shows the saturated VAPEX models.

74

Figure ‎3-16: The schematic of the oil saturation set-up

Transfer Cell

Wat

er In

ject

ion

Oil

Inje

ctio

nSyringe Pump

75

Figure ‎3-17: Oil saturated sand packs

76

3.3.1.4 Permeability measurement

To calculate permeability, pressure drops at the injection and production points were

recorded. Then, the Darcy equation was used to measure the permeability. For this

purpose, the oil was injected at different flow rates, and, at each time, the stabilized

pressure drop was recorded to measure permeability. This procedure was repeated five

times and the average permeability value was measured for each test.

To confirm the value obtained by this method, the following equation proposed by

Carmen-Kozeny and modified by Panda and Lake (Faruk, 2007) was employed:

………………………………………………………….. (3.1)

where DP is the particle size, is porosity, is skewness, σ is variance, CP is the

coefficient of variance, and is tortuosity.

PPPP dDDfDD

3

0

3

1

………...……………………………...……….……. (3.2)

PPPP dDDfDD

2

0

2

….………………………………………….………….. (3.3)

P

PD

C

……………………………………………………….……...…………….. (3.4)

Kozeny-Carman-based models are the most common and oldest models used for

estimating permeability. These models treat porous media as bundles of capillary tubes of

equal length and constant cross section. Kozeny derived the equation to predict

permeability by solving the Navier-Stokes equation for all capillary tubes passing

222

22332

1172

13

p

ppp

C

CCDk

77

through a point (Krause, 2009).Carman (1937) modified the Kozeny equation to its new,

more recognized form. Kozeny-Carman-based models are mostly used by researchers to

estimate permeability (Krause, 2009).

3.3.2 VAPEX experiments

In this study, a total of eighteen VAPEX experiments were carried out. Different solvents

were injected in two VAPEX models with different drainage heights. Once the models

were saturated, the solvent injection line was connected to the top connection ports of the

VAPEX models. The solvent was injected at constant pressure from the gas cylinders to

DFMs and then to the VAPEX models at a pre-specified constant pressure. The flow

rates and total injected solvent volumes were recorded by the DFMs. For each VAPEX

test, the solvent was injected to the physical models at the operating pressure while the

production pressure was atmospheric pressure and the solvent and oil production was

monitored carefully. The pressure at the production point was implemented after which

the connection between the injection and production well was visually observed.

Once the oil was produced through the BPR, it was collected in the separators. By

reading from the calibrated visual separators, the cumulative produced oil was recorded

regularly during the course of the experiments. The produced gas was separated, and

then, from the top valves on each of the separators, the produced gas was passed through

the WTMs to measure its total volume. During the experiments, the produced oil samples

were collected at the production point in small oil containers. Next, the weight of the

produced oil was recorded with a high precession Mettler TOLEDO electric balance.

Then, the samples were kept for 7 days at atmospheric pressure, and the final weight of

each sample was recorded to find the dissolved solvent amount.

78

In addition, a Canon EOS T3i digital camera, in conjunction with a fluorescent light

source, was used to take digital images of the solvent chamber and its evolution at

different times during the tests. These images were further used for Image Analysis (IA)

purposes.

The compositional analysis of the heavy oil samples collected from the separators was

obtained by using the simulated distillation method. Furthermore, the density and

viscosity of the produced oil sample from each VAPEX model was measured for each

test.

A continuous presence of an operator was needed during the experiments for manually

recording some of the above-mentioned data. The solvent-leaching gravity-drainage

process is a very slow recovery process; therefore, this part of the experiments took about

7 to 21 days, depending on the size of the model and the type of solvent used for each

experiment.

The termination time for each test was considered the time at which a stabilized oil

production rate was monitored when the gas production rate was significantly high. At

this stage, solvent injection was shut down and the injection valves were closed. Next, the

models were depressurized and a blow down process was initiated. The production valves

were kept open until the pressure in the VAPEX models reached atmospheric pressure

and no more oil and solvent were produced.

79

3.3.3 Residual oil saturation and asphaltene content measurement

At this stage of the experiments, the connection lines at the injection and production ports

of the VAPEX models were opened, and the models were set to horizontal position. The

VAPEX models were then disassembled carefully. From the horizontal position, and

because of the special design of the VAPEX models, each slab could be taken apart from

the main steel frame separately while the other slab remained in place. Once the top slab

was removed, four different samples were collected from four different locations of each

VAPEX model. These sand pack samples were picked to locally cover different parts of

the sand packs. Sample 1 was collected near the injection point; sample 2 was collected

from the transition zone; sample 3 was collected in the oil zone between the transition

zone and the production point, while sample 4 was collected near the production point.

Figure 3-18 shows the different sand pack sample locations. Finally, the residual oil

saturation and asphaltene content of each sample were measured.

80

Figure ‎3-18: Sample locations, (a) Small model, (b) Large model

81

3.3.3.1 Residual oil saturation measurement

To measure the oil saturation in each sample, oil was separated from the sand. The

weight of each oil sample was measured individually, and, by knowing the oil density

and the specific gravity of the sand, the volume for each component was calculated.

The setup shown in Figure 3-19 was used to separate the oil from the sand. Toluene was

added to each sample. Then, the mixture was passed through Whatman No. 2-filter paper.

The filter paper was mounted on a Buchner funnel that was sealed on a Buchner Flask.

The jar was connected to vacuum pump, and, then, toluene was added gradually to the

sample until no oil was observed in the sand. The mixture of drained oil and toluene was

collected in the Buchner Flask. The collected mixture was kept in the air bath until the

toluene evaporated from the mixture. Finally, the weight of the remaining oil was

recorded to calculate residual oil saturation.

3.3.3.2 Asphaltene content measurement

The asphaltene content of each sample was measured using the standard ASTM D2007-

03 method. The precipitant used here was n-pentane. n-pentane was added to the oil

sample and stirred thoroughly. Then, the mixture was filtered through 0.2 μm Whatman

No. 5 filter paper as shown in Figure 3-20. The n-Pentane was added to the oil mixture on

the filter paper and was stirred continuously; this process was continued until clean liquid

drainage was monitored from the filter paper. Afterward, the asphaltene precipitant on the

filter paper was kept in the air bath for one day to dry completely. The final weight of the

asphaltene precipitate was recorded to measure the asphaltene content of each sample.

82

Figure ‎3-19: Schematic of the set up used to separate the oil from the sand

E-1

P-1

Mixture of Toluene

and Sand sample

Filtrate collects here

Rubber Bung

E-2

Filter Paper

Buchner

Funnel

Buchner

Flask

Rubber

Tubing Vacuum

Pump

Extractor

hood

Oil

sample/

toluene

mixture

83

Figure ‎3-20: Schematic of the set up used to measure the asphaltene content of the oil samples

E-1

P-1

Mixture of n-Pentane

and oil sample

Filtrate collects here

Rubber Bung

E-2

Filter Paper

Buchner

Funnel

Buchner

Flask

Rubber

Tubing

Vacuum

Pump

Extractor

hoodAsphaltene

precipitate

Electric

balance

Asphaltene

precipitate

84

3.3.4 Cleaning

After each test, the VAPEX models and all the connection points, lines, valves, and

fittings were disassembled and cleaned to be ready for the next set of experiments. The

lines and connection points to the VAPEX models were removed and the models were set

to the horizontal position. Then, the steel cover protectors, the Plexiglas slabs, and the

gaskets were removed. After taking the required samples, the sand was discarded from

the VAPEX models to a dumping container. Because of the fine grains and the residual

oil in place, the cleaning procedure was cumbersome, especially for the large model.

However, once the models were unpacked, they were set back to the vertical position.

Next, the second steel cover protectors, the Plexiglas slabs, and the gaskets were

removed. The steel frame and all the piping and valves were cleaned with toluene before

being dried with pressurized air. The Plexiglas slabs were cleaned with non-corrosive

(kerosene and conventional glass cleaners) solvents. The same procedure was used to

clean the separators.

The total time required for running each test, including preparation, experimental runs

and cleaning was at least one month for each test. In some cases, unexpected leakages or

equipment failure could extend this time to two months.

85

4. CHAPTER 4: EXPERIMENTAL RESULTS AND

DISCUSSION

A total of18 experiments were conducted using different solvents. Propane, CO2,

methane, butane, a mixture of propane/CO2 (70%/30%), and a mixture of

propane/methane (70%/30%) were considered as respective injection solvents to carry

out the VAPEX experiments. The experiments were carried out at temperature of 21°C

and a pressure of 110 to 850 kPa. Silica sand number 530 was used for packing the

models. The summary of operating conditions for the VAPEX experiments is provided in

Table 4-1.

At times, there were sudden oil production rate fluctuations. This can be due to pressure

disturbance during gas injection. However, in the case of propane injection (pure and

mixture), this was more severe; however, different researchers have also observed these

sudden production rate changes (Yazdani and Maini, 2005 and Ahmadloo et al., 2011). In

fact, Ahmadloo et al. (2011) suggested that the fluctuations could be due to the re-

imbibition of the oil phase in swept zones of porous media. Furthermore, re-imbibition is

more significant in larger models due to faster drainage. Along these lines, Yazdani and

Maini (2005) suggested that this fluctuation could be the result of asphaltene precipitation

near production points. In such instances, the production port blockage caused by

asphaltene deposition will lead to a surge of oil that may cause fluctuations. After

measuring the asphaltene content of the injected oil and different samples from different

86

physical model locations, noticeable asphaltene precipitation was observed specifically

near the production ports.

The produced oil from each model was analyzed to measure its density and viscosity.

Following each test, a produced oil compositional analysis was carried out. On the one

hand, propane injection in the small model had the highest ultimate oil recovery factor.

On the other hand, the tests with pure CO2 injection have the lowest recovery factor.

However, the propane/CO2 mixture injection shows significant results in both models.

Different parameters are investigated during the test in the small and large models using

different solvents. Recovery factor, flow rate, solvent utilization factor (SUF), viscosity,

density, molecular weight, and hydrocarbon content of produced oil are described in

more detail in the following sections. Then, the results for asphaltene content experiments

are demonstrated. In the last section, a comprehensive image analysis (IA) is carried out

for the tests in order to monitor the solvent chamber velocity and drainage height in more

detail.

87

Table ‎4-1: Operating conditions of the VAPEX experiments

Test

No.

Model

height

(cm)

Solvent

Porosity

(%)

Permeability

(D)

Pressure

(kPa)

Temperature

(°C)

Oil

Density

(kg/m3)

Oil

Viscosity

(mPa.s)

1 47.5 propane 38.7 7.90 700 21 971.53 5650

2 24.5 propane 36.9 5.32 700 21 971.53 5650

3 47.5 propane 39.3 6.51 700 21 971.53 5650

4 24.5 methane 40.7 5.12 850 21 971.53 5650

5 47.5 methane 41.8 5.88 850 21 971.53 5650

6 24.5 CO2 42.1 6.11 850 21 971.53 5650

7 47.5 CO2 42.6 6.70 850 21 971.53 5650

8 24.5 propane/ CO2 41.5 5.63 850 21 971.53 5650

9 47.5 propane/ CO2 42.2 5.79 850 21 971.53 5650

10 24.5 butane 42.4 9.63 140 21 971.53 5650

11 24.5 butane 42.1 8.69 110 21 971.53 5650

12 47.5 butane 42.3 9.08 110 21 971.53 5650

13 24.5 propane 42.2 8.78 700 21 971.53 5650

14 47.5 propane 43.1 9.12 700 21 971.53 5650

15 24.5 propane/CO2 41.8 8.64 850 21 971.53 5650

16 47.5 propane/CO2 42.4 8.87 850 21 971.53 5650

17 24.5 propane/methane 42.0 8.50 850 21 971.53 5650

18 47.5 propane/methane 42.1 9.23 850 21 971.53 5650

88

4.1 VAPEX performance

4.1.1 Effect of drainage height

In this section, different parameters between the two physical models are compared.

Small model results are graphed alongside large model results.

4.1.1.1 Recovery factor and produced oil rate

4.1.1.1.1 Propane injection

Figure 4-1 shows the recovery factor after injecting propane as the solvent in the physical

models. Here, the ultimate recovery factors in the small and large models were

approximately the same and about 75% of original oil in place. However, it should be

mentioned that two more tests were conducted prior to these tests, and it was found that

connection between injection and production wells would significantly affect the results.

This disparity will be discussed in more detail later in this chapter.

The produced oil flow rate is also shown in Figure 4-2. It was found that stabilized

drainage rate was higher in the large model due to greater drainage height and effect of

gravity drainage. As can be seen, the stabilized drainage rate for the small model was

about 0.22 mL/min and 0.50 mL/min for the large model. Some fluctuations were

observed for the drainage rate during the experiments, which was also reported by some

other researchers (Ahmadloo et al., 2012 and Yazdani, 2007). The possible causes for

these fluctuations are a pressure drop in the pressure regulator or the backpressure

regulator and asphaltene precipitation may also cause such fluctuations. After conducting

the asphaltene precipitation experiments, asphaltene precipitation at different locations

was observed.

89

Time (h)

0 20 40 60 80 100 120 140

Rec

ov

ery F

acto

r (%

OO

IP)

0

20

40

60

80

100

Small Model

Large Model

Figure ‎4-1: The recovery factor after propane injection in VAPEX models

90

Time (hr)

0 20 40 60 80 100 120 140

Pro

duced

Oil

Rat

e (m

L/m

in)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Large Model

Small Model

Figure ‎4-2: The produced oil rate after propane injection in VAPEX models

91

4.1.1.1.2 Methane injection

Figure 4-3 shows the recovery factor after injecting methane as the solvent in the physical

models. Of note, the ultimate recovery factors in the small and large model were 36%

and 32% of original oil in place, respectively. As the figure demonstrates, the recovery

factor increases steadily for the small and large models. In addition, the process was

significantly slower than in the case of propane injection, specifically for the small

model. Moreover, the first solvent breakthrough was observed later in comparison to the

propane injection, which is due to the low solubility of methane in the operating injection

pressure.

The produced oil flow rates in the small and large physical models are presented in

Figure 4-4. Furthermore, the flow rate is relatively higher for the larger model with

greater drainage height, while production rate fluctuations were observed with methane

injection, specifically in the small model. It was found that stabilized drainage rate in the

small model was about 0.027 mL/min, while the stabilized drainage rate in the large

model with greater drainage height was about 0.057 mL/min. Considering these results, it

was found that pure methane in the operating pressure of the experiments was not a good

choice for injection solvent.

92

Time (h)

0 100 200 300 400 500

Rec

ove

ry F

acto

r (%

OO

IP)

0

10

20

30

40

Small Model

Large Model

Figure ‎4-3: The recovery factor after methane injection in VAPEX models

93

Time (h)

0 50 100 150 200 250 300 350

Pro

duc

ed o

il ra

te (

mL

/min

)

0.00

0.02

0.04

0.06

0.08

0.10

Large Model

Small Model

Figure ‎4-4: The produced oil rate after methane injection in VAPEX models

94

4.1.1.1.3 CO2 injection

Figure 4-5 shows the recovery factor after injecting CO2 as the solvent in the physical

models. Here, the ultimate recovery factor increased steadily in both models, and it was

almost the same in both physical models and found to be about 36% of original oil in

place. In addition, the process was significantly slow, specifically for the small model.

The produced oil flow rates are also shown in Figure 4-6. As can be seen, the stabilized

flow rate was significantly higher for the large model with greater drainage height due to

the gravity drainage. The stabilized drainage rates were 0.012 mL/min and 0.028 mL/min

for the small and large models, respectively. As with the methane injection, the results

showed that pure CO2 injection is not a good choice as an injection solvent. The low

production for this case may be due to the low injection pressure, which was limited

because of VAPEX model specifications.

4.1.1.1.4 Butane injection

As shown in Figure 4-7, the ultimate recovery factors for both models were almost the

same, and they were approximately 57% of original oil in place. Moreover, the process

seemed faster in the small model compared to butane injection in large model. This

disparity may be the result of well configuration and the shorter distance between the

injection and production well in the small model. However, compared to propane

injection, the process was significantly slower during butane injection, which might be

due to the low vapour pressure of butane. The produced oil flow rate is also shown in

Figure 4-8. As expected, the produced oil flow rate was higher in the large model. The

stabilized drainage rate was about 0.32 mL/min in the large model and about 0.14

mL/min in the small model. As it will be explained later in this chapter, the slow rate of

the process resulted in asphaltene precipitation especially close to the injection wells.

95

Time (h)

0 100 200 300 400 500 600

Rec

ove

ry F

acto

r (%

OO

IP)

0

5

10

15

20

25

30

Small Model

Large Model

Figure ‎4-5: Recovery factor after CO2 injection in the VAPEX models

96

Time (h)

0 100 200 300 400 500

Pro

duc

ed o

il ra

te (

mL

/min

)

0.00

0.01

0.02

0.03

0.04

Large Model

Small Model

Figure ‎4-6: Produced oil rate after CO2 injection in the VAPEX models

97

Time (h)

0 20 40 60 80 100 120 140

Rec

over

y F

acto

r (%

OO

IP)

0

10

20

30

40

50

60

70

Small Model

Large Model

Figure ‎4-7: Recovery factor after butane injection in the VAPEX models

98

Time (h)

0 20 40 60 80 100 120 140

Pro

du

ced

Oil

Rat

e (m

L/m

in)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Large model

Small model

Figure ‎4-8: Produced oil rate after butane injection in the VAPEX models

99

4.1.1.1.5 Propane/CO2 injection

Figure 4-9 shows the recovery factor after injecting propane/CO2 as the solvent in the

physical models. Here, the ultimate recovery factors for both models were almost the

same, and they were about 54% of original oil in place. The performance of the VAPEX

process was significantly improved compared to pure CO2 injection, specifically for the

small model. The results proved the suitability of CO2 as a carrier gas for solvents such as

propane and butane in the VAPEX process.

As can be seen in Figure 4-10, the flow rate was significantly higher in the large model,

and it was found to be around 0.33 mL/min. The observed flow rate in the small model

was about 0.15 mL/min.

Because of the pressure constraints of the physical model, a mixture of 30% CO2 and

70% propane was injected as the solvent. The performance trend was close to the case of

pure propane injection. By increasing the volume percent of CO2, the vapour pressure of

the mixture will be increased, meaning it will exceed the maximum pressure tolerance of

the physical models used in these tests. However, new models can be employed to

perform a sensitivity analysis on the recovery performance of solvent mixtures with

different volume percentages of CO2.

100

Time (h)

0 20 40 60 80 100 120 140

Rec

ov

ery F

acto

r (%

OO

IP)

0

10

20

30

40

50

60

Small

Large

Figure ‎4-9: Recovery factor after propane/CO2mixture injection in the VAPEX models

101

Time (h)

0 20 40 60 80 100 120 140

Pro

du

ced

Oil

Rat

e (m

L/m

in)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Large model

Small model

Figure ‎4-10: Produced oil rate after first propane/CO2 injection in VAPEX models

102

4.1.1.1.6 Propane/methane injection

Figure 4-11 shows the recovery factor after injecting propane/methane mixture as the

solvent in the VAPEX physical models. In this case, the ultimate recovery factor was

higher in the small model, and it was about 48% of original oil in place. On the other

hand, the ultimate recovery factor achieved in the large model was about 40% of original

oil in place. The performance of the VAPEX process was significantly improved

compared to pure methane injection. The results proved the suitability of methane as a

carrier gas for solvents such as propane and butane in the VAPEX process.

As can be seen in Figure 4-12, the flow rate was significantly higher in the large model,

and it was found to be around 0.25 mL/min. The observed flow rate in the small model

was about 0.13 mL/min.

Because of the pressure constraints of the physical model, a mixture of 30% methane and

70% propane was injected as the solvent. By increasing the volume percent of methane,

the vapour pressure of the mixture would be increased; therefore, to monitor the

performance of the process with such a solvent new physical models with higher pressure

tolerance would be required.

103

Time (h)

0 20 40 60 80 100 120 140

Rec

ov

ery F

acto

r (%

OO

IP)

0

10

20

30

40

50

60

Small

Large

Figure ‎4-11: Recovery factor after propane/methane mixture injection in the VAPEX models

104

Time (h)

0 20 40 60 80 100 120 140 160

Pro

du

ced

Oil

Rat

e (m

L/m

in)

0.0

0.1

0.2

0.3

0.4

0.5

Large model

Small model

Figure ‎4-12: Produced oil rate after propane/methane mixture injection in the VAPEX models

105

4.1.1.2 Solvent utilization factor (SUF)

During the experiments, the amount of injected solvent was recorded by DFMs for

various solvents. The solvent utilization factor (SUF) at any time during the experiments

is the ratio of the net oil production to the total injected volume of solvent. This

parameter was calculated with equation (4.1). The results are shown in more detail in the

following sections.

)(

)(

mLsolventinjectedofvolumeTotal

mLproductionoilNetSUF ............................................................ (4.1)


As Figure 4-13 demonstrates, the SUF is higher in the large model compared to the

results obtained for the small model. In fact, the SUF increases gradually until the final

breakthrough of the gas, after which there would be a great amount of solvent production

with less oil produced. The results showed that up-scaling the VAPEX process did not

result in solvent loss although the distance between the injection and production wells

was significantly increased.


Figure 4-14 indicates that SUF is higher in the small model in the case of methane

injection. In effect, the SUF increases until the final breakthrough of the gas. At that time,

there is a sharp decrease in the SUF for both the small and large models. At this time, the

process is no longer efficient because there is little oil production with the amount of

solvent injected. The SUF curves can also be used as an indicator of the final shut-in time

for the VAPEX tests. However, some fluctuations may occur due to the sudden pressure

drops in the BPR lines, which may cause a solvent production increase.

106

Time (h)

0 20 40 60 80 100 120 140

SU

F (

mL

Oil

Pro

d./

mL

Sol.

Inj.

)

5.0x10-4

10-3

1.5x10-3

2.0x10-3

2.5x10-3

3.0x10-3

3.5x10-3

Small Model

Large Model

Figure ‎4-13: Solvent utilization factor (SUF) after propane injection in VAPEX models

107

Time (h)

50 100 150 200 250 300 350 400 450

SU

F (

mL

Oil

Pro

d./

mL

Sol.

Inj.

)

0

2.0x10-4

4.0x10-4

6.0x10-4

8.0x10-4

10-3

1.2x10-3

1.4x10-3

Small Model

Large Model

Figure ‎4-14: Solvent utilization factor (SUF) after methane injection in VAPEX models

108


The SUF is also shown in Figure 4-15. As can be seen in this figure, the SUF is higher in

the small model. In fact, the SUF increases until the final breakthrough of the solvent.

Then, there would be a sharp decrease in the SUF for both the small and large models. At

this time, the process is no longer efficient, as there is little oil production with the

solvent injected. The low injection pressure of pure CO2 resulted in low efficiency of the

process in the small model with small drainage height.


The SUF after butane injection is shown in Figure 4-16. As can be seen in this figure, the

SUF is higher in the small model, specifically at the earlier stages of the experiments.

This can be due to the lower vapour pressure of butane and the longer distance of the

injection and production wells in the large model. In fact, the results show that up-scaling

the VAPEX process in the case of butane injection might result in some additional

solvent loss. However, the increase rate of SUF is higher in the large model compared to

the small model, and the total SUF is greater in the large physical model.


Figure 4-17 indicates that SUF is higher in the large model compared to the small model

for the case of propane/CO2 mixture injection. As can be seen in this figure, the total SUF

obtained in the large model is about 0.0024 (mL oil prod./mL of gas inj.), while the total

SUF for the small model is about 0.0019 (mL oil prod./mL of gas inj.). In fact, the SUF

increases slightly until the final breakthrough of the gas in the large and small physical

models.

109

Time (h)

0 50 100 150 200 250 300 350 400 450 500

SU

F (

mL

Oil

Pro

d./

mL

So

l. In

j.)

0

10-4

2x10-4

3x10-4

4x10-4

5x10-4

Small Model

Large Model

Figure ‎4-15: Solvent utilization factor (SUF) after CO2 injection in VAPEX models

110

Time (h)

0 20 40 60 80 100 120 140 160

SU

F (

mL

Oil

Pro

d./

mL

Sol.

Inj.

)

2x10-4

4x10-4

6x10-4

8x10-4

10-3

1x10-3

1x10-3

2x10-3

2x10-3

2x10-3

Small Model

Large Model

Figure ‎4-16: Solvent utilization factor (SUF) after butane injection in VAPEX models

111

Time (h)

0 20 40 60 80 100 120 140 160

SU

F (

mL

Oil

Pro

d./

mL

Sol.

Inj.

)

4.0x10-4

6.0x10-4

8.0x10-4

10-3

1.2x10-3

1.4x10-3

1.6x10-3

1.8x10-3

2.0x10-3

2.2x10-3

2.4x10-3

Small Model

Large Model

Figure ‎4-17: Solvent utilization factor (SUF) after propane/CO2 injection in VAPEX models

112


The SUF obtained in the small and large models after injecting propane/methane mixture

is shown in Figure 4-18. As can be seen in this figure, the SUF is almost the same at

earlier stages of the experiments for small and large models. However, the increase rate

of SUF is higher for the large model toward the end of the experiments, and the SUF

increases until the final breakthrough of gas for both models. The total SUF for the large

model is about 0.0022 mL oil prod./mL of gas inj. while the total SUF achieved in the

small model is about 0.0019 mL oil prod./mL of gas inj..

4.1.1.3 Viscosity, density, molecular weight, and hydrocarbon components for the

produced oil

In this section, the effect of solvent injection on viscosity, density, molecular weight, and

the hydrocarbon components of original injection oil in the small and large models are

demonstrated.


Tables 4-2 and 4-3 show the compositional analysis results of the produced heavy oil

after propane injection in the small and large models, respectively. Viscosity, molecular

weight, and produced oil density are also presented. Propane injection has significantly

diluted the original oil and reduced its viscosity. The density and the molecular weight

are also decreased due to the extraction of some of the components. Figure 4-19 shows

the compositional analysis of the produced oil for the small and large models. The

comparison of the hydrocarbon components with the original oil shows an increase in the

mol% of lighter hydrocarbons. However, the lighter components have increased more in

the small model than in the large model.

113

Time (h)

0 20 40 60 80 100 120 140 160

SU

F (

mL

Oil

Pro

d./

mL

So

l. I

nj.

)

0

2.0x10-4

4.0x10-4

6.0x10-4

8.0x10-4

10-3

1.2x10-3

1.4x10-3

1.6x10-3

Small Model

Large Model

Figure ‎4-18: Solvent utilization factor (SUF) after propane/methane injection in VAPEX models

114

Table ‎4-2: Compositional analysis result of the produced heavy oil after propane injection in small model


C1 0.0 C31 1.10

C2 0.0 C32 0.96

C3 12.49 C33 0.72

C4 0.0 C34 0.81

C5 0.0 C35 0.83

C6 0.0 C36 0.80

C7 0.0 C37 0.63

C8 0.0 C38 0.53

C9 2.45 C39 0.78

C10 9.67 C40 0.81

C11 10.88 C41 0.45

C12 4.58 C42 0.46

C13 2.64 C43 0.70

C14 2.73 C44 0.79

C15 3.31 C45 0.41

C16 2.90 C46 0.40

C17 3.00 C47 0.47

C18 3.04 C48 0.37

C19 2.49 C49 0.36

C20 2.19 C50 0.39

C21 2.52 C51 0.39

C22 1.54 C52 0.37

C23 1.80 C53 0.35

C24 1.65 C54 0.29

C25 1.61 C55 0.29

C26 1.47 C56 0.29

C27 1.45 C57 0.31

C28 1.42 C58 0.27

C29 1.18 C59 0.27

C30 1.10 C60+ 6.31


oil kg/m 853.50=

115

Table ‎4-3: Compositional analysis result of the produced heavy oil after propane injection in large model


C1 0.0 C31 1.03

C2 0.0 C32 1.05

C3 5.27 C33 0.84

C4 0.0 C34 0.82

C5 0.0 C35 0.80

C6 0.0 C36 0.72

C7 1.00 C37 0.66

C8 1.72 C38 0.72

C9 6.32 C39 0.64

C10 12.23 C40 0.64

C11 8.58 C41 0.56

C12 3.34 C42 0.57

C13 2.99 C43 0.58

C14 3.25 C44 0.56

C15 3.31 C45 0.55

C16 3.08 C46 0.50

C17 3.35 C47 0.40

C18 2.85 C48 0.43

C19 2.71 C49 0.40

C20 2.44 C50 0.35

C21 2.20 C51 0.35

C22 1.99 C52 0.35

C23 1.79 C53 0.30

C24 1.67 C54 0.29

C25 1.57 C55 0.30

C26 1.57 C56 0.28

C27 1.50 C57 0.25

C28 1.30 C58 0.26

C29 1.25 C59 0.27

C30 1.08 C60+ 6.17


oil kg/m 938.17=

116

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C1

0C

11

C1

2C

13

C1

4C

15

C1

6C

17

C1

8C

19

C2

0C

21

C2

2C

23

C2

4C

25

C2

6C

27

C2

8C

29

C3

0C

31

C3

2C

33

C3

4C

35

C3

6C

37

C3

8C

39

C4

0+

Mole

%

0

2

4

6

8

10

12

14

16

Injection oil

Produced oil after propane injection in small model

Produced oil after propane injection in large model

Figure ‎4-19: Compositional analysis of the produced oil after propane injection

117

Likewise, Luo et al. (2005) found that dissolution of propane can significantly reduce the

heavy oil viscosity in the VAPEX process. However, they observed the formation of a

multilayer solvent-heavy oil system, in which the top layer is a solvent-enriched, liquid

phase heavy oil with the dissolved solvent comprising the middle layer, while the bottom

layer mainly consists of heavy components.



after methane injection in the small and large models, respectively. Viscosity, molecular

weight, and produced oil density are also shown. It can be seen that methane injection has

not significantly diluted the original oil. Moreover, viscosity reduction is much less than

during propane injection. This can be due to the low solubility of methane in heavy oil at

the operating pressure. Regardless, the density and the molecular weight change are not

promising either. However, more viscosity reduction was observed in the small model

and, consequently, more dilution occured. Figure 4-20 shows the compositional analysis

of the produced oil for the small and large models. Ultimately, the comparison of the

hydrocarbon components with the original oil does not show any significant change in

the hydrocarbon component of the produced oil for each model.

118

Table ‎4-4: Compositional analysis result of the produced heavy oil after methane injection in small model


C1 0.0 C31 1.22

C2 0.0 C32 1.18

C3 0.0 C33 0.81

C4 0.0 C34 0.90

C5 0.0 C35 0.99

C6 0.0 C36 0.95

C7 0.0 C37 0.70

C8 0.0 C38 0.59

C9 4.02 C39 0.99

C10 10.26 C40 0.98

C11 11.53 C41 0.54

C12 5.50 C42 0.62

C13 3.15 C43 0.84

C14 3.13 C44 0.83

C15 3.80 C45 0.46

C16 3.25 C46 0.45

C17 3.50 C47 0.60

C18 3.55 C48 0.53

C19 2.81 C49 0.45

C20 2.63 C50 0.43

C21 2.93 C51 0.43

C22 1.77 C52 0.38

C23 1.97 C53 0.35

C24 1.86 C54 0.34

C25 1.95 C55 0.35

C26 1.59 C56 0.30

C27 1.64 C57 0.30

C28 1.64 C58 0.34

C29 1.34 C59 0.29

C30 1.27 C60+ 6.79


oil kg/m 969.28=

119

Table ‎4-5: Compositional analysis result of the produced heavy oil after methane injection in large model


C1 0.0 C31 1.25

C2 0.0 C32 0.78

C3 0.0 C33 1.28

C4 0.0 C34 0.82

C5 0.0 C35 1.00

C6 0.0 C36 1.05

C7 0.00 C37 0.63

C8 0.00 C38 0.67

C9 3.94 C39 0.98

C10 9.26 C40 0.90

C11 11.00 C41 0.65

C12 5.15 C42 0.53

C13 3.42 C43 0.91

C14 3.32 C44 0.88

C15 3.77 C45 0.47

C16 3.45 C46 0.45

C17 3.75 C47 0.53

C18 3.50 C48 0.51

C19 3.02 C49 0.45

C20 2.66 C50 0.43

C21 3.02 C51 0.44

C22 1.73 C52 0.42

C23 2.17 C53 0.35

C24 1.88 C54 0.35

C25 1.97 C55 0.32

C26 1.68 C56 0.35

C27 1.62 C57 0.34

C28 1.74 C58 0.28

C29 1.39 C59 0.31

C30 1.30 C60+ 6.90


oil kg/m 970.11=

120

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

C37

C38

C39

C40+

Mo

le %

0

2

4

6

8

10

12

14

16

Produced oil after methane injection in small model

Injection oil

Produced oil after methane injection in large model

Figure ‎4-20: Compositional analysis of the produced oil after methane injection

121



after CO2 injection in the small and large models, respectively. Viscosity, molecular

weight, and produced oil density are shown in the above-mentioned tables. It is apparent

that CO2 injection has not significantly diluted the original oil. In fact, viscosity reduction

is much less than what happened during propane injection. This can be due to the low

solubility of CO2 in heavy oil at the operating pressure of the experiments. However, the

viscosity reduction in the large model is more prominent than in the small model.

Furthermore, changes in density and molecular weight are not promising either. Figure 4-

21 shows the compositional analysis of the produced oil for the small and large models.

Here, a comparison of the hydrocarbon components with the original oil does not show

any significant change in the hydrocarbon component of the produced oil for each model.



after butane injection in the small and large models, respectively. Viscosity, molecular

weight, and produced oil density are also presented in those tables. A significant viscosity

reduction was observed after injecting butane. However, viscosity reduction is less than

what happened during propane injection. On the other hand, the viscosity reduction in the

large model is more prominent than in the small model. Figure 4-22 shows the

compositional analysis of the produced oil for the small and large models. The

comparison of the hydrocarbon components with the original oil shows an increase in the

mol% of lighter hydrocarbons. However, the lighter components have increased more in

the small model than in the large model.

122

Table ‎4-6: Compositional analysis result of the produced heavy oil after CO2 injection in small model


C1 0.0 C31 1.35

C2 0.0 C32 1.18

C3 0.0 C33 0.81

C4 0.0 C34 0.78

C5 0.0 C35 1.09

C6 0.0 C36 0.96

C7 0.0 C37 0.64

C8 0.0 C38 0.69

C9 4.02 C39 0.91

C10 10.38 C40 1.08

C11 10.69 C41 0.54

C12 5.11 C42 0.57

C13 3.35 C43 0.84

C14 3.25 C44 0.88

C15 3.84 C45 0.46

C16 3.41 C46 0.45

C17 3.81 C47 0.52

C18 3.37 C48 0.51

C19 3.20 C49 0.44

C20 2.63 C50 0.43

C21 2.80 C51 0.43

C22 1.90 C52 0.41

C23 2.15 C53 0.39

C24 1.86 C54 0.37

C25 1.82 C55 0.31

C26 1.72 C56 0.30

C27 1.78 C57 0.33

C28 1.55 C58 0.29

C29 1.43 C59 0.29

C30 1.14 C60+ 6.56


oil kg/m 962.88=

123

Table ‎4-7: Compositional analysis result of the produced heavy oil after CO2 injection in large model


C1 0.0 C31 1.22

C2 0.0 C32 1.18

C3 0.0 C33 0.81

C4 0.0 C34 0.82

C5 0.0 C35 1.05

C6 0.0 C36 0.99

C7 0.0 C37 0.67

C8 0.0 C38 0.62

C9 4.45 C39 1.01

C10 9.95 C40 0.97

C11 10.96 C41 0.53

C12 5.16 C42 0.57

C13 3.23 C43 0.84

C14 3.28 C44 0.81

C15 3.83 C45 0.52

C16 3.32 C46 0.44

C17 3.70 C47 0.52

C18 3.59 C48 0.51

C19 2.98 C49 0.44

C20 2.63 C50 0.43

C21 2.98 C51 0.43

C22 1.71 C52 0.41

C23 2.32 C53 0.34

C24 1.85 C54 0.33

C25 1.78 C55 0.35

C26 1.50 C56 0.34

C27 1.87 C57 0.30

C28 1.55 C58 0.28

C29 1.42 C59 0.28

C30 1.27 C60+ 6.66


oil kg/m 961.80=

124

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

C37

C38

C39

C40+

Mo

le %

0

2

4

6

8

10

12

14

16

Injection oil

Produced oil after CO2 injection in small model

Produced oil after CO2 injection in large model

Figure ‎4-21: Compositional analysis of the produced oil after CO2 injection

125

Table ‎4-8: Compositional analysis result of the produced heavy oil after butane injection in small model


C1 0.0 C31 1.01

C2 0.0 C32 0.94

C3 0.0 C33 0.92

C4 0.0 C34 0.80

C5 0.0 C35 0.89

C6 0.0 C36 0.70

C7 0.91 C37 0.63

C8 1.30 C38 0.69

C9 6.10 C39 0.66

C10 11.78 C40 0.66

C11 8.55 C41 0.52

C12 3.56 C42 0.59

C13 3.05 C43 0.57

C14 3.29 C44 0.52

C15 3.24 C45 0.47

C16 3.24 C46 0.50

C17 3.30 C47 0.47

C18 3.00 C48 0.41

C19 2.65 C49 0.33

C20 2.40 C50 0.39

C21 2.34 C51 0.33

C22 2.12 C52 0.28

C23 1.76 C53 0.25

C24 1.64 C54 0.32

C25 1.68 C55 0.32

C26 1.62 C56 0.27

C27 1.46 C57 0.25

C28 1.34 C58 0.44

C29 1.23 C59 0.22

C30 1.18 C60+ 5.73


oil kg/m 934.54=

126

Table ‎4-9: Compositional analysis result of the produced heavy oil after butane injection in large model


C1 0.0 C31 1.24

C2 0.0 C32 1.20

C3 0.0 C33 0.83

C4 0.0 C34 0.79

C5 0.0 C35 1.01

C6 0.0 C36 1.01

C7 0.0 C37 0.68

C8 0.0 C38 0.59

C9 4.91 C39 0.94

C10 9.83 C40 0.87

C11 11.19 C41 0.54

C12 4.89 C42 0.53

C13 3.13 C43 0.57

C14 3.11 C44 0.85

C15 3.72 C45 0.83

C16 3.31 C46 0.45

C17 3.39 C47 0.53

C18 3.44 C48 0.54

C19 3.04 C49 0.53

C20 2.48 C50 0.44

C21 2.85 C51 0.44

C22 1.75 C52 0.42

C23 2.17 C53 0.39

C24 1.75 C54 0.38

C25 1.82 C55 0.35

C26 1.78 C56 0.35

C27 1.67 C57 0.31

C28 1.61 C58 0.32

C29 1.40 C59 0.32

C30 1.19 C60+ 7.33


oil kg/m 965.92=

127

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

C37

C38

C39

C40+

Mole

%

0

2

4

6

8

10

12

14

16

18

20

Injection oil

Produced oil after butane injection in small model

Produced oil after butane injection in large model

Figure ‎4-22: Compositional analysis of the produced oil after butane injection

128



after a mixture of propane/CO2 injection in the small and large models, respectively. On

the same tables, viscosity, molecular weight, and produced oil density are also shown. It

is apparent that proapne/CO2 injection has significantly diluted the original oil, and the

viscosity of the original oil is decreased drastically. In fact, viscosity reduction is much

more than what happened during pure CO2 injection. However, the viscosity reduction in

the large model is more prominent than in the small model. Viscosity of produced oil is

decreased to 1160 mPa.s and 1480 mPa.s in the large and small models, respectively.

Figure 4-23 shows the compositional analysis of the produced oil for the small and large

models.



after propane/methane injection in the small and large models, respectively. Viscosity,

molecular weight, and produced oil density of the produced oil after propane/methane

injection are also presented in those tables. It is apparent that propane/methane injection

has diluted the original oil and the viscosity of original oil is decreased from 5650 mPa.s

to 2080 and 2380 mPa.s in the large and small physical models, respectively. In fact,

viscosity reduction is less than what happened during propane injection, but the dilution

has significantly improved compared to pure methane injection. However, the viscosity

reduction in the large model is more significant than in the small model. Figure 4-24

shows the compositional analysis of the produced oil for the small and large models.

129

Table ‎4-10: Compositional analysis result of the produced heavy oil after propane/CO2 injection in small

model


C1 0.0 C31 1.26

C2 0.0 C32 1.21

C3 0.0 C33 0.86

C4 0.0 C34 0.81

C5 0.0 C35 1.04

C6 0.0 C36 0.99

C7 0.0 C37 0.65

C8 0.0 C38 0.62

C9 0.0 C39 0.93

C10 9.73 C40 0.73

C11 13.17 C41 0.71

C12 6.79 C42 0.53

C13 3.58 C43 0.84

C14 3.15 C44 0.83

C15 3.69 C45 0.48

C16 3.30 C46 0.48

C17 3.42 C47 0.56

C18 3.51 C48 0.54

C19 3.01 C49 0.47

C20 2.61 C50 0.45

C21 2.86 C51 0.44

C22 1.74 C52 0.41

C23 2.17 C53 0.40

C24 1.86 C54 0.38

C25 1.85 C55 0.35

C26 1.70 C56 0.34

C27 1.66 C57 0.34

C28 1.65 C58 0.33

C29 1.41 C59 0.32

C30 1.24 C60+ 7.58


oil kg/m 954.48=

130

Table ‎4-11: Compositional analysis result of the produced heavy oil after propane/CO2 injection in large

model


C1 0.0 C31 1.07

C2 0.0 C32 0.96

C3 12.79 C33 0.40

C4 0.0 C34 0.38

C5 0.0 C35 0.87

C6 0.0 C36 0.33

C7 0.0 C37 0.32

C8 0.0 C38 0.27

C9 1.99 C39 0.26

C10 8.71 C40 0.18

C11 11.95 C41 0.18

C12 6.88 C42 0.18

C13 3.04 C43 0.13

C14 2.66 C44 0.13

C15 3.01 C45 0.13

C16 2.66 C46 0.13

C17 2.80 C47 0.12

C18 2.93 C48 0.08

C19 2.43 C49 0.07

C20 2.13 C50 0.07

C21 2.29 C51 0.07

C22 1.50 C52 0.07

C23 1.73 C53 0.07

C24 1.50 C54 0.06

C25 1.55 C55 0.05

C26 1.40 C56 0.05

C27 1.31 C57 0.05

C28 1.38 C58 0.05

C29 1.11 C59 0.05

C30 1.00 C60+ 9.45


oil kg/m 944.46=

131

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

C37

C38

C39

C40+

Mole

%

0

2

4

6

8

10

12

14

16

18

20

Injection oil

Produced oil after propane/CO2 injection in small model

Produced oil after propane/CO2 injection in large model

Figure ‎4-23: Compositional analysis of the produced oil after propane/CO2 injection

132

Table ‎4-12: Compositional analysis result of the produced heavy oil after propane/methane injection in

small model


C1 0.0 C31 1.13

C2 0.0 C32 1.11

C3 2.71 C33 0.81

C4 0.0 C34 0.78

C5 0.0 C35 0.98

C6 0.0 C36 0.95

C7 0.0 C37 0.61

C8 1.74 C38 0.69

C9 1.42 C39 0.84

C10 9.67 C40 0.81

C11 11.85 C41 0.51

C12 7.31 C42 0.53

C13 3.15 C43 0.78

C14 2.75 C44 1.21

C15 3.25 C45 0.03

C16 3.23 C46 0.45

C17 3.19 C47 0.51

C18 3.11 C48 0.48

C19 2.94 C49 0.46

C20 2.38 C50 0.43

C21 2.74 C51 0.43

C22 1.52 C52 0.37

C23 2.11 C53 0.34

C24 1.69 C54 0.34

C25 1.72 C55 0.34

C26 1.56 C56 0.30

C27 1.61 C57 0.31

C28 1.55 C58 0.33

C29 1.32 C59 0.29

C30 1.18 C60+ 7.15


oil kg/m 963.10=

133

Table ‎4-13: Compositional analysis result of the produced heavy oil after propane/methane injection in

large model


C1 0.0 C31 1.11

C2 0.0 C32 1.09

C3 1.51 C33 0.79

C4 0.0 C34 0.75

C5 0.0 C35 0.96

C6 0.0 C36 0.92

C7 0.0 C37 0.59

C8 2.74 C38 0.66

C9 2.44 C39 0.82

C10 10.67 C40 0.79

C11 12.79 C41 0.48

C12 7.29 C42 0.50

C13 3.05 C43 0.76

C14 2.85 C44 1.19

C15 3.35 C45 0.00

C16 3.13 C46 0.43

C17 3.17 C47 0.50

C18 3.05 C48 0.47

C19 2.89 C49 0.46

C20 2.36 C50 0.42

C21 2.71 C51 0.42

C22 1.48 C52 0.37

C23 2.09 C53 0.33

C24 1.66 C54 0.33

C25 1.71 C55 0.34

C26 1.54 C56 0.29

C27 1.59 C57 0.29

C28 1.53 C58 0.33

C29 1.30 C59 0.28

C30 1.16 C60+ 6.26


oil kg/m 961.10=

134

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C1

0C

11

C1

2C

13

C1

4C

15

C1

6C

17

C1

8C

19

C2

0C

21

C2

2C

23

C2

4C

25

C2

6C

27

C2

8C

29

C3

0C

31

C3

2C

33

C3

4C

35

C3

6C

37

C3

8C

39

C4

0+

Mo

le %

0

2

4

6

8

10

12

14

16

18

Injection oil

Produced oil after propane/ methane injection in small model

Produced oil after propane/ methane injection in large model

Figure ‎4-24: Compositional analysis of the produced oil after propane/methane injection

135

4.1.2 Effect of solvent type

In this section, the results obtained for each model after injecting different solvents are

graphed together to investigate the effect of solvent type on VAPEX process recovery

performance.

4.1.2.1 Small model

4.1.2.1.1 Recovery factor and produced oil rate

Figure 4-25 shows the effect of solvent type on recovery factor after utilizing the VAPEX

process in the small model. As can be seen, the recovery factor is significantly higher

during propane injection, and the ultimate recovery factor was found to be about 80% of

original oil in place. The second best solvent was found to be the mixture of propane and

CO2 with an ultimate recovery factor of about 60% of original oil in place. Butane

seemed to show high recovery performance, and the ultimate recovery factor achieved

after injecting butane was also about 60%, however the process was observed to be

slower compared to propane and propane/CO2 injection. On the other hand, injecting pure

methane and CO2 did not show promising results and the process was extremely slow.

Figure 4-26 shows the effect of solvent type on the produced oil rate. The same trend as

the recovery factor can be seen for different solvents. In short, the highest production rate

was observed for propane injection, while the lowest production rate was observed for

pure methane and pure CO2 injection.

136

Time (h)

0 100 200 400 420 440

Rec

ov

ery F

acto

r (%

OO

IP)

0

20

40

60

80

Propane

Propane/ CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-25: Effect of the solvent type on recovery factor in small model

137

Time (h)

0 100 400 450

Pro

du

ced

Oil

Rat

e (m

L/m

in)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Propane

Propane/CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-26: Effect of the solvent type on produced oil rate in small model

138

4.1.2.1.2 Solvent utilization factor (SUF)

Figure 4-27 shows the effect of solvent type on SUF in the small model. As expected, the

highest SUF was observed for the case of propane injection, which shows the efficiency

of the process after using propane as the injection solvent. The total SUF was about

2.3×10-3

(mL Oil Prod./mL Sol. Inj.) for propane injection. Taking into account the

importance of solvent inventory, these results confirm the suitability of propane as an

injection solvent.

4.1.2.1.3 Viscosity, density, molecular weight, and hydrocarbon components for the

produced oil

Table 4-14 shows the effect of solvent type on viscosity, density and molecular weight of

the produced oil. The highest viscosity reduction was achieved using propane as the

solvent. In fact, the viscosity of original oil was diluted from 5650 mPa.s to 999 mPa.s

after injecting propane, while the produced oil viscosity was found to be 1480, 2380 and

2960 mPa.s after injecting propane/CO2, propane/methane, and butane, respectively.

However, injecting pure methane and CO2 did not result in a noticeable heavy oil

dilution.

Figure 4-28 shows the effect of solvent type on the hydrocarbon components of the

produced oil. It was observed that the amount of lighter hydrocarbons in the produced oil

was highest for the propane injection. This shows that heavier hydrocarbons can be

extracted after injecting propane as the solvent. Comparing the results after injecting

methane and CO2 with the carbon number of the hydrocarbons in the original oil reveals

that no significant extraction occurred after injecting the above-mentioned gases as the

injection solvent.

139

Time (h)

0 100 200 400

SU

F (

mL

Oil

Pro

d./

mL

So

l. I

nj.

)

0

5.0x10-4

10-3

1.5x10-3

2.0x10-3

2.5x10-3

Propane

Propane/ CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-27: Effect of solvent type on solvent utilization factor (SUF) for small model

140

Table ‎4-14: Produced oil properties for the small model

Solvent Viscosity (mPa.s) Density (kg/m3) Molecular weight

Propane 999 853.50 509

Methane 4730 969.28 507

CO2 5010 962.88 505

Butane 2960 934.54 490

Propane/CO2 1480 954.48 525

Propane/methane 2380 963.10 508

141

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C1

0C

11

C1

2C

13

C1

4C

15

C1

6C

17

C1

8C

19

C2

0C

21

C2

2C

23

C2

4C

25

C2

6C

27

C2

8C

29

C3

0C

31

C3

2C

33

C3

4C

35

C3

6C

37

C3

8C

39

C4

0+

Mole

%

0

2

4

6

8

10

12

14

16

18

20

Injection oil






Produced oil after propane/methane injection in small model

Figure ‎4-28: Effect of solvent type on hydrocarbon components in small model

142

4.1.2.2 Large model

4.1.2.2.1 Recovery factor and produced oil rate

Figure 4-29 shows the effect of solvent type on recovery factor in the large model. The

same trend as the small model was observed in the large model after injecting propane as

the injection solvent. As can be seen, the recovery factor is significantly higher during

propane injection and the ultimate recovery factor was found to be about 80% of original

oil in place. The second best solvents were found to be the mixture of propane and CO2

and pure butane with ultimate recovery factors of about 60% of original oil in place.

However, injecting pure methane and CO2 did not show promising results and the process

was extremely slow.

Figure 4-30 shows the effect of solvent type on the produced oil rate. The same trend as

the recovery factor can be seen for different solvents. The highest stabilized drainage rate

was observed for propane injection, which was about 0.50 mL/min, while the lowest

production rate was observed for pure CO2 injection. The stabilized drainage rates were

0.33 mL/min, 0.25 mL/min and 0.32 mL/min for propane/CO2, propane/methane, and

butane injection.

143

Time (h)

0 100 200 300 400 500

Rec

ov

ery F

acto

r (%

OO

IP)

0

20

40

60

80

Propane

Propane/ CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-29: Effect of the solvent type on recovery factor in large model

144

Time (h)

0 100 400 500

Pro

duced

Oil

Rat

e (m

L/m

in)

0.0

0.2

0.4

0.6

0.8

1.0

Propane

Propane/CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-30: Effect of the solvent type on produced oil rate in large model

145

4.1.2.2.2 Solvent utilization factor (SUF)

Figure 4-31 shows the effect of solvent type on SUF in the large model. More or less the

same trend as the small model was observed in the large model for various solvents. As

expected, the highest SUF was observed for the case of propane injection, which shows

the efficiency of the process after using propane as the injection solvent. The total SUF

was about 2.9×10-3

(mL Oil Prod./mL Sol. Inj.) for propane injection. These results show

that up-scaling the process did not result in additional solvent loss, and it was even

observed that the VAPEX process was significantly improved.

4.1.2.2.3 Viscosity, density, molecular weight, and hydrocarbon components for the

produced oil

Table 4-15 shows the effect of solvent type on viscosity, density, and molecular weight of

the produced oil. The heavy oil dilution was more prominent in the large model, and the

viscosity of original oil was diluted from 5650 mPa.s to469 mPa.s after injecting propane,

while the produced oil viscosity was found to be 1160, 2080, and 3220 mPa.s after

injecting propane/CO2, propane/methane, and butane, respectively. However, injecting

pure methane and CO2 did not result in a noticeable heavy oil dilution.

Figure 4-32 shows the effect of solvent type on the hydrocarbon components of the

produced oil. The same behaviour as the small model was observed after utilizing various

solvents in the large physical model.

146

Time (h)

0 100 200 350 400 450

SU

F (

mL

Oil

Pro

d./

mL

Sol.

Inj.

)

0

5.0x10-4

10-3

1.5x10-3

2.0x10-3

2.5x10-3

3.0x10-3

3.5x10-3

Propane

Propane/ CO2

CO2

Butane

Methane

Propane/ methane

Figure ‎4-31: Effect of solvent type on the solvent utilization factor (SUF) for large model

147

Table ‎4-15: Produced oil properties for the large model

Solvent Viscosity (mPa.s) Density (kg/m3) Molecular weight

Propane 469 938.17 501

Methane 5520 970.11 509

CO2 4910 961.80 504

Butane 3220 965.92 522

Propane/CO2 1160 944.46 611

Propane/methane 2080 961.10 504

148

Carbon number

C1

C2

C3

C4

C5

C6

C7

C8

C9

C1

0C

11

C1

2C

13

C1

4C

15

C1

6C

17

C1

8C

19

C2

0C

21

C2

2C

23

C2

4C

25

C2

6C

27

C2

8C

29

C3

0C

31

C3

2C

33

C3

4C

35

C3

6C

37

C3

8C

39

C4

0+

Mole

%

0

2

4

6

8

10

12

14

16

18

20

Injection oil






Produced oil after propane/methane injection in small model

Figure ‎4-32: Effect of solvent type on hydrocarbon components in large model

149

4.2 Residual oil saturation

The procedure for residual oil saturation measurements was explained in detail in Chapter

3. As explained, different samples were taken from different locations of the small and

large physical models. The saturation profiles are presented in Figures 4-33 and 4-34 for

the small and large models, respectively.

It was observed that residual oil saturations close to the injection well were very low for

all the solvents. However, the lowest residual oil saturation was obtained after injecting

propane for both the small and large models. The residual oil saturation for sample

location 1 was 4.3% and 5.1% for the small and large models, respectively. On the other

hand the highest residual oil saturation was observed at the bottom of the physical models

and close to production wells. The highest residual oil saturation was found to be 80.4%

in the small model and 88.9% in the large model for the case of CO2 injection.

The residual oil saturations were the lowest at the top of the models and close to the

injection points because the solvents were injected from the top injection point and the

diluted oil was drained downward by gravity and solvent flooding.

150

Residual oil saturation (%)

0 20 40 60 80 100

Hei

ght

(cm

)

0

5

10

15

20

25

Propane

Methane

CO2

Butane

Propane/CO2

Propane/methane

Injector

Producer

Figure ‎4-33: Residual oil saturation profile for various solvents in the small model

151

Residual oil saturation (%)

0 20 40 60 80 100

Hei

ght

(cm

)

0

5

10

15

20

25

30

35

40

45

50

Propane

Methane

CO2

Butane

Propane/CO2

Propane/methane

Injector

Producer

Figure ‎4-34: Residual oil saturation profile for various solvents in the large model

152

4.3 Asphaltene precipitation

The asphaltene content of each sample was measured using the standard ASTM D2007-

03 method. The precipitant used was n-Pentane. The experimental procedure was

explained in Chapter 3. Figure 4-35 shows some of the asphaltene precipitate on the filter

paper after the experiments. The mass of the dried particulate on the filter paper, m2, was

compared to the original mass of the heavy oil sample, m1, to determine the asphaltene

mass percent:

%100%1

2

m

mAsphaltenewt ...……………………………………………….…. (4.2)

In order to investigate the asphaltene precipitation in more detail, various samples were

taken after each experiment from various locations of the VAPEX models. Figure 4-36

shows the schematic of the locations of each heavy oil sample in the physical models. In

the next sections, the corresponding graphs for each sample are provided.

153

Figure ‎4-35: Asphaltene precipitate after conducting the asphaltene measurement tests

154

Figure ‎4-36: Schematic of the locations of each heavy oil samples in the physical models

155

4.3.1 Effect of drainage height

4.3.1.1 Propane injection

Figure 4-37 shows the asphaltene precipitation for different locations in the small and

large models following propane injection. As can be seen, the amount of asphaltene

precipitation is almost the same for both physical models. However, at the injection point,

the asphaltene content is slightly more in the large model; this might be due to the contact

time between the solvent and the heavy oil system. The contact time, which is an

effective parameter on asphaltene precipitation, is longer for the large model than for the

small model. Furthermore, more asphaltene precipitation occurs at the injection point and

at the solvent/heavy oil interfaces in comparison to the location close to the production

point. As can be seen, the asphaltene precipitation was about 40% and 31.5% close to the

injection wells for the large and small models, respectively. The minimum asphaltene

precipitation was observed to be about 23% at location #4, which was close to the

production wells. As discussed earlier in Chapter 2, there is a phase change during the

VAPEX process when the solvent diffuses into the oil at the solvent oil interface. During

this phase change, there will be change in temperature, pressure, and concentration at the

contact interface. This proves the difference regarding the amount of asphaltene

precipitation at the injection and production points. It was also found that the texture of

precipitated asphaltene on the filter paper changed at different locations. For instance,

asphaltene precipitants close to the injection points were brittle; however, precipitants

close to the production points were more ductile.

156

Asphaltene weight percent

0 10 20 30 40 50

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model

Figure ‎4-37: Effect of drainage height on asphaltene precipitation at different locations in the small and

large models after propane injection

157

4.3.1.2 Methane injection


large models after methane injection. The same trend as the propane injection was

observed; the amount of asphaltene precipitation is slightly more in the large model,

which might be due to the contact time between the solvent and the heavy oil system. The

highest asphaltene precipitation was found to be about 48% at location #1 in the large

model.

4.3.1.3 CO2 injection


large models after CO2 injection. The trend for this solvent was slightly different from

what was observed for propane and methane. In fact, the amount of asphaltene

precipitation is slightly more in the large model at various locations except for location 2

in the small model. At this point, significant asphaltene precipitation of about 63% was

observed in the small model, which indicates the low recovery factor of that specific test.

4.3.1.4 Butane injection


large models after butane injection. More asphaltene precipitation was observed in the

large model with higher drainage rate and longer distance between the injection and

production wells. The asphaltene precipitation was highest at location #1, and it was

38.7% and 31.3% for the large and small models, respectively.

158


0 10 20 30 40 50 60

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model


large models after methane injection

159


0 10 20 30 40 50 60 70

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model


large models after CO2 injection

160


0 10 20 30 40 50

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model


large models after butane injection

161

4.3.1.5 Propane/CO2 injection

As can be seen in Figure 4-41, the same trend as the propane injection was observed in

the case of propane/CO2 mixture injection. The asphaltene precipitation was slightly

more in the large model. The highest asphaltene precipitation was about 38.6% at

location #1 for the large model. However, the highest asphaltene precipitation for the

small model was observed at the injection point, which was about 29.7%. The minimum

asphaltene precipitation was about 22% for both models and was observed at the

production points. These results confirm the in-situ upgrading of heavy oil by injecting a

mixture of propane and CO2.

4.3.1.6 Propane/methane injection

As can be seen in Figure 4-42, the same trend as the propane injection was observed in

the case of propane/methane mixture injection. The highest asphaltene precipitation was

about 39.1% at location #1 for the large model. The minimum asphaltene precipitation

was observed at location #4 it was about 22%.

It was also found that the asphaltene precipitants close to the injection wells were brittle;

however, precipitants close to the production wells were more ductile.

162


0 10 20 30 40 50

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model

Figure ‎4-41: Effect of drainage height on asphaltene precipitation at different locations in small and large

models after propane/CO2 injection

163


0 10 20 30 40 50

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Small model

Large model

Figure ‎4-42: Effect of drainage height on asphaltene precipitation at different locations in small and large

models after propane/methane injection

164

4.3.2 Effect of solvent type

In this section, the results obtained for each model after injecting different solvents are

graphed together to investigate the effect of solvent type on asphaltene precipitation in

each physical model.

4.3.2.1 Small model

Figure 4-43 shows the results of the asphaltene content measurement test after using

different solvents in the small model. As can be seen, the highest asphaltene precipitation

was achieved after injecting methane, and it was about 41.7%. Generally, injecting CO2

showed the least asphaltene precipitation and, consequently, the least heavy oil dilution.

The asphaltene precipitation after injecting CO2 at location #1 was about 22.5%. It should

be mentioned that the low injection pressure for CO2 could be a reason for this low

dilution. It was expected that the difference in asphaltene precipitation for butane and

propane injection would be more prominent; however, the slow process of butane

injection resulted in some excessive asphaltene precipitation. Comparing the results for

the propane injection with the mixture of propane/CO2, it can be seen that there will be

less asphaltene precipitation at different locations of the physical models. It was also

observed by Javaheri and Abedi (2013) that by adding CO2 to pure propane, less

asphaltene precipitation would be observed. Moreover, it was observed that adding

methane, would also results in less asphaltene precipitation compared to pure propane

injection.

165


0 10 20 30 40 50 60 70

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Propane

Methane

CO2

Butane

Propane/ CO2

Propane/ methane

Figure ‎4-43: Effect of solvent type on asphaltene precipitation at different locations in the small model

166

4.3.2.2 Large model

Figure 4-44 shows the results of the asphaltene content measurement test after using

different solvents in the large model. This time, the overall amount of asphaltene

precipitation was highest at location #1 for the case of methane injection, and it was

about 48%. This means that more dilution and in-situ upgrading of heavy oil was

achieved by injecting propane. The trend for various solvents was almost the same as

what was observed in the small model. The asphaltene precipitation for location #1 was

found to be 40%, 39.1%, 38.7%, and 38.6% for propane, propane/ methane, butane, and

propane/CO2, respectively.

Figure 4-45(a) shows the asphaltene precipitation close to the injection points in one of

the tests. Here, severe asphaltene precipitation was observed, and, as mentioned earlier,

the texture was more brittle. Asphaltene streaks were observed in most of the tests,

especially at the solvent/oil interface. In Figure 4-45(b) asphaltene streaks in the sand

pack can be seen after opening the Plexiglas plate at the end of one of the experiments.

167


0 10 20 30 40 50 60

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1

Propane

Methane

CO2

Butane

Propane/ CO2

Propane/ methane

Figure ‎4-44: Effect of solvent type on asphaltene precipitation at different locations in the large model

168

(a)

(b)

Figure ‎4-45: (a) Asphaltene precipitation close to the injection point, (b) Asphaltene streaks on the sand

pack at the end of experiments

169

4.4 Image analysis (IA)

As mentioned in Chapter 3, the physical models were designed so the solvent chamber

evolution could be monitored. To this end, a digital camera was used to take pictures of

the physical models at different times during the experiments. These images were used to

analyse the chamber evolution over time and also to measure the displacement efficiency

for each solvent in VAPEX models.

For this purpose, software with a graphical user interface was coded to analyse the

images from the small and large physical models. Software was specifically coded for

this purpose because of the limitations of commercial IA softwares. Such softwares need

a specific resolution and zooming for the test images. However, because of the

limitations in the laboratory with the large models, the angle and zooming of the images

changed for certain images.

The software was coded using C# programming with Microsoft Visual Studio 2012. The

interface of the coded software is shown in Figure 4-46.

The images are input to the software and the coordinates are fixed based on the model

dimensions. Then, an approximate interface of the solvent/heavy oil system can be

manually selected. Next, the software detects the interface based on the colour change

and determines the best polynomial for the detected interface. In addition, the swept zone

area and the drainage height will be shown at each time. Figure 4-47and 4-48 show these

steps for the small and large models respectively for a sample image.

170

Figure ‎4-46: The interface of the coded software for IA

171

(a)

(b)

(c)

Figure ‎4-47: The procedure for conducting IA in the small model: (a) The coordinates of the image are

specified, (b) The interface curve is defined, and (c) The oil and solvent zones are schematically reprinted

by the software

172

(a)

(b)

(c)

Figure ‎4-48: The procedure for conducting IA in the large model: (a) The coordinates of the image are

specified, (b) The interface curve is defined, and (c) The oil and solvent zones are schematically reprinted

by the software

173

Figures 4-49 and 4-50 show the post-propane-injection chamber evolution for the small

and large models, respectively. These pictures are processed with the developed IA

software. In these pictures, the solvent and oil zones appear distinctively during the

experiments. The untouched zone is shown with the darker colour, while the swept zone

is shown with light grey color. Of note, a similar shape is observed in both physical

models. The chamber forms and develops toward the sidewalls and then moves

downward with reduced available drainage height. It was also observed that the solvent

chamber descended slightly faster on the right wall because of the location of the

injection well, which was closer to the left wall until the end of the experiments when it

was fully developed and reached the bottom boundary of the physical models. As these

figures show, the solvent and oil interface is not a smooth straight line. Therefore, the

best curve was applied to model the solvent/oil interface. Then, the area of the swept

zone was calculated to be used to calculate the sweep efficiency for each solvent in the

small and large models. For this purpose, equation (4.3) was introduced, and it was

assumed that vertical sweep efficiency was equal to one.

dA EERF ……………………………………………………………..………….. (4.3)

where RF is the recovery factor, EA is the areal sweep efficiency, and Ed is the sweep

efficiency.

174

Figure ‎4-49: Solvent chamber evolution in small model after propane injection

175

Figure ‎4-50: Solvent chamber evolution in large model after propane injection

176

The results for the sweep efficiency are provided in Figures 4-51 and 4-52 for the small

and large models, respectively. It was found that the sweep efficiency in the small model

was very close to the large model for each solvent. However, from the solvent type point

of view, the highest sweep efficiency was achieved for the case of propane injection,

which was about 0.86. Injecting butane resulted in high sweep efficiency in both models,

and it was about 0.72. Additionally, injecting propane/CO2 and propane/methane

mixtures showed promising sweep efficiency. However, the sweep efficiency of

propane/CO2 mixture was slightly higher than the propane/methane mixture. On the other

hand, the sweep efficiency of pure CO2 injection was higher than methane injection. The

lowest sweep efficiency was achieved in the large model after injecting methane, which

was 0.38.

By monitoring the slight change of sweep efficiency for each solvent during the course of

the experiments, it was found that the sweep efficiency decreases after the solvent

breakthrough and during the first stages of chamber evolution.

177

Time (h)

0 20 40 60 80 100 200 250 300 350 400 450

Sw

eep E

ffic

iency

(E

d)

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Propane

Butane

CO2

Methane

Propane/CO2

Propane/methane

Figure ‎4-51: Sweep efficiency of various solvents in the small model

178

Time (h)

0 20 40 60 80 100 120 140 160 350 400 450 500

Sw

eep E

ffic

iency

(E

d)

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Propane

Butane

CO2

Methane

Propane/CO2

Propane/methane

Figure ‎4-52: Sweep efficiency of various solvents in the large model

179

4.5 Effect of injection-production wells connection

One of the key parameters to implement a successful VAPEX process is to control the

profiles of the vapour chamber, which will result in high areal sweep efficiency. To

achieve this goal, an optimum well configuration and connection between the injection

and production wells is desirable. After conducting experiments no. 2, 3, 8, and 9 with

propane and propane/CO2 mixture, some disparities with the results in literature were

observed. To further investigate these conflicting results, another injection scenario was

followed to observe the effect of connection between the injection and production wells.

Therefore, tests no. 13, 14, 15, and 16 were carried out as the second injection scenario.

In the following sections, the observations are presented for these tests. In the first

scenario (tests no. 2, 3, 8, and 9), the solvent was injected into the physical models at the

operating pressure while the production pressure was atmospheric pressure, and the

solvent and oil production was monitored carefully. Once, the solvent breakthrough was

monitored the production pressure was set to the operating pressure to eliminate the

pressure difference and start producing due to the gravity drainage. For the second

scenario (tests no. 13, 14, 15, and 16), the pressure at the production point was

implemented after that the connection between the injection and production well was

visually observed. It took more injection time and a larger amount of solvent was

produced before exerting the back-pressure at the production well.

4.5.1 Small model

Figure 4-53 shows the recovery factor after injecting propane and propane/CO2 for two

injection scenarios in the small VAPEX model. The first distinctive difference between

the results was that following the first injection scenario, the process was extremely slow.

180

However, the ultimate recovery factor was almost the same for both injection scenarios.

The ultimate recovery factor was about 75% of original oil in place after injecting

propane as the solvent in the small model, and the ultimate recovery factor after injecting

propane/CO2 was about 60% of original oil in place. The major conflict was observed

while comparing the drainage rates for the small and large models. As presented in

Figure 4-54, the drainage rate was found to be higher in the small model with smaller

drainage height. This disparity may be the result of well configuration and the poor

connection between the injection and production wells. Therefore, the solvent chamber

moved faster in the small physical model in comparison to the large model with a greater

drainage height. The stabilized drainage rate after propane injection was observed to be

about 0.12 mL/min for the first injection scenario and about 0.22 mL/min after

implementing the second injection scenario. For the case of propane/CO2 injection, the

stabilized drainage was found to be about 0.08 mL/min and 0.15 mL/min for the first and

second injection scenarios, respectively. Hence, it was observed that stabilized drainage

rate was increased approximately two times for the second injection scenario. After

measuring the asphaltene content of different samples from different locations of the

physical models, it was observed that less asphaltene deposition occurred after propane

and propane/CO2 injection in the small model following the second injection scenario.

Better connection between the injection and production wells decreased the recovery

process time significantly, which directly affected the asphaltene deposition in both the

small and large models. The results for the asphaltene content at different locations of the

small model are presented in Figure 4-55.

181

Time (h)

0 20 40 60 80 100 120

Rec

ov

ery F

acto

r (%

OO

IP)

0

20

40

60

80

100

Propane after first injection scenario

Propane after second injection scenario

Propane/CO2 after first injection scenario

Propane/CO2 after second injection scenario

Figure ‎4-53: Effect of connection establishment between the injection and production wells on the recovery

factor in the small model

182

Figure ‎4-54: Effect of connection establishment between the injection and production wells on the

produced oil rate in the small model

Time (h)

0 20 40 60 80 100 120

Pro

duce

d O

il R

ate

(mL

/min

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4



Propane/ CO2 after first injection scenario

Propane/ CO2 after second injection scenario

183


0 10 20 30 40 50 60

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1






asphaltene precipitation in the small model

184

Fluid properties of the produced fluid were measured after each test to observe the effect

of different injection scenarios and the type of injection solvent on the produced fluid.

Table 4-16 showsviscosity, molecular weight, and oil density of the produced oil after

injecting propane and propane/CO2 for the two injection scenarios. The results show two

effective parameters as mentioned before: the type of the injection solvent used and the

connection establishment between the injection and production. Propane

injectionsignificantly diluted the original oil and reduced its viscosity in both injection

scenarios. However, the decrease in the viscosity is more noticeable in the second

injection scenario.For instance, for the case of propane injection, the first injection

scenario resulted in the reduction of original oil viscosity from 5650 mPa.s to 1235 mPa.s

while it decreased from 5650 mPa.s to 999 mPa.s for the second injection scenario.

Figure 4-56 shows the chamber evolution after implementing the first injection scenario.

The chamber forms and develops toward the sidewalls and then moves downward with

reduced available drainage height. It was also observed that the solvent chamber

descended faster on the left wall until the end of the experiments when it was fully

developed and reached the bottom boundary of the physical models. The chamber

evolution after implementing the second injection scenario in the small model was

presented in Figure 4-49 earlier in this chapter.

185

Table ‎4-16: Produced oil properties

Test

No.

Physical

Model

Solvent

Viscosity

(mPa.s)

Density

(kg/m3)

Molecular

weight

2 Small Propane 1235 957.51 493

8 Small Propane/CO2 1950 954.48 501

13 Small Propane 999 853.50 509

15 Small Propane/CO2 1480 944.46 505

186

Figure ‎4-56: Solvent chamber evolution in small model after propane injection (first injection scenario)

187

4.5.2 Large model

After following the second injection scenario to establish a more confident connection

between the injection and production wells, the results were different and higher

productions rates were observed in the larger model with greater drainage height. Figure

4-57 shows the recovery factor after injecting propane and propane/CO2as the solvent in

the large physical model. As can be seen in this figure, the effects of connection between

the injection and production wells were more noticeable in the large model. This can be

due to the specific well configuration used in this study and the longer distance between

the injection and production wells. The ultimate recovery factor of about 50% of original

oil in place was observed after injecting propane as the solvent for the first injection

scenario while utilizing the second injection scenario resulted in a recovery of 80% of

original oil in place and a significantly faster process. The same trend was observed when

a mixture of propane/CO2 was used as the solvent for the VAPEX process. In the case of

propane/CO2 injection, the ultimate recovery factor was found to increase from 42% of

original oil in place for the first injection scenario to 52% for the second injection

scenario. The stabilized drainage rate after propane injection was found to increase from

0.04 mL/min to 0.50 mL/min after injecting propane in the large model. The stronger

connection between the injection and production wells slightly improved the recovery

performance of the process in the small model, but it significantly boosted the process in

the large model. For the case of propane/CO2 injection, the stabilized drainage rate was

found to increase from 0.02 mL/min for the first injection scenario to 0.33 mL/min for

the second injection scenario. The results for produced oil drainage rate are presented in

Figure 4-58.

188

Time (h)

0 50 100 150 200 250 300 350 400 450

Rec

over

y F

acto

r (%

OO

IP)

0

20

40

60

80



Propane/CO2 after first injection scenario

Propane/CO2 after second injection scenario

Figure ‎4-57: Effect of connection establishment between the injection and production wells on the recovery

factor in the large model

189

Time (h)

0 100 200 500 600

Pro

duce

d O

il R

ate

(mL

/min

)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9






produced oil rate in the large model

190

The results for the asphaltene content at different locations of the large model are

presented in Figure 4-59. It was observed that less asphaltene deposition occurred after

propane and propane/CO2 injection in the large model following the second injection

scenario.

Viscosity, molecular weight, and oil density of the produced oil after injecting propane

and propane/CO2 for the two injection scenarios in the large VAPEX model are presented

in Table 4-17. Like other parameters that was discussed earlier, the connection between

the wells affected the produced oil properties more significantly in the large model

compared to the small model. For the case of propane injection, the first injection

scenario resulted in the reduction of original oil viscosity from 5650 mPa.s to 644 mPa.s,

while it decreased from 5650 mPa.s to 469 mPa.s for the second injection scenario.

The shape of the chamber after utilizing the second injection scenario in test no. 9 is

presented in Figure 4-50, while the shape of the chamber after following the first

injection scenario is presented in Figure 4-60.

191


0 10 20 30 40 50

Sam

ple

num

ber

s

Sample 4

Sample 3

Sample 2

Sample 1






asphaltene precipitation in the large model

192

Table ‎4-17: Produced oil properties

Test

No.

Physical

Model

Solvent

Viscosity

(mPa.s)

Density

(kg/m3)

Molecular

weight

3 Large Propane 644 952.75 507

9 Large Propane/CO2 1500 953.71 506

14 Large Propane 469 938.17 469

16 Large Propane/CO2 1160 954.41 503

193

Figure ‎4-60: Solvent chamber evolution in large model after propane injection (first injection scenario)

194

4.6 Scale-up:

Butler and Mokrys (1989) carried out VAPEX experiments in Hele-Shaw cells and found

that there is square root functionality between the stabilized drainage rate and the

medium permeability, drainage height, and physical properties of oil and solvent. They

assumed that there is complete miscibility of solvent and bitumen, and they also

neglected the convection term and mechanical dispersion coefficients. Based on their

findings, they proposed equation (4.4) to predict the produced flow rate after

implementing the VAPEX process:

so HNSkgQ 22 ………………………………………………………………… (4.4)

In this equation, Ns is the VAPEX dimensionless number, which accounts for the oil-

solvent properties and is defined by equation (4.5):

max

min

1C

Cs

smix

sss dC

C

DCN

………………………………………………………… (4.5)

In these equations, Q is the stabilized drainage rate per unit length of the horizontal well,

k is permeability, g is acceleration due to gravity, φ is porosity, ΔSo is change in oil

saturation, Δρ is density difference between solvent and bitumen, Cs is solvent

concentration, Ds is diffusivity of solvent in bitumen, and μmix is the viscosity of the

mixture at solvent concentration.

Later, it was found by Das and Butler (1994, 1998) that the above equation

underestimates the production rate in porous media. To consider the effect of porous

195

media, they introduced the effective diffusion coefficient, Deff, and cementation factor, Ω.

Therefore, equations (4.4) and (4.5) were modified as follows:

so HNSkgQ 22 …………………………………………………….………… (4.6)

where

max

min

1C

Cs

smix

effs

s dCC

DCN

………………………………...……………………… (4.7)

and

seff DD ………………………………………………………………………… (4.8)

In equation (4.8), λ is the mass transfer enhancement coefficient.

Equation (4.6) can be rearranged as follows:

so NSgkHQ 2 …...…...……………………………………….………… (4.9)

The second term on the right hand side of equation (4.9), so NSg2 , is constant for a

specific oil-solvent system at constant pressure and temperature. Therefore, for two

different sand pack models with different drainage heights, the following equation can be

derived:

1

2

1

2

kH

kH

Q

Q…………………………………………………………………... (4.10)

196

Equation (4.10) can be used to upscale the drainage rate obtained in a model with smaller

drainage height to one with a larger drainage height. However, it was found later by

several researchers that this up-scaling equation still cannot predict the drainage rate

(Yazdani, 2007). Based on the results that were obtained during experiments with various

models with different drainage heights, Yazdani (2007) showed that this equation

underestimates the drainage rate. Equation (4.10) was modified by him and the following

equation was proposed:

1

2

1

2

1

2

k

k

H

H

Q

Qn

…………………………………………….....……………... (4.11)

The exponent n in equation (4.11) is in the range of 1.10 to 1.30, while this exponent is

0.50 in Butler’s equation. In order to find the correct value of exponent n, various values

of n = 0.50, 1.10, and 1.30 were used to predict the drainage rate. For this purpose,

equation (4.11) was rearranged to equation (4.12), and the two terms on each side of this

equation were measured for different solvents. To graphically present the results,

equations (4.13) and (4.14) were used; hence, the results are presented graphically in

Figures 4-61 to 4-63. The subscript, L stands for the large physical model, and the

subscript S stands for the small physical model.

3.13.1

SS

n

S

S

LL

n

L

L

kH

Q

kH

Q

………...……………..………………..………... (4.12)

3.1

LL

n

L

LL

kH

QR

………...……………..………………………………..……... (4.13)

197

3.1

SS

n

S

SS

kH

QR

………...…………………………..……...…………………... (4.14)

It can be seen in Figure 4-61 that Butler’s model significantly under-predicted the

drainage rate for all types of solvents used in this study. However, the results obtained

based on Yazdani’s model are more accurate, and the data points are closer to the

prediction line. The results in Figure 4-62 showing that exponent n = 1.1 still

underestimates the actual drainage rate, but as is presented in Figure 4-63, exponent n =

1.3 resulted in over estimating the experimental drainage rates. Therefore, exponent n =

1.2 was chosen, and the results obtained based on this value are graphed in Figure 4-64. It

was found that the experimental results match the prediction based on this new value, and

the best results were obtained with n = 1.2.

Considering the fact that there is a linear relationship between stabilized drainage rate, Q,

and 3.1kH n , and knowing the best value for exponent n is 1.2, the following equations

can be found for different solvents based on the results presented in Figure 4-65.

For propane:

3.12.10334.0 kHQ ................................................................................................ (4.15)

For propane/CO2 mixture:

3.12.10227.0 kHQ ................................................................................................ (4.16)

198

For butane:

3.12.10217.0 kHQ ................................................................................................ (4.17)

For propane/methane mixture:

3.12.10174.0 kHQ ............................................................................................... (4.18)

199

Rs (h×104)

0.0 0.1 0.2 0.3 0.4 0.5

RL (

h×104)

0.0

0.1

0.2

0.3

0.4

0.5

Propane

Methane

CO2

Propane/CO2

Butane

Propane/Methane

n=0.5

Figure ‎4-61: The results obtained for up-scaling the stabilized drainage rate based on the proposed

exponent by Butler (1994), (n=0.5). The dotted line is the drainage rate prediction based on Butler’s model;

the data points for different solvents are the experimental results obtained in this study.

200

Rs (h×104)

0.00 0.01 0.02 0.03 0.04 0.05

RL

(h×

10

4)

0.00

0.01

0.02

0.03

0.04

0.05

Propane

Methane

CO2

Propane/CO2

Butane

Propane/Methane

n=1.1


exponent by Yazdani (2007), (n=1.1). The dotted line is the drainage rate predicted based on Yazdani’s

model; the data points for different solvents are the experimental results obtained in this study.

201

Rs (h×104)

0.000 0.005 0.010 0.015 0.020 0.025 0.030

RL

(h×

10

4)

0.000

0.005

0.010

0.015

0.020

0.025

0.030

Propane

Methane

CO2

Propane/CO2

Butane

Propane/Methane

n=1.3


exponent by Yazdani (2007), (n=1.3). The dotted line is the drainage rate predicted based on Yazdani’s

model; the data points for different solvents are the experimental results obtained in this study.

202

Rs (h×104)

0.00 0.01 0.02 0.03 0.04 0.05

RL (

h×10

4)

0.00

0.01

0.02

0.03

0.04

0.05

Propane

Methane

CO2

Propane/CO2

Butane

Propane/Methane

n=1.2

Figure ‎4-64: The results obtained for up-scaling the stabilized drainage rate. The dotted line is the drainage

rate predicted based on n=1.2; the data points for different solvents are the experimental results obtained in

this study.

203

Figure ‎4-65: Linear regression for the results obtained for different solvents in the small and large physical

models

3.12.1 kH

60 80 100 120 140 160 180 200

Q(c

m2/h

)

1

2

3

4

5

6

7

Propane

Propane/CO2

Butane

Propane/Methane

204

4.7 Dimensionless VAPEX number, Ns calculation:

As described earlier in this chapter, Butler’s model includes a dimensionless number, Ns,

which is also called the VAPEX number. This number accounts for oil-solvent properties

and can be calculated by rearranging equations (4.6) and (4.7) to generate the following

equation:

HSkg

QdC

C

DCN

o

C

Cs

smix

sss

8

1 2max

min

…………………...………………..… (4.19)

The experimental results were used to calculate the VAPEX number using equation

(4.19). The results were graphed versus the drainage height to investigate the effect of

drainage height on VAPEX number. As can be seen in Figure 4-66, VAPEX number

increases with increasing drainage height. This was also observed by other researchers

(Ahmadloo 2012, Yazdani 2007), and it can be concluded that VAPEX number is

dependent on the drainage height and oil-solvent properties. However, it should be noted

that in equation 4.19 by changing the values for drainage height the other parameters are

changing too, which will result in the final value for Ns. Comparing different solvents

used for these experiments, the highest values for VAPEX number were achieved after

injecting propane, and VAPEX numbers for butane and propane/CO2 were also relatively

high and close to each other. However, the lowest values were obtained after injecting

pure methane and pure CO2.

205

H (m)

0.20 0.25 0.30 0.35 0.40 0.45 0.50

Ns

(dim

ensi

onle

ss)

0

5.0x10-5

10-4

1.5x10-4

2.0x10-4

2.5x10-4

Propane

Methane

CO2

Propane/CO2

Butane

Propane/Methane

Figure ‎4-66: Effect of drainage height and solvent type on dimensionless VAPEX number, Ns

206

5. CHAPTER 5: PVT STUDIES AND NUMERICAL

SIMULATION

5.1 Viscosity and density measurement

Viscosity and density of the heavy oil directly affect the amount of solvent dissolved in

the heavy oil; therefore, in order to tune the PVT model against the experimental data

viscosity and density of heavy oil used in these experiments were calculated at different

temperatures. The measurements are provided in Figure 5-1.

5.2 Vapour pressure

During solvent injection, vapour pressure is a key factor that significantly affects the

VAPEX performance. It has been found that the optimum VAPEX performance would be

obtained if the solvent is injected close to its vapour pressure. Therefore, CMG’s

WinpropTM

package (Computer Modelling Group Ltd., 2011) was utilized to calculate the

vapour pressure for various solvents used in this study. The results are provided in Table

5-1, and P-T two phase envelopes are presented in Figure 5-2.

207

Temperature, °C

15 20 25 30 35 40 45 50 55

Den

sity

, kg/m

3

950

955

960

965

970

975

Vis

cosi

ty, m

Pa.

s

0

1000

2000

3000

4000

5000

6000

Density

Viscosity

Figure ‎5-1: Densities and viscosities of the heavy oil used in this study at various temperatures and

atmospheric pressure

208

Table ‎5-1: Vapour pressure of solvents used in this study at 21 °C

Solvent Vapour pressure at T = 21°C (kPa)

CO2 5766.9

Methane NA

Propane 757.4

Butane 116.6

Propane/CO2 1156.5

Propane/methane 1204.1

209

Temperature (°C)

-120 -100 -80 -60 -40 -20 0 20 40 60 80 100

Pre

ssure

(kP

a)

0

1000

2000

3000

4000

5000

6000

7000

P-T diagram for Propane (70%)/Methane (30%) mixture

P-T diagram for Propane (70%)/CO2 (30%) mixture

Experimental operating conditions

Figure ‎5-2: Two-phase envelopes for propane/CO2 and propane/methane mixtures

210

5.3 Solubility measurement

In this study, separate experiments were carried out to measure the solubility of pure

solvents used for the VAPEX experiments. The experiments to measure the solubility of

propane, butane, CO2, and methane were carried out at a temperature of 21°C and various

pressures for each solvent. The solubility measurement experimental set-up consists of a

stainless steel cylinder with a volume of 196.4 cm3, a solvent tank, a pressure regulator,

digital pressure gauges, DFMs, valves, and tubing. The schematic diagram of the

experimental set-up is presented in Figure 5-3. To carry out the solubility measurement

experiments, first, 20cm3 heavy oil was added into the cylinder. Then, the cylinder was

sealed with a cylinder cap and vacuumed with the vacuum pump. Secondly, the solvent

was introduced into the vacuumed cylinder through the digital flow meters until the

pressure inside the cylinder reached the operating pressure. A pressure regulator was set

on the solvent tank to maintain the pressure at the operating pressure. Therefore, solvent

was gradually injected into the cylinder, and the total volume was carefully recorded by

the digital flow meters for each solvent. The process was carried out for several days to

make sure that the operating pressure remained constant while no further solvent was

being injected. At the end of each test, the amount of free solvent was subtracted from the

total amount of injected solvent to obtain the dissolved amount of solvent in the heavy

oil. Results for various solvents at their respective operating pressures are presented in

Figure 5-4.

211

Figure ‎5-3: Schematic of the experimental set-up used for solubility measurement tests

Solvent

cylinder

Pressure

regulator

Digital

flow meter

Heavy oil/

solvent

cell

Digital

pressure

gauge

Vacuum

pump

212

Pressure (kPa)

0 100 200 300 400 500 600 700 800

Solu

bil

ity (

wt%

)

0

10

20

30

40

50

Propane

(a)

Pressure (kPa)

100 200 300 400 500 600 700 800 900

Solu

bil

ity (

wt%

)

0

2

4

6

8

10

Methane

(b)

Pressure (kPa)

100 200 300 400 500 600 700 800 900

Solu

bil

ity (

wt%

)

0

2

4

6

8

10

CO2

(c)

Pressure (kPa)

20 40 60 80 100 120 140 160

Solu

bil

ity (

wt%

)

0

5

10

15

20

25

30

35

Butane

(d)

Figure ‎5-4: Solubility of (a) propane, (b) methane, (c) CO2, and (d) butane at 21°C

213

5.4 Solvent volume fraction in heavy oil for VAPEX tests

As mentioned in Chapter 3, several samples were taken from the produced oil for both

the small and large models during the VAPEX experiments. The initial weights of the

samples were recorded, and, then, the samples were kept at room pressure and

temperature for 5-7 days. Then the final weight of each sample was recorded at the room

temperature and pressure. The difference between the weights of each sample was used to

estimate the dissolved volume of the solvent.

The results are presented in Figures 5-5 and 5-6 for the small and large models,

respectively. The highest solvent volume fraction was observed to be 0.35 in the case of

propane injection in the large model. The propane volume fraction was about 0.30 in the

small model. The lowest solvent fraction was observed for methane and CO2 in both

physical models.

In addition, it was observed that solvent volume fraction is higher in the large model with

greater drainage height due to excessive contact time and area between the solvent and

heavy oil. Muhammad (2012) and Yazdani (2007) reported the same trend for the effect

of drainage height on the solvent mass fraction.

214

Time (h)

0 100 200 400

Solv

ent

volu

me

frac

tion, C

s

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Propane

Methane

CO2

Butane

Propane/CO2

Propane/methane

Figure ‎5-5: Solvent volume fraction in the produced oil from the small model for various solvents at 21°C

215

Time (h)

0 100 200 400 500

Solv

ent

volu

me

frac

tion, C

s

0.0

0.1

0.2

0.3

0.4

Propane

Methane

CO2

Butane

Propane/CO2

Propane/methane

Figure ‎5-6: Solvent volume fraction in the produced oil from the large model for various solvents at 21°C

216

5.5 Numerical simulation

In this section, the results of the simulation study of the VAPEX process are presented.

The main goal was to achieve the best history-match of the VAPEX experiments, which

were explained in the previous chapter. In this research, CMG’s STARSTM

package

(Computer Modelling Group Ltd., 2011) was utilized to carry out the numerical

simulation studies.

To construct the PVT model, CMG’s WinpropTM

package was utilized to tune the

equation of state based on the heavy oil-solvent system used in this study and the

experimental data presented earlier in this chapter. Then, the developed PVT model was

exported to STARSTM

to simulate the VAPEX experiments.

5.5.1 Model construction

Two lab-scale 3D simulation models were developed to simulate the experimental

conditions for this study. Each 3D model represents one of the VAPEX physical models,

which were used for the experimental studies. The detailed information about the

simulation models is provided in Tables 5-2 and 5-3 for the small and large models,

respectively. The injection wells were located at the top layer, and the production wells

were located at the bottom layer. Therefore, for the small models, grid numbers i =12 × j

= 1 to 5 × k = 1 were perforated for the injection well and grid numbers i =12 × j = 1 to

5 × k = 24 were perforated to represent the production well. For the case of the large

model, grid numbers i =22 × j = 1 to 5 × k = 1 were perforated for the injection well and

grid numbers i =22 × j = 1 to 5 × k = 47 were perforated to represent the production

well. The injection and production wells were perforated in a way to accurately represent

the experimental models (Figures 5-7 and 5-8).

217

Table ‎5-2: Properties of small simulation model

Grid type Cartesian

Number of grids in i-direction 20

Number of grids in j-direction 5

Number of grids in k-direction 24

Number of grid blocks 2400

Grid thickness (cm) 1

Porosity (%) 42

Permeability (i, j and k-directions) (D) 9

Temperature (°C) 21

Original oil in place (cm3) 1008

Solvent Propane

Oil saturation (%) 100

No. of injection wells 1

No. of production wells 1

218

Table ‎5-3: Properties of large simulation model

Grid type Cartesian

Number of grids in i-direction 40

Number of grids in j-direction 5

Number of grids in k-direction 47

Number of grid blocks 9400

Grid thickness (cm) 1

Porosity (%) 43

Permeability (i, j and k-directions) (D) 10

Temperature (°C) 21

Original oil in place (cm3) 4042

Solvent Propane

Oil saturation (%) 100

No. of injection wells 1

No. of production wells 1

219

(a)

(b)

Figure ‎5-7: (a) 2D view of the simulated model with the injection and production wells for the small

physical model, (b) 3D view of the simulated model with the injection and production wells for the small

physical model

220

(a)

(b)

Figure ‎5-8: (a) 2D view of the simulated model with the injection and production wells for the large

physical model, (b) 3D view of the simulated model with the injection and production wells for the large

physical model

221

5.5.2 Injection and production wells’ constraints

As discussed earlier, two horizontal wells were designed in the model as the injection and

production wells. In order to simulate the experimental conditions and to carry out the

VAPEX process, certain well constraints had to be defined for the injection and

production wells.

For the injection well, two operating constraints were defined and implemented. One of

the constraints was the maximum surface injection rate, and the other constraint was the

maximum bottom-hole pressure. The maximum pressure was set close to the dew point of

propane as the experimental conditions.

For the production well, the pressure constraints were selected in a way to represent the

experimental conditions and to establish the connection between the injection and

production well. For this purpose, the first constraint was the minimum bottom-hole

pressure, which was close to atmospheric pressure. Based on the experimental results, the

time required to establish the connection between the injection and production wells was

considered for the first pressure constraint. The second operating constraint was

introduced after the breakthrough time, and it was close to the injection pressure with

about 5kPa pressure difference as the experimental conditions.

5.5.3 History matching

Once the simulation models were built, the models were tuned to match the experimental

results. The only matching parameters in this study were dispersion coefficient and

relative permeability curves. These two parameters were changed to achieve the best

match between the simulation results and experimental observations.

222

Figure 5-9 shows the results obtained for the case of propane injection in the small

model. The results show a good match between the simulated values of recovery factor

and the experimental results. The average absolute error was observed to be about 5.9%.

Figure 5-10 shows the results obtained after injecting propane into the large model. The

number of grids in the large model was considerably greater, and the elapsed time for

running the simulation in the large model was significantly greater. Therefore, it was

more time consuming to get a better match in the large model. The average absolute error

was observed to be about 14.8% in the large model. In both the small and large models,

there was a very promising match between the experimental and simulation results during

the injection/production wells’ connection establishment. However, after the first

breakthrough of the solvent, there is a difference between the experimental and

simulation results, which is more noticeable in the large model. This can be due to the

sudden increase in the solvent production rate and change from single-phase production

to two-phase production. The same issue has been addressed by several researchers

during simulating the VAPEX process (Xu, et al. 2012, Rahnema et al. 2008, and

Yazdani 2007). They have reported that this difference after the solvent breakthrough can

be more noticeable when the distance between the injector and producer increases. Xu et

al. (2012) found that the entire-grid oil dilution by the numerical simulator can be another

reason for this discrepancy; while in the experiments the solvent may find a path between

the injector and producer and therefore the dilution takes place along this path. They

suggested that using finer grades might overcome this issue; however this might result in

significantly longer simulation time, numerical instability, and convergence problems. In

the case of butane injection, the simulation and experimental results were perfectly

223

matched even after the solvent breakthrough (Figure 5-11). This can be due to the low

rate of diffusion and relatively slower production rate compared to propane injection. The

experimental and simulation results were shown for CO2 and methane injection in

Figures 5-12(a) and 5-12(b), respectively. In the case of propane/CO2 and

propane/methane injection, the results showed the same trend as the pure propane

injection. That is, there was an over estimation of recovery factor by the simulated model

after the solvent breakthrough and under estimation of recovery factor towards the end of

the experiments. These results are graphed in Figures 5-13(a) and 5-13(b). The chamber

growth was also monitored during the simulation, is shown in Figure 5-14. It was found

that the simulation results were fairly consistent with the chamber growth observed

during the experiments.

The error analysis showed that the average absolute errors were 8.25%, 9.21%, 13.29%,

12.12%, and 22.57% for butane, CO2, methane, propane/CO2, and propane/methane,

respectively.

224

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Experimental

Simulation

Figure ‎5-9: Experimental and simulation results for the recovery factor after injecting propane in the small

model

225

Time (h)

0 20 40 60 80 100 120 140

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Experimental

Simulation

Figure ‎5-10: Experimental and simulation results for the recovery factor after injecting propane in the large

model

226

Time (h)

0 10 20 30 40 50 60 70 80 90 100 110

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Experimental

Simulation

Figure ‎5-11: Experimental and simulation results for the recovery factor after injecting butane in the small

model

227

Time (h)

0 50 100 150 200 250 300 350 400 450

Rec

over

y F

acto

r(%

OO

IP)

0

10

20

30

40

Experimental

Simulation

(a)

Time (h)

0 50 100 150 200 250 300

Rec

over

y F

acto

r(%

OO

IP)

0

10

20

30

40

Experimental

Simulation

(b)

Figure ‎5-12: Experimental and simulation results for the recovery factor after injecting (a)CO2 and (b)

methane in the small model

228

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Experimental

Simulation

(a)

Time (h)

0 20 40 60 80 100 120

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Experimental

Simulation

(b)

Figure ‎5-13: Experimental and simulation results for the recovery factor after injecting (a) propane/CO2 and

(b) propane/methane mixtures in the small model

229

(a)

(b)

Figure ‎5-14: Chamber evolution after 26 h in (a) simulated small model, (b) laboratory model

230

5.5.4 Effect of well configurations

The well configurations for the experimental studies were explained in detail in previous

chapters. In order, to investigate the effect of well spacing on the VAPEX process, a

series of simulation runs was carried out with various well spacing. For the second well

configuration, the injection well was considered to be 16 cm above the production well;

therefore, grid numbers i =12 × j = 1 to 5 × k = 8 were perforated for the injection well

while the production well was at the bottom of the model as it was in the first well

configuration. For the third well configuration, the injection well was considered to be 4

cm above the production well; therefore, grid numbers i =12 × j = 1 to 5 × k = 20 were

perforated for the injection well while the production well was at the bottom of the model

as it was in the first and second well configurations. For the forth configuration, the

injection well and production wells were located at the right corners of the models right

above each other (i.e., injection well: i =20× j = 1 to 5 × k = 1, and production well: i

=20× j = 1 to 5 × k = 24). Finally, the fifth well configuration was the injection well at

the right top corner and the production well at the left bottom corner of the model (i.e.,

injection well: i =20× j = 1 to 5 × k = 1 and production well: i =1× j = 1 to 5 × k = 24).

This sensitivity analysis was carried out in the small model as the simulation in the large

model elapsed over a significantly longer time.

Figure 5-15 depicts the results obtained for all the well configurations used in this study

for injecting propane in the small model. The fifth well configuration resulted in the

highest drainage rate and highest recovery factor. This can be due to the increased

distance between the injection and production wells, which resulted in greater contact

area between the solvent and heavy oil.

231

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

Frist well configuration

Second well configuration

Third well configuration

Forth well configuration

Fifth well configuration

Figure ‎5-15: Effect of well configuration on the recovery factor. For the first well configuration, the

injection well is located at the top of the model and 24 cm above the production well; for the second well

configuration, the injection well is located 16 cm above the production well; for the third well

configuration, the injection well is 4 cm above the production well; for the forth well configuration, the

injection well is 24 cm above the production well; and for the fifth well configuration, the injection well is

at the right top corner of the model; the production well is at the left bottom corner.

232

The greater contact area would result in better mixing between the solvent and heavy oil,

which would reduce the viscosity of the heavy oil. At the same time, the fifth

configuration has the advantage of great drainage height. When the injection well was

located 16 cm above the production well and in the pay zone, high production rate and

recovery factor were observed. However, when the well spacing was decreased to 4 cm,

the production rate and the recovery factor were drastically decreased. It can be

concluded, then, that the optimum mixing of solvent and heavy oil was achieved in the

fifth well configuration, and the first and second well configurations also showed

promising production rates and recovery factors.

5.5.5 Effect of permeability

Figure 5-16 shows the effect of permeability on the recovery factor after injecting

propane in the small model. Increasing the permeability increased the recovery factor.

The impact was more noticeable after the first breakthrough of the solvent and when the

two-phase production started from the production well.

5.5.6 Effect of grid thickness

The increase in the grid thickness in the j-direction significantly influenced the recovery

performance of the VAPEX after injecting propane. The results are shown in Figure 5-17,

which shows how changing the grid thickness from 1 cm to 8 cm reduced the production

rate and recovery factor.

5.5.7 Effect of time step

The change in the time step did not affect the results, as CMG STARSTM

chooses the

optimum time step based on the range set by the user. Therefore, as the results in Figure

233

5-18 show, no significant change occurred from selecting various time steps. However, it

should be mentioned that setting larger minimum time steps resulted in convergence error

by the software.

234

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

k = 10 D

k = 100 D

k = 150 D

Figure ‎5-16: Effect of permeability on the recovery factor after injecting propane

235

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

y = 1 cm

y = 4 cm

y = 8 cm

Figure ‎5-17: Effect of grid thickness on the recovery factor after injecting propane

236

Time (h)

0 20 40 60 80 100

Rec

over

y F

acto

r(%

OO

IP)

0

20

40

60

80

100

t = 1e-8 d

t = 1e-6 d

t = 1e-10 d

t = 1e-12 d

Figure ‎5-18: Effect of time step change on recovery factor after injecting propane

237

6. CHAPTER 6: SOFT COMPUTING APPROACH

“As complexity increases precise statements lose meaning and meaningful statements lose precision.”___

L. A. Zadeh

Soft computing techniques are tolerant of imprecision and partial truth, and inductive

reasoning is a key factor in this technique. Artificial neural network (ANN) is one of the

components of soft computing science, and its basis is like what a human brain does to

process tasks. The recent progress and success of utilizing artificial neural networks

(ANN) to solve various complicated engineering problems has drawn attention to its

potential applications in the petroleum industry.

ANN has been successfully utilized in several areas, such as permeability prediction, well

testing, PVT properties prediction, identification of sandstone lithofacies, improvement

of gas well production, prediction and optimization of well performance, and integrated

reservoir characterization (Mohammadpoor et al. 2010, 2011, 2012).

ANN is a system of interconnected parallel neurons that takes the input data and

multiplies it by connection weights. A bias value is added, and, then, the result is entered

into the transfer functions. In the case of supervised learning algorithms, the products of

transfer functions are compared with desired targets. If the product is not in the

acceptable range of error, then the initial weights and biases will be changed to match the

desired target.

Figure 6-1 shows a schematic of a neural network model. The parameter I is the input, the

parameter w is the weight, parameter b is the bias, and n is net input for the transfer

function f. Then, the output O is defined by equation (6.2).

238

I1

I2

I3

IN

w1

w2

w3

wN

∑

b

nf O

Figure ‎6-1: Schematic of an artificial neural network

239

bwIwIwIwIn NN ...332211......................................................................... (6.1)

nfO ....................................................................................................................... (6.2)

Each ANN has an input layer and one or more hidden layers. The input layer includes the

input neurons, and the hidden layers include hidden neurons and transfer functions.

Different types of transfer functions and the learning algorithm will be discussed in more

detail later in this chapter.

6.1 Data handling procedures

ANN is highly dependent on the input and output data. The accuracy and the total

training time are directly related to the number and type of input variables. Therefore, a

comprehensive study is essential prior to developing an ANN model. Data handling

procedures include two main steps: (1) data acquisition, and (2) data pre-processing.

6.1.1 Data acquisition

In order to construct a successful ANN model, choosing the most potent inputs is of

critical importance. Ineffectual inputs may complicate the training procedure and result in

imprecise predictions. After conducting a comprehensive literature review on the

available experimental studies and considering the observations during the experimental

results obtained in this study, five different parameters were considered as the inputs for

training the ANN model. In this study, drainage height, heavy oil viscosity, solvent type,

permeability, and porosity were considered as the inputs to predict the stabilized drainage

rate as the output of the ANN model.

240

To successfully model the complex relationships between the input and output

parameters, a large number of data sets is required to train and test the ANN model. In

order to gather an appropriate number of data sets, the experimental results from this

study were combined with the available experimental results in the literature, and a total

of 223 data sets was collected to develop the ANN model. These data sets were divided

into two categories: one category included 155 data sets for training and validating the

network and the second category consisted of 68 data sets for testing the trained network.

It should be mentioned that the data sets considered for the training procedure must cover

the whole input and output data range. Figure 6-2 shows the data distribution for training

and testing data sets for the input parameters used in this study. The collected data sets

and the sources of data are presented in Table B-1 to B-9 in Appendix B. Table 6-1

shows the input and output parameters and some general properties of data sets used for

this study.

6.1.2 Data normalization

In order to decrease the ANN training time, some data pre-processing such as data

normalization should be carried out. Different types of inputs with different data ranges

and distributions are typically present in the data sets, so data normalization will

significantly reduce this variation. For this purpose, equation (6.3) was used for the input

and output data sets. The approach used for scaling the network inputs and targets was to

normalize the mean and standard deviation of the training and testing data sets.

minmax

min.

XX

XXX norm

…………………………………………………………...……. (6.3)

241

H (cm)

0 20 40 60 80 100

Q (

mL

/h)

0

50

100

150

200

250

Training data sets

Testing data sets

(a)

Pinj

(kPa)

0 500 1000 1500 2000 2500 3000 3500 4000 4500

Q (

mL

/h)

0

50

100

150

200

250

Training data sets

Testing data sets

(b)

Porosity (%)

20 25 30 35 40 45

Q (

mL

/h)

0

50

100

150

200

250

Training data sets

Testing data sets

(c)

k (D)

0 100 200 300 400 500 600 700 800 900 1000 1100 1200

Q (

mL

/h)

0

50

100

150

200

250

Training data sets

Testing data sets

(d)

(cp)

0 50x103 100x103 150x103 200x103 250x103

Q (m

L/h)

0

50

100

150

200

250

Training data sets

Testing data sets

(e)

Figure ‎6-2: Data distribution for training and testing sets; stabilized drainage rate vs. (a) height (cm), (b)

injection pressure (kPa), (c) porosity (%), (d) permeability (D), and (e) viscosity (mPa.s)

242

Table ‎6-1: Data range for various input and output parameters used in this study

Variable Minimum Maximum Mean

Standard

deviation

Input

Drainage

height (cm)

7.5 100.5 30.5 19.1

Injection

pressure

(kPa)

69.0 4227.0 599.3 771.4

Oil viscosity

(mPa.s)

1390 225000 31239 47849

Permeability

(D)

3.0 1123.0 265.0 283.5

Porosity (%) 43.1 20.5 36.3 2.7

Output

Drainage

rate (mL/h)

0.02 218.10 35.02 45.74

243

where X denotes the input and output parameters. The subscript max refers to the

maximum and min refers to the minimum value of the variable. The new normalized

variable, Xnorm., takes the range from zero to 1 for all the parameters.

6.2 Neural network development

The number of layers, the interconnections between the layers, and the number of

processing neurons per layer define the ultimate architecture of a neural network model.

Hence, developing an optimal network based on the mentioned variables is not an easy

task and requires following some rules to reduce the number of iterations. There are

different types of supervised and un-supervised ANN architectures utilized for different

science and engineering problems. Among all types of available networks, the multiple-

layer feed-forward back-propagation (BP) architecture is the most widely used neural

network for petroleum engineering applications. This type of network is capable of

representing non-linear functional mappings between inputs and outputs

(Mohammadpoor et al. 2010, 2012). It has been observed by several researchers that a

two-layer BP model with a sigmoid function in the hidden layer and linear function in the

output layer can fit any finite mapping problem (Xu, 2012, Beale et al., 2010, Salahshoor

et al., 2012).

The BP network employed in this study was composed of four hidden layers, and each

layer included one transfer function. There are various types of transfer functions. Three

of the most common functions used are sigmoid, linear and hard limit functions. The

transfer function that was utilized in the hidden layer was a sigmoid function, which is

defined by equation (6.4).

244

jnje

O

1

1……………………………….…………………………………...……. (6.4)

where O is the output of each neuron, and n is the sum weighted inputs and bias.

Equation (6.5) can be used to calculate n:

j

N

i

iijj bOwn

1

……………………………….……………….…………...……. (6.5)

In equation (6.5), N is the number of neurons in each layer, bj is a bias parameter, and wij

is the weight between node j of layer l to node i of layer l-1. The term bias is utilized to

minimize the number of iterations and develop a constant offset.

As mentioned earlier, the linear transfer function was used for the output layer:

jj nO ……………………………….……………………………….………...……. (6.6)

After initializing the network weights and biases, the network is prepared for conducting

the training procedure. Once the training starts, the weights and biases are adjusted

iteratively to minimize average squared error between the network outputs and the

desired targets. The root mean square error (MSE) defined in equation (6.7) is called the

network performance function in BP networks.

n

i

ipi OOn

MSE1

2

,

1...………………..……………..……………………………. (6.7)

where Op is the predicted output of the network and O is the initial target of the network.

Various training algorithms such as Scaled Conjugate Gradient, Gradient Descent, and

Levenberg-Marquardt (LM), which utilize the gradient of network performance function

to adjust the weights, are available. The LM training function is widely accepted because

245

of its robustness and fast training convergence. The goal in the LM algorithm is to

minimize the following equation, which is, in fact, the sum of the squares of deviation

(Salahshoor et al., 2012):

m

j

j xrxf1

2

2

1……………………………………………………………………. (6.8)

where x is a vector and rj denotes the jth residual function. Then, the LM algorithm uses

equation (6.9) as an iteration formula to conduct the search procedure.

iii xfHdiagHxx

1

1 ……..……………………………………………. (6.9)

where H is the Hessian matrix defined by equation (6.10), ixf is the first difference,

and xf is defined by equation (6.11):

m

j

jj

TxrxrxJxJxfH

1

22…………………………………….…. (6.10)

m

j

jj xrxrxf1

…………………………………………………...………... (6.11)

The last step is to define the number of neurons in the hidden layer, which will

significantly affect the ultimate performance of the neural network model. There is no

proven rule about the optimum number of neurons in the hidden layers. Lawrence et al.

(1996) found that the network size is dependent on: 1) the complexity of the

approximation function, 2) the range of data sets distribution, and 3) the size of the

network compared to required size for an optimal solution. A low number of neurons will

result in under-fitting, which will consequently give high training and generalization

246

error. A large number of neurons will result in over-fitting, which will decrease the

training error, but the generalization error will be high. The best way to find the optimum

number of neurons is to start with a reasonable value and monitor the results using cross-

validation technique to find the optimum number of hidden neurons. For the first guess,

there are some rules of thumb that can be used as a guideline, but these rules do not

consider the nature of the problem or the quality and number of data sets. Among these

rules, the followings are more often cited (Heaton, 2008); however, it should be

emphasized that these rules can be used only as a starting point guideline:

The number of hidden neurons should be between the input and output layers’

size.

The number of hidden neurons should be less than the size of the input layer.

sizelayeroutputsizelayerinputneuronshiddenofNumber 3

2

The above-mentioned rules were used as initial guidelines for training BP networks in

this study; therefore, the basic approach to constructing the optimum network was trial

and error, and, then, the results for various network topologies were utilized to find the

optimum network. In this study, MATLAB (R2012a) software from Mathworks was

utilized to train and test the BP networks. Figure 6-3 presents an example of the training

procedure, in which training, validation, and testing stages are shown. In this figure, the

results for each trial are provided for the predicted flow rate versus the actual flow rate

fed to the network.

247

Figure ‎6-3: An example of network training procedure; plot of: (a) predicted outputs by network for

training data sets, (b) predicted outputs by network for validation data sets, (c) predicted outputs by

network for testing data sets, and (d) predicted outputs by network for the whole group of data sets

chosen for training procedure

0 0.2 0.4 0.6 0.8 1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Actual flow rate (normalized Q)

Pred

icte

d f

low

ra

te (

no

rm

ali

zed

Q)

(a) Training: R=0.98268

Flow rate, Q

Fit

y=x

0 0.2 0.4 0.6 0.8

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8


Pred

icte

d f

low

ra

te (

no

rm

ali

zed

Q)

(b) Validation: R=0.92013

Flow rate, Q

Fit

y=x

0 0.2 0.4 0.6 0.8

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8


Pred

icte

d f

low

ra

te (

no

rm

ali

zed

Q)

(c) Test: R=0.84419

Flow rate, Q

Fit

y=x

0 0.2 0.4 0.6 0.8 1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Pred

icte

d f

low

ra

te (

no

rm

ali

zed

Q)

(d) All: R=0.948

Flow rate, Q

Fit

y=x

248

Training, validation, and testing data sets were chosen randomly among the 155 data sets

for training inputs. Regression analysis was carried out, and R and RMSE values were

monitored for each trial to plan the next network topology. A summary of some of the

various trials for network topologies and the results obtained are presented in Table 6-2.

After constructing each network, the untouched portions of actual data (testing category)

were used to simulate the developed network. For each network, the results were

graphed, and error analysis was carried out to monitor the accuracy of the prediction. To

carry out the error analysis, RMSE (Eq. 6.12) and correlation coefficient (R-coefficient)

(Eq. 6.13) were calculated for the training and testing data sets.

n

i

ipi OOn

RMSE1

2

,

1...……………….………..……………..………………. (6.12)

n

i

n

i

pipi

n

i

pipi

OOOO

OOOO

R

1 1

2

,

2

1

,

………..………...……………....………………. (6.13)

Table 6-2 shows that the optimum network was found to be topology #1 with four hidden

layers and 20 neurons on the first hidden layer,15 neurons on the second hidden layer, 10

neurons on third hidden layer, and 5 neurons on the fourth hidden layer. The schematic of

the developed BP network is presented in Figure 6-4. The number of inputs and output,

number of neurons on each hidden layer, and transfer functions are schematically

presented in this figure.

The network predictions for the training data sets are presented in Figure 6-5.

249

Table ‎6-2: Summary of the results for some selected training and testing trials

No. Network topology R (Training) RMSE

(Training) R (Testing)

RMSE

(Testing)

1 I-20-15-10-5-O 0.9616 0.0543 0.8177 0.1094

2 I-10-10-O 0.9556 0.0650 0.6233 0.2067

3 I-15-10-O 0.9552 0.0629 0.8187 0.1409

4 I-10-10-5-O 0.9528 0.0645 0.8264 0.0819

5 I-10-O 0.9395 0.0689 0.6901 0.1751

6 I-15-15-15-O 0.9367 0.0782 0.7025 0.1532

7 I-20-20-O 0.9319 0.0699 0.6997 0.1878

8 I-10-10-10-5-O 0.9316 0.0756 0.7175 0.1483

9 I-20-O 0.9297 0.0953 0.5159 0.3341

10 I-10-20-15-O 0.9167 0.0924 0.7278 0.2173

11 I-15-10-10-O 0.9125 0.0821 0.7636 0.1483

12 I-15-O 0.8994 0.1053 0.6762 0.2798

13 I-10-15-10-O 0.8839 0.0715 0.5788 0.1294

14 I-20-25-O 0.8683 0.1025 0.7410 0.1313

15 I-15-15-10-5-O 0.8122 0.1327 0.3324 0.5179

250

Figure ‎6-4: Schematic of the developed BP network; there are 20 neurons on the first hidden layer and 15

neurons on the second hidden layer. The transfer functions used for hidden layers were log sigmoid

functions, and linear transfer function was used for output layer

251


0.0 0.2 0.4 0.6 0.8 1.0 1.2

Pre

dic

ted f

low

rat

e (n

orm

aliz

ed Q

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Flow rate, Q

y=x

Fit

Figure ‎6-5: Output of the developed network vs. the actual data after simulating the model with training

data sets

252

The cross plot of network outputs versus training data sets shows a perfect correlation

with the y = x line. In order to test the validity and accuracy of the model, the outputs

were simulated using the untouched portion of data, and the results are shown in Figure

6-6.

As mentioned earlier, the calculations are based on the number of nodes, transfer

functions on hidden layers, weights for the nodes, and the biases. The matrices of weights

and biases for the hidden layers of the developed network are provided in Appendix C.

These matrices can be used to regenerate the developed BP model and utilize the model

to predict the production rate.

6.3 Sensitivity analysis

As mentioned previously, the selected input variables had been found by several

researchers to be the key parameters governing the ultimate performance of the VAPEX

process. However, the degree of dependency of drainage height on each of these

parameters has yet to be determined. In this research, the developed ANN model was

utilized to conduct a sensitivity analysis on the input variables. For this purpose, the term

relevance factor, r, was employed to study the significance of the five input variables on

the heavy oil production rate. The greater the absolute value of the r-factor for a specific

input, the more significant an effect it would have on the output. On the other hand, the

positive value of r-factor shows the positive impact of the input parameter on the output,

while the negative value of r-factor implied the negative impact of the input parameter on

the output. A very small value of r-factor indicates the negligible impact of the input

parameter on the output. Therefore, the r-factor for each input was calculated based on

the available experimental data using equation (6.14) (Chen et al., 2014):

253


0.0 0.2 0.4 0.6 0.8 1.0 1.2

Pre

dic

ted f

low

rat

e (n

orm

aliz

ed Q

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Flow rate, Q

y=x

Fit

Figure ‎6-6: Output of the developed network vs. the actual data after simulating the model with testing data

sets

254

n

i

n

i

ii

n

i

ii

QQII

QQII

QIr

1 1

22

1, …………………………………………………. (6.14)

where I is the input parameter, I is the average value of the input parameter, iQ is the

production rate, and Q is the average value of the production rate.

The results obtained for the relevance factor for each parameter are shown in Figure 6-7.

The highest relevance factor was found to be 0.5751 for the permeability; this was even

higher than the value for drainage height, which was found to be 0.4653. As mentioned in

Chapter 4, Butler proposed that both permeability and drainage height have a square root

relationship with production rate. However, it was found later by Yazdani (2007), that

production rate has a higher dependency on drainage height. This was further proved by

the experimental results obtained in this study, and it was explained in detail in Chapter

4. However, the results obtained based on the sensitivity analysis, which has taken into

account a significantly wider range of data, showed that there is a higher dependency

between the production rate and permeability than previously thought. This can be

another reason for the underestimation of the production rate using Butler’s equation. The

r-factor for the porosity was very low at 0.00001. This shows that porosity did not have

any significant effect on the production rate. The r-factor for viscosity was found to be -

0.1969. As expected, the viscosity had a significant negative impact on the production

rate.

255

Input parameters

Injec

tion pres

sure

Porosit

y

Permeb

ility

Viscosit

y

Drainag

e heig

ht

Rel

evan

cy f

acto

r (r

)

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

0.0824

0.00001

-0.1969

0.4653

0.5751

Figure ‎6-7: Relevancy (r) factor for various parameters to the production rate

256

6.4 Comparison of results

As discussed in the previous chapter, based on the experimental results, the optimum

value for exponent n for VAPEX scale-up was found to be 1.2. Also, it was mentioned

that Yazdani (2007) found out that exponent n should be between 1.2 and 1.3. During his

experiments, the following correlations were developed based on Butler equation and

new observations for exponent n:

kHQ 26.1017.0 ……………………………………………….…………….... (6.15)

kHQ 13.10288.0 ……………………..…………………………………….... (6.16)

The drainage rate was then calculated using these equations, and the results were graphed

versus the actual testing data sets to check the accuracy of these correlations. Moreover,

the obtained equations based on the experiments in this study were also used, and error

analysis was carried out on the results. R-coefficient and RMSE were calculated for the

testing data sets after utilizing the above-mentioned correlations.

The results after implementing equations (6.15) and (6.16) are presented in Figures6-8

and 6-9, respectively. The same procedure was followed for testing equations (4.15) to

(4.18), and the results are presented in Figures 6-10 to 6-13. In addition, a summary of

the results for the error analysis is provided in Table 6-3.

257

Actual flow rate, Q (cm2/h)

0 20 40 60 80 100 120

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

20

40

60

80

100

120

Figure ‎6-8: Plot of predicted stabilized drainage rate by eq. 6.15 versus actual data sets for testing

258


0 20 40 60 80 100 120

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

20

40

60

80

100

120


259


0 20 40 60 108 109 110

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

10

20

30

40

50140

150

Flow rate, Q

y=x

Fit


260


0 20 40 60 108 109 110

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

10

20

30

40

5095

100

105

Flow rate, Q

y=x

Fit


261


0 20 40 60 108 109 110

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

10

20

30

40

50

96

98

100

Flow rate, Q

y=x

Fit


262


0 20 40 60 108 109 110

Pre

dic

ted f

low

rat

e, Q

(cm

2/h

)

0

10

20

30

76

78

80

Flow rate, Q

y=x

Fit


263

Table ‎6-3: Error analysis for various techniques to predict drainage rate

Prediction method R RMSE

BP network 0.8177 0.1094

Eq. 6.15 0.7097 14.8966

Eq. 6.16 0.6968 11.6893

Eq. 4.15 0.7135 18.2123

Eq. 4.16 0.7134 12.3721

Eq. 4.17 0.7135 11.8365

Eq. 4.18 0.7132 9.4831

264

7. CHAPTER 7: CONCLUSIONS AND

RECOMMENDATIONS

7.1 Conclusions

An extensive experimental study involving injecting various solvents in two large-scale

visual physical models was carried out. Various parameters were recorded during the

experiments to investigate the effect of drainage height and solvent type on the VAPEX

process. A comprehensive database was gathered, and the following major conclusions

were drawn:

1. Propane showed promising recovery factor results in both physical models, while

butane injection also showed acceptable results in terms of ultimate recovery

performance. The ultimate recovery factor after injecting propane was found to be

about 75% of original oil in place in the small and large models.

2. Although pure CO2 and methane injection did not show acceptable recovery

performance, CO2 and methane were found to be good carrier gases, while

propane/CO2 and propane/methane mixtures significantly improved recovery

performance. In the case of propane/CO2 injection, an ultimate recovery factor of

54% of original oil in place was observed in the VAPEX models. On the other hand,

after injecting propane/methane mixture, an ultimate recovery factor of 48% of

original oil in place was observed in both the small and large VAPEX models.

3. The main effect of drainage height was observed while comparing the results for

stabilized drainage rates in the small and large physical models. The stabilized

drainage rates were significantly higher in the large model with greater drainage

265

height, which proves the prominent effect of drainage height on the VAPEX process.

For instance, the stabilized drainage rates after injecting propane were found to be

0.22 mL/min and 0.50 mL/min in the small and large models, respectively.

4. The efficiency of propane as an injection solvent was further confirmed by comparing

the solvent utilization curves for various solvents used in this study.

5. It was observed that residual oil saturations close to the injection wells were very low

for all the solvents. Moreover, the lowest residual oil saturation was obtained after

injecting propane for both small and large models. The residual oil saturation for

sample location 1 was 4.3% and 5.1% for the small and large models, respectively.

On the other hand the highest residual oil saturation was observed at the bottom of the

physical models and close to production wells. The highest residual oil saturation was

found to be 80.4% in the small model and 88.9% in the large model for the case of

CO2 injection.

6. Using various solvents, it was observed that more asphaltene precipitation occurred

close to the injection points and at the oil/solvent interface. Comparing the textures of

the asphaltene precipitants from different locations of the models, it was found that

the precipitants close to the injection points where more brittle, while the precipitants

close to the production points were more ductile.

7. The amount of asphaltene precipitation in the large model was slightly greater due the

longer path between the injection and production wells and the longer contact time

between the oil and solvent.

266

8. After comparing the asphaltene precipitation in the small and large models, it was

observed that in the case of propane injection, more asphaltene precipitation was

observed in different physical model locations.

9. The image analysis on the chamber evolution showed that the highest sweep

efficiency was observed after injecting propane, followed by butane, propane/CO2,

propane/methane, CO2, and methane.

10. It was found that establishing a sound connection between the injection and

production wells would significantly affect the ultimate performance of the VAPEX

process. Poor connections between the injection and production wells resulted in

drastically lower production rates and slow rate processes. For instance, the stabilized

drainage rate after propane injection was found to increase from 0.04 mL/min to 0.50

mL/min after injecting propane in the large model by improving the connections

between injection and production well at the beginning of the VAPEX experiment.

11. Further analysis of the experimental results obtained in this study showed that

Butler’s equation, which states square root proportionality between the drainage

height and drainage rate, significantly under predicts the drainage rate. However, it

was found that results proposed by Yazdani (2007) showed better proportionality

between the drainage height and drainage rate in the VAPEX process. The

experimental results obtained in this study indicated that drainage rate is proportional

to the drainage height raised to the power of 1.2 in the VAPEX process.

12. VAPEX number, Ns, increased with increasing drainage height, and it was concluded

that VAPEX number was dependent on the drainage height and oil-solvent properties.

Comparing various solvents used for these experiments, the highest values for

267

VAPEX number were achieved after injecting propane, and VAPEX numbers for

butane and propane/CO2 were also relatively high and close to each other. However,

the lowest values were obtained after injecting pure methane and pure CO2.

13. The experiments were simulated numerically, and satisfactory history matching was

achieved. The major difference between the experimental and simulation results was

observed after the first breakthrough of the solvent.

14. Injection and production wells’ configurations significantly affected the recovery

performance of the VAPEX process. It was observed that longer distance between the

injection and production wells alongside the drainage height will increase the

production rate in VAPEX.

15. It was found that current empirical correlations fail to predict the drainage rate in the

VAPEX process, and they are very limited to the oil-solvent conditions under which

they were initially developed. Hence, a new soft computing-based approach was

utilized to develop a universal model to predict the drainage rate in the VAPEX

process. The estimated drainage rates with the new model showed high accuracy and

a wider range of applicability.

268

7.2 Recommendations

1. The physical models used in this study had maximum injection pressure limitations.

Therefore, in order to evaluate the suitability of various mixtures of propane and other

carried gases, new models with higher pressure tolerance can be designed.

2. In order to obtain more data for various drainage heights and get more accurate height

dependency correlations, new physical models with various drainage heights can be

designed and employed for VAPEX experiments.

3. To further investigate the effect of asphaltene formation and precipitation in the

VAPEX process, a molecular and compositional analysis on the asphaltene

precipitation in the physical models can be carried out.

4. The experimental results on the VAPEX process are very limited so far; therefore, the

ANN model can be further improved by introducing a wider range of data sets in the

future.

5. Other soft computing techniques such as Genetic Algorithm (GA) and Fuzzy Logic

can be incorporated alongside ANN to optimize the developed model in the future.

These soft computing techniques can be utilized to improve the results obtained by

commercial simulators to enhance the performance of the numerical simulation of the

VAPEX process.

269

REFERENCES

Abukhalifeh, H.: Determination of concentration-dependent dispersion of propane in

vapour extraction of heavy oil, Ryerson University, Toronto, ON, Canada, 2010.

Abukhalifeh, H., Upreti, S. R., Lohi, A.: Effect of drainage height on concentration-

dependent propane dispersion in VAPEX, The Canadian Journal of Chemical

Engineering, Volume 90, No. 2, pp. 336-341, 2011.

Adams, D. M.: Experiences with waterflooding Lloydminster heavy-oil reservoirs, JPT,

Volume 34, No. 8, pp. 1643-1650, 1982.

Afsahi, B. and Kantzas, A.: Effect of the presence of sand on solvent diffusion in

bitumen, SPE 2005-092, Calgary, Alberta, Canada, 2005.

Ahmadloo, F.: Investigation of interplay of capillarity, drainage height, and aqueous

phase saturation on mass transfer phenomena in heavy oil recovery by VAPEX

process, University of Regina, Regina, SK, Canada, 2012.

Ahmadloo, F., Asghari, K., Henni, A. and Freitag, N. P.: Experimental results and

analytical modeling of solvent leaching gravity drainage phenomenon in heavy oil

reservoirs, SPE 144542, Denver, Colorado, USA, 30 October-2 November, 2011.

Alkindi, A., Muggeridge, A. and Al-Wahaibi, Y.: Experimental investigation into the

influence of convective dispersion and model height on oil drainage rates during

VAPEX, SPE 129169, Muscat, Oman, 30 April 11-13, 2010.

Allen, J. C.: Gaseous solvent heavy oil recovery, US Patent 1027851, 1974.

270

Allen, J. C.: Multiple solvent heavy oil recovery method, US Patent 3954141, 1976.

Alvarado, V. and Manrique, E.: Enhanced oil recovery: An update review, Energies

Journal, Volume 3, No. 9, pp. 1529-1575, 2010.

Association for the Study of Peak Oil and Gas USA, http://peak-oil.org/.

Ayub, M.: The role of capillarity in the VAPEX process, Journal of Petroleum Science

and Technology, Volume 27, No. 10, pp. 1020-1032, 2009.

Ayub, M. and Tuhinuzzaman, M.: The role of capillarity in the VAPEX process, SPE

2007-075, Calgary, Alberta, Canada, June 12 – 14, 2007.

Azin, R., Kharrat, R., Ghotbi, C. and Vossoughi, S.: Applicability of the VAPEX process

to Iranian heavy oil reservoirs, SPE 92720, Bahrain, 12-15 March 2005.

Azin, R., Kharrat, R., Vossoughi, S., and Ghotbi, C.: Study of the VAPEX process in

fractured physical systems using different solvent mixtures, Oil & Gas Science

and Technology, Volume63, No. 2, pp. 219-227, 2008.

Babadagli, T., Sahin, S., Kalfa, U., Celebioglu, D., Karabakal, U. and Topguder, N. N.:

Development of heavy oil fractured carbonate Bati Raman Field: Evaluation of

steam injection potential and improving ongoing CO2injection, SPE 115400,

Denver, Colorado, USA, September 21-24, 2008.

Badamchizadeh, A.: Use of CO2 in VAPEX, experimental and modeling study,

University of Calgary, Calgary, AB, Canada, 2013.

271

Bagci, A. S.: Experimental and simulation studies of SAGD process in fractured

reservoirs, SPE 99920-MS, Tulsa, Oklahoma, April 22–26, 2006.

Baviere, M., Glenat, P. and Plazanet, V.: Improvement of the efficiency/cost ratio of

chemical EOR processes by using surfactants, polymers, and alkalis in

combination, SPE Reservoir Engineering Journal, Volume 10, No. 3, pp. 187-

193, 1995.

Beale, M.H., Hagan, M.T. and Demuth, H. B.: Neural Network Toolbox User’s Guide,

MA: Mathworks, 2010.

Beckman, J. W.: The action of bacteria on mineral oil, Industrial Engineering Chemical

News, 4 November 1926.

Blevins, T. R. and Billingley, R. H.: The ten-pattern steamflood, Kern River field,

California, JPT, Volume 27, No. 12, pp. 1505-1514, 1975.

Boustani, A. and Maini, B. B.: The role of diffusion and convective dispersion in vapour

extraction process, JCPT, Volume 40, No.4, pp. 68-77, 2001.

Bursell, C. G. and Pittman, G. M.: Performance of steam displacement in the Kern River

field, JPT, Volume 27, No. 8, pp. 997-1004, 1975.

Butler R. M. and Jiang Q.: Improved VAPEX performance using widely spaced

horizontal injectors and producers, SPE 97-174, Regina, Saskatchewan, October

19-22 1997.

272

Butler, R. M. and Jiang, Q.: Improved recovery of heavy oil by VAPEX with widely

spaced horizontal injectors and producers, JCPT, Volume 39, No.1, pp. 48-56,

2000.

Butler, R. M. and Jiang, Q.: Selection of well configuration in VAPEX process, SPE

37145, Calgary, AB, November 18-20, 1996.

Butler, R. M., Mokrys, I. J.: A new process (VAPEX) for recovering heavy oils using hot

water and hydrocarbon vapour, JCPT, Volume 30, No. 1, pp. 97-106, January

1991.

Butler, R .M. and Mokrys, I. J.: Solvent analogue model of steam-assisted gravity

drainage”, AOSTRA Journal of Research, Volume 5, No. 1, pp. 17-32, 1989.

Canadian Association of Petroleum Producers, Available at: http://www.capp.ca/rce/wp-

content/uploads/2013/10/Map_Western-Canada_NEW_Jan10.jpg

Carman, P. C.: Fluid flow through granular beds, Transactions of the Institution of

Chemical Engineers, Volume 15, pp. 150-167, 1937.

Chatzis, I. and Morrow, N. R.: Correlation of capillary number relationships for

sandstone, SPE Journal, Volume 24, No. 5, pp. 555-562, 1984.

Chen, G., Fu, K., Liang, Z., Sema, T., Li, C., Tontiwachwuthikul, C. and Idem, R.: The

genetic algorithm based back propagation neural network for MMP prediction in

CO2-EOR process, Fuel, Volume 126, pp. 202-212, 2014.

Chu, C.: State-of-the-art review of fireflood field projects, JPT, Volume 34, No. 1, pp.

19-36, 1982.

273

Chugh, S., Baker, S. and Telesford, A.: Mainstream options for heavy oil: Part I – cold

production, JCPT, Volume 39, No. 4, April 2000.

Clark, B.: Heavy oil, extra-heavy oil and bitumen – unconventional oil, working

document of the NPC Global Oil and Gas study, July, 2007.

Cooke, C. E., Williams R. E. and Kolodzie P. A.: Oil recovery by alkaline waterflooding,

JPT, Volume 26, No. 12, pp. 1365-1374, 1974.

Cuthiell, D., McCarthy, C., Kissel, G. and Cameron, S.: The role of capillarity in

VAPEX, SPE 2006-073, Calgary, Alberta, Canada, 2006.

Das S. K.: In situ recovery of heavy oil and bitumen using vapourized hydrocarbon

solvents, University of Calgary, Calgary, Canada, 1995.

Das S. K.: VAPEX: An efficient process for the recovery of heavy oil and bitumen, SPE


Das, S. K. and Butler, R. M.: Diffusion coefficients of propane and butane in Peace River

bitumen, Canadian Journal of Chemical Engineering, Vol. 74, No.6, pp. 985-992,

1996.

Das, S. K. and Butler, R. M.: Mechanism of the vapour extraction process for heavy oil

and bitumen, Journal of Petroleum Science and Engineering, Vol. 2, pp. 43-59,

1998.

Das, S. K. and Butler, R. M.: VAPEX process in a packed cell using butane as a solvent,

SPE HWC94-97; Calgary, Alberta, Canada, March 1994.

274

Derakhshanfar, M.: Effects of waterflooding, solvent injection, and solvent convective

dispersion on vapour extraction (VAPEX) heavy oil recovery, University of

Regina, Regina, SK, Canada, 2012.

Derakhshanfar, M., Kharrat, R. and Rostami, B.: Effect of mixed gas solvent injection on

performance of the VAPEX process in an Iranian heavy oil sample, CIPC2009-

094, Calgary, Alberta, Canada, June 16-18, 2009.

Dong, M.: Simulation of enhanced heavy oil recovery by chemical flooding, PTRC

Report, December 2008.

Dong, M, Dullien, F.A.L., and Dai, L. and Li, D.: Immiscible displacement in the

interacting capillary bundle model: Part II applications of model and comparison

of interacting and non-interacting capillary bundle models, Transport in Porous

Media, Volume63, pp. 289-304, 2006.

Dunn, S. G., Nenniger, E. H. and Rajan, V. S. V.: A study of bitumen recovery by gravity

drainage using low temperature soluble gas injection, Can. J. Chem. Eng.,

Volume 67, pp. 978-91, 1989.

Dusseault, M. B. and Geilikman, M. B.: Practical requirements for sand production

implementation in heavy oil applications, SPE 30250, Calgary, Alberta, Canada,

June 19-21, 1995.

Dusseault, M. B., Ormis, A. S. and El-Sayed, S.: Heavy oil production enhancement by

encouraging sand production, SPE 59276, Tulsa, Oklahoma, USA, 3–5 April.

275

El-Haj, R., Lohi, A. and Upreti, S. R.: Experimental determination of butane dispersion

in vapour extraction of heavy oil and bitumen, Journal of Petroleum Science and

Engineering, Volume 67, pp. 41-47, 2009.

Emmons, F. R., Hudspeth, L. D., Clancy, J. P., Zornes, D. R. and Philcox, J. E.: Nitrogen

management at the East Binger Unit using an integrated cryogenic process, SPE

15591, New Orleans, 5–8 October, 1986.

Etminan, S. R., Maini, B. B., and Kharrat, R.: The role of connate water saturation in

VAPEX process, JCPT, Volume47, No. 2, pp. 8-12, 2008.

Fanchi, J. R.: Feasibility of reservoir heating by electromagnetic irradiation, SPE 20483,

September 1990.

Farouq Ali, S. M. and Meldau, R. F.: Current steamflood technology, JPT, Volume 31,

No. 10, pp. 1332-1342, 1979.

Farouq Ali, S. M.: A current appraisal of in-situ combustion field tests, JPT, Volume 24,

No. 4, pp. 477-486, 1972.

Faruk, C.: Reservoir formation damage, fundamentals, modeling and mitigation and

assessment, second edition, Gulf Professional Publishing, Burlington, MA, USA,

pp. 129-132, 2007.

Fatt, I.: Pore structure of sintered glass from diffusion and resistance measurements, The

Journal of Physical Chemistry, Volume 63, pp. 751-752, May 1959.

Forth, R., Slevinsky, B., Lee, D. and Fedenczuk, L.: Application of statistical analysis to

optimize reservoir performance, JCPT, Volume 35, No. 8, 1996.

276

Frauenfeld, T. W. and Lillico, D. A.: Solvent-assisted method for mobilizing viscous

heavy oil, US Patent 5899274A,1999.

Friedrich, K.: Effect of a non-condensable gas on VAPEX process, University of

Waterloo, Waterloo, Ontario, Canada, 2005.

Goodlet, G. O., Honarpour, F. T., Chung, F. T., Sarathi, P. S.: The role of screening and

laboratory flow studies in EOR process evaluation, SPE 15172, Billings,

Montana, USA, 1986.

Grane, F. E. and Gardner, G. H. F.: Measurements of transverse dispersion in granular

media, Journal of Chemical and Engineering Data, Volume 6, No. 2, pp. 283-

287, 1961.

Grogan, A. T., Pinczewski, V. W., Ruskauff, G. J. and Orr, F. M.: Diffusion of CO2 at

reservoir conditions: models and measurements, SPE Reservoir Engineering


Haghighat, P.: Impact of asphaltene precipitation on VAPEX performance, University of

Calgary, Calgary, AB, Canada, 2008.

Hayduk, W., Castaneda, R., Bromfield H. and Perras, R. R.: Diffusivities of propane in

normal paraffin- chlorobenzene, and butanol solvents, AIChE Journal, Volume

19, pp. 859-861, 1973.

Hayduk, W. and Cheng, S. C.: Review of relation between diffusivity and solvent

viscosity in dilute liquid solutions, Chem. Eng. Sci., Volume 26, pp.635-646,

1971.

277

Hayduk, W. and Minhas, B. S.: Correlations for prediction of molecular diffusivities in

liquids, Can. J. Chem. Eng., Volume 60, pp. 295-299, April 1982.

Heaton, J.: Introduction to neural networks for Java, Heaton Research Inc., Ch. 5, 2008,

available online at: http://www.heatonresearch.com/online/introduction-neural-

networks-java-edition-2

Hiss, T. G. and Cussler, E. L.: Diffusion in high viscosity liquids, AIChE Journal,

Volume 19, pp. 698-703, 1973.

Islam, M. R.: Emerging technologies in enhanced oil recovery, Energy Sources, Volume

21, pp. 97-111, 1999.

Islam, M. R., Wadadar, S. S. and Bansal, A.: Enhanced oil recovery of Ugnu tar sands of

Alaska using electromagnetic heating with horizontal wells, SPE 22177,

Anchorage, Alaska, May 29-31, 1991.

James, L. A.: Mass transfer mechanisms during the solvent recovery of heavy oil,

University of Waterloo, Waterloo, Ontario, Canada, 2009.

James, L. A. and Chatzis, I.: Details of gravity drainage of heavy oil during vapour

extraction, SCA 2004-23, Abu Dhabi, UAE, October 5 – 9, 2004.

James, L. A. and Chatzis, I.: Mass transfer coefficients in vapour extraction (VAPEX),

SPE 2007-199, Calgary, Alberta, Canada, June 12 – 14, 2007.

James L. A, Rezaie N. and Chatzis I.: VAPEX, warm VAPEX, and hybrid VAPEX—the

state of enhanced oil recovery for in situ heavy oils in Canada, SPE 2007-200,

Calgary, Alberta, Canada, June 12-14 2007.

278

Javaheri, M. and Abedi, J.: Natural convection in CO2-based VAPEX, SPE 165352,

Monterey, California, USA, April 19-25 2013.

Jia, X., Zeng, F., and Gu, Y.: A novel solvent injection technique for enhanced heavy oil

recovery: Cyclic production with continuous solvent injection, SPE 165455,

Calgary, AB, June 11-13, 2013.

Jia, X., Zeng, F., and Gu, Y.: One-dimensional mathematical modelling of vapour

extraction (VAPEX),SPE 157938, Calgary, AB, June 12-14, 2012.

Jia, X., Zeng, F., and Gu, Y.: Pressure pulsing cyclic solvent injection (PP-CSI): A new

way to enhance the recovery of heavy oil through solvent-based enhanced oil

recovery techniques, SPE 166453, New Orleans, LA, 30 September-2 October,

2013.

Jiang, Q.: Recovery of heavy oil and bitumen using VAPEX process in homogeneous and

heterogeneous reservoirs, University of Calgary, Calgary, AB, Canada, 1997.

Jiang, Q., and Butler, R. M.: Selection of well configuration in VAPEX process, SPE

37145, Calgary, AB, November 18-20, 1996.

Kapadia, R. A., Upreti, S. R., Lohi, A. and Chatzis, I.: Determination of gas dispersion in

vapour extraction of heavy oil and bitumen, Journal of Petroleum Science and

Engineering, Volume 51, pp. 214–222, 2006.

Karmaker, K. and Maini, B. B.: Applicability of vapour extraction process to problematic

viscous oil reservoirs, SPE 84034, Denver, Colorado, USA, October 2003.

279

Kok, M. V., Yildirim, Y., Akin, S.: Application of vapour extraction (VAPEX) process

on carbonate reservoirs, Energy Sources: Part A, Volume 31, pp. 377–86, 2009.

Krause, M. H.: Modelling sub-core scale permeability in sandstone for use in studying

multiphase flow of CO2 and brine in core flooding experiments, Stanford

University, San Francisco, USA, 2009.

Kumar, M. and Beatty, F. D.: Cyclic steaming in heavy oil diatomite, SPE 29623,

Bakersfield, California, March8–10, 1995.

Lawrence, S., Giles, C. L., and Tsoi A. C.: What size neural network gives optimal

generalization? Convergence properties of back-propagation, Institute of

Advanced Computer Studies, University of Maryland, MD, USA, 1996.

Luhning, R. W., Das, S. K., Bakker, J., Engleman, J. R., Wong, S., Fisher, L. J.,

Grabowski, J., Sullivan, L. A. and Boyle, H. A.: Full scale VAPEX process—

Climate change advantage and economic consequences, JCPT, Volume 42, No. 2,

pp. 29-34, 2003.

Luo, P.: Asphaltene precipitation and its effects on a solvent-based heavy oil recovery


Luo, P. and Gu, Y.: Effects of asphaltene content and solvent concentration on heavy oil

viscosity, SPE PS2005-354, Calgary, Alberta, Canada, 1-3 November 2005.

Luo, P., Yang, C., Tharanivasan, A. K. and Gu, Y.: In situ upgrading of heavy oil in a

solvent- based heavy oil recovery process, SPE 2005-098, Calgary, Alberta,

Canada, June 7 – 9, 2005.

280

Maerker, J. M. and Gale, W. W.: Surfactant flood process design for Loudon, SPE 20218,

Tulsa, Oklahoma, April 22-25, 1990.

Mai, A. and Kantzas, A.: Heavy oil waterflooding effects of flow rate and oil viscosity,

SPE 2007-144, Calgary, Alberta, Canada, June 12-14, 2007.

Manik, J., Ertekin, T., and Kohler, T. E.: Development and validation of a compositional

coal bed simulator, JCPT, Volume 41, No. 4, pp. 39-45, 2002.

Martin, F. D. and Taber, J. J.: Carbon dioxide flooding, JPT, Volume 44, No. 4, pp. 396-

400, 1992.

Maudgalya, S., Knapp, R. M., and McInerney, M. J.: Microbial enhanced-oil-recovery

technologies: A review of the past, present, and future, SPE 106978, April 2007.

Mayer, E. H., Berg, R. L., Carmichael, J. D. and Weinbrandt, R. M.: Alkaline injection

for enhanced oil recovery: A status report, JPT, Volume 35, No. 1, pp. 209-221,

1983.

Mendoza, H. A., Finol, J. J. and Butler, R. M.: SAGD, pilot test in Venezuela, SPE

53687, Caracas, Venezuela, April 21-23, 1999.

Mendoza, H. and Herrera, L.: Controllability and observability issues for SAGD

intelligent control in Venezuela, SPE 69699, Porlamar, Margarita Island,

Venezuela, March 12-14, 2001.

Miller, K. A.: State of the art of Western Canadian heavy oil water flood technology, SPE

2005-251, Calgary, Alberta, Canada, 7-9 June 2005.

281

Miller, K. A.: Use and misuse of heavy oil and bitumen viscosity data, SPE 95-93, Banff,

Alberta, Canada, 14-17 May 1995.

Miller, K. A., Nelson, L. A. and Almond, R. M.: Should you trust your heavy oil

viscosity measurement? JCPT, Volume 45, No. 4, pp. 42-48, 2006.

Moghadam, S.: Theoretical and physical modeling of vapour extraction heavy oil

recovery, University of Regina, Regina, SK, Canada, 2007.

Mohammadpoor, M., Qazvin Firouz, A. and Torabi, F.: Implementing simulation and

artificial intelligence tools to optimize the performance of the CO2sequestration in

coalbed methane reservoirs, SPE 151307-MS, Orlando, Florida, USA, February7-

9 2012.

Mohammadpoor, M., Shahbazi, K., and Torabi, F.: A new methodology for prediction of

bottomhole flowing pressure in vertical multiphase flow in Iranian oil fields using

artificial neural networks (ANNs), SPE 139147-MS, Lima, Peru, December 1-3,

2010.

Mokrys, I. J. and Butler, R. M.: In-situ upgrading of heavy oils and bitumen by propane

deasphalting: The VAPEX process, SPE 25452, Oklahoma City, Oklahoma, USA,

March 1993.

Moritis, G.: EOR dips in U.S. but remains a significant factor, Oil & Gas Journal,

Volume 26, September 1994.

Muhamad, H.: Optimal control of vapour extraction of heavy oil, Ryerson University,

Toronto, ON, Canada, 2012.

282

Nakutnyy, P. and Renouf, G.: Cold flow production database, PTRC Report, December

2009.

National Energy Board: Canada’s Energy Future 2013, Energy supply and demand

projections to 2035, Available at: http://www.neb-one.gc.ca/clf-

nsi/rnrgynfmtn/nrgyrprt/nrgyftr/2013/nrgftr2013-eng.pdf

Nghiem, L. X., Kohse, B. F., Farouq Ali, S. M. and Doan, Q.: Asphaltene precipitation:

Phase behavior modeling and compositional simulation, SPE 59432, Yokohoma,

Japan, April 2000.

Nghiem L. X., Sammon, P. H. and Kohse B. F.: Modelling asphaltene precipitation and

dispersive mixing in the VAPEX process, SPE 66361, Houston, Texas, February

2001.

Oballa, V. and Butler, R. M.: An experimental study of diffusion in the bitumen-toluene

system, JCPT, Volume 28, No.2, pp. 63-69, 1989.

Oduntan, A. R., Chatzis, I., Smith, J. and Lohi, A.: Heavy oil recovery using the VAPEX

process: Scale-up issues, SPE 2001-127, Calgary, Alberta, Canada, June 2001.

Oglesby, K. D., Blevins, T. R., Rogers, E. E. and Johnson, W. M.: Status of the 10-

pattern steamflood, Kern River Field, California, JPT, Volume 34, No. 10, pp.

2251-2257, 1982.

Ovalles, C., Fonseca A., Lara, A., Alvarado, V., Urrecheaga, K., Ranson A. and

Mendoza, H.: Opportunities of downhole dielectric heating in Venezuela: Three

283

case studies involving medium, heavy and extra-heavy crude oil reservoirs, SPE

78980, Calgary, Alberta, November 4-7, 2002.

Ovalles, C., Vallejos, C., Vasquez, T., Martinis, J., Perze-Perez, A., Cotte, E.,

Castellanos, L. and Rodriguez H.: Extra-heavy crude oil downhole upgrading

process using hydrogen donors under steam injection conditions, SPE 6992,

Margarita Island, Venezuela, March 12-14, 2001.

Perkins, T. K. and Johnston, O. C.: A review of diffusion and dispersion in porous media,

SPE Journal, Volume 3, No. 1, pp. 70-80, March 1963.

Pourabdollah, K. and Mokhtari, B.: The VAPEX process, from beginning up to date,

Fuel, Volume107, pp. 1-33, May 2013.

Pourabdollah, K., Zarringhalam Moghaddam, A., Kharrat, R. and Mokhtari, B.: Study of

asphaltene and metal upgrading in the vapour extraction (VAPEX) process,

Energy and Fuels, Volume24, pp. 4396-4401, 2010.

Pourabdollah, K., Zarringhalam Moghaddam, A., Kharrat, R. and Mokhtari, B.:

Improvement of heavy oil recovery in the VAPEX Process using Montmorillonite

Nanoclays, Oil& Gas Science and Technology, Volume 66, No. 6, pp. 1005-

1016, 2011.

Prats, M.: Thermal recovery, monograph series, SPE 7, 161, 14, Dallas, 1982.

Rahnema H., Kharrat R. and Rostami B.: Experimental and numerical study of vapour

extraction process (VAPEX) in heavy oil fractured reservoir, SPE 2008-116,

Calgary, Alberta, Canada, 17- 19 June 2008.

284

Renner, T. A.: Measurement and correlation of diffusion coefficients for CO2 and rich-

gas applications, SPE Reservoir Engineering Journal, Volume 3, No. 2, pp. 517-

523, 1988.

Restine, J. L.: Effect of preheating on Kern River field steam drive, JPT, Volume 35, No.

3, pp. 523-529, 1983.

Restine, J. L., Graves, W. G. and Elias, R. Jr.: Infill drilling in a steamflood operation:

Kern River Field, SPE Reservoir Engineering Journal, Volume 2, No. 2, pp. 243-

248, 1987.

Rezaei, N.: Experimental investigations in improving the VAPEX performance for

recovery of heavy oil and bitumen, University of Waterloo, Waterloo, ON,

Canada, 2010.

Riazi, M. R.: A new method for experimental measurement of diffusion coefficients in

reservoir fluids, Journal of Petroleum Science and Engineering, Volume 14, pp.

235-250, 1996.

Roopa, I. and Dawe, R. A.: VAPEX—Gravity-assisted film drainage for heavy oil

recovery, Journal of Petroleum Science and Technology, Volume 25, No. 5, pp.

645-658, 2007.

Rostami, B., Etminan, S. R., Soleimani, A. and Kharrat, R.: Effect of capillarity and

surface tension on the performance of VAPEX process, SPE 2007-039, Calgary,

Alberta, Canada, June 12 – 14, 2007.

285

Sahin, S., Kalfa, U. and Celebioglu, D.: Bati Raman field immiscible CO2 application-

status quo and future plans, SPE 106575-PA, Buenos Aires, Argentina, April 15-

18, 2008.

Salahshoor, K., Hajisalehi, M. H. And Haghighatsefat, M.: Nonlinear model

identification and adaptive control of CO2 sequestration process in saline aquifers

using artificial neural networks, Applied Soft Computing, Volume 12, No. 11, pp.

3379-3389, 2012.

Salama, D. and Kantzas, A.: Experimental observations of miscible displacement of

heavy oils with hydrocarbon solvents, SPE PS2005-388, Calgary, Alberta,

Canada, 2005.

Schmidt, T., Leshchyshyn, T. H.; and Puttagunta, V. R.: Diffusivity of CO2 into reservoir

fluids, Paper 82 33-100, Petroleum Society of CIM, Calgary, Alberta, Canada,

June 6-9, 1982.

Sedaee Sola, B. and Rashidi, F.: Application of the SAGD to an Iranian carbonate heavy-

oil reservoir, SPE 100533, Anchorage, Alaska, May 8-10, 2006.

Shin, H. and Polikar, M.: Optimizing the SAGD Process in three Canadian oil-tar sands

area, SPE 95754, Dallas, TX, October9-12, 2005.

Shutang, G., Huabin, L. and Hongfu, L.: Laboratory investigation of combination of

alkali/surfactant/polymer technology for Daqing EOR, SPE Reservoir

Engineering Journal, Volume 10, No. 3, pp. 194-197, 1995.

286

Smith, G. E.: Waterflooding heavy oils, SPE 24367, Casper, Wyoming, USA, May 18-

21, 1992.

Speight, J.: Handbook of petroleum product analysis, Wiley and Sons, 2001

Speight, J. G.: Petroleum asphaltenes Part 1: Asphaltenes, resins and the structure of

petroleum, Oil & Gas Science and Technology, Volume 59, No. 5, pp. 467-477,

2004.

Speight, J. G.: The chemistry and technology of petroleum, 4th edition, CRC press, page

217, 2007.

Stephen, C. M., Domier, D. B. and MacDermont, R. N.: Waterflood optimization of the

Buffalo Coulee Bakken heavy oil pool of South western Saskatchewan, SPE

30285, Calgary, Alberta, Canada, June 19-21, 1995.

Taber, J. J.: Environmental improvements and better economics in EOR operations, In

Situ Journal, Volume 14, No. 4, December 1990.

Taber, J. J.: The use of flue gas for the enhanced recovery of oil, Presented at the

Symposium “EOR by Gas Injection”, Copenhagen, Denmark, September 14,

1988.

Talbi, K.: Evaluation of CO2 based VAPEX process for recovery of heavy oil and

bitumen, University of Calgary, Calgary, AB, Canada, 2005.

Talbi, K. and Maini, B. B.: Evaluation of CO2based VAPEX process for the recovery of

bitumen from tar sand reservoirs, SPE 84868, Kuala Lumpur, Malaysia, October

2003.

287

Tam, S. P. Y.:VAPEX experiments in an annular packing of glass beads and the

numerical simulation of VAPEX using COMSOL, University of Waterloo,

Waterloo, ON, Canada, 2007.

Torabi, F., Yadali Jamaloei, B., Stengler, M. B. and Jackson, D. E.: The evaluation of

CO2-based vapour extraction (VAPEX) process for heavy-oil recovery, J. Petrol.

Explor. Prod. Technol., Volume 2, pp. 93–105, 2012.

Tremblay, B. and Forshner, K.: Modeling of sand production from wells on primary

recovery, JCPT, Volume 37, No. 3, pp. 41-50, 1998.

Tremblay, B., Sedgwick, G. and Forshner, K.: Simulation of cold production in heavy-oil

reservoirs: wormhole dynamics, SPE Reservoir Engineering, Volume 12, No. 2,

pp. 110-117, 1997.

Tremblay, B., Sedgwick, G. and Vu, D.: CT imaging of wormhole growth under solution-

gas drive, SPE 54658-PA, Tulsa, Oklahoma, USA, April 19-22, 1998.

Tuhinuzzaman, M.: The role of capillarity in the VAPEX process, University of Regina,

Regina, SK, Canada, 2006.

Turta, A., Ayasse, C., Najman, J., Fisher, D. and Singhal, A.: Laboratory investigation of

gravity-stable waterflooding using toe-to-heel displacement: Part I: Hele Shaw

model results, SPE 78988, Calgary, Alberta, Canada, November 4 – 7, 2002.

Turta, A., Goldman, J., Singhal, A. and Zhoa, L.: Toe- to-heel waterflooding Part II: 3D

laboratory test results, SPE 84077, Denver, Colorado, USA, October 5 – 8, 2003.

288

Upreti, S. R., Lohi, A., Kapadia, R. A. and El-Haj, R.: Vapour extraction of heavy oil and

bitumen: A review, Energy and Fuels, Volume 21, pp. 1562-1574, 2007.

Upreti, S. R. and Mehrotra, K. A.: Diffusivity of CO2, CH4, C2H6, and N2, in Athabasca

bitumen, Can. J. Chem. Eng., Volume 80, pp. 116-125, 2002.

Wen Y. W. and Kantzas, A.: Monitoring bitumen-solvent interactions with low-field

nuclear magnetic resonance and X-ray computer-assisted tomography, Energy

and Fuels, Volume 19, pp. 1319–1326, 2005.

White, D. and Moss, J. T.: Thermal recovery methods, Pennwell Publishing, Tulsa,

Oklahoma, 1983.

Wilke, C. R. and Chang, P.: Correlation of diffusion coefficients in dilute solutions,

AIChE Journal, Volume 1, pp. 264, 1955.

Williams, L. L., Fong, W. S. and Kumar, M.: Effects of discontinuous shales on

multizone steamflood performance in the Kern River field, SPE-73174-PA, New

Orleans, September 27–30, 2001.

Wong, F. Y., Anderson, D. B., O’Rourke, J. C., Rea, H. Q. and Scheidt, K. A.: Meeting

the challenge to extend success at the Pikes Peak steam project to areas with

bottom water, SPE J. of Reservoir Evaluation and Engineering, Volume 6, No. 3,

pp. 156-167, 2003.

Xu, S.: Automatic history match and upscaling study of VAPEX process and its

uncertainty analysis, University of Regina, Regina, SK, Canada, 2012.

289

Xu, S., Zeng, F., Gu, Y. and Knorr, K. D.: Upscaling study of vapour extraction process

through numerical simulation, SPE 157784, Calgary, Alberta, Canada, June 12-

14, 2012.

Xu, W.: The role of interfacial tension in the VAPEX process, University of Regina,

Regina, SK, Canada, 2006.

Yang, C. and Gu, Y.: Effects of heavy oil/solvent interfacial tension on gravity drainage

in the vapour extraction (VAPEX) process, SPE PS2005-406, Calgary, Alberta,

Canada, 2005.

Yannimaras, D. V. and Tiffin, D. L.: Screening of Oils for in-situ combustion at reservoir

conditions by accelerating-rate calorimetry, SPE Reservoir Engineering, Volume

10, No. 1, pp. 36-39, 1995.

Yazdani, A.: Physical and numerical modeling of permeability and drainage height

effects in VAPEX, University of Calgary, Calgary, 2007.

Yazdani, A. and Maini, B. B.: Effect of drainage height and grain size on production rates

in the VAPEX process: Experimental Study, SPE Reservoir Engineering, Volume

8, No. 3, pp. 205-213, 2005.

Yildirim, Y.: Application of VAPEX (vapour extraction) process on carbonate reservoirs,

The Middle East Technical University, 2003.

Zhoa, L. and Turta, A.: Toe-to-heel waterflooding: Field scale numerical simulation, SPE

89380, Tulsa, Oklahoma, USA, April 17 – 21, 2004.

290

Zhang, H.: Physical modeling of heavy oil recovery in a vapour extraction (VAPEX)


ZoBell, C. E.: bacteriological process for treatment of fluid-bearing earth formations, U.

S. patent, No. 2413278, 1946.

ZoBell, C. E.: Bacterial release of oil from sedimentary materials, Oil and Gas Journal,

Volume 46, No.13, 1947.

291

Appendix A

Table A-1: Production method versus heavy oil resource (1) (after Clark, 2007)

Production

method or

resource

Open-pit mining

Cold-production

horizontal wells

& multilaterals

Waterflood

Cold production

with sand

(CHOPS)

Status Commercial Commercial Commercial Commercial

Shallowest (<50

m) Only solution No No No

Shallow (50 to 100

m)

Possible but

economically

limited

No No No

Medium depth

(100 to 300 m) No

Unlikely unless

very low viscosity

or high solution

gas along with

high permeability

Unlikely unless

very low viscosity

and high

permeability

Unlikely, may

require solution

gas, but may be

possible

Intermediate depth

(300 to 1,000 m) No

Requires low

viscosity with

solution gas or

high formation

temperature and

high permeability

Requires low

viscosity and/or

high formation

temperature

Requires

unconsolidated

formation and

generally requires

solution gas

Deep (>1,000 m) No

Requires low

viscosity with

solution gas or

high formation

temperature and

high permeability

Requires low

viscosity and/or

high formation

temperature

Unlikely because

requires

unconsolidated

formations

Arctic No Maybe Maybe Disposal of sand

and water an issue

Offshore No Maybe Yes, North Sea Disposal of sand

and water an issue

Carbonate No No No No

Thin beds (<10 m

thick)

Can be mined if

near surface and

thin overburden

Maybe Maybe Yes

Highly laminated

Can be mined if

near surface and

thin overburden

Yes, if

multilaterals can

penetrate multiple

layers

Maybe for vertical

wells Yes

292

Table A-2: Production method versus heavy oil resource (2)(after Clark, 2007)

Production

method or

resource

Cyclic Steam

Stimulation Steamflood SAGD

Solvent without

heat or steam

Status Commercial Commercial Commercial Pilot test

Shallowest (<50

m) No No No No

Shallow (50 to 100

m) No No No

Possible, but

unproven

Medium depth

(100 to 300 m)

No, unless good

sealing Caprock

No, unless good

sealing Caprock

Yes, if good

vertical and

horizontal

permeability and

pay zone> 10m

Unproven, needs

good vertical and

horizontal

permeability

Intermediate depth

(300 to 1,000 m)

Yes, but deep

zones need higher

temperature steam

&are less economic

Yes, but deep

zones need higher

temperature steam

& are less

economic

Yes, but deep

zones need higher

temperature steam

& are less

economic

Unproven, needs

good vertical and

horizontal

permeability

Deep (>1,000 m)

No, needs high

temperature and

high-pressure

steam and too

much heat losses to

overburden through

injection wellbore

No, needs high

temperature and

high-pressure

steam and too

much heat losses

to overburden

through injection

wellbore

No, needs high

temperature and

high-pressure

steam and too

much heat losses

to overburden

through injection

wellbore

Possible, but

unproven

Arctic

Maybe if

permafrost can be

managed

Maybe if

permafrost can be

managed

Maybe if

permafrost can be

managed

Possible, but

unproven

Offshore

No, too much heat

loss in riser to

ocean water

No, too much heat

loss in riser to

ocean water

No, too much heat

loss in riser to

ocean water

Possible, but

unproven

Carbonate No No No Unknown

Thin beds (<10 m

thick)

Possible with

horizontal wells

No, needs at least

10 m bed, heat

losses to

overburden are too

great

No, needs at least

10 m bed great

Possible, but

unproven

Highly laminated Possible with

horizontal wells

May be possible

with horizontal

wells, but

unproven

No, need at least

10mbed Unlikely

293

Table A-3: Production method versus heavy oil resource (3) (after Clark, 2007)

Production

method or

resource

Solvent with heat

or steam

Fire flood with

vertical wells (~20

API oil only)

Fire flood with

vertical and

horizontal wells

Downhole steam

generation (CSS,

flood, SAGD)

Status Pilot test Commercial Pilot test Experimental

Shallowest (<50

m)

No No No No

Shallow (50 to 100

m)

Unknown No Unknown No

Medium depth

(100 to 300 m)

Unproven, needs

good vertical and

horizontal

permeability

Possible Unknown Tested but

commercially

unproven

Intermediate depth

(300 to 1,000 m)

Unproven, needs

good vertical and

horizontal

permeability

Yes Possible Possible, but

unproven

Deep (>1,000 m) Unknown Possible Possible, but

unproven

Unknown, greater

depth means need

high steam

pressure &

temperature

Arctic Unproven, must

manage permafrost

issue

Possible, but

unproven

Possible, but

unproven

Possible, but

unproven

Offshore Unlikely Possible, but

unproven

Possible, but

unproven

Possible, but

unproven

Carbonate Unknown Unknown Unknown Possible, but

unproven

Thin beds (<10 m

thick)

Possible, but

unproven

Unknown Unlikely Possible, but

unproven

Highly laminated Unknown Unknown Unlikely Possible, but

unproven

294

Table A-4: Production method versus heavy oil resource (4)(after Clark, 2007)

Production method or

resource

Electric, induction or

RF heating

Supercritical fluids

(e.g. CO2)

Biotechnology

Status Pilot test Experimental Research

Shallowest (<50 m) No No, needs higher

reservoir pressure

Unknown

Shallow (50 to 100 m) Possible, limited field

successes in isolated

cases

No, needs higher

reservoir pressure

Unknown

Medium depth (100 to

300 m)

Possible, limited field

successes in isolated

cases

No, needs higher

reservoir pressure

Unknown

Intermediate depth (300

to 1,000 m)

Possible, but unproven Unknown Unknown

Deep (>1,000 m) Possible, but unproven Unknown Unknown

Arctic Possible, but unproven Unknown Unknown

Offshore Possible, but unproven Unknown Unknown

Carbonate Possible, but unproven Unknown Unknown

Thin beds (<10 m thick) Possible, but unproven Unknown Unknown

Highly laminated Possible, but unproven Unknown Unknown

295

Table A-5: Technology versus production method (1)(after Clark, 2007)

Technology

or production

method

Simulations

and modeling

Geomechanics Downhole

sampling

In situ

viscosity

Fluid

Characterization

Cold-

production

horizontal &

multilaterals

High High High High High

Waterflood High Medium High High High

Cold

production

with sand

(CHOPS)

Medium High High High High

Cyclic steam

stimulation

(CSS)


Steamflood

with surface

burners


SAGD High High High High High

Solvent

without heat

or steam


Solvent with

heat or steam

High Medium High High High

Fire flood

with vertical

wells (~20

API oil only)


Fire flood

with vertical

and

horizontal

wells


Downhole

steam

generation

(CSS,

steamflood,

SAGD)


Electric,

induction, or

RF heating

downhole


Supercritical

fluids


Biological Unknown Unknown High High High

296


Technology or

production

method

Flow

assurance Drilling

Well

placement Multilaterals Cementing

Cold-

production

horizontal &

multilaterals

High High High High Low

Waterflood High Medium Medium Low Low

Cold

production

with sand

(CHOPS)

High Low Low Low Low

Cyclic steam

stimulation

(CSS)

High Medium Low Medium High

Steamflood

with surface

burners

High Medium Medium Low High

SAGD High High High Low High

Solvent

without heat

or steam

High High High Low Low

Solvent with

heat or steam High High High Low Medium

Fire flood with

vertical wells

(~20 API oil

only)

High Medium Low Low High

Fire flood with

vertical and

horizontal

wells

High High High Low High

Downhole

steam

generation

(CSS,

steamflood,

SAGD)

High Medium Low to High Medium High

Electric,

induction, or

RF heating

downhole

High High High Low to

medium High

Supercritical

fluids High Medium Unknown Unknown High

Biological High Medium Unknown Unknown Low

297


Technology or

production

method

High

temperature

completions

High

temperature,

long life

pumps

Pumps with

high sand and

solids

capability

Sand control Monitoring

and Control

Cold-

production

horizontal &

multilaterals

Low Low High High High

Waterflood Low Low Medium High High

Cold

production

with sand

(CHOPS)

Low Low High Low Medium

Cyclic steam

stimulation

(CSS)

High High Medium High High

Steamflood

with surface

burners


SAGD High High Medium High High

Solvent

without heat

or steam

Medium Low Medium High High

Solvent with

heat or steam High Medium Medium High High

Fire flood with

vertical wells

(~20 API oil

only)


Fire flood with

vertical and

horizontal

wells

High Medium Medium High High

Downhole

steam

generation

(CSS,

steamflood,

SAGD)


Electric,

induction, or

RF heating

downhole


Supercritical

fluids High High Medium High High

Biological Low Low Medium High High

298


Technology or

production

method

Devices for

downhole flow

control

Distributed

temperature

Downhole

pressure

High

temperature

electronics &

sensors

(>200°C)

Downhole

multiphase

flow sensors

Cold-

production

horizontal &

multilaterals

High Low Medium Low High

Waterflood High Low High Low High

Cold

production

with sand

(CHOPS)

Low Low Low Low Low

Cyclic steam

stimulation

(CSS)

High High High High Medium

Steamflood

with surface

burners


SAGD High High High High High

Solvent

without heat

or steam

High High Medium Low Low

Solvent with

heat or steam High High High

Medium to

High Low

Fire flood

with vertical

wells (~20 API

oil only)

Medium High High High Low

Fire flood

with vertical

and horizontal

wells

Low High High High Low

Downhole

steam

generation

(CSS,

steamflood,

SAGD)


Electric,

induction, or

RF heating

downhole


Supercritical

fluids Unknown High High Medium High

Biological Unknown Low Low Low Unknown

299


Technology or

production

method

Microseismic

while

fracturing

Cross-well

EM for fluid

saturation

Cross-well

seismic for

gas saturation

Through-

casing fluid

monitoring

Composition

monitoring for

in situ

upgrading

Cold-

production

horizontal &

multilaterals

Low Low Low Low Low

Waterflood Low High Low Medium Low

Cold

production

with sand

(CHOPS)

Low High Medium High Low

Cyclic steam

stimulation

(CSS)

Medium Medium High Low Low

Steamflood

with surface

burners

High High High High Low

SAGD Medium High Medium High Low

Solvent

without heat

or steam

Low Low Medium Low Medium

Solvent with

heat or steam Medium Low to High High Medium Medium

Fire flood

with vertical

wells (~20 API

oil only)

Low Low to High High Medium High

Fire flood

with vertical

and horizontal

wells

Low Unknown High Medium High

Downhole

steam

generation

(CSS,

steamflood,

SAGD)

Medium Low to High High Low to

Medium Low

Electric,

induction, or

RF heating

downhole

Low Unknown Low High High

Supercritical

fluids Medium Unknown Unknown High High

Biological Unknown Low Low High High

300


Technology or

production

method

Surface

multiphase flow

sensors

4D surface

seismic

Fluids separation

and disposal

Produced- solids

separation

Cold-production

horizontal &

multilaterals

High Medium High Medium

Waterflood High High High Medium

Cold production

with sand

(CHOPS)

High Medium High High

Cyclic steam

stimulation (CSS) High High High Medium

Steamflood with

surface burners High High High Medium

SAGD High Medium High Medium

Solvent without

heat or steam High Medium High Medium

Solvent with heat

or steam High Medium High Medium

Fire flood with

vertical wells (~20

API oil only)

High High High High

Fire flood with

vertical and

horizontal wells

High High High High

Downhole steam

generation (CSS,

steamflood,

SAGD)


Electric,

induction, or RF

heating downhole


Supercritical

fluids High High High Medium

Biological High Low Unknown Medium

301

Appendix B

Table B-1: The experimental data on VAPEX experiments conducted by different researchers

No. Researcher

Solvent/Inj. Pressure

(kPa)

Φ (%) k (D) μ(cp) H (cm) Q (mL/h)

1 Das, 1994 C4/343 35.00 830.00 130000 21.9 19.20

2 Das, 1994 C4/ 343 36.00 217.00 130000 21.9 9.50

3 Das, 1994 C4/ 343 37.00 43.50 130000 21.9 3.60

4 Das, 1994 C4/ 343 35.00 43.50 130000 21.9 4.60

5 Das, 1994 C4/ 343 36.00 27.00 130000 21.9 2.50

6 Das, 1994 C4/ 343 37.00 830.00 10000 21.9 42.00

7 Das, 1994 C4/ 343 35.00 830.00 10000 21.9 39.70

8 Das, 1994 C4/ 343 36.00 830.00 10000 21.9 37.60

9 Das, 1994 C4/ 343 37.00 217.00 10000 21.9 24.00

10 Das, 1994 C4/ 343 35.00 217.00 10000 21.9 25.10

11 Das, 1994 C4/ 343 36.00 43.50 10000 21.9 14.40

12 Das, 1994 C4/ 343 36.00 43.50 10000 21.9 15.60

13 Das, 1994 C4/ 343 37.00 43.50 10000 21.9 16.70

14 Das, 1994 C4/ 343 37.00 27.00 10000 21.9 5.80

15 Butler, 1996 C4/ 240 35.00 220.00 7400 22.9 21.50

16 Jiang, 1997 C4/ 212 35.00 217.00 7000 22.9 21.80

17 Jiang, 1997 C4/ 212 36.00 217.00 7000 22.9 36.20

18 Jiang, 1997 C4/ 212 35.00 217.00 7000 22.9 40.70

19 Jiang, 1997 C4/ 212 36.00 217.00 7000 22.9 39.30

20 Jiang, 1997 C4/ 212 35.00 217.00 7000 22.9 21.90

21 Jiang, 1997 C4/ 212 36.00 217.00 7000 22.9 50.20

22 Jiang, 1997 C4/ 212 35.00 217.00 7000 22.9 31.20

23 Jiang, 1997 (C4/N2)/ 239 35.00 220.00 7400 22.9 11.00

24 Jiang, 1997 (C4/N2)/ 239 35.00 220.00 7400 22.9 14.57

25 Jiang, 1997 (C4/N2)/ 239 35.00 43.00 7400 22.9 8.50

26 Jiang, 1997 C4/ 308 35.00 43.00 7400 22.9 19.40

302

Table B-2: The experimental data on VAPEX experiments conducted by different researchers (Cont'd)

No. Researcher

Solvent/Inj.Pressure

(kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

27 Jiang, 1997 C4/ 308 35.00 217.00 7400 22.9 21.50

28 Jiang, 1997 C4/ 308 35.00 43.00 7400 22.9 21.70

29 Jiang, 1997 (C4/C1)/ 308 35.00 217.00 7400 22.9 18.70

30 Jiang, 1997 C4/ 308 35.00 43.00 7400 22.9 15.50

31 James, 2004 C4/114 30.00 74.00 85000 32.5 6.60

32 James, 2004 C4/ 114 30.00 68.00 85000 40.1 12.00

33 James, 2004 C4/114 30.00 66.00 85000 54.5 14.40

34 James, 2004 C4/114 30.00 76.00 85000 60.2 18.60

35 James, 2004 C4/114 38.00 285.00 85000 92.0 12.00

36 James, 2004 C4/114 38.00 350.00 85000 23.7 6.60

37 Talbi, 2003 (CO2/C3)/ 250 35.00 640.00 3300 30.5 73.15

38 Talbi, 2003 (C3/C1)/ 250 35.00 640.00 3300 30.5 76.45

39 Talbi, 2003 (C3/C1)/ 600 35.00 640.00 3300 30.5 71.43

40 Talbi, 2003 (CO2/C3)/ 600 35.00 640.00 3300 30.5 115.02

41 Yazdani, 2007 C4/ 240 34.10 220.00 18000 7.5 4.00

42 Yazdani, 2007 C4/ 240 36.80 330.00 18000 7.5 5.00

43 Yazdani, 2007 C4/ 240 36.50 640.00 18000 7.5 7.00

44 Yazdani, 2007 C4/ 240 34.10 220.00 18000 15.0 11.00

45 Yazdani, 2007 C4/ 240 36.80 330.00 18000 15.0 14.00

46 Yazdani, 2007 C4/ 240 36.50 640.00 18000 15.0 20.00

47 Yazdani, 2007 C4/ 240 34.10 220.00 18000 30.0 20.00

48 Yazdani, 2007 C4/ 240 36.80 330.00 18000 30.0 25.00

49 Yazdani, 2007 C4/ 240 36.50 640.00 18000 30.0 38.00

50 Yazdani, 2007 C4/ 240 34.10 220.00 18000 30.0 34.00

51 Yazdani, 2007 C4/ 240 36.80 330.00 18000 30.0 40.00

52 Yazdani, 2007 C4/ 240 36.50 640.00 18000 30.0 60.00

53 Yazdani, 2007 C4/ 240 34.10 220.00 18000 60.1 71.00

303


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

54 Yazdani, 2007 C4/ 240 36.80 330.00 18000 60.1 97.00

55 Yazdani, 2007 C4/ 240 36.50 640.00 18000 60.1 135.00

56 Yazdani, 2007 C4/ 240 34.10 220.00 18000 100.5 90.00

57 Yazdani, 2007 C4/ 240 36.80 330.00 18000 100.5 120.00

58 Yazdani, 2007 C4/ 240 36.50 640.00 18000 100.5 160.00

59 Yazdani, 2007 C4/ 240 34.10 220.00 18656 60.1 67.80

60 Yazdani, 2007 C4/ 240 36.80 330.00 18656 60.1 86.10

61 Yazdani, 2007 C4/ 240 36.50 640.00 18656 60.1 123.30

62 Tuhinuzzaman, 2006 C4/ 240 40.00 13.00 5800 35.6 6.00

63 Tuhinuzzaman, 2006 C4/ 240 40.00 13.00 14400 35.6 4.00

64 Xu, 2006 C4/ 204 40.00 13.00 150000 30.48 20.00

65 Xu, 2006 C4/ 188 40.00 13.00 150000 30.48 4.00

66 Zhang, 2006 C4/ 240 36.80 441.30 38347 10.0 6.15

67 Zhang, 2006 C4/ 240 37.50 132.00 38347 10.0 3.76

68 Etminan, 2007 C4/ 240 35.18 10.00 18600 15.2 7.00

69 Etminan, 2007 C4/ 240 33.29 10.00 18600 15.2 6.00

70 Tam, 2007 C4/ 103 38.00 1123.00 23200 100.0 69.00

71 Tam, 2007 C4/ 103 38.00 1123.00 23200 100.0 75.00

72 Tam, 2007 C4/97 39.00 300.00 23200 100.0 39.48

73 Tam, 2007 C4/ 93 39.00 300.00 23200 100.0 33.60

74 Tam, 2007 C4/ 110 38.00 1123.00 23200 100.0 129.00

75 Zhang, 2007 C3/ 800 36.20 438.00 38347 10.0 1.29

76 Zhang, 2007 C3/ 800 36.80 158.00 38347 10.0 0.88

77 Zhang, 2007 C3/ 800 37.40 418.00 38347 10.0 2.91

78 Zhang, 2007 C3/ 800 36.50 417.00 38347 10.0 4.93

79 Zhang, 2007 C3/ 800 36.50 122.00 38347 10.0 2.95

80 Zhang, 2007 C3/ 800 35.80 424.00 38347 10.0 3.63

304


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

81 Zhang, 2007 C3/ 800 35.30 410.00 38347 10.0 2.21

82 Zhang, 2007 C3/ 800 36.70 143.00 38347 10.0 9.14

83 Zhang, 2007 C3/ 800 36.10 118.00 38347 10.0 1.99

84 Azin, 2008 (C3/C1)/ 1069 39.50 830.00 58770 30.0 19.90

85 Azin, 2008 (C3/C1)/ 1655 39.50 830.00 58770 30.0 29.08

86 Azin, 2008 (C3/C1)/ 1069 39.50 830.00 58770 30.0 27.35

87 Azin, 2008 (C3/C1)/ 689 39.50 830.00 58770 30.0 43.37

88 Haghighat, 2008 C3/ 814 36.42 3.00 2050 30.0 1.23

89 Haghighat, 2008 C3/ 850 34.24 3.00 2050 30.0 1.09

90 Haghighat, 2008 (C3/Toluene)/ 850 34.71 3.00 2050 30.0 2.12

91 Haghighat, 2008 (C3/Toluene)/ 850 36.44 3.00 2050 30.0 1.06

92 Haghighat, 2008 C3/ 750 37.70 3.00 2050 30.0 1.07

93 Haghighat, 2008 C4/ 240 39.93 3.00 2050 30.0 1.40

94 Moghadam, 2008 C3/ 800 32.50 310.00 11900 10.0 14.55

95 Moghadam, 2008 C3/ 800 32.90 103.00 11900 10.0 3.70

96 Moghadam, 2008 C3/ 800 33.10 96.00 11900 10.0 1.36

97 Moghadam, 2008 C3/ 800 35.40 49.00 11900 10.0 3.28

98 Moghadam, 2008 C3/ 800 35.70 25.00 11900 10.0 1.88

99 Moghadam, 2008 C3/ 800 36.30 16.00 11900 10.0 1.30

100 Talbi, 2008 (CO2/C3)/ 1814 35.40 640.00 4500 30.5 70.35

101 Talbi, 2008 (C3/C1)/ 1814 35.20 640.00 4500 30.5 77.05

102 Talbi, 2008 (CO2/C3)/ 1814 35.00 640.00 4500 30.5 73.10

103 Talbi, 2008 (C3/C1)/ 4227 35.30 640.00 4500 30.5 70.92

104 Talbi, 2008 (CO2/C3)/ 4227 35.10 640.00 4500 30.5 97.00

105 Talbi, 2008 (CO2/C3)/ 4227 35.10 640.00 18600 30.5 52.92

106 Talbi, 2008 (CO2/C3)/ 2848 35.15 640.00 18600 30.5 43.10

107 Talbi, 2008 (CO2/C3)/ 1469 35.20 640.00 18600 30.5 34.28

108 Talbi, 2008 (C3/C1)/ 4227 35.20 640.00 18600 30.5 36.33

305


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

109 Talbi, 2008 (C3/C1)/ 2848 35.00 640.00 18600 30.5 45.28

110 Talbi, 2008 (C3/C1)/ 1469 35.10 640.00 18600 30.5 44.00

111 Talbi, 2008 CO2/ 4227 35.30 640.00 18600 30.5 23.31

112 Derakhshanfar, 2009 C3/ 860 35.70 40.00 21000 15.3 7.00

113 Derakhshanfar, 2009 (C3/C1) 1480 36.60 40.00 21000 15.3 5.50

114 Derakhshanfar, 2009 (C3/C1)/ 2859 37.00 40.00 21000 15.3 3.50

115 Derakhshanfar, 2009 (C3/CO2)/ 2859 36.90 40.00 21000 15.3 4.50

116 Luo, 2009 C4/ 240 36.80 441.00 24137 10.0 4.06

117 Luo, 2009 C4/ 240 37.50 132.00 24137 10.0 1.85

118 Luo, 2009 C3/ 800 36.50 122.00 24137 10.0 2.95

119 Luo2009 C3/ 918 36.70 143.00 24137 10.0 9.14

120 Luo, 2009 C3/ 800 32.50 310.00 12900 10.0 17.82

121 Luo, 2009 C3/ 800 32.90 103.00 12900 10.0 4.29

122 Luo, 2009 C3/ 800 35.40 49.00 12900 10.0 3.48

123 Luo, 2009 C3/ 800 36.30 16.00 12900 10.0 1.18

124 Alkindi, 2010 Ethanol/ 198 39.00 43.00 1390 30.0 0.48

125 Alkindi, 2010 Ethanol/ 198 39.00 43.00 1390 30.0 0.75

126 Alkindi, 2010 Ethanol/ 198 39.00 43.00 1390 30.0 1.03

127 Alkindi, 2010 Ethanol/ 198 39.00 43.00 1390 15.0 0.41

128 Alkindi, 2010 Ethanol/ 198 39.00 43.00 1390 15.0 0.52

129 Abukhalifeh, 2010 C3/ 791 38.00 204.00 225000 25.0 29.46

130 Abukhalifeh, 2010 C3/ 791 38.00 204.00 225000 35.0 34.38

131 Abukhalifeh, 2010 C3/ 791 38.00 204.00 225000 45.0 38.46

132 Abukhalifeh, 2010 C3/ 791 37.80 102.00 225000 25.0 23.52

133 Abukhalifeh, 2010 C3/ 791 37.80 102.00 225000 35.0 26.52

134 Abukhalifeh, 2010 C3/ 791 37.80 102.00 225000 45.0 31.32

135 Abukhalifeh, 2010 C3/ 791 37.60 51.00 225000 25.0 15.60

306


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

136 Abukhalifeh, 2010 C3/ 791 37.60 51.00 225000 35.0 17.34

137 Abukhalifeh, 2010 C3/ 791 37.60 51.00 225000 45.0 20.34

138 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 61.72

139 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 68.58

140 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 72.24

141 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 76.80

142 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 79.56

143 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 80.04

144 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 80.04

145 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 81.84

146 Rezaei, 2010 C5/69 37.10 780.80 40500 36.0 132.96

147 Rezaei, 2010 C5/69 30.50 148.80 40500 36.0 14.28

148 Rezaei, 2010 C5/69 20.50 19.10 40500 36.0 67.26

149 Rezaei, 2010 C5/69 28.00 119.10 40500 35.8 64.92

150 Rezaei, 2010 C5/69 29.40 147.10 40500 36.0 69.72

151 Rezaei, 2010 C5/69 30.10 132.40 40500 36.0 74.40

152 Rezaei, 2010 C5/69 32.10 143.70 40500 36.0 83.22

153 Rezaei, 2010 C5/69 37.10 830.00 40500 41.9 111.56

154 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 61.72

155 Rezaei, 2010 C5/69 37.10 830.00 5400 41.9 169.62

156 Rezaei, 2010 C5/69 37.10 220.00 5400 41.9 96.90

157 Rezaei, 2010 C5/69 37.10 830.00 40500 41.9 119.76

158 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 70.38

159 Rezaei, 2010 C5/69 37.10 830.00 5400 41.9 184.68

160 Rezaei, 2010 C5/69 37.10 220.00 5400 41.9 104.70

161 Rezaei, 2010 C5/69 37.10 830.00 40500 41.9 176.94

162 Rezaei, 2010 C5/69 37.10 830.00 40500 41.9 170.52

307


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

163 Rezaei, 2010 C5/69 37.10 830.00 5400 41.9 208.92

164 Rezaei, 2010 C5/69 37.10 830.00 5400 41.9 218.1

165 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 100.56

166 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 104.22

167 Rezaei, 2010 C5/69 37.10 220.00 5400 41.9 128.94

168 Rezaei, 2010 C5/69 37.10 220.00 5400 41.9 119.34

169 Rezaei, 2010 C5/69 37.10 830.00 40500 41.9 121.158

170 Rezaei, 2010 C5/69 37.10 220.00 40500 41.9 45.72

171 Rezaei, 2010 C5/69 37.10 830.00 5400 41.9 208.5

172 Rezaei, 2010 C5/69 37.10 220.00 5400 41.9 109.26

173 Ahmadloo, 2012 C4/ 240 37.10 6.46 10541 24.5 0.32

174 Ahmadloo, 2012 C4/ 240 37.10 6.46 10541 47.5 0.62

175 Ahmadloo, 2012 C4/ 240 30.30 5.19 10541 24.5 0.02

176 Ahmadloo, 2012 C4/ 240 30.30 5.19 10541 47.5 0.05

177 Ahmadloo, 2012 C4/ 240 35.60 5.62 10541 24.5 0.04

178 Ahmadloo, 2012 C4/ 240 35.60 5.62 10541 47.5 0.17

179 Ahmadloo, 2012 C4/ 240 37.10 6.46 10541 24.5 0.80

180 Ahmadloo, 2012 C4/ 240 37.10 6.46 10541 47.5 2.80

181 Ahmadloo, 2012 C4/ 240 30.30 5.19 10541 24.5 0.90

182 Ahmadloo, 2012 C4/ 240 30.30 5.19 10541 47.5 4.00

183 Ahmadloo, 2012 C4/ 240 35.60 5.62 10541 24.5 0.60

184 Derakhshanfar, 2012 C3/800 33.80 9.20 11900 10.0 2.07

185 Derakhshanfar, 2012 C3/800 35.20 8.30 11900 10.0 1.38

186 Derakhshanfar, 2012 C3/800 34.60 10.80 11900 10.0 1.61

187 Derakhshanfar, 2012 C3/800 34.30 10.10 11900 10.0 1.17



308


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

190 Muhamad, 2012 C3/ 689 38.50 439.20 14500 25.0 1.46

191 Muhamad, 2012 C3/ 689 38.00 220.00 14500 25.0 1.17

192 Muhamad, 2012 C3/ 689 37.80 97.40 14500 25.0 0.62

193 Muhamad, 2012 C3/ 689 37.60 44.40 14500 25.0 0.49

194 Muhamad, 2012 C4/ 192 38.00 204.00 14500 25.0 0.30

195 Muhamad, 2012 C4/ 200 38.00 204.00 14500 25.0 0.38

196 Muhamad, 2012 C4/ 208 38.00 204.00 14500 25.0 0.48

197 Muhamad, 2012 C4/ 214 38.00 204.00 14500 25.0 0.54

198 Badamchizadeh, 2013 (CO2/C3)/ 1974 35.70 640.00 15000 30.5 4.00



201 Badamchizadeh, 2013 C3/ 784 35.90 640.00 15000 30.5 15.00

202 Jia, 2013 C3/ 800 35.46 4.50 8411 10.0 4.15

203 Jia, 2013 C3/ 800 35.47 4.23 8411 10.0 5.51

204 Jia, 2013 C3/ 800 35.79 4.22 8411 10.0 3.28

205 Jia, 2013 C3/ 800 35.17 4.20 8411 10.0 3.36

206 Jia, 2013 C3/ 800 35.83 4.79 8411 10.0 3.46

207 Jia, 2013 C3/ 800 35.66 4.20 8411 10.0 1.88

208 Jia, 2013 C3/ 800 35.75 5.05 8411 10.0 1.94

209 Jia, 2013 C3/ 800 35.88 4.75 8411 10.0 2.81

210 Jia, 2013 C3/ 800 36.00 5.20 5875 10.0 2.75

211 Jia, 2013 C3/ 800 35.60 5.60 5875 10.0 11.99

212 This study C3/ 800 42.20 8.78 5650 24.5 13.20

213 This study C3/ 800 43.10 9.12 5650 45.5 30.00

214 This study (C3/CO2)/ 850 41.80 8.64 5650 24.5 9.00

215 This study (C3/CO2)/ 850 42.40 8.87 5650 45.5 19.80

216 This study (C3/C1)/850 42.00 8.50 5650 24.5 7.80

217 This study (C3/C1)/850 38.50 439.20 5650 45.5 15.00

309


No. Researcher

Solvent/Inj.

Pressure (kPa)

Φ (%) k (D) μ (cp) H (cm) Q (mL/h)

218 This study C4/ 140 42.10 8.69 5650 24.5 8.40

219 This study C4/ 140 42.30 9.08 5650 45.5 19.20

220 This study CO2/ 850 42.10 6.11 5650 24.5 0.72

221 This study CO2/ 850 42.60 6.70 5650 45.5 1.68

222 This study C1/ 850 40.70 5.12 5650 24.5 1.62

223 This study C1/ 850 41.80 5.88 5650 45.5 3.42

310

Appendix C

The matrix of weights from input parameters to the first hidden layer nodes (iw 1, 1):

[3.8342 2.6339 -1.7228 -1.2711 0.28176;

2.6878 1.68 -3.5471 0.51937 1.8157;

-2.9572 2.6228 2.6736 -1.0331 1.5031;

-2.8597 -0.33989 2.7232 -0.19416 -3.0003;

0.5961 3.7079 -3.1228 -0.91768 -1.2655;

2.7566 0.52986 -2.0731 -2.0107 3.7189;

-4.1331 1.348 0.084948 2.8174 0.45116;

-0.032271 0.094057 0.091646 -4.9785 -1.1014;

-0.7139 -2.6992 0.094485 0.84579 4.3104;

2.238 -1.0346 -3.2367 2.4626 2.0406;

0.81488 -0.78138 2.6493 -1.6253 4.658;

-1.6764 0.57618 -0.88634 4.6013 1.4561;

-3.6315 -2.1886 -1.3978 -1.8416 -2.178;

-3.8408 0.060839 3.4376 -2.0979 0.29411;

1.6419 2.3604 2.6563 -1.2682 1.3958;

4.0527 -2.9136 0.9637 -0.9541 1.49;

-0.60968 -1.1512 0.35623 -4.5634 -1.2164;

-4.7813 -0.1938 -2.4813 0.45669 1.6036;

3.4817 2.0315 -2.4783 3.2969 1.5177;

-0.91766 3.1251 -3.0631 -0.20696 -1.2271]

311

The matrix of weights from the first hidden layer nodes to the second hidden layer nodes

(iw 2, 1):

[2.6318 -2.0514 -0.80709 2.3836 0.093621 0.5588 1.2372 0.64943 -2.3911 -0.58394 -

0.4018 0.19941 0.10041 -0.079929 -0.7413 0.051628 2.1146 2.6368 1.2221 -0.67989;

-2.1714 -0.77674 1.3303 0.96678 2.0044 1.295 0.95583 -1.8506 0.37636 0.30933 2.0122

-2.0375 0.17967 1.8269 -0.64497 -0.29409 1.6621 1.6449 -0.58706 2.0361;

-0.8789 -0.82931 -0.33006 2.4585 0.25847 0.59257 -1.3263 -1.4221 0.56028 -1.246

0.48434 2.6359 2.567 0.33864 2.1691 2.0583 2.2656 0.75952 1.062 -0.69716;

-0.92178 1.8881 2.0184 -1.9775 0.88015 1.6529 1.9165 0.28952 1.7327 1.8705 2.0931 -

0.18077 1.1225 1.6229 -0.11249 1.4712 -1.5148 0.12491 1.7676 0.34114;

1.1154 -1.3387 -0.11071 -0.30272 -0.69325 2.3929 -1.6084 -0.21867 -1.5819 -0.43299

2.2543 2.148 1.1971 -1.1345 1.5195 0.43341 -2.5531 -0.78669 1.445 1.334;

-0.6284 1.5918 1.1481 0.45122 2.8126 -1.2392 -2.1094 -0.66093 -1.1668 0.91009

0.60498 1.0852 1.1834 2.3248 1.9419 0.26604 -2.0877 0.20785 1.944 -0.90947;

1.9587 -1.7722 -0.50812 -2.319 -1.6095 2.4513 -2.3669 0.54037 1.172 -2.1491 -1.3738 -

0.32042 1.3534 -0.053012 1.4311 0.18885 1.3234 0.96526 -0.77075 -1.2046;

0.62395 0.67625 -0.077309 -1.9163 -0.99189 -0.15776 0.45703 1.1481 2.2203 -0.80612

2.4831 -1.2634 0.97829 2.1727 0.37317 1.1661 0.28517 2.7068 2.6351 -2.3245;

-0.19732 -0.71551 0.083086 2.4232 0.52199 1.2527 -0.89368 0.56382 -3.0429 -1.4215

1.8027 2.7168 1.9995 -0.91605 0.83622 0.58416 -0.93651 0.70624 -1.9017 0.62462;

-0.21302 2.4763 0.62008 -0.52747 0.030618 0.797 2.4542 1.807 1.8153 -1.1298 -1.2273

-1.2086 -0.5148 2.7781 -1.5728 -1.1492 0.15256 1.4514 -1.657 -1.3266;

1.8093 -0.62835 -0.010542 0.43175 1.8556 0.85602 0.27024 -0.69986 2.3253 -1.7988

2.4142 -0.8184 1.6639 1.9188 -2.342 2.1695 0.63335 -1.358 0.21704 0.25233;

-0.62302 0.44494 -2.5205 -1.9005 2.4151 -0.72544 -0.87981 -0.62693 1.8809 1.7572

1.2932 0.39357 2.3716 -1.08 0.43669 -2.2839 0.65154 -0.50548 -0.29694 -1.2114;

1.1464 -0.41652 2.1002 1.4334 2.0287 1.1927 -0.48917 0.49421 1.0455 -2.5653 -

0.41121 2.4319 1.081 -1.8942 1.4026 0.12931 0.4948 -1.018 -2.2073 -0.99635;

-1.6046 -2.2245 1.677 -1.6962 -1.182 -0.065035 0.18135 -1.6376 -0.28411 -1.846

1.9622 -0.38106 -0.12648 1.2878 1.5887 0.27783 -1.1348 2.5239 -1.8249 1.6319;

2.1892 -1.2199 -2.3435 -0.72842 -1.6442 -0.37102 -2.2672 1.6081 -1.4481 1.3056 -

0.63831 -1.0642 -1.881 1.7183 0.70666 -0.12495 -1.1613 0.21133 -2.2819 -1.6896]

312

The matrix of weights from the second hidden layer nodes to the third hidden layer nodes

(iw 3, 2):

[1.8746 -0.89305 1.744 2.1698 2.0149 -1.8158 1.7097 0.30429 -0.73645 -2.1121 1.9202 -

1.5021 -2.3995 -1.3412 1.0464;

-0.24969 0.20851 2.7371 -0.55734 -0.68594 1.1659 0.32128 2.8112 -1.838 2.5849

2.6554 -1.9633 0.22612 -1.0195 -1.7012;

0.43593 0.83728 -0.44766 -2.7619 0.57858 -2.6619 -0.81414 2.6012 0.7144 -0.19577

1.5826 -1.5387 2.2225 2.0047 1.0532;

-1.8505 -0.069614 -0.3922 -3.2289 -0.026621 -0.28225 1.1239 1.8144 -2.575 -0.63929

1.371 -2.6888 0.49397 2.7586 -0.91565;

-2.2247 -2.1071 1.843 -1.7703 -0.424 1.0666 1.3701 1.5006 2.304 0.41289 -2.1081 -

2.446 2.0402 1.1134 1.5589;

-1.317 2.6514 -2.464 0.66646 -2.5365 0.51663 -0.14216 -0.37108 2.7571 2.5638

0.81146 -0.14786 -1.3037 1.8148 0.50924;

0.77755 -0.15352 -2.0285 2.3663 1.8349 3.2685 0.64157 -2.4044 1.1428 1.3576 2.2002

-0.58845 -1.5505 -1.2971 -0.72603;

0.060605 -1.2487 1.0722 1.1642 -1.6616 1.8614 -1.9825 3.2714 -0.28992 -2.7026 1.157

-0.2449 1.3993 3.3683 1.0698;

2.281 -1.8388 1.848 -1.2109 -1.8779 1.9401 0.89893 1.8977 1.8982 2.0604 2.5457

0.62028 -0.30926 2.14 1.6572;

-2.5277 0.3542 2.4269 0.83718 -0.91481 0.094006 -2.7763 -2.1455 -2.7416 -1.3561

1.7942 -1.3213 -0.50729 -1.2123 -2.0684]

313

The matrix of weights from the third hidden layer nodes to the fourth hidden layer nodes

(iw 4, 3):

[3.3118 -0.029973 -1.7086 2.6421 2.5819 -3.2451 -0.90743 -0.85697 -2.0989 -0.48801;

-2.6397 1.4878 0.224 -3.3185 -1.46 -0.72539 -2.6905 -1.2204 1.3719 -3.0077;

1.2133 -2.7204 -0.63364 0.38662 -3.9798 -2.9172 0.30322 -0.55424 3.3309 0.32613;

2.9842 0.6636 -0.95366 -0.45138 1.6837 -3.0945 -0.55417 -3.4008 1.3252 3.8397;

3.1591 2.6323 -0.21436 -1.425 -3.2038 0.66739 2.93 0.084489 2.0293 -0.72111]

The matrix of weights from the fourth hidden layer nodes to the fifth hidden layer nodes

(iw 5, 4):

[1.6703 1.0973 -0.65058 -2.1238 -0.56007]

314

The matrix of biases for the first hidden layer nodes (b 1):

[-5.1636;

-4.4262;

4.0208;

3.755;

-3.0667;

-2.0715;

1.8703;

1.2979;

1.3744;

-0.33229;

0.5332;

-1.3046;

-1.1322;

-1.688;

3.1306;

2.49;

-3.9999;

-4.1369;

3.4163;

-5.4958]

315

The matrix of biases for the second hidden layer nodes (b 2):

[-5.662;

-2.255;

-1.6396;

-5.5168;

-4.4382;

-3.6276;

0.68071;

-4.7479;

-3.9795;

-0.29637;

-2.7047;

-2.4574;

0.91688;

-0.10902;

7.9531]

316

The matrix of biases for the third hidden layer nodes (b 3):

[-4.817;

0.2902;

-5.5327;

2.8447;

-0.32041;

-3.2198;

0.14043;

-4.2837;

-2.6078;

4.1015]

The matrix of biases for the forth hidden layer nodes (b 4):

[-1.5436;

7.8936;

1.9165;

1.289;

0.84813]

The matrix of biases for the output layer node (b 5):

[0.75002]

experimental, numerical, and soft computing-based …

Documents