Modelling and Simulation of Gas adsorption in

Shale Gas Reservoirs

JOHN SENAM FIANU

The thesis is submitted in partial fulfilment of the requirements for the award of the degree

of

Doctor of Philosophy
of the

University of Portsmouth

March 2019
Declaration

Whilst registered as a candidate for the above degree, I have not been registered for any
other research award. The results and conclusions embodied in this thesis are the work of the
named candidate and have not been submitted for any other academic award

John Senam Fianu

March 2019

i
Acknowledgment

I wish to express my profound gratitude to my supervisor’s Dr Jebraeel Gholinezhad and Dr

Mohamad Hassan for believing in me and going beyond and above to guide me throughout
my academic journey. They showed great trust in handing me this opportunity even in the
face of stiff competition for this funded scholarship. It has been a pleasure working for both
of you and I hope I have repaid a little of the faith you placed in me. Your support and guidance
have been invaluable throughout my study.

Special thanks also go to my lovely wife, Sammuella Fianu whose patience has been tested so
many times over the course of my studies. You have always supported me and comforted me
when I have not always been at my best.

To my two lovely children Selasie and Fafali, without the smiles you put on my face every day,
I doubt I would have been successful at this. I would also like to thank my parents, brothers
and sisters for always loving me and supporting me.

Finally, I would like to thank all of my friends and colleagues at room A0.23/3 and all the
wonderful people who have always believed in me and contributed in many ways to shape
the person I am today.

The research project has been fully funded by the University of Portsmouth and all the work
undertaken carried out at this premises.

ii
Dedication

To my Dad, Livingstone Kwasi Fianu.

iii
Abstract

Gas adsorption accounts for a large portion of total gas in place in most shale gas reservoirs.
However, the mechanism of contribution of gas adsorption to total gas recovery from shale
gas reservoirs is hardly understood. Gas adsorption can be expressed both as a function of
pressure and temperature. In most studies of gas adsorption, it is only expressed at a single
temperature by using Langmuir isotherm. Very often, erroneous calculations are obtained for
gas in place and prediction of production performance since the isotherm used might not
represent the true reservoir temperature. Very few adsorption models have addressed gas
adsorption as a function of both pressure and temperature with majority of the models
currently used in reservoir simulation and gas in place calculation ignoring the dependence of
gas adsorption on temperature. Therefore, the use of temperature-dependent models for
shale gas adsorption is crucial not only because it accounts for the effect of temperature in
gas adsorption/desorption but also for conducting numerical reservoir simulation where the
need for thermal stimulation could be explored as an enhanced recovery mechanism in shale
gas reservoirs.

On the other hand, in material balance calculations for unconventional reservoirs such as coal
bed methane and shale gas, temperature-dependent gas adsorption models can be
incorporated in analytical methodologies to predict useful information such as gas in place
calculations and future production performance analysis once data is obtained for adsorption
capacities at several temperatures. This would ensure accurate representation of gas
adsorption throughout the reservoir and better predictions of gas in place, pressure and
future production performance.

This thesis presents new methodology for incorporating temperature-dependent gas

adsorption models into material balance calculations for unconventional gas reservoirs and
also explores the use of microwave heating as a thermal stimulation strategy for enhancing
gas recovery in shale gas reservoirs. A dual porosity –dual permeability model is developed
for the system of shale gas with the account of both viscous and Knudsen diffusion in the
matrix. This is coupled with the microwave heating by solving Maxwell’s equation for the
electric field using a finite difference time domain methodology. Furthermore, a coupled
iv
electromagnetic –thermal model is developed to investigate the production of gas from shale
gas reservoir using microwave heating as a novel enhanced gas recovery technique.
Simulation results indicate higher production of gas when microwave heating is used through
the elevation of formation temperature around the waveguide prompting desorption of gas
into the well compared with when no thermal stimulation is used. The findings from this work
can provide a better insight into modelling gas adsorption in shale formations and open
avenues for obtaining higher ultimate estimated recoveries from these unconventional
reservoirs through the use of microwave heating.

v
Contents
Publications by the Candidate ................................................................................................... xi
Chapter 1 Introduction ............................................................................................................... 1
1.1 Motivation ...................................................................................................................... 1
1.1.1 Adsorption Phenomenon .......................................................................................... 4
1.1.2 Adsorption in Shale Rock Reservoirs ....................................................................... 4
1.2 Problem statement and Research Gaps ............................................................................ 5
1.3 Research Questions .......................................................................................................... 6
1.4 Aim of study .................................................................................................................... 6
1.5 Objectives ........................................................................................................................ 6
1.6 Methodology .................................................................................................................... 7
1.7 Thesis Layout ................................................................................................................... 8
Chapter 2 .................................................................................................................................. 10
Modelling of Optimised Gas Adsorption in Shale Gas Reservoirs with Advanced Error
Analysis.................................................................................................................................... 10
2.1 Introduction .................................................................................................................... 10
2.1.1 Types of gas adsorption Isotherm ........................................................................... 12
2.2 Pure Component Adsorption Models............................................................................. 14
2.2.1 Langmuir Isotherm.................................................................................................. 14
2.2.2 BET Isotherm .......................................................................................................... 15
2.2.3 Potential Theory Approach ..................................................................................... 16
2.2.4 Dubinin – Radushkevich (D-R) and Astakhov (D-A) Equations ........................... 18
2.3 Multicomponent Adsorption Models ............................................................................. 20
2.3.1 Extended Langmuir Model ..................................................................................... 22
2.3.2 Ideal Adsorbed Solution Theory (IAST) ................................................................ 23
2.3.3 Vacancy Solution Model......................................................................................... 26
2.4 Methodology .................................................................................................................. 32
2.4.1 Procedure for the Estimation of Parameters ........................................................... 32
2.4.2 Error Function Analysis .......................................................................................... 32
Average Relative Error Function (ARE) ......................................................................... 32
Sum of Squared Error Function (SSE)............................................................................. 33
Sum of Absolute Error Function (SAE)........................................................................... 33
Marquardt’s Percent Standard Deviation (MPSD) .......................................................... 33
Hybrid Fractional Error Function (HYBRID) ................................................................. 33

vi
 Sum of Normalised Error (SNE)...................................................................................... 34
Sum of Normalised Error Procedure................................................................................ 34
2.5 Experimental Data ..................................................................................................... 35
2.6 Results and Discussion .................................................................................................. 36
2.6.1 Single component modelling of New Albany Shale ............................................... 36
3.2 Multi-Component Modelling of Ethane and Methane Adsorption on Activated Carbon
(Szepesy and Illes, 1963) ................................................................................................. 42
2.7 Summary ........................................................................................................................ 47
Nomenclature ....................................................................................................................... 48
Chapter 3 .................................................................................................................................. 51
Temperature-Dependent Gas Adsorption Modelling in Shale Gas Reservoirs ....................... 51
3.1 Introduction .................................................................................................................... 51
3.2 Temperature-dependent Langmuir volume or Saturation loading ................................. 52
3.3 Temperature-dependent Langmuir Models.................................................................... 55
3.4 Modified Vacancy Solution ........................................................................................... 56
3.5 Simplified Local Density (SLD) .................................................................................... 57
3.6 Bi-Langmuir Model ....................................................................................................... 58
3.7 Exponential Model ......................................................................................................... 59
3.8 Materials and Methods ................................................................................................... 63
3.8.1 Shale Gas Data ........................................................................................................ 63
3.9 Results and Discussion .................................................................................................. 64
3.10 Summary ...................................................................................................................... 77
Nomenclature ....................................................................................................................... 77
Chapter 4 .................................................................................................................................. 79
Material Balance Calculations for Unconventional Gas Reservoirs Using Temperature-
Dependent Gas Adsorption Models ......................................................................................... 79
4.1 Introduction .................................................................................................................... 79
4.2 P/Z plot methodology used for unconventional gas reservoirs ...................................... 80
4.2.1 King’s Method ........................................................................................................ 80
4.2.2 Jensen and Smith Modified Method ....................................................................... 81
4.2.3 Seidle Method – Modified King Method ................................................................ 82
4.2.4 Ahmed and Roux Method ....................................................................................... 83
4.2.5 Moghadam et al., 2011 Method .............................................................................. 84
4.2.6 Firanda Method ....................................................................................................... 85
4.3 New Method – Temperature-Dependent Adsorption Models ....................................... 86
vii
 4.4 Methodology .................................................................................................................. 91
4.4.1 Average Reservoir Pressure .................................................................................... 93
4.4.2 Prediction of Reservoir Production Performance ................................................... 95
4.5 Results and Discussion .................................................................................................. 98
4.5.1 OGIP and Average Pressure ................................................................................. 105
4.5.2 Reservoir performance prediction......................................................................... 110
4.6 Summary ...................................................................................................................... 112
Nomenclature ..................................................................................................................... 113
Chapter 5 ................................................................................................................................ 115
Numerical Simulation of Gas Adsorption in Shale Gas Reservoirs ...................................... 115
5.1 Introduction .................................................................................................................. 115
5.2 Mathematical Formulation of Isothermal Model in Shale Gas Reservoir ................... 120
5.2.1 Mass Accumulation Term ..................................................................................... 124
5.2.2 Flow Vector Term ................................................................................................. 125
5.2.3 Source and Sink Term ........................................................................................... 128
5.3 Discretization and Implementation .............................................................................. 131
5.3.1 Newton Method .................................................................................................... 132
5.3.2 Solution Method of Coupled Nonlinear PDE ....................................................... 136
5.3.3 Validation of Numerical Method with Analytical method .................................. 137
5.4 Results and Discussion ................................................................................................ 140
5.4.1 Simulation Study I ................................................................................................ 140
5.4.2 Effect of Gas Adsorption ...................................................................................... 144
5.4.3 Effect of Matrix and Fracture Porosity ................................................................. 147
5.4.4 Effect of Matrix and Fracture Permeability .......................................................... 149
5.5 Heat Transfer through a Porous Media ........................................................................ 152
5.5.1 Fully Implicit Method for the Heat Equation ....................................................... 156
5.5.2 Temperature –Dependent Variables ..................................................................... 156
5.5.3 Gas Viscosity ........................................................................................................ 157
5.5.4 Z Factor Correlation .............................................................................................. 158
5.6 Simulation Study II ...................................................................................................... 159
5.7 Summary ...................................................................................................................... 165
Nomenclature ..................................................................................................................... 167
Chapter 6 ................................................................................................................................ 170
Thermal Stimulation of Shale Gas Reservoirs Using Microwave Heating ........................... 170

viii
 6.1 Introduction .................................................................................................................. 170
6.2 Microwave Heating ...................................................................................................... 172
6.3 Modelling microwave heating ..................................................................................... 178
6.3.1 Maxwell equations ................................................................................................ 179
6.3.2 Dielectric properties and constitutive relations..................................................... 180
6.3.3 Transverse Electromagnetism (TEM), Transverse Electric (TE) and Transverse
Magnetic (TM) ............................................................................................................... 183
6.4 Finite Difference Time Domain (FDTD) Method ....................................................... 183
6.4.1 Courant stability criterion ..................................................................................... 185
6.4.2 Update Equations .................................................................................................. 185
6.4.3 Microwave penetration depth ............................................................................... 186
6.5 Numerical Modelling of a 2D Reservoir ..................................................................... 188
6.6 Results and Discussion ................................................................................................ 192
6.6.1 Effect of microwave heating on gas production ................................................... 192
6.6.2 Temperature distribution ....................................................................................... 193
6.6.3 Effect of microwave frequency ............................................................................. 195
6.6.4 Effect of microwave heating on free gas content.................................................. 199
6.7 Conclusion: .................................................................................................................. 201
Nomenclature ..................................................................................................................... 203
Chapter 7 ................................................................................................................................ 205
Summary, Conclusion and Future Work ............................................................................... 205
7.1 Conclusion:.............................................................................................................. 205
7.1.1 Review and Modelling in Shale Gas Reservoirs (Chapter 2) ............................... 205
7.1.2 Temperature Dependent Gas Adsorption Modelling in Shale gas Reservoirs
(Chapter 3) ..................................................................................................................... 206
7.1.3 Material Balance Calculations for Unconventional Gas Reservoirs Using
Temperature Dependent Gas Adsorption Models (Chapter 4) ...................................... 207
7.1.4 Numerical Simulation of Gas adsorption in Shale Gas Reservoirs (Chapter 5) ... 207
7.1.5 Thermal stimulation of shale Gas Reservoirs using Microwave Heating (Chapter 6)
........................................................................................................................................ 208
7.2 Future Work ................................................................................................................. 208
References .............................................................................................................................. 210
Appendix 2 ......................................................................................................................... 226
Appendix 3 ......................................................................................................................... 232
Appendix 4 ......................................................................................................................... 241

ix
Appendix 5 ......................................................................................................................... 242
Appendix 6 ......................................................................................................................... 248

x
 Publications by the Candidate

 Peer Reviewed Publications

 Fianu J, Gholinezhad J, Hassan M. Comparison of temperature-dependent gas

adsorption models and their application to shale gas reservoirs. Energy & Fuels. 2018
Apr 2;32(4):4763-71.

 Fianu, J. S., Gholinezhad, J., & Sayed, M. H. (2019). Comparison of single, binary and
temperature dependent adsorption models based on error function analysis. Journal
of Oil, Gas and Petrochemical Sciences, 2(2), 77-91.

 Fianu, J., Gholinezhad, J., Hassan, M., 2019. Heliyon Application of temperature-
dependent adsorption models in material balance calculations for unconventional
gas reservoirs. Heliyon 5, e01721. doi:10.1016/j.heliyon.2019.e01721
 Book Publications

 Gholinezhad J, Fianu JS, Hassan MG. Challenges in modelling and simulation of shale
gas reservoirs. Springer International Publishing; 2018 Feb.

 Other Publications

 Mustanen D, Fianu J, Pucknell J. Account of hydraulically fractured onshore wells in

the UK and seismicity associated with these wells. Journal of Petroleum Science and
Engineering. 2017 Sep 1; 158:202-21.
 Mohamed Hasan,Michael Kenmore, John Fianu , Amir Gharavi. Hydraulic Fracturing ,
a UK prospective. Article published in London petrophysical society Newsletter.

xi
 Chapter 1 Introduction
1.1 Motivation
Due to the high adsorption rate of most shale gas reservoirs, critically analysing and
understanding the adsorption processes, as well as its contribution to the overall recovery of
gas, is crucial. Also, by investigating the gas adsorption phenomenon, one can gain an
understanding of well performance, shale characterisation and the optimisation of fracture
design in shale gas reservoirs (Yu et al.,2015; Lane et al., 1991, 1989).

Shale gas is increasingly becoming an abundant source of energy for the world. It has already
revolutionised the energy industry in the USA. Recent advances and knowledge in the
production of shale gas have made it possible for it to be produced cheaply, and as a
consequence, have contributed to gas prices falling relative to oil prices. Different countries
in the world are eagerly advancing their quest to exploit this source of energy that some years
ago was deemed unfavourable in exploring for oil and gas.

As an unconventional reservoir, it is characterised by low permeability, which makes the

economic development of this resource risky, requiring higher investment. To reduce this risk,
a combination of horizontal drilling and hydraulic fracturing has allowed for easy access to
the natural gas by providing a permeable pathway through which the gas can flow.

Other unique features of this unconventional resource present challenges to its development.
Large amounts of gas are adsorbed on the shale matrix as well as being stored in the pore
networks of the shale. The presence of natural fractures also plays an important role in the
production of natural gas from shale gas reservoirs. Where these fractures are present, there
must be an extensive network of intersecting fractures to be able to contribute towards
production. Diffusion in the nano-pore network of the shale and the concept of non-Darcy
flow are among several other characteristics of shale gas reservoirs that pose a challenge to
its development.

A large surface area of shale/coal means that more gas can be adsorbed compared to that
stored by compression in the pore matrix. Production forecast in shale gas reservoirs without
the inclusion of adsorbed gas will be less accurate than forecast made when the effects of gas
adsorption are included (Lane et al., 1991, 1989).

1
Adsorption plays an important role in accounting for the original gas in place (OGIP) of shale
gas reservoir. The desorbed gas also contributes towards production, even though for some
shale plays, like the Barnett1, the contribution of gas desorption towards production is small
and only relevant at the later stage of the well’s life (Cipolla et al., 2010).

According to Cipolla et al. (2010), the contribution from adsorbed gas makes up a low
percentage of the overall gas recovery due to the ultra-low permeability of the rock. He also
observed that the impact of desorption decreases when the network fracture spacing is
smaller using the Barnett shale play. Therefore, it is expected that gas desorption will play a
small role in well performance, and this will usually occur at the later life of the well when
pressures in the tight matrix have become very low. Frantz et al. (2005) also concluded that
in the Barnett shale, desorption is not significant for well performance.

The contribution of adsorption to shale gas performance was further analysed by Yu et al.
(2015). This was done by conducting experimental measurements of methane adsorption
from the Marcellus shale2. They concluded that the adsorption of methane on the Marcellus
shale deviated from the Langmuir isotherm but obeyed the Brauner, Emmet and Teller (BET)
isotherm. This implies that gas adsorption in the Marcellus shale reservoir behaves like multi-
layer adsorption rather than mono-layer adsorption assumed for the Langmuir isotherm.
Contribution from the BET isotherm model shows an increase in overall gas recovery
compared to use of a Langmuir isotherm.

Heller and Zoback (2014) noted the critical effect of adsorption of methane on the production
capacity of the reservoir. The adsorption of methane is controlled by the composition and
microstructure of the rock. The adsorbed phase is also pressure dependent and with a
depleting reservoir, the adsorbed phase contributes to additional gas production. Heller and
Zoback (2014) pointed out that certain factors such as desorption pressure, kinetics and
effective stresses affect the contribution of desorbed gas to the total production.
Furthermore, Heller and Zoback (2014) studied the adsorption of methane and Carbon

1
Barnett Shale is a geological formation located in the North central Texas primarily for shale
gas drilling
2
Marcellus Shale is a sedimentary rock that stretches through Pennsylvania to West virgin also
noted for shale gas drilling
2
Dioxide (CO2) on various shale samples in order to understand adsorption in gas shales. Their
studies focused on measuring the adsorptive capacities of four shale samples as well as
samples of carbon, illite and kaolinite. Their experiments proved that in all the samples, CO 2
proved to have better affinity to shale, and that the adsorptive capacity of shale with CO 2 is
indeed two to three times larger than that with methane.

Mengal and Wattenbarger (2011) studied the effect of adsorbed gas in shales using a set of
techniques previously employed for the evaluation of tight gas reservoirs. This included
reviewing previous literature that utilised end of transient time and boundary dominated flow
methods to estimate original gas in place. Mengal and Wattenbarger (2011) included the
concept of adsorbed gas in their studies. By including adsorption into their study, they
reasoned that at very late times, the well is more likely to undergo boundary dominated flow,
and it is at this point that desorption from shale normally takes place. By reviewing analytical
solutions from previous research works aimed at estimating OGIP (Anderson et al., 2010;
Ibrahim et al., 2006; Ibrahim and Wattenbarger, 2003), Mengal and Wattenbarger (2011)
included desorbed gas in all of the methods that ignored gas adsorption effects. By applying
these combined models to field data, they were able to analyse results with and without
adsorption. The results of their research showed that desorption plays an important role in
determining the OGIP calculations, and it also affects the recovery factor of the field. Ignoring
desorption will therefore lead to inaccurate estimates of OGIP and recovery factors. They
were able to show that OGIP estimates increased by 30% when adsorbed gas was included in
the estimation; however, recovery factor estimates were found to have decreased by 17%
when adsorbed gas was included in the calculation.

Modelling gas adsorption is therefore crucial when conducting numerical simulation studies.
The adsorption capacity of shale gas reservoirs depends on both the pressure and
temperature. Several isotherm models have been explored to represent the adsorption of
gas onto shale and coal. Amongst which is the Langmuir isotherm (Langmuir, 1916), which
has been widely used to explore the pressure dependence of gas onto shale.

The Langmuir isotherm assumes that the adsorbed gas behaves as an ideal gas under
isothermal conditions. Hence there is a dynamic equilibrium, at constant temperature and
pressure, between the adsorbed and non-absorbed gas.

3
Modelling of temperature dependence has been largely ignored in most studies, with only a
handful of researchers focussing on the pressure and temperature dependency of adsorbed
gas on the shale.

1.1.1 Adsorption Phenomenon

Adsorption is often seen as exothermic process where the gas molecules or atoms adhere to
the surface of a rock due to the imbalance in the surface energy. This process often results in
a thin layer of the adsorbate on the surface of the adsorbent. The adsorbate is referred to as
the gas molecules while adsorbent is the surface or rock that the gas adheres to. Hence
adsorption is often referred to as a surface phenomenon. The gas that adsorbs onto the rock
surfaces can also desorb from the surface especially at very low pressures or high
temperature. This is often refrred to as desorption. The process of adsorption and desorption
is reversible due to the weak “Van Der Waals” attraction forces that occurs between the gas
moleculs and the rock surface.

There are two main types of adsorption known as physiosorption and chemisorption.
Physiosorption is a reversible process which occurs when molecular forces (Van Der Waals)
between adsorbate and adsorbent is relatively weak. This type of adsorption normally has a
low enthalpy 25-45KJ/mol which means the adsorbate has low kinetic energy and the process
can occur at low temperature (Sing et al., 1985).

Chemical adsorption is an irreversible process and occurs when there is a chemical attraction
force between the adsorbate and adsorbent which result in strong chemical bond between
molecules. Unlike physiosorption, it occurs within a single layer between the two phases and
this indicate a stronger adsorption in terms of attraction forces and result in high adsorption
entropy (200-400KJ/mol) , it can occur at high temperature and hence requires an activation
energy for starting the adsorption process (Ho and McKay, 1998).

1.1.2 Adsorption in Shale Rock Reservoirs

Shale reservoirs are naturally fractured and possess layered matrix composed of micropores.
High amount of gas is sorbed onto the surface of these micropores and the rest is trapped
free gas. Gas in shale is found in three ways: it can can be within the pore spaces, held within
natural rock fractures or adsorbed onto the matrix of the rock. The gas held in natural
fractures in produced directly while the adsorbed gas is released when downhole pressure is

4
decreased. The structure of pore spaces in shale provides great surface area which allow shale-pores
to adsorb high amount of natural gas. Various production models (Javadpour et al., 2007) are
used for gas production built on the assumptions that production is measured by a constant
diffusivity coefficient from spherical bodies connected via large fracture network.

1.2 Problem statement and Research Gaps

Most of the published studies on gas adsorption in shale gas reservoirs have mainly
considered adsorption as a function of pressure, with the Langmuir isotherm widely used as
the adsorption model in these applications ( Arri et al., 1992; Mengal and Wattenbarger,
2013; Yu et al., 2015, 2014; Yu and Sepehrnoori, 2014; Zhang et al., 2015). One of the reasons
for this is the ease with which the Langmuir model can be incorporated into most reservoir
simulators. However, with this choice of adsorption model, there can be no examination of
thermal stimulation as an enhanced recovery method.

Only a handful of researchers have considered thermal stimulation in shale gas reservoirs or
(until recently) the use of microwave heating as a heat source. The choice of temperature
dependent adsorption model has been made with little consideration to which model gives
the best accuracy in modelling gas adsorption.

Furthermore, where microwave heating has been used, the mathematical model of the shale
gas has failed to consider the non-Darcy effect of flow in shale gas reservoirs and hence does
not accurately represent the physical system of shale gas reservoirs.

Still further, Firanda (2011), King (1993), Moghadam, Jeje, & Mattar ( 2011), Seidle & Arri
(1990) and Singh (2013) consider Langmuir isotherms in material balance simulators without
considering the effect of temperature on gas adsorption in predicting recovery of gas.

Therefore, the focus of this research will be to address these gaps by first reviewing different
adsorption models that incorporate temperature to find the best model for a set of shale gas
adsorption data and second, developing a coupled electromagnetic-thermal model for shale
gas reservoirs. The work will also focus on developing a new methodology for incorporating
temperature dependent gas adsorption model into a material balance simulator for
unconventional gas reservoirs.

5
1.3 Research Questions
In the course of this research, the choice of methodology, as well as the results obtained, have
been guided by the following questions:

 How well can gas adsorption models predict shale gas adsorption capacities, and what
models are best for single, multi-component and temperature dependent gas
adsorption on shale gas reservoirs?
 Can temperature dependent gas adsorption models be used in analytical material
balance calculations for unconventional gas reservoirs?
 What contribution can gas adsorption make to overall total production?
 Can a temperature dependent model be used in place of the Langmuir model in
numerical simulations involving thermal stimulation, and which temperature
dependent model gives the best accuracy in modelling gas adsorption?
 What factors influence production in a shale gas involving thermal stimulation, and
what thermal stimulation strategies exist for shale gas reservoirs?
 Is microwave heating suitable for enhanced shale gas recovery, and what factors
influences microwave heating in shale gas recovery?

1.4 Aim of study

The principal aim of this thesis is to examine the modelling of shale gas reservoir considering
the use of temperature-dependent adsorption models, inorder to give accurate
representation of adsorption phenomenon. The thesis also examines the simulation of gas
adsorption production in shale gas reservoirs with microwave heating as enhanced gas
recovery method (EGR).

1.5 Objectives
 Review applications of gas adsorption models in shale gas reservoirs. Different gas
adsorption models are able to predict differently on gas adsorption data. The choice
of adsorption model would therefore depend on how well the model is able to fit the
adsorption data and whether it can be used in single or multicomponent systems.
 Develop a new methodology for predicting gas in place, average pressure and future
gas production, using material balance calculations for unconventional gas reservoirs.

6
 The new methodology should be able to account for gas adsorption at multiple
temperatures.
 Develop a numerical simulation model to analyse the contribution of gas adsorption
to shale gas reservoirs.
 Investigate the application of thermal stimulation in shale gas reservoirs, using
microwave heating as the heat source.

1.6 Methodology
In order to address these questions, several methodologies have been used to provide
answers to these questions. Both analytical and numerical models have been used, some
sourced from literature sources, whilst others are new methods developed to address these
fundamental questions.

Several types of nonlinear regression techniques have been proposed and used to identify an
optimal model for gas adsorption on various adsorbents. The techniques involve minimising
error distribution between the experimental data and the predicted results. Thus, several
error functions have been used in the non-linear regression, to try to optimise the adsorption
model for any particular set of experimental data. A more statistically robust method
involving sum of normalised error (SNE) has been used throughout the adsorption modelling
to obtain an optimised adsorption model. These error functions include sum of squared error
(SSE), absolute relative error (ARE), sum of absolute error (SAE), Marquardt’s percent
standard deviation (MPSD) and hybrid fractional error (HYBRID).

Analytical material balance methods have been used to determine gas in place, average
pressures and future production predictions for unconventional gas reservoirs. Several
methodologies have been proposed for material balance in unconventional reservoirs, with
earlier methodologies focused on King’s Method. However, the choice of methodology used
throughout this thesis for analytical calculations of material balance follows an approach by
Ahmed and Roux (2006),with modifications made to include temperature dependent
adsorption models.

Numerical modelling was chosen as the preferred method in those parts of the thesis relating
to simulations of shale gas. This choice is obvious because of the set of nonlinear partial
differential equations involved in describing the physical system. A combination of Newton’s
7
method and iterative methods was used to solve the system of equation concerning the shale
gas system. The finite difference method was used to discretize the medium. In terms of
solving the electromagnetic equations, finite difference time domain methodology was used.

The whole set of equations involving shale gas systems, heat equations and Maxwell
equations has been coupled together to describe a coupled electromagnetic –thermal model
for shale gas systems. A segregated approach was employed in the solution of the systems
of equations to arrive at the final solutions. Convergence limits have all been incorporated
into the simulation program written in MATLAB. The detailed methodology is presented in
the relevant chapters of this thesis.

1.7 Thesis Layout

Following the compilation style PhD thesis guidance, each chapter has been presented as a
standalone with dedicated, specific objectives. However, the chapters work in combination
to present the modelling and simulation of gas adsorption in shale gas reservoirs.

Chapter 2: This chapter provides a complete overview of gas adsorption in shale gas
reservoirs. It primarily provides details of most used adsorption models in the petroleum
industry by focussing on single component models and multi component models. The chapter
ends by applying the adsorption models discussed to a set of adsorption data to investigate
their predictability and accuracy.

Chapter 3: An overview of temperature dependent modelling and its application is the main
theme of this chapter. Firstly, the chapter goes into detail by examining the different forms
of temperature dependent adsorption models and why they are relevant in modelling gas
adsorption. It also discusses the limitations of using the Langmuir isotherm in modelling gas
adsorption. Finally, a complete shale gas adsorption dataset sourced from literature provides
the basis for modelling temperature dependent gas adsorption.

Chapter 4: New methodology that incorporates temperature dependent gas adsorption

modelling into material balance calculations is introduced in this chapter. A review of
different mathematical methods for material balance calculations in unconventional gas
reservoirs is conducted, with particular emphasis on Ahmed and Roux’s methodology. A

8
comparison is made of the new methodology with the Ahmed and Roux method for gas in
place calculations and future production results.

Chapter 5: This chapter discusses the use of numerical simulation in shale gas reservoirs first
by comparing models with and without gas adsorption in a mathematical model of shale gas.
The role of several factors, such as matrix porosity and fracture permeability, is also
examined. Validation of the proposed mathematical model for shale gas system is shown,
using results from earlier literature. Finally, the chapter introduces the concept of thermal
stimulation by coupling the heat equation with the shale gas system. The source term is left
as generic and production results are examined with and without a heat source.

Chapter 6: This chapter reviews the concept of thermal stimulation in shale gas reservoirs.
Microwave heating is chosen as the heat source for simulation work involving thermal
stimulation of shale gas reservoirs. The chapter presents mathematical formulations of
Maxwell’s equation and solution methods. Finite difference time domain as a numerical
method for solving the set of Maxwell equation is used and later coupled with the heat
equation and the shale gas model developed in previous chapter. Several factors that
influence microwave heating and gas recovery, such as microwave frequency, are also
examined.

Chapter 7: Finally, a summary of all the preceding chapters and their results is discussed in
this chapter. Recommendations for future work are also provided in this chapter.

9
 Chapter 2
Modelling of Optimised Gas Adsorption in Shale Gas
Reservoirs with Advanced Error Analysis

2.1 Introduction
Shale gas reservoirs are characterised by gas adsorption on shale matrix and free gas stored
within the pores of the matrix. Both free gas and adsorbed gas make up a large portion of the
original gas in place (OGIP) of these reservoirs, with gas adsorption estimated to be about 20-
85% of the total gas in place (Curtis, 2002). Gas adsorption plays an important role in the
estimations of the overall gas in place which, and so, in turn, is crucial when developing these
resources for future production. Langmuir isotherm has remained one of the most popular
models used in representing the relationship between the amount of gas adsorbed and
pressure. However, several other models have also been developed that can also represent
the adsorption process in most of these shale reservoirs (Ahmadpour et al., 1998; Bazan et
al., 2008; Charoensuppanimit et al., 2016; L. Chen et al., 2017; Clarkson et al., 2013; Clarkson
and Haghshenas, 2013; Guo et al., 2017; Sandoval et al., 2017; Tang et al., 2017; Y. Wang et
al., 2015). To ensure accurate representation of the amount of gas adsorbed, these models
need to be evaluated and compared with the experimental data for the gas adsorption in the
shale matrix. Since each shale rock might show unique properties, it may not be possible to
select a single model to represent the adsorption process in all the shale formations. For
instance, it has been reported that the Brunauer, Emmet and Teller (BET) model represents
the adsorption process in Marcellus shale better than Langmuir isotherm, based on different
samples within the formation (Yu et al., 2015). However, this may not necessarily be the case
for other shales.

Gas adsorption modelling involves applying a set of different adsorption models to acquired
experimental shale gas adsorption data. These models can be grouped as single component
systems or multi-component systems. Under the single component system, a single gas such
as methane is used as the adsorbed gas on shale. The advantage of using single component
models is that they are very simple to use in the calculation of adsorbed gas amount. This is
especially useful when conducting numerical simulations involving the calculation of
10
adsorption in shale gas reservoirs. For this reason, single component models can be found in
a variety of reservoir simulations of shale gas systems. However, this simplification is not valid
because in most cases, the formation gas is a mixture containing more than one component
(Wang et al., 2015). In shale gas systems, methane, carbon dioxide (CO2) and other gases can
be found, therefore, modelling the gas adsorption in such systems requires adsorption
models capable of addressing the multi-component gas mixture present.

There are lots of factors that can account for the adsorption capacity of methane on shale.
These factors include, but are not limited to, the total organic content (TOC), the level of
thermal maturity of the shale, Kerogen content, pressure and temperature. Experimental
studies have suggested that a plot of adsorbed gas quantity versus TOC will show a
proportional relationship , with high TOC of shale leading to high adsorption capacity
(Chalmers and Bustin, 2008; Lu et al., 1995; Ross and Bustin, 2007; Ross and Marc Bustin,
2009; Zhang et al., 2012). Low reservoir pressure will correspond to a much lower adsorbed
quantity due to the fact that higher binding energy is required for gas adsorption (Guo et al.,
2013; Raut et al., 2007).

Several works on adsorption modelling have been conducted without taking into
consideration the choice of error function used in optimising the adsorption model
(Chareonsuppanimit et al., 2012; Clarkson et al., 2013; Fianu et al., 2018; Tang et al., 2017;
Yang et al., 2017). This often results in only one set of adsorption constants for the adsorption
models being used, without any serious interrogation into how accurately the set of
adsorption constants fits the adsorption model to experimental data. According to Sreńscek-
Nazzal et al., (2015), very few detailed studies have been conducted that compare the
accuracy of the error functions used in modelling gas adsorption and the accuracy of the
predicted isotherm parameters. No study, however, has compared different error functions
on modelling gas adsorption in shale gas reservoirs. In minimising the difference between the
experimental data and the predicted results from the adsorption models, several error
functions have been proposed and applied to predict optimal isotherms, including sum of
square error (SSE), average relative error (ARE), sum of absolute error (SAE), Marquardt’s
percent standard deviation (MPSD) and hybrid fractional error (HYBRID) (Allen et al., 2003;
Ho et al., 2002; Kumar and Porkodi, 2007; Sreńscek-Nazzal et al., 2015).

11
2.1.1 Types of gas adsorption Isotherm
Gas adsorption involves gas molecules adhering to the surfaces of an adsorbent. This process
can be represented by many different isotherms. These isotherms can be basically classified
into 6 groups defined by the International Union of Pure and Applied Chemistry (IUPAC) for
physical adsorption. Due to the weak Van der Waal forces that characterise the interaction of
the gas molecules, gas adsorption is normally referred to as physical adsorption and hence a
reversible process allowing for both adsorption and desorption.

The following descriptions of each of the isotherm types have been given by Sing et al., (1985)
and are graphically represented in Figure 2.1.

Type I is described as the isotherm for microporous adsorbents in which the pore size is
smaller than the molecular diameter of the sorbate molecule (Ruthven, 1984). They have
relatively small external surfaces; examples being activated carbons and molecular zeolites.

Type II is associated with non–porous or macroporous absorbents. This is normally referred

to as BET isotherms. For non-porous adsorbents, the adsorption mechanism is monolayer
formation where the adsorbate is assumed to adhere on a single layer followed by multi-
layer adsorption in which case multiple layers are assumed to be present for the adsorption
process (Clarkson and Haghshenas, 2013) (See Figure 2.2). At low pressure, it becomes
concave and then turns linear when it reaches one third of the isotherm as pressure decreases
indicating monolayer coverage completion. We can also see from Figure 2.1 that the graph
then turns to be convex indicating multilayer formation.

The Type III isotherm is convex over its entire range and therefore does not exhibit a point B.
Point B indicates the end of monolayer coverage and the beginning of multilayer adsorption.
This type of isotherm is common in nonporous and microporous adsorbents, the isotherm
indicates unrestricted multilayer formation process, it forms because the interactions
between adsorbate’s molecules are stronger than the interactions between adsorbent and
adsorbate. The isotherm for types II and III occurs for wide-ranging pore sizes. Type II and III
isotherms show a continuous progression from monolayer to multilayer adsorption, with
Ruthven (1984) noting the high capacity of adsorption at very high temperatures being due
to capillary condensation that occurs in pores of increasing diameter.

12
Type IV has a hysteresis loop which is associated with capillary condensation taking place in
mesopores. As shown in Figure 2.1, the initial part of the isotherm follows the same path as
type II where monolayer to multilayer adsorption takes place. Type IV occur in many
mesoporous industrial adsorbents. Type IV isotherm is similar to Type II isotherm at low
relative pressures; however, Type IV isotherm starts decreasing when high pressure is
reached and beyond that the slope becomes horizontal to a constant adsorption value.

Type V according to Sing et al., (1985) is the least common, and it is related to type III isotherm
where the interaction between the adsorbent and adsorbate is weak. The isotherm is convex
to relative pressure, yet Type V isotherm reaches a plateau at high relative pressure. This type
of isotherm is common in mesoporous or microporous adsorbents.

Type VI isotherms show a stepwise multi-layer adsorption on a non-uniform porous surface,

with the sharpness of the steps dependent on the system and the temperature.

Figure 2.1: Types of adsorption isotherm (Yang , 2013)

When the adsorption model can only be applied to a single component system, such as
methane, it is normally classified under single/pure component modelling, otherwise referred
to as multi-component models. A brief summary of all the models discussed in the preceeding

13
sections under both single component and multi-component models have been summarised
in Table 2.1.

2.2 Pure Component Adsorption Models

2.2.1 Langmuir Isotherm
One of the key assumptions of Langmuir isotherm is that there must be a homogeneous
surface and no interaction between the neighbouring molecules. This is, however, a difficult
concept to apply even in coal or shale systems, because their internal organic matter is
chemically heterogeneous (Clarkson and Haghshenas, 2013) . Also, the monolayer volume is
assumed to be invariant to temperature. According to Clarkson et al., (1997) , several studies
have shown this assumption to be inaccurate, since the monolayer amount do vary with
temperature. Langmuir isotherm is given by the formula below

VL P
V ,
P  PL ………………………………………………(2.1)

V = volume of adsorbed gas at pressure P ,

VL = Langmuir volume or maximum gas adsorption at infinite pressure and

PL = Langmuir pressure corresponding to one half of the Langmuir volume. In order to

convert the gas content from scf / ft 3 to scf / ton the bulk density of shale is needed.

The assumptions under which Langmuir isotherm was developed make its limitation obvious
especially when applied to real systems; however, it has been found to provide good
approximations to experimental data on microporous materials (Bell and Rakop, 1986;
Clarkson et al., 1997; Mavor et al., 1990).

Due to the presence of multiple gas in shale reservoirs, there is competitive adsorption with
the gases competing to adhere on the same adsorption site. Carbon dioxide (CO2) is found to
have the most affinity to adsorption on coal and shales. Predicting accurate gas in place,
reserves and rates becomes even more important when there is the presence of
multicomponent gas. An extension of the Langmuir model known as Extended Langmuir has
been proposed, to accurately predict adsorption in the presence of other gases.

14
2.2.2 BET Isotherm
Stephen Brunauer, P.H. Emmet and Edward Teller developed in 1938 the BET isotherm, in
which they assumed that the adsorption layers on the surface of the organic carbon were
infinite. Unlike Langmuir isotherm, which assumed monolayer adsorption, BET isotherm
extended Langmuir’s application to include a multilayer adsorption.

With several key assumptions, such as homogeneous surface, no lateral interactions between
molecules, and number of layers becoming infinite at saturation pressure, BET isotherm is
considered a better fit to describe the adsorption processes in shale gas reservoir. BET
isotherm occurs in a non-porous or a macroporous material (Kuila and Prasad, 2013). The
general form of the BET isotherm can be given as

  P
n
 P 
n 1

1   n  1    n   
P
VmC
PO   PO   PO   ,………………………………….(2.2)
V  P   n 1 
P
1  1   C  1 P  C  P  
PO    
PO  PO 

Vm = maximum adsorption gas volume when the entire absorbent surface is being covered
with a complete monolayer

C = constant related to the net heat of adsorption

PO = saturation pressure of the gas

n = maximum number of adsorption layers.

When n =1, the equation will be reduced to the Langmuir isotherm and when n =  , the
equation reduces to

VmCP
VL  ………………………………………………(2.3)
  C  1 P 
 Po  P  1  
 Po 

The BET isotherm has been described as successful only for a limited range of data and has
also been found to be inaccurate when used to make mixture predictions (Danner and
Wenzel, 1969; Kaul, 1984).

15
Yu et al., (2015) investigated the contribution of adsorption on Marcellus shale by analysing
several experiments of methane adsorption and found that gas adsorption in Marcellus shale
deviated from the Langmuir isotherm but obeyed the BET isotherm. Although the BET
matched the experimental data more accurately than the Langmuir model, other adsorption
models, since proven to be better fits to shale/coal experimental data sets such as Dubinin-
Astakhov (D-A) and the Vacancy Solution Model (VSM), were not tested. Figure 2.2 shows a
comparison between Langmuir and BET models for gas adsorption.

Figure 2.2: Comparison of Langmuir and BET models for gas adsorption

2.2.3 Potential Theory Approach

The Potential theory concept, first introduced by Polanyi in 1914, explains that the surface
force field is represented by an equipotential contour above the surface, and that the space
between each contour corresponds to a definite adsorbed volume. Thus the cumulative
volume of the adsorbed space, W , is a function of the chemical potential of the gas,  (Yang,
1997) ( See figure 2.3). According to this theory, adsorption can be measured through the
equilibrium between the chemical potential of a gas near the surface and the chemical
potential of the gas from a large distance away. In terms of application, the Polanyi theory

can be applied at much higher P / Po (between 0.1-0.8 MPa) compared with that of BET
(between 0.05-0.35 MPa)

W  f    ………………………………………(2.4)
16
Figure 2.3: Schematic representation of adsorption by Polanyi potential theory (modified
from Yang,1997)

Polanyi referred to this function as the characteristic curve; he assumed that the adsorption
potential is independent of temperature. Thus, it is possible to predict the adsorption of gases
onto the solids at different temperatures once the characteristic curve has been defined at
one temperature. The characteristic curve, therefore, represents the relationship between
the adsorption potential and the distance from the solid surface.

The adsorption potential is related to the vapour pressure by

P 
  RT ln  o  ………………………………………….. (2.5)
P

17
2.2.4 Dubinin – Radushkevich (D-R) and Astakhov (D-A) Equations
One of the widely used equations for describing experimental data of the adsorption of gases
on microporous solids is the D-R equation. This equation was proposed by Dubinin and
Radushkevich for solids with homogeneous structure of micropores with later extensions to
non-homogeneous microporous structures by Dubinin –Astakhov equations.

The D-R adsorption model is a pore-filling model which does not assume monolayer surface
coverage (Sakurovs et al., 2007) ( See figure 2.4). Dubinin and Radushkevish’s theory of
micropore volume filing was based on the Polanyi concept of a characteristic curve (Gil and
Grange, 1996). Dubinin used adsorption of zeolites to prove his theory of volume filling
mechanism (Do, 1998).The D-R equation is expressed as

  RT P 2 
W  Wo exp    ln s   ……………………………….(2.6)
   E P  

W = volume adsorbed volume,

Wo = micropore volume

E = energy of adsorption

 = affinity of the sorbent for the gas.

A more general form of the equation is the D-A equation written in the form

  RT P m 
W  Wo exp    ln s   ……………………………………………(2.7)
   E P  

When m is equal to 2, the D-A equation reduces to the D-R equation. The additional
parameter m allows for some flexibility of modelling (Clarkson and Haghshenas, 2013;
Kapoor et al., 1990), compared with the two-parameter D-R equation.

Dubinin equations can be used to represent adsorption data at different temperatures. The
characteristic energy is independent of pressure, and a plot of the fractional loading versus
the adsorption potential for different temperatures will collapse into one curve, called the
characteristic curve (Do, 1998). Talu and Myers (1988) noted that the D-R equation is not

18
thermodynamically consistent in Henry's law region (limit of zero loading). When the pressure
reduces to zero, the D-R equation is not able to return to Henry’s Law.

Kapoor et al., (1989) modified the D-R equation to include Henry’s law limits that are
applicable to pressure ranges from 0 to saturation pressure. The modification of the D-R
equation makes it useful for prediction of multicomponent adsorption equilibria.

Figure 2.4: monolayer (top) and pore filling mechanism (bottom) of Langmuir and Dubinin
models (Clarkson and Hagshenas, 2013)

The D-R method provides a better fit to adsorption data than the Langmuir method, even
though they both have the same number of parameters (Clarkson and Bustin, 2000).

According to Sakurovs et al., (2007), D-R cannot be used to model the adsorption of gases at
temperatures or pressures where the gases are supercritical and the saturation pressure is
undefined. For instance, methane’s critical temperature is 190.6K and hence in shale gas
reservoirs, methane is expected to be in the supercritical state. Methane will therefore not
exhibit a saturated vapour pressure under supercritical conditions.

Sakurovs et al., (2007) proposed a general DR equation that replaced the saturation pressure
term with an adsorbed density and gas density term in order for the isotherm to be applied
to supercritical conditions. The modified isotherm can, therefore, be applied to a much wider

19
temperature and pressure range. Sakurovs et al., (2007) used experimental data from
Australian bituminous coals to fit the modified D-R equation.

2.3 Multicomponent Adsorption Models

Many models have been proposed for modelling the prediction of mixed gas adsorption on
coal and shale. These models usually use single component isotherm data to predict
multicomponent adsorption. Two of the simplest models used are the Extended Langmuir
Model (EL) and the Ideal Adsorbed Solution Theory (IAST). However, EL models are found to
be the simplest model in predicting multicomponent adsorption.

Due to the complexities of multi-component adsorption isotherm tests, such as difficulties in

measurement and time consumption, predictions are rather made from single component
adsorption isotherms, and the adsorption capacities of each component in the mixture are
predicted at a temperature and pressure. This process, according to Wang (2016), involves
using the single component adsorption model for fitting the experimental data and to obtain
model parameters. The parameters are then applied to the prediction models to calculate the
multicomponent mixture adsorption equilibrium. Figure 2.5 shows the procedure for
obtaining binary prediction from pure component modelling.

20
Figure 2.5: Illustration showing prediction of binary gas adsorption from pure component
isotherms (Modified from Clarkson and Bustin, 2000)

21
2.3.1 Extended Langmuir Model
An extension to the Langmuir isotherm was developed by Markham and Benton ( 1931) and
called the Extended Langmuir Model. This model has been used widely in the prediction of
multicomponent adsorption. Most commercial reservoirs prefer the use of this model, due
to its simplicity and ease of use. This model extends the Langmuir model to include a multi-
component system, by taking into consideration the partial pressures and molar composition.

One of the main critiques of this model is the issue of thermodynamic inconsistency. A
thermodynamically consistent model implies that the sorption limit must be equal for all the
components (Clarkson and Haghshenas, 2013). Therefore, the Langmuir sorption constants
for the pure components must be equal for thermodynamic consistency (Arri et al., 1992).

The equation below is used to represent the EL model for multi-component systems ( Arri et
al., 1992)

n VLi ( Pg yi )
Va   n
……………………………………………..(2.8)
1
i 1
PLi   ( y j Pg )
j 1 PLj

VLi = Langmuir volume constant for pure component I, (SCF/Ton)

PLi = Langmuir pressure constant for pure component i , (psia)

yi = Gas phase composition of component I, (fraction)

Pg = Gas phase pressure, (psia)

Vi = Adsorbed volume of component I, (SCF/Ton)

The EL method works well especially in CBM reservoirs when the compositional changes are
minimal. The majority of the gas in such reservoirs is methane and there is little effect of
compositional changes. However for shale gas reservoirs, due to compositional changes as a
result of primary depletion and enhanced recovery phases, the use of EL method will fall short
of accurately description of the desorption process. According to Manik et al., (2002), the EL
method is a purely empirical formulation that ignores thermodynamic equilibrium between
gas components in the free and adsorbed gas phases.
22
Arri et al., (1992) modified a reservoir simulator to use EL isotherm for describing the
equilibrium relationship between the free and sorbed gas. By measuring the sorption of
various binary gas compositions at different pressures and fixed temperature on coal samples,
they were able to show that the use of EL isotherm provided a reasonable correlation to the
data. However Arri et al., (1992) realised that at very high pressures, the EL isotherm loses its
accuracy and under-predicted the binary sorption data. The extended Langmuir equation is
found to be inadequate overall in predicting adsorption equilibrium, even though it offers
simplicity in adsorption calculations (Tien, 1994).

2.3.2 Ideal Adsorbed Solution Theory (IAST)

Ideal adsorbed solution theory can be used to predict binary adsorption equilibrium for
various mixtures from pure component adsorption data. This theory was first proposed by
Myers and Prausnitz in 1965. For multi-component adsorption prediction, it has quickly
established itself as one of the favoured methods. One key assumption under which the IAST
was derived is that the adsorbed mixture behaves like an ideal adsorbed solution. This is
similar to Raoult's law for a bulk solution.

pyi  poi ( ) xi ………………………………………………(2.9)

p o i Is the vapour pressure of the pure component I, at the same spreading pressure and
same temperature T, as the adsorbed mixture.

xi = sorbed phase gas mole fraction

 = spreading pressure, where the spreading pressure is defined as the reduction in the
surface tension of the surface as the adsorbate spreads over the surface (Ruthven, 1984). The

relationship between p
o
i and i is expressed as

pi o
A na ( p)
 i*  i 
RT 0
p
dp ……………………………………….(2.10)

na ( p) = pure component isotherm

A = specific surface area of the adsorbent.

23
The condition below needs to be satisfied for both adsorbed mole fractions and mole
fractions of the free gas.

y
j 1
i  1 …………………………………………………(2.11)

x
j 1
i  1 ………………………………………………………..(2.12)

Total adsorbed gas in the mixture is given as

nc
1 x
  io ,……………………………………………….. (2.13)
nt i 1 n
i

The amount of each component adsorbed in the mixture is given as

ni  nt xi ……………………………………………………… (2.14)

Any pure component adsorption isotherm could be used in the above equation to evaluate
the spreading pressure. Clarkson and Bustin (2000) argued that for IAS fitting to experimental
data, a more accurate pure component isotherm should be applied.

The successful calculations of IAS depend on the correct fitting of the single component data
in regions of low pressure and at very high pressures where pure hypothetical pressure lies
(Do, 1998). Errors in these regions would lead to significant errors in multicomponent
calculations.

One key advantage of the use of this theory is the fact that it can allow any type of single
component isotherm to be used for the prediction of the multicomponent equilibria, as long
as it fits the experimental data over the range of pressures (Richter et al., 1989).

Several researchers (Hall et al., 1994; Stevenson et al., 1991) have found that when compared
with the extended Langmuir isotherm, the IAS theory is more accurate for mixed gas
adsorption, especially on dry coal .

The model does not require the use of any mixture data and is independent of the actual
mode of physical adsorption (O’Brien and Myers, 1985). The theory is based on the

24
hypothesis that both the free gas and adsorbed gas phase equilibrium is similar to the vapour-
liquid equilibrium (Manik et al., 2002).

There are several models which have been proposed for modelling adsorption that can be
said to be either thermodynamically consistent or inconsistent. Most of these models have
been utilised in coalbed methane reservoirs; more recently, a lot of these models have been
applied to shale gas reservoirs.

For ideal and non-ideal gas mixtures, as well as homogeneous and heterogeneous surfaces,
IAST method has been found to be the preferred method in studying adsorption equilibria,
due to the fact that it is thermodynamically consistent (Ahmadpour et al., 1998; Chen et al.,
2011).The equation below describes the thermodynamic equilibrium between adsorbed and
free gas phases, using the ideal adsorbed phase assumption (Rajput, 2016). Fugacity may be
used to account for non-ideality of the gas phase where the activity coefficient is considered
to be greater than one.

Pg yii  Pi oi o xi a ……………………………………………. (2.15)

Where

Pg = gas phase pressure (psia)

yi =Gas phase composition of component i (fraction)

i = Fugacity coefficient of pure component i in the gas phase, (dimensionless)

Pi o = Standard state pressure of pure component i in the gas phase (psia)

i o = Fugacity coefficient of pure component i in the gas phase at standard condition

(dimensionless)

xi a = Molar composition of component i in adsorbed phase (fraction)

Chen et al., (2011) studied adsorption equilibria of pure methane and ethane gases and their
binary mixtures on activated carbon by comparing their isotherm data, using the sips equation
for pure component isotherm and the extended sips equation, as well as IAST, for binary
25
equilibria. They found out that IAST performed better when predicting data on highly
heterogeneous carbon samples.

Successful calculations of the IAS require single component data to be fitted accurately for
low pressure and high-pressure regions where the hypothetical pressure lies. According to Do
(1998), large errors will occur in multicomponent calculations if these regions are in error.

2.3.3 Vacancy Solution Model

The vacancy solution model by Suwanayuen and Danner (1980) treats the adsorbed phase as
a mixture of adsorbed species and their vacancies (Kaul, 1984). That is, it assumes two
solutions in the system, made up of the gas phase and the adsorbed phase. The surface is
considered to be made up of a vacancy (species v ) and adsorbed species (species 1) (Yang,
1997). The vacancy is defined as “vacuum entity occupying a space that can be filled by an
adsorbate molecule” (Kaul, 1984) . The vacancies are an imaginary entity with the same size
as the adsorbate. In order to account for the non-ideality of the system, the activity coefficient
obtained from pure component data is used. The VSM has been found to be applicable to all
gas adsorption systems (Suwanayuen and Danner, 1980); in view of this, its application could
be said to be suitable for shale and coal bed methane systems. There are however limited
applications of this model in shale gas systems.

Using Wilson equation to define the activity coefficient, the Wilson – VSM isotherm equation
can be obtained for the single component as

 n    1  (1   v1 )    v1 (1   v1) (1  1v ) 
P 1   1v  exp     ………… (2.16)
 b1 1     1v  (1   v1 )   1  (1   v1 ) 1v  (1  1v ) 

With the provision of adsorption data, the four parameters n1 , b1 , 1v , v1 could be
determined through non-linear regression analysis. P refers to pressure whilst  is the
fractional coverage.

The activity coefficient for i and v is given by the Wilson equation below

 
   X i  ik 
ln  k  1  ln   X j  kj      ……………………………. (2.17)
 j  i   X j  ij 
 j 

26
Vacancy solution model can be used for both single component and multicomponent
adsorption isotherms.

Using a set of data on activated carbon and zeolites, Suwanayuen and Danner (1980)
successfully predicted adsorption phase diagrams for binary mixtures, using the VSM. By
ignoring the adsorbate-adsorbate interactions, excellent predictions were obtained for binary
adsorption of light hydrocarbons such as methane and ethane.

VSM has proved to be successful when applied to predict multicomponent adsorption

isotherms, especially for high surface coverage interactions. This was demonstrated by Kaul
(1984) when he reviewed different isotherm models in predicting both single and
multicomponent adsorption isotherms. For low coverage multicomponent data prediction,
Kaul (1984) found that both IASM and VSM were sufficient; however, the VSM provided
accurate predictions of high coverage multicomponent data from both single and binary data
using different adsorbate on two different adsorbents involving molecular sieve and activated
carbon. Low surface coverage implies that the adsorbate molecules are less likely to interact
with each other and hence the adsorbed phase can be treated as ideal, whereas the opposite
occurs for high surface coverage.

The vacancy solution theory was criticised by Talu et al., (1988) for its inconsistency in the
binary selectivity predictions in Henry’s law limit. Bhatia and Ding (2001) noted that the error
in the model by Suwanayuen and Danner (1980) was the result of the incorrect treatment of
vacancies in the surface equation. They proposed a thermodynamically consistent approach
that is instead based on the mass-action principle.

For multicomponent adsorption calculations, the general form of the VSM is given as

nis ,  i 3 a 
i yi P   x n s , exp( 3i  1) exp  i  ………………………….(2.18)
s s
i i m
nm bi  RT 

The fugacity coefficient is set to unity for gas adsorption at moderate pressures. This

equation is normally solved by trial and error to obtain yi and nms .

27
Table 2.1: Summary of gas adsorption models used in coal/shale gas reservoirs

MODEL DESCRIPTION ADVANTAGES/LIMITATION REFERENCE

Langmuir First developed by Irving Langmuir in It cannot be applied to the real system due to shale Langmuir 1916.
Isotherm 1916. The model postulates that gas systems having heterogeneous surface however
behaves as an ideal gas under isothermal good approximations have been found for
conditions. Key assumption being a microporous materials. It assumes that the Langmuir
homogeneous surface and no interaction volume is temperature invariant but this is not the
between neighbouring sites. Monolayer case. At higher pressures, the model fails to apply
coverage is also assumed. adequately.

BET Model Developed by Brunaur, Emmet and Teller Not applicable for multicomponent systems can only Brunauer ,Emmet
in 1938.This model assumes that the be used at a single temperature. It is only suitable for and Teller 1938
adsorption layers on the surface of the ideal systems.
organic carbon were infinite. Unlike
Langmuir, multilayer adsorption coverage
is assumed. It has proved useful in the
determination of surface area of
materials.

28
Potential Introduced by Polanyi in 1914. It assumes
Mostly applicable at higher P/Po, unlike Langmuir Polanyi 1914
Theory gas molecules move according to and BET. P=pressure, Po =vapour pressure. The
potentials like that of gravity or electric
characteristic curve allows for successful prediction
field. of adsorption of gases at different temperatures.
Makes an assumption about ideal gas. Used in the
Uses the concept of the characteristic
design of adsorption systems and for binary gas
curve, which is independent of
predictions under different temperatures, thereby
temperature. Adsorption is measured
saving time and cost of experiments.
through an equilibrium between the
chemical potential of a gas near the
surface and that of the gas from a
distance away.

D-R and D-A Used for describing experimental data of It is useful for representing assumption at different Dubinin (1975)
Model the adsorption of gases on microporous temperature.it has also being found to be
solids.it is a pore filling model which does thermodynamically inconsistent in Henry’s region.
not assume monolayer coverage

MULTI-COMPONENT MODELS

29
Extended An extension of the Langmuir model This model has been criticised for not being Markham and
Langmuir developed by Markham and Benton 1931 thermodynamic consistent. For thermodynamic Benton (1931)
Model for binary gas adsorption. This model consistency, the sorption limit must be equal for all
considers the partial pressures and molar the components. Most preferred model for use due
composition of the adsorbates in its to its simplicity and ease of use in commercial
calculation. Purely empirical formulation simulators
that ignores thermodynamic equilibrium
between the gas components in the free
and adsorbed gas phase.

Ideal Adsorbed First proposed by Myers and Prausnitz in Does not require the use of any mixture data and it Myers and
Solution 1965 for the prediction of binary is independent of the actual mode of physical Prausnitz (1965)
adsorption equilibrium for various adsorption. It can allow any type of single
mixtures from pure component component isotherm to be used for the prediction of
adsorption data. Derived from the fact multicomponent equilibria. It however assumes an
that the adsorbed gas behaves like an ideal gas system for its formulation. Requires
ideal adsorbed solution similar to Raoult's rigorous thermodynamic calculations
law for bulk solutions.

Vacancy Suwanayuen and Danner developed the Has been criticised by Talu et al 1988 for its Suwanayuen and
Solution Model vacancy solution model which inconsistency in the binary selectivity prediction in Danner (1980)
basically treats the adsorbed phase as a Henry's law limit. Suitable for predicting both single
mixture of adsorbed species and their and multi-component adsorption isotherms. Kaul
vacancies. The vacancies are an imaginary 1984 found that it was insufficient for low coverage
entity with the same size as the multicomponent predictions. Very successful for
adsorbate. Accounts for non-ideality of multi-component predictions at high surface
the system by introducing an activity coverage interactions.
coefficient.

30
TEMPERATURE DEPENDENT MODELS

Exponential An improved Langmuir model used in Capable of predicting a wide range of gas adsorption Ye ,Chen,Pan,Xia
Model evaluating gas adsorption capacity of at different temperatures with no need to have and Ding (2016)
shale under various pressures and different Langmuir volume constants at different
temperatures. The constant Langmuir temperatures. Only applicable to single component
volume has been modified to be a systems and upon the assumptions of an ideal gas.
function of temperature

Bi-Langmuir Also an improved Langmuir model; Useful for predicting gas adsorption at several Lu,Li and Watson
Model describes gas adsorption on an adsorbent temperatures. Takes into account the heterogeneity (1995)
as having two discrete sharp peaks of of the adsorbent by accounting for gas adsorption on
adsorption energy distribution. One term the clay minerals and kerogen. The Langmuir volume
describes gas adsorption on clay minerals constant is still assumed to be temperature
while the other accounts for gas independent just like the Langmuir isotherm. Only
adsorption on kerogen. The authors useful for single component system
noted that temperature dependence is
often ignored in the Langmuir equation
even though it plays an important role in
determining the amount of gas adsorbed
on shale.

D-R and D-A Used for describing experimental data of It is useful for representing assumption at different Dubinin(1975)
Model the adsorption of gases on microporous temperature.it has also being found to be
solids.it is a pore filling model which does thermodynamically inconsistent in Henry’s region.
not assume monolayer coverage

31
2.4 Methodology
2.4.1 Procedure for the Estimation of Parameters
Several nonlinear regression techniques have been used to identify the optimal isotherm for
gas adsorption on various adsorbents. These nonlinear regression methods involve
minimizing the error distribution between the experimental data and the predicted results
(Kumar, 2006). For the selection of optimum isotherm, the use of nonlinear regression has
been found to be the best (Ho, 2004). The method of least squares, is a key technique used
by many researchers in predicting the optimum isotherm (Kumar, 2006; Kumar et al., 2008)
and has also been found to be the most widely used (Kumar, 2007; Kumar and Porkodi, 2006).
Some other widely used nonlinear regression methods include hybrid fractional error function
(HYBRID), Marquardt’s percent standard deviation (MPSD), sum of the squared (SSE), sum of
the absolute errors (SAE) and absolute relative error (ARE) (Foo and Hameed, 2009; Ho, 2004;
Kumar, 2006; Kumar et al., 2008; Porter et al., 1999; Wong et al., 2004). In this study Excel
solver have been used for non-linear regresion with an objective function defined based on
the formula of the error function. This approach is highly dependent on the value of the initial
guess chosen. To improve the accuracy of the selection of the initial guess, several
computations have been done for the same exercise to gain some confidence on the choice
of parameters obtained. The second approach has been to compare the result of the
parameters obtained for other adsorption models using similar error function and if they are
close to each other , to conclude on the accuracy of the final parameters obtained and of the
choice of initial guess chosen.

2.4.2 Error Function Analysis

Average Relative Error Function (ARE)
Absolute average deviation (AAD%) or Average relative error (ARE) has been used extensively
by various researchers to find the optimum model for a set of adsorption data on coal and
shale (Charoensuppanimit et al., 2015; Clarkson, 2003; Clarkson et al., 1997; Clarkson and
Haghshenas, 2013; Kapoor et al., 1989). Even though it has been reported to have a tendency
to under- or overestimate the experimental data, it minimizes the fractional error distribution
across the sample concentration range (Foo and Hameed, 2009; Kapoor and Yang, 1989).

32
  N  (nical  ni exp )  
   n  
  i exp   
% AAD  100* ABS  i 1   …………………… (2.19)
 N 

 


nical = calculated adsorbed concentration,

ni exp = experimental adsorption data,
N = number of data points for the isotherm.

Sum of Squared Error Function (SSE)

This is probably the most common of all the error functions. It is given as

2
 N 
SSE    nical  ni exp  …………………………(2.20)
 i 1 

Sum of Absolute Error Function (SAE)

This is very similar to Sum of Squared Error and is given as

N
SAE   | nical  ni exp | ……………………………… (2.21)
i 1

Marquardt’s Percent Standard Deviation (MPSD)

This is often seen as an ideal error function in most adsorption studies (Foo and Hameed,
2009; Sreńscek-Nazzal et al., 2015). It can be expressed as

2
1 N  nical  ni exp 
MPSD  100 
N  r i 1  ni exp
 …………….. (2.22)


Hybrid Fractional Error Function (HYBRID)

At low-pressure values, the HYBRID function improves the overall fitting of the model to the
experimental data, compared to some of the other error functions, such as SSE.

100 N   nical  ni exp  

 2

HYBRID  
N  r i 1  ni exp 
……………………….. (2.23)
 

33
Sum of Normalised Error (SNE)
Non-linear regression is mostly preferred to linear regression due to inherent bias resulting
from linearization (Ho et al., 2002; Porter et al., 1999). Porter et al., (1999), proposed the use
of statistically robust method of sum of normalised error. They argued that due to the
different set of isotherm parameters produced by the different error criteria, results can be
obtained by finding normalised results for each parameter set for each isotherm model and
combining them. The procedure involves obtaining the value of errors for each error function
for each set of isotherm constants and dividing by the maximum errors for that error function
(Nazzal et al., 2015). Each parameter is obtained by minimising the error functions across the
gas pressures, by using Microsoft Excel Solver add-in.

Sum of Normalised Error Procedure

The sum of normalised errror procedure has been summarised in steps 1 to 10 of Figure 2.6.
in Appendix 2 , a sample calculation is provided using the steps shown in Figure 2.6.

Figure 2.6 : Procedure for calculating SNE

34
2.5 Experimental Data
Obtaining experimental data for shale gas modelling is a challenging task due to the low
adsorption capacities and the range of pressure and temperature under which experiments
are conducted (Sandoval et al., 2017). Reports of inconsistent results from measurements
observed at higher pressures, and the lack of available quality database, have also contributed
to challenges in this area. In particular, very limited experimental binary gas adsorption data
on shale are available. Due to this scarcity of available data, adsorption capacity on activated
carbon is used by many researchers to model the performance of an adsorption model in
predicting binary mixtures of gas adsorption capacities on shale. Fitzgerald et al., (2006) argue
that experimental uncertainty in the use of activated carbon is lower compared to coal, which
has a similar structure to shale. Also, because coal/ shale has a more complex structure pore
to activated carbon, the adsorption on activated carbon can serve as a reference for more
complicated adsorption/desorption on coal/shale (Ren et al., 2017).

Model comparison is essential in the study of their capabilities and limitations when fitting
pure component isotherms, as well as for predicting multi-component systems. Adsorption
data about Methane, Carbon dioxide and Nitrogen on New Albany shale have been obtained
from the literature (Chareonsuppanimit et al., 2012) to model single component gas
adsorption. New Albany shale is one of a number of organic-rich shales of the upper Devonian
and lower Mississippian age in North America. Its formation consists of brown , black and
green shale with minor beds of dolomite and sandstone (Chareonsuppanimit et al., 2012).
The formation is also thought to be a major source of oil shale. The adsorption data for New
Albany shale have been provided in Appendix 2. The results obtained for the single
component modelling is based on these data. For the purpose of showing the performance
of the different binary adsorption models, data about activated carbon from Szepesy and Illes
(1963) have been used. For temperature dependent models, pure component data about
shale obtained at several temperatures in Green River shale (Zhang et al., 2012) have been
used to evaluate the different models and their performance. Data relating to the Green River
shale have been provided in Appendix 2.

35
2.6 Results and Discussion
2.6.1 Single component modelling of New Albany Shale
Table 2.2-2.5 the error analysis involving the use of SNE. The values obtained by the use of
SNE have been compared to identify parameters of the isotherms that can provide the most
accurate match to the measured data on New Albany shale. The bold numbers represent the
minimum SNE for each of the isotherms and their associated optimum parameter set for
different gas adsorption is shown in Appendix Table 11-16.

The Langmuir isotherm parameters for New Albany shale for Methane, Carbon dioxide and
Nitrogen were obtained using non-linear regression technique and shown in Table 11-16 of
Appendix 2. Similar values can be easily observed for the different error functions used in that
analysis. The SNE values are very similar for the different gas adsorption, with the exception
of methane, which showed a much higher SNE for SAE compared with the other error

functions. Also, the Langmuir parameter constants VL and b are quite similar in magnitude.
Overall, we can see that the Langmuir isotherm provided a good fit to the New Albany
datasets. ARE provided the best match parameters for Langmuir isotherm of Methane and
CO2, whilst SSE gave the best match for the case of nitrogen adsorption using Langmuir
isotherm. Overall, the Langmuir isotherm is recommended for modelling the experimental
data for New Albany shale for methane and CO2. As can be observed in Figure 2.7, the
Langmuir isotherm fits very well with the experimental data, regardless of which error
function is chosen. The BET isotherm constants and error analysis using the different error
functions are shown in Table 2.3 and in Appendix 2 Table 2.12 for methane, CO2 and nitrogen
adsorption. It can also be observed that the parameters for BET for all the error functions are
very similar, with only slight variations. Likewise, the SNE values are very much similar.
Comparing the SNE values, it can be concluded that SSE for methane and ARE for CO2 and
nitrogen provided the best BET fit for the experimental data for New Albany shale.

Non–linear modelling of D-A equation with different error functions also showed similar
values for the D-A constants (see Appendix 2 Table 2.13). However, for nitrogen adsorption,
MPSD showed a much higher value for Wo compared with the rest of the error functions.

Comparing the SNE also shows that ARE is a much better fit for methane gas adsorption using
D-A, whereas HYBRID error function showed the closest fit for CO2 adsorption and nitrogen
36
adsorption. Observing Figures 2.6 - 2.9 shows that D-A model was however not the best for
modelling methane, CO2 and nitrogen adsorption on New Albany shale.

Appendix 2 Table 2.14 and Table2.5 presents the VSM parameters and the error analysis using
different error functions. Overall, just like previous adsorption models, the model parameters
for VSM were very similar to all the error functions used. In terms of the SNE comparison, it
is noticeable that SSE showed consistently higher values for methane, CO2 and nitrogen
adsorption on shale, whereas the remaining SNE for the other error functions were generally
quite similar. MPSD was the best error function to be used for methane adsorption, whilst
SAE was found to be best for CO2 adsorption. ARE was also found to be most suited for
modelling nitrogen adsorption using VSM on New Albany shale.

Of the overall 12 different results for SNE calculations, SSE and Hybrid produced parameter
sets showing the minimum sum of normalised error in only 2 of the results. ARE provided 6
minimum SNE out of the total 12, proving to be the most consistent error function to be used
in shale gas adsorption modelling. Finally, SAE and MPDS provided only a single result showing
minimum SNE for all the results generated. From observation of Tables 2.2 - 2.5 and Figures
2.7 - 2.10, it can also be deduced that Langmuir isotherm provides the best overall fit for the
data on all of the gas adsorption involving methane and CO2, with the exception of nitrogen
adsorption, where D-A proved to be the best fit for the experimental data. Adsorption
modelling for CO2 was difficult to achieve especially at very higher pressures. This implies
that, the different adsorption models might struggle to represent accurately CO2 adsorption
in the New Albany shale. Figure 2.8 confirms that BET isotherm cannot be used for modelling
CO2 adsorption on New Albany shale data, even though there was a better match for methane
and nitrogen adsorption using BET. Due to the inherent heterogeneity of shale which might
be different for different shale reservoirs, data obtained for one shale should not be assumed
to be representative for all shale reservoirs. In that respect, the results of this thesis are only
applicable to the data obtained about shale from these reservoirs. These data have been
provided in Appendix 2. Adsorption models might predict differently with different data.

37
Table 2.2: SNE for Non-Linear Langmuir isotherm

SSE ARE SAE MPSD HYBRID

Methane
SNE 2.0456 2.0083 5 2.1427 2.2411
Carbon Dioxide
SNE 4.2552 3.6307 4.0848 4.2752 4.2320
Nitrogen
SNE 3.9504 4.0152 4.1666 4.0152 4.2265

0.14

0.12
Adsorption amount, mmol/g

0.1

0.08

0.06

0.04

0.02

0
0 2 4 6 8 10 12 14
Pressure, Mpa
Methane SSE HYBRID CO2
SAE MPSD Nitrogen ARE

Figure 2.7: Experimental data for methane, CO2 and nitrogen for Langmuir isotherm
obtained by SSE, ARE, SAE, MPSD and HYBRID

38
 0.14

0.12
adsorbed amount,mmol/g

0.1

0.08

0.06

0.04

0.02

0
0 2 4 6 8 10 12 14
Pressure,Mpa
Methane CO2 Nitrogen SSE
ARE SAE MPSD HYBRID

Figure 2.8: Experimental data for methane, CO2 and nitrogen for BET isotherm obtained by
SSE, ARE, SAE, MPSD and HYBRID

0.14

0.12
Adsorbed amount,mmol/g

0.1

0.08

0.06

0.04

0.02

0
0 2 4 6 8 10 12 14
Pressure,Mpa

Nitrogen SSE ARE SAE

MPSD HYBRID CO2 Methane

Figure 2.9: Experimental data for methane, CO2 and nitrogen for Dubinin Astakhov obtained
by SSE, ARE, SAE, MPSD and HYBRID

39
 0.14

0.12
Adsorbed amount, mmol/g

0.1

0.08

0.06

0.04

0.02

0
0 2 4 6 8 10 12 14
CO2 SSE ARE SAE
Pressure,Mpa
MPSD HYBRID Nitrogen Methane

Figure 2. 10: Experimental data for methane, CO2 and nitrogen for VSM obtained by SSE,
ARE, SAE, MPSD and HYBRID

40
Table 2.3: SNE for Non-Linear BET

SSE ARE SAE MPSD HYBRID

Methane
SNE 4.668 4.7349 4.7410 4.8365 4.7780
CO2
SNE 4.9481 3.9526 4.8955 4.9129 4.9076
Nitrogen
SNE 4.0081 3.9433 4.2248 3.9650 4.1983

Table 2.4: SNE for Non-Linear Dubinin-Astakhov isotherm

SSE ARE SAE MPSD HYBRID

Methane
SNE 3.5769 3.5619 4.7807 3.5619 3.8472
CO2
SNE 4.3903 4.5538 4.7416 4.5539 4.3059
Nitrogen
SNE 3.4763 3.9055 3.8840 3.9055 3.14481

Table 2.5: SNE for Non-Linear Vacancy solution model

SSE ARE SAE MPSD HYBRID

Methane
SNE 4.6122 4.1126 4.4328 4.1071 4.1482
CO2
SNE 4.6145 3.9830 3.6623 3.9830 3.7010
Nitrogen
SNE 4.9796 3.98022 3.9921 3.9884 4.0611

41
3.2 Multi-Component Modelling of Ethane and Methane Adsorption on
Activated Carbon (Szepesy and Illes, 1963)
Adsorption data for methane and ethane on activated carbon at a temperature of 293.15 k
have been reported by Szepesy and Illes (1963). Pressures for pure component adsorption
data were extended up to 124 KPa. The use of pure adsorption models to represent
adsorption data is significant when modelling shale gas reservoir simulation. Very often,
Langmuir equation has been used to represent pure adsorption data due to the ease with
which it can represent the data, and also because of its use in numerical reservoir simulators.
In order to conduct binary gas adsorption modelling, results from pure components are used
to obtain adsorption prediction. For this study, VSM and Langmuir's isotherm has been used
to conduct the pure component adsorption modelling, and later used in carrying out the
binary adsorption prediction. The results from single component modelling using Langmuir
and VSM are shown in Tables 2.6 and 2.7.

Table 2.6: SNE for binary mixture of methane and ethane (Langmuir isotherm)

SSE ARE SAE MPSD HYBRID

Methane
SNE 4.9668 4.9169 4.9358 4.9358 4.9235
Ethane
SSE ARE SAE MPSD HYBRID
SNE 4.2434 4.5727 4.3382 4.5728 4.2024

In both Appendix 2 Table 2.15 and 2.16, the individual Langmuir and VSM parameters are very
similar, irrespective of the error function that was used. For methane adsorption, the sum of
normalised error was minimum for ARE compared to the rest of the error function when
Langmuir isotherm was used as the adsorption model (See Table 2.6). ARE also proved to be
the best fit for ethane adsorption when modelling with VSM (See Table 2.7). The HYBRID error
function was however found to fit the model better when using Langmuir isotherm, whilst it
was found to be a worse fit for methane adsorption using VSM. MPSD was the most
appropriate error function to be used in methane adsorption when using VSM. The results
from Tables 2.6 and 2.7 also indicate that VSM would be the preferred adsorption model to
be used in modelling single component adsorption for this data set.

42
Table 2.7: SNE for binary mixture of methane and ethane (Vacancy solution model)

SSE ARE SAE MPSD HYBRID

Methane
SNE 1.7534 1.6184 1.6436 1.6092 5
Ethane
SSE ARE SAE MPSD HYBRID
SNE 4.0883 3.4805 3.5592 3.6697 3.6697

Binary gas adsorption modelling can now be undertaken once pure component adsorption
model fitting has been completed. Throughout this study, EL, IAS and the VSM have been
used. The different multi-component adsorption models have already been discussed in the
previous section. Figures 2.10 and 2.11 show the predictions of the binary gas-phase diagrams
for each of the models at a pressure of 101 Kpa. The phase diagrams shown in Figures 2.10
and 2.11 show the plots of mole fraction of both ethane and methane in the adsorbed phase
versus mole fractions in the free gas phase (non-sorbed), while Figure 2.12 is the plot of mole
fraction of methane in sorbed phase versus the total gas adsorption. The predictions show
that each of the multi-component adsorption models was able to fit closely to the
experimental binary adsorption data. The Extended Langmuir showed the worst fit compared
with both the Ideal adsorbed solution and the Vacancy solution model. The Vacancy solution
model was, however, able to fit more accurately for predicted equilibrium compositions in
Figures 2.10 - 2.12. EL predicted more ethane in the sorbed phase in Figure 2.10 and more
methane in the free gas phase, as shown in Figure 2.11 more than what the experimental
data showed.

43
 0.8

0.7
Mole fraction of Ethane in Sorbed

0.6

0.5

Data
0.4
EL

0.3 IAS
VSM
0.2

0.1

0
0.4 0.5 0.6 0.7 0.8 0.9 1
Mole fraction of Ethane in Gas Phase

Figure 2.10: Predicted equilibrium composition diagram showing free gas phase versus the
sorbed phase for Ethane

0.9
Mole fraction of Methane in Sorbed

0.8

0.7

0.6
Data
0.5
EL
0.4
IAS
0.3
VSM
0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6
Mole fraction of Methane in Gas Phase

Figure 2.11: Predicted equilibrium composition diagram showing free gas phase versus the
sorbed phase for methane

44
 3.5

2.5
Ntotal mmol/g

2 VSM
Data

1.5 IAS
EL

0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6
X

Figure 2.12: Total volume of mixtures adsorbed at pressures of 101 Kpa for Case 1

For model calculations of EL and IAS, the free gas phase compositions have been inputted and
the adsorbed phase mole fraction predicted. However, for VSM, the adsorbed mole fraction
was inputted and the free mole gas phase was calculated. To be able to express the relative
adsorption of components within an adsorption system, separation factor calculations are
useful (Hartman et al., 2011). Separation factor calculations also help to understand gas
storage mechanism especially in cases of carbon capture and sequestration. Most often, the
higher the separation factor, the stronger the relative adsorption compared with other
components (Ruthven, 1984). In a binary system of carbon dioxide and methane, carbon
dioxide is expected to exhibit a much higher separation factors indicating an affinity to the
shale than methane.

One other use of separation factor calculations is to help decide on which gas component to
be used in enhanced gas recovery method. Competitive adsorption means gas with a smaller
relative adsorption to the shale will be released, whereas those with a stronger relative

45
adsorption will be adsorbed. Since methane has a much smaller relative adsorption compared
with other heavier gas components, enhanced gas recovery method such as injecting heavier
components into shale gas reservoirs could help to produce more of the adsorbed methane.
In this study, all the models predict a higher selectivity ratio or separation factor for ethane
than for methane (see Figure 2.14). According to Ruthven (1989), the separation factor
measures the affinity of the adsorbent for component i relative to the component j . This
can be expressed as

( x y )i
 ij 
( x y) j

25
yC2H6=0.519: 101 Kpa

20
Ethane Separation factors

15

10

EL IAS VSM

Figure 2.14: Separation factor calculations for mixtures for case 1 corresponding to a 0.519-
mole fraction of ethane.

The separation factor for the Extended Langmuir (EL) shows a constant value compared to
the other models that show variable separation factors. This is because, for the EL model, the
separation factor is not a function of pressure or composition (Ruthven,1984). Figure 2.14

46
shows the separation factor calculations for a mixture corresponding to 0.519 mole fraction
of ethane.

2.7 Summary
This chapter has critically reviewed the different adsorption models used in the prediction of
methane adsorption in shale gas reservoirs by grouping them under single and multi-
component adsorption models. It has further investigated the application of adsorption
models on shale gas reservoirs using a statistically robust method. Comparison was made
between different categories of adsorption models, allowing for the choice of adsorption
model to be selected based on evaluating different error functions, such as SSE, SAE, ARE,
MPSD and HYBRID.

By adopting a more statistically robust approach using the sum of normalised errors to
minimise the error distributions for each adsorption model adopted, an optimised adsorption
model was selected and used in the modelling of gas adsorption in shale gas reservoirs.

Based on the results from the analysis, ARE proved to be the most accurate in terms of
predicting the minimum sum of normalised error, and would therefore provide the best fit to
the experimental data. In all, 6 of the results provided minimum SNE for ARE calculations,
whereas 2 of the results showed minimum SNE for both SSE and Hybrid. Overall, the Langmuir
model gave the most accurate predictions for single component modelling, compared with
other models. In Binary mixture studies, VSM proved to give accurate results and fitted the
data more appropriately compared with IAS and EL models. ARE proved to be the best fit for
ethane adsorption when modelling with VSM. The HYBRID error function was however found
to fit the model better when using Langmuir isotherm, whilst it was found to be a worse fit
for methane adsorption using VSM.

47
Nomenclature

V = the volume of adsorbed gas

P = Gas pressure

VL = Langmuir volume or maximum gas adsorption at infinite pressure

PL = Langmuir pressure corresponding to one half of the Langmuir volume

Vm = maximum adsorption gas volume

C = constant related to the net heat of adsorption

PO = the saturation pressure of the gas

n = maximum number of adsorption layers

W = volume adsorbed volume

Wo = micro pore volume

E = energy of adsorption

 = affinity of the sorbent for the gas

VLi = Langmuir volume constant for pure component I, (SCF/ton)

PLi = Langmuir pressure constant for pure component i , (psia)

yi = Gas phase composition of component I, (fraction)

Pg = Gas phase pressure, (psia)

Vi = Adsorbed volume of component I, (SCF/ton)

p o i = vapour pressure of the pure component

xi = sorbed phase gas mole fraction

 = spreading pressure

na ( p) = pure component isotherm

A =specific surface area of the adsorbent.

48
Pg = gas phase pressure (psia)

yi =Gas phase composition of component i (fraction)

i = Fugacity coefficient of pure component i in gas phase, (dimensionless)

Pi o = Standard state pressure of pure component i in gas phase (psia)

i o = Fugacity coefficient of pure component i in gas phase at standard condition

(dimensionless)

xi a = Molar composition of component i in adsorbed phase (fraction)

nical = calculated adsorbed concentration,

ni exp = experimental adsorption data

N = number of data points for the isotherm.

r = number of parameters in adsorption model

i , j = gas components

n1 =maximum number of moles of i in surface phase

 = fractional coverage

v1 , 1v = Wilsons parameters for interaction between vacancy and adsorbate

b1 = Henry’s law constant for component one

i =fugacity coefficient of I in bulk gas mixture

 is = activity coefficient of I in adsorbed phase vacancy solution

nms =total number of moles of mixture in phase

nis , =maximum number of i in surface phase

nms , = maximum number of moles of I in surface phase

49
ai =partial molar surface area of i

T = temperature of adsorption system

R = Universal gas constant

r = Number of isotherm parameters, dimensionless

50
 Chapter 3
Temperature-Dependent Gas Adsorption Modelling in Shale
Gas Reservoirs

3.1 Introduction
There are different factors controlling the adsorption capacity of methane on shale. These
include, but are not limited to, total organic carbon, pressure, temperature, thermal maturity,
moisture and shale clay minerals (Guo et al., 2013). Total Organic Carbon (TOC) is seen as a
primary factor that contributes to the adsorption capacity of shale. TOC refers mainly to the
amount of organic carbon present in shale and is a key indicator of the ability of shale to
generate hydrocarbons. Previous experimental studies have reported high TOC of shale,
resulting in a higher capacity for the adsorbed gas (Chalmers and Bustin, 2008; Lu et al., 1995;
Ross and Bustin, 2007; Ross and Bustin, 2009; Zhang et al., 2012).
The adsorption capacity of shale can also be affected by pressure. Higher reservoir pressure
corresponds to a higher adsorption capacity. Raut et al., (2007) point out that with low
pressure, higher binding energy is required for gas adsorption, and hence the adsorbed
content decreases compared with when the pressure is very high. In terms of production from
shale gas reservoirs, the low matrix permeability of shale reservoirs implies that considerable
time will be needed for the average pressure to drop to a level where gas desorption becomes
significant, by which time the economic shut in limits of the well must have been reached
(Cipolla et al., 2010) Therefore, dependence on pressure depletion alone for ultimate
recovery of gas adsorbed on the shale matrix will contribute less towards overall gas
production in the well.
One other decisive factor contributing to the adsorption capacity is temperature. The process
of shale gas adsorption results in an exothermic reaction. Thereby, shale gas adsorption
capacity is expected to decrease with increasing temperature. The temperature dependence
of adsorption capacity is greatly controlled by the isosteric heat, which also depends on the
surface coverage. Many experimental studies have been conducted on the influence of
temperature on adsorbed shale gas content (Charoensuppanimit et al., 2015; S. Chen et al.,

51
2011; Clarkson, 2003; Gasparik et al., 2015; Guo, 2013; Hu, 2014; Ji et al., 2014; Lu et al., 1995;
Rexer et al., 2013; Ye et al., 2016)
Guo et al., (2013) reported that adsorbed gas content increased with increasing pressure
under isothermal conditions, whereas the gas content decreased under isobaric conditions
for increases in temperature. Guo et al., (2013) obtained adsorption isotherms at different
temperatures and also at different pressure points for shale samples of the Ordos Basin,
China. Their studies also noted that, for low pressure and temperature zones, pressure had a
much greater effect on the adsorbed gas content. For high pressure and temperature zones,
however, the effect of temperature on the adsorbed gas content was dominant.
The amount of gas in a shale gas reservoir is affected by the existence of geothermal
gradients. There are temperature differentials at different depths in the reservoir, and due to
the effect of temperature on the gas adsorbed, the gas in place will differ at each depth.
Temperatures are usually at near critical or supercritical regions of the adsorbed gas, hence
adsorption models should be able to provide reliable predictions of the temperature
dependence of adsorption in these regions (Charoensuppanimit et al., 2015).

3.2 Temperature-dependent Langmuir volume or Saturation loading

Since gas adsorption is an exothermic reaction, the heat produced is normally dissipated
towards the immediate surroundings. The ratio of the change that occurs in the adsorbate’s
enthalpy to the change in the adsorbed amount is normally referred to as the heat of
adsorption. The adsorption equilibrium constant, b(T ) normally represented by the Van’t
Hoff expression, is temperature-dependent (Zhu et al., 2006):
 H ads 
b  b0 exp    ……… (3.1)
 RT 
H ads = adsorption enthalpy change for a component.
Langmuir isotherm remains one of the most commonly used adsorption models in coal/shale
gas simulation, with most commercial simulators, such as CMG and Schlumberger Eclipse,
implementing gas adsorption modelling using Langmuir isotherm.
The expression for the adsorption on the microporous adsorbent is given as
bp
V  VL ……….. (3.2)
1  bp

where VL refers to the theoretical maximum Langmuir volume.

52
This equation can be written in terms of saturation loading as the temperature-independent
form used in this study:
bp
Langmuir 3: V  qso (3.2a)
1  bp

where qso is the saturation loading similar to the Langmuir volume.

While the adsorption equilibrium constant b is temperature-dependent, as can be seen in

equation (3.1), the Langmuir volume or saturation loading for the Langmuir isotherm is
considered constant and does not change with temperature. This assumption has been made
by several researchers. Zhu et al., (2006) in their studies of zeolites stated that “saturation
loading should be constant and independent of operating temperature”. Do (1998)
conducted studies with the temperature-dependent form of the Sips isotherm. The results
from the fitting of the Sips isotherm to several temperatures showed that the saturation
loading at different temperatures was the same as the reference temperature, indicating the
temperature independence of the saturation loading. Furthermore, a new adsorption model
was proposed by Do and Do (1997) for heterogeneous solids and tested with isotherm data
for various adsorbates on activated carbon and zeolites. They concluded that the thermal
expansion parameter for the saturation capacity is very small, implying that the saturation
loading is independent of temperature.
Other researchers, however, have strongly supported the idea of temperature dependence
of the saturation loading (Koresh, 1982). According to Malek and Farooq (1996), the
temperature dependence of saturation loading accounts for the thermal expansion of the
adsorbed phase, leading to decreasing monolayer coverage at very high temperatures.
Various temperature-dependent equations have been proposed for saturation loading, but
the exact functional form and the thermodynamic consistency of the parameters have not
been completely validated (Helminen et al., 2000a). Hwang et al., (1997) proposed the
dependence of the saturation loading on temperature as
a2
qs  a1  …………………. (3.3)
T
An empirical equation to describe the temperature-dependent saturation loading was given
(Cochran and Danner, 1985; Malek and Farooq, 1996) as

 q 
qs  qso exp  1  ………….. (3.4)
 T 
53
with q1  0 if the saturation loading is not temperature-dependent.

qso is temperature-independent constant.

There have been other works on petroleum applications, including methane adsorption on
coal and shale. Ji et al., (2014) conducted experiments on shale samples from the south-
eastern Ordos Basin in China to ascertain the effects of TOC, maturity, clay mineral and
thermodynamic conditions, such as pressure and temperature, on the methane sorption
capacity. A positive correlation was found to exist between the sorption capacity of the
samples and the TOC, whereas the clay mineralogy content of the samples had minimal effect
on their adsorption capacity. Ji et al., (2014) developed a computational algorithm that
allowed for the calculation of methane sorption capacity as a function of TOC, temperature
and pressure. They found that with an increase in temperature, the Langmuir pressure
increases exponentially, while the Langmuir volume undergoes a linear decrease. Ye et al.,
(2016) also proposed an exponential relation to describe the temperature dependence of the
Langmuir volume, based on analyzing experimental data for gas adsorption capacity on
different shales. The dependence of temperature of the Langmuir volume has been
attributed to the isosteric heat of adsorption by several researchers (Ji et al., 2014; Lu et al.,
1995; Rexer et al., 2013; Ye et al., 2016; Zhang et al., 2012) .
The effects of temperature on the adsorption capacity of shale have also being explored by
Guo et al., (2013). According to them, shale adsorption capacity decreases with increase in
temperature. They showed that the desorption curve showed hysteresis for a particular set
of adsorption data, and that this was due to thermodynamic reasons, such as the isosteric
heat capacity of the adsorption process being greater than that of the desorption process.
The adsorption isotherm curve is described by the Langmuir equation, whereas the
desorption isotherm is described by an equation proposed by Ma et al., (2014):

VLbp
V  C1 …………………………..(3.5)
1  bp

Where C1 is described as the residual adsorption capacity.

According to Guo et al., (2013), isosteric heat and isothermal adsorption capacity show a good
linear relationship, given as

54
qst  a3n  b3 ……………………………..(3.6)

By determining the isosteric heat from the Clausius-Clapeyron equation, adsorption

isotherms can be predicted from different temperatures. The Clausius-Clapeyron equation is
given as

1 dp q
 st 2 ……………………………………(3.7)
p dT RT

Integrating and rearranging will result in

ln p  qst / RT  c3 ……………………………………..(3.8)

Where c3  c2  ln po

They proposed the formula below to be used for predicting adsorption isotherm at different
temperatures

 a n  b3 a3n  b3 
p2  exp  ln p1  3   ……………………….(3.9)
 RT1 RT2 

a3,b3 are obtained from the linear relation between the isosteric heat capacity and

adsorption capacity, n . p2 refers to the pressure under the condition of T2 .

This model therefore allows the calculation of adsorption capacity using the isosteric heat
curve and adsorption /desorption model at a single temperature. Guo et al., (2013) argue
that since the production of shale gas is due to desorption, including desorption models into
production forecast and numerical simulation will ensure accurate estimations.

3.3 Temperature-dependent Langmuir Models

Equation (3.1), (3.3) and (3.4) can be applied to the classical Langmuir isotherm to obtain
several temperature-dependent Langmuir Models. These equations can be expressed as
follows

Langmuir 1: V   a1 
a2   bp  ……… (3.10)
 T   1  bp 

 q   bp 
Langmuir 2: V  qso exp  1    ………………. (3.11)
 T   1  bp 
55
  H ads 
Where b  b0 exp   .
 RT 
Due to the heterogeneity of most adsorbent’s, the assumption of a homogenous surface by
Langmuir isotherm does not apply. Shale is considered to be an extremely heterogeneous
adsorbent (Chen et al., 2017). In this case, the use of an adsorption model such as Freundlich
equation combined with Langmuir equation can be applied, by which the adsorbent’s
heterogeneity can be successfully represented (Helminen et al., 2000). However, very few
works have been found concerning the application of combined Langmuir and Freundlich
equation in modelling shale gas adsorption data. An expression for the Langmuir–Freundlich
equation is given (Helminen et al., 2000) as

qsbp n
Langmuir Freundlich 3: q  ……… (3.12)
1  bp n
The parameter n represents the surface heterogeneity and has been added to the Langmuir
equation without any proper physical meaning (Do, 1998; Helminen et al., 2000a; Ruthven,
1984). Higher values of n (greater than one) represent a highly heterogeneous adsorbent.
By combining the temperature-dependent saturation equations (3.3) and (3.4) with equation
(3.12), the following temperature-dependent Langmuir–Freundlich equations can be
proposed, involving the temperature-dependent parameter b .

  bp  ……….. ( 3.13)
n
Langmuir Freundlich 1: V   a1  a2  
 T   1  bp n 

 q   bp 
n
Langmuir Freundlich 2: V  qso exp  1   n 
……….. ( 3.14)
 T   1  bp 

3.4 Modified Vacancy Solution

Clarkson (2003) discusses the suitability of the Dubinin equations to microporous solids and
points out that for Dubinin-Polanyi theory, part of its additional suitability to coal is its
prediction of the temperature dependence of adsorption. Thus, by developing a new model
that combines the vacancy solution theory and Dubinin-Polanyi theories, multicomponent
adsorption predictions could be made, and also predictions of adsorption capacity at different
temperatures. Clarkson (2003) uses the Dubinin–Astakhov (D-A) equation to generate activity
coefficients as a function of the degree of pore filling for the pure component gas. For
multicomponent gas adsorption at different temperatures, this new model of adsorption

56
offers the ability to predict multicomponent adsorption data from pure component data
collected at a single temperature. The plot of  vs A or characteristics curve is temperature
invariant, which implies that once E and n are derived at a single temperature, different
temperatures could be easily derived. Thus once D is obtained from regression of the single
isotherm data, and an activity coefficient as a function of the degree of pore filling is obtained,
D can be recalculated for both component gases at the new temperature , assuming E is

temperature invariant .

Eo 
D …………………………………………… (3.15)
RT

E  Eo  …………………………………………..(3.16)

P 
A  RT ln  o  ……………………………………. (3.17)
P

Eo is the characteristic energy of adsorption for standard vapour, J mol

 is similarity coefficient which is dimensionless.

A is specific surface area of adsorbent

n is constant in Dubinin-Astakhov equation.

 is saturation loading, dimensionless

Clarkson (2003) applied this new model to only one experimental binary gas adsorption data
set. Applications of this model to other, varied data sets would help to ascertain the
robustness of this new adsorption model.

3.5 Simplified Local Density (SLD)

The simplified local density model has been successfully applied in many studies and in
particular extended to application in shales by Chareonsuppanimit et al., (2012). They argue
that this method offers distinct advantages in providing predictions for a spectrum of
adsorption phenomena, from adsorption to permeability, on coal and shale. For instance, Pan
and Connell (2007) integrated the swelling model to account for adsorption-induced swelling
of coals. These advantages, according to Chareonsuppanimit et al., (2012), are crucial for

57
more realistic reservoir simulations. Charoensuppanimit et al., (2015) improved the
temperature dependency of the SLD model by incorporating a new temperature dependency
for the adsorbed phase volume that was based on the volume–expansivity approach
proposed by Do (1998). By comparing results from the new temperature dependency model
to those of the original SLD, significant improvement could be observed over a larger range
of temperature. The modified SLD model is therefore suitable for modelling supercritical and
near critical gas adsorption. The adsorbed phase volume was related to temperature by the
expression

Vads T
dV

Vads ,0
V
    dT ……………………………………… (3.18)
T0

Vads  Vads ,0 exp( (T  T0 )) ,…………………………………………(3.19)

Where Vads is the adsorbed phase volume, Vads ,0 is the adsorbed phase volume at reference

temperature T0 and  is the thermal expansion coefficient.

For reservoir simulation, adsorbed phase densities are commonly required, and
Charoensuppanimit et al., (2015) showed that the prediction of the adsorbed phase densities
by the new model was much closer to that of methane and nitrogen, compared to other
models that predicted the densities.

3.6 Bi-Langmuir Model

Lu et al., (1995) claimed that Langmuir isotherm is insufficient in describing the adsorption in
Devonian shale gas reservoirs. They noted that temperature dependence is very often ignored
but plays an important role, especially during thermal stimulation as a recovery method. In
fact, both pressure and temperature are major factors in determining the amount of
adsorbed gas on the shale. Investigating adsorption as a function of pressure and temperature
showed that the effects of temperature are significant (Lu et al., 1995). They reported a
general reduction in the adsorption capacity of shale at increased temperatures. Bi-Langmuir
model was used to reconcile the data obtained from samples that were used in the
experiment. Although Langmuir model might be suitable to represent gas/shale isotherms
measured at a single temperature, Bi-Langmuir model was thought to be more suitable for
representation of multiple temperatures. Lu et al., (1995) argue that the assumption of
58
homogeneous adsorbent in the Langmuir model may not be suitable for a gas/shale system,
due to the fact that different materials, such as clay minerals and kerogen, may contribute to
gas desorption.
Bi – Langmuir model, written in the following form, is used to describe gas adsorption on
an adsorbent having two discrete, sharp peaks of adsorption energy distribution. One term
of the equation describes the gas adsorption on clay minerals, whilst the other term accounts
for gas adsorption on kerogen.
N ads k T  p k T  p
 f1 1  1  f1  2
Nm 1  k1 T  p 1  k 2 T  p
……………. (3.20)
k T   koT 1/2 exp   E / RT  ……………………………….(3.21)

Where,
f1 = fraction of adsorption site

k1,2
= adsorption equilibrium constant
N ads =amount of adsorbed gas per unit volume adsorbent

N m = Amount of adsorbed gas at monolayer coverage

E1 = Adsorption energy

3.7 Exponential Model

Ye et al., (2016) conducted similar studies on the temperature dependence of gas adsorption.
Their studies proposed a new, improved Langmuir model that was used in evaluating gas
adsorption capacity of shale under various pressures and temperatures. Gas adsorption is
considered as an exothermic process and hence increasing temperature will restrain the gas
adsorption process, thus causing a general reduction in the amount adsorbed as temperature
increases (Lu et al., 1995; Ye et al., 2016).

The constant Langmuir volume ( VL ) was modified to be a function of temperature and

thereby an exponential model was proposed that relates VL to temperature as

VL  Vs exp   DT T  ……………………. (3.22)

The final exponential model derived, expressing a temperature-dependent Langmuir model,

can be defined as

59
 Vs exp   DT T  p
V …………….. (3.23)
p  T  A exp B 
  
T 
Vs = theoretical maximum adsorption capacity

DT =Reduction coefficient related to temperature increase

The coefficients Vs , DT , A , B in equation (3.23) can be determined by fitting the

experimental adsorption data.

An exponential relationship between temperature and Langmuir volume can be observed
from Figure 3.1, where at increasing temperature, the Langmuir volume decreases
exponentially.

Figure 3.1: The Langmuir volume constant vs Temperature for typical shale samples (Ye et
al., 2016)

Ye et al., (2016) analysed published gas sorption data on shales from the USA, China, and
Canada and concluded that an improved Langmuir model that incorporates the effect of
temperature is able to describe shale gas sorption capacity under the effect of temperature
and pressure. Ji et al., (2012) compared the Exponential model with the linear model and the
classical Langmuir model. The comparison was based on the use of average relative error.
Their results showed the exponential model as the most efficient model with the least error
when predicting gas adsorption at different temperatures.
Table 3.1 and Table 3.2 summarises the overall temperature dependent models derived in
this chapter. Modified Vacancy Solution (MVS) and SLD have not been included in the Tables

60
because it was not considered further in the analysis section of this thesis. It was felt that
these two models required computationally complex solutions that would have made their
subsequent inclusion in numerical studies difficult.

Table 3.1: Langmuir and Langmuir Freundlich Isotherm

Model Name Model Equations Equation Fitting

No. parameters

Langmuir 3  q bp  2a qso , H ads ,

V   so 
 1  bp 
b0

Langmuir Freundlich 3  q bp n  7 qso , n , H

V   so n 
 1  bp  , b0

Table 3:2: Variations of temperature-dependent Langmuir equations

Model Name Model Equations Equation Fitting

 H  No. parameters
b  b0 exp   
 RT 
Langmuir 1
V   a1  2
a   bp  5 a1 , a2 ,
 
 T   1  bp 
H ads , b0
Langmuir 2  q   bp  6 qso , q1 ,
V  qso exp  1   
 T   1  bp 
H , b0

Langmuir Freundlich 1   bp  8 a1 , a2 , n
n
V   a1 
a2

T   1  bp n 
   H ads , b0
Langmuir Freundlich 2  q   bp  9 qso , q1 , n
n
V  qso exp  1   n 
 T   1  bp 
H ads , b0
Other Temperature dependent Models

61
 Exponential Model Vs exp   DT T  p 12 Vs , DT , A ,
V
p  T  A exp B 
 T    B

Bi-Langmuir Model N ads k T  p k T  p 10 N m , k1 , k2 ,

 f1 1  1  f1  2
Nm 1  k1 T  p 1  k 2 T  p
E1 , E2 , f1
where k (T )  kiT 1/2
exp  Ei RT 

Selecting a model that is capable of predicting most accurately the adsorption data at
different temperatures becomes essential when isotherm estimations are needed at
temperatures for which no adsorption data exist. If validated, based on the measurements
made in laboratory conditions for certain temperatures, such a model can be used to predict
gas adsorption behavior at any other reservoir temperature.
The use of temperature-dependent Langmuir volume/saturation loading when describing gas
adsorption avoids the repeated matching of Langmuir constants for each isotherm (Ye et al.,
2016). This can lead to more accurate representation of gas adsorption in the reservoir, and
also makes it easier to be applied in reservoir simulation where temperature effects are
considered.
Furthermore, evaluating gas adsorption as a function of both pressure and temperature using
different temperature-dependent adsorption models will allow for the possibility of
employing thermal stimulation techniques to enhance the recovery of gas from shale
reservoirs.
Incorporating temperature –dependent adsorption model has already being explored by
several authors (Lin et al., 2015; Wang, 2016). Wang (2016) included Bi-Langmuir model into
an unconventional simulator for thermal stimulation purposes. Their work showed that by
using a temperature-dependent adsorption model, the potential to increase gas recovery
rose significantly through thermal stimulation. This was achieved by altering the shale gas
adsorption/desorption behavior through the elevation of formation temperature.

62
3.8 Materials and Methods
3.8.1 Shale Gas Data
Experimental data of samples obtained from different shale gas reservoirs have been
compiled and adsorption isotherms have been obtained for 11 different samples at different
temperatures, ranging from 25°C to 90°C. Four main reference sources have been used to
collect the data used in this thesis (Guo, 2013; Ji et al., 2014; Lu et al., 1995; Zhang et al.,
2012). Lu et al., (1995) obtained samples from the Devonian shale with reported mineral
compositions consisting mainly of clay minerals and organic materials, principally kerogen,
with compositions ranging from 0 to 15%, depending on the sample (Luffel et al., 1996). Two
shale samples from the Gas Research Institute and one illite sample were used in their
adsorption measurements, at four different temperatures, of 25°C, 37.7°C, 50°C and 60°C.
Zhang et al., (2012) collected 2 different sets of samples of organic rich shales. These included
the Green River Formation (Eocene, Utah) and Woodford shale (Upper Devonian) for the first
set of samples, with the second set collected from the Barnett shale cores. Methane sorption
capacity for these different samples were obtained at three different temperatures, of 35.4°C,
50.4°C and 65.4°C. Samples from the Ordos Basin in China were collected and used to obtain
methane adsorption measurements at different temperatures, of 30°C, 40°C, 50°C, 60°C and
90°C (Guo, 2013; Ji et al., 2014).
A similar method to that used in Chapter 2 section 2.4 has been adopted for identifying the
optimal model for gas adsorption, using the temperature-dependent form of Langmuir
isotherm. From the results in chapter 2, it was identified that Absolute Relative Error (ARE) or
Absolute Average Deviation (AAD%) provided the most consistent best fit for the
experimental data. Hence this error function has been used in evaluating the temperature
dependent models. It can also be recognised that several researchers (Guo, 2013; Ji et al.,
2014, Lu et al., 1995; Zhang et al.,2012) have opted for the use of AAD% over other error
functions consistently over time. AAD% was found to be most accurate for majority of the
data used in this thesis, thus the rest of the research have been conducted using AAD% to
optimise the different adsorption models.
The percentage average relative error has been defined as

63
  N  (nical  ni exp )  
   n  
  i exp   
% AAD  100* ABS  i 1  
 N 
 

Where,
nical is the calculated adsorbed concentration,
ni exp is the experimental adsorption data,

N is the number of data points for the isotherm

3.9 Results and Discussion

Tables 3.3 and 3.4 display the results for the regression carried out on all 11 shale samples
used in the study. The results from the regression has been analysed using the Absolute
Average Deviation (AAD %) error approach to determine which models gives the least error.
The model with the least absolute average deviation error can be said to be the best model
for representing the adsorption capability of the sample. Six different models based on the
temperature-dependent Langmuir volume, and three models of temperature-independent
Langmuir volume, have been compared (see Table 3.3 and 3.4).
In Table 3.3, absolute average deviations have been obtained for models with temperature-
dependent Langmuir volume. By comparing these models, it can be ascertained that the
Langmuir–Freundlich 2 and the Exponential model were able to give the least error for a
majority of the data. Langmuir-Freundlich model 2 predicted most accurately for shale
samples from Green river formation, Barnett shale Tarrant A3, Blakely 1 and Ordos Basin A
and B in China. On the other hand, three of the shale samples, from Woodford shale, Ordos
Basin YY33-2, Ordos Basin YY34-1 and Antrim-7, were most accurately predicted by the
Exponential model.
Langmuir model 1 failed to predict most accurately the gas adsorption for any of the shale
samples evaluated. The AAD% for Langmuir 1 was notably higher in most of the results when
compared with the rest of the temperature-dependent models, ranging from 0.49 % -7.48 %.
Similarly, Langmuir 2 could predict only one of the samples (CSW2) most accurately.
However, when the Freundlich model was combined with Langmuir model, the results of the
modelling improved, with Langmuir-Freundlich 2 model giving a very small average absolute

64
deviation error for five of the samples. This could also be attributed to the extra parameter
present in the Langmuir–Freundlich model. The more parameters that are present in the
model, the greater the likelihood of the model predicting more accurately. Furthermore, the
Langmuir-Freundlich model is better able to describe the heterogeneity of the shale sample
than the classic Langmuir model.
Similar comparison has been made for the models with temperature-independent Langmuir
volume. These models include Langmuir 3, Langmuir-Freundlich 3 and the Bi-Langmuir model
(see Table 3.4). Comparing the values obtained for all three models, it can be observed that
Langmuir-Freundlich 3 model provided the least error for six of the shale samples, compared
with three and two for Langmuir 3 and Bi-Langmuir model respectively.
Most of the shale data have been adequately described by the combination of both the
Langmuir and Freundlich models. Thus Langmuir-Freundlich 2 and Langmuir-Freundlich 3
proved to be better models in describing the adsorption process in the shale formations. Since
most of these shales are deemed to be highly heterogeneous, with mixtures of clay minerals
like quartz and kerogen compositions of type I or II, the combined Langmuir-Freundlich
models are best for modelling gas adsorption. The Exponential model also proved to fit
adequately most of the data compared with the Langmuir models. Hence the choice of model
would depend on how highly heterogeneous the shale sample is, and since Langmuir model
assumes a homogeneous surface, the preferred choice of combined Langmuir and Freundlich
would best describe the adsorption process in most shales.
In order to compare how well the temperature-dependent and -independent forms of the
Langmuir volume models compare to each other when applied to the same datasets, Table
3.3 was combined with Table 3.4 and the model with the least error was chosen to be the
most accurate representation of the gas adsorption data on the shale samples. Seven of the
data sets were accurately described by the temperature-independent Langmuir volume
models, compared to only four by the temperature-dependent models. However, the error
difference was minimal.
Due to the large number of graphs generated for the representation of the modelling process,
only results showing Woodward shale have been presented. The best fitting parameters for
all the models have been provided in Appendix 3 (see Table 3.5-3.12) for checking the models’
applicability to the data.

65
 2
The R method to measure the level of accuracy of the fit for Woodford shale shows very
close approximation to one, meaning all the models could adequately fit the gas adsorption
2
process in Woodford shale (see Figure3.2-Figure3.9). This was not the case when the R was
evaluated for the rest of the data (see Table 3.5-3.12 in Appendix 3), as some of the models
could not adequately describe the gas adsorption process. In particular, models like Langmuir
2
1 and 2 failed to describe adequately for some of the adsorption data, as the R shows in
Tables 3.5 to 3.12.

66
 Langmuir 1
0.025
Amounts adsorbed, m3/kg

0.02
35.4°
0.015 50.4°C
65.4°
0.01 35.4°C
65.4°C
0.005
50.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.2: Modelling methane adsorption using Langmuir 1 model on experimental data of
Woodford shale

Langmuir 2
0.025

0.02
Amounts adsorbed,m3/kg

35.4°C
0.015 50.4°C
65.4°C
0.01 35.4°C
50.4°C
65.4°C
0.005

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.3: Modelling Methane adsorption using Langmuir 2 model on experimental data
from Woodford shale

67
 Langmuir-Freundlich 1
0.025
Amounts adsorbed,m3/kg

0.02

35.4°C
0.015
50.4°C
65.4°C
0.01
35.4°C

0.005 50.4°C
65.4°
0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.4: Modelling methane adsorption using Langmuir-Freundlich 1 model on

experimental data from Woodford shale

Langmuir-Freundlich 2
0.025
Amounts adsorbed , m3/kg

0.02

35.4°C
0.015
50.4°C
65.4°C
0.01
35.4°C
50.4°C
0.005
65.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.5: Modelling methane adsorption using Langmuir-Freundlich 2 model on

experimental data from Woodford shale

68
 0.025
Exponential

0.02
Amounts adsorbed,m3/kg

0.015 35.4°C
50.4°C
65.4°C
0.01
35.4°C
50.4°C
0.005 65.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.6: Modelling methane adsorption using Exponential model on experimental data
from Woodford shale

Bi-Langmuir
0.025

0.02
Amounts adsorbed,m3/kg

35.4°C
0.015
50.4°C
65.4°C
0.01
35.4°C
50.4°C

0.005 65.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.7: Modelling methane adsorption using Bi-Langmuir model on experimental data
from Woodford shale

69
 Langmuir 3
0.025

0.02
Amounts adsorbed, m3/kg

35.4°C
0.015
50.4°C
65.4°C
0.01
35.4°C
50.4°C
0.005
65.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.8: Modelling Methane adsorption using Langmuir 3 model on experimental data
from Woodford shale

Langmuir-Freundlich 3
0.025
Amounts adsorbed,m3/kg

0.02
35.4°C
0.015 50.4°C
65.4°C
0.01 35.4°C
50.4°C
0.005
65.4°C

0
0 2 4 6 8 10 12 14 16
Pressure,Mpa

Figure 3.9: Modelling Methane adsorption using Langmuir Freundlich 3 model on

experimental data from Woodford shale

70
Table 3.3: Absolute Average Deviation (AAD%) for models with temperature-dependent volume

Sample Reference Langmuir 1 Langmuir 2 Langmuir-Freundlich Langmuir- Exponential No.of Data

1 Freundlich 2
Green River Formation Zhang et al 2012 0.49 % 0.48 % 2.66 % 0.07 % 0.33 % 33
Woodford Shale Zhang et al 2012 2.68 % 2.67 % 2.99 % 3.72 % 1.13 % 40
Barnett Shale Tarrant A3 Zhang et al 2012 1.11 % 1.13 % 0.32 % 0.22 % 2.18 % 39
Barnett Shale Blakeley 1 Zhang et al 2012 9.28 % 9.23 % 4.07 % 4.00 % 7.82 % 42

Ordos Basin YY33-2 Ji et al 2014 1.53 % 2.74 % 5.75 % 2.25 % 0.97 % 35

Ordos Basin YY34-1 Ji et al 2014 6.95 % 10.63 % 7.57 % 2.00 % 0.76 % 35
Ordos Basin A Guo 2014 5.57 % 6.81 % 0.46 % 0.25 % 5.76 % 42
Ordos Basin B Guo 2014 7.14 % 6.30 % 0.63 % 0.45 % 0.94 % 39
Ordos Basin C Guo 2014 7.48 % 5.74 % 0.11 % 5.82 % 0.26 % 36
CSW2 Lu et al 1995 0.971 % 0.32 % 2.60 % 1.84 % 0.85 % 40
Antrim-7 Lu et al 1995 0.976 % 0.83 % 1.71 % 0.77 % 0.66 % 40

71
Table 3.4: Absolute Average Deviation (AAD %) for models with temperature-independent Langmuir volume

Sample Reference Langmuir 3 Langmuir-Freundlich 3 Bi-Langmuir No.of Data

Green River Formation Zhang et al 2012 0.42 % 0.06 % 2.01 % 33

Woodford Shale Zhang et al 2012 2.68 % 1.03 % 4.12 % 40

Barnett Shale Tarrant A3 Zhang et al 2012 1.05 % 0.08 % 1.30 % 39

Barnett Shale Blakeley 1 Zhang et al 2012 9.21 % 4.09 % 3.56 % 42

Ordos Basin YY33-2 Ji et al 2014 1.79 % 1.92 % 8.96 % 35

Ordos Basin YY34-1 Ji et al 2014 6.48 % 2.79 % 8.71 % 35

Ordos Basin A Guo 2014 3.82 % 0.09 % 6.46 % 42

Ordos Basin B Guo 2014 6.16 % 0.40 % 16.07 % 39

Ordos Basin C Guo 2014 2.38 % 5.79 % 5.28 % 36

CSW2 Lu et al 1995 0.69 % 2.31 % 0.71 % 40

Antrim-7 Lu et al 1995 0.73 % 0.22 % 0.14 % 40

72
Temperatures can also be extrapolated to conditions that might be outside laboratory
conditions, to review how the models might predict gas adsorption. This is only possible
because of the use of temperature-dependent models. Extrapolation to temperatures outside
laboratory conditions is useful especially in simulation studies where thermal application is
involved. For enhanced shale gas recovery studies, the use of temperature dependent-models
is key for prediction of gas where temperature variation can significantly affect the amount
of gas adsorbed. Undertaking simulation studies would require validating the models before
extrapolation to higher temperatures. The experimental adsorption data for Sample YY33-2
at 30°C, 40°C, 50°C and 60.4°C have been used for the establishment and calibration of the
models.
Since extrapolation is necessary especially for numerical simulation involving thermal
application, this study have used isotherm data at 70°C for YY33-2 to get some idea about the
importance of model selection in extrapolation of temperature. Both temperature-
dependent Langmuir volume models and temperature-independent Langmuir volume
models have been extrapolated to 70°C to confirm how well the models predict gas
adsorption for validation and extrapolation purposes (See Figure3.10 and Figure3.11).
The Exponential model from Figure 10 shows a much more accurate fit of the extrapolation
to 70°C for the shale data of YY33-2. This implies that subsequent extrapolation is more likely
to give better approximation to gas adsorption at temperatures outside laboratory
conditions. With the exception of the Exponential model that gave a better validation, both
Langmuir 1 and 2 gave an exaggerated prediction, whereas the combined Langmuir and
Freundlich models’ prediction was close to the experimental data.
For temperature-independent Langmuir volume models, the validation of the extrapolation
with the actual data was not very accurate. All three models either under-predicted or over-
predicted the results (see Figure 3.11). This could be because, for the shale sample YY33-2,
the temperature-independent Langmuir volume models do not describe the adsorption
process well, compared with the temperature-dependent Langmuir volume models.
The results from Figures 3.10 and 3.11 show that care must be taken in the choice of model
and the representation of temperature effects on adsorption, due to the different predictions
offered by the models. Although some of the results are not exactly far off from the range of
data at 70°C, caution is still expected to be exercised in the use of these models, as results

73
might be an over- or under-estimation of the actual adsorbed gas at temperatures for which
there are no experimental data.

74
 0.006 Temperature Dependent Langmuir Volume Models Extrapolation to 70°C

0.005

0.004
Adsorbed Amount, m3/kg

30°C

0.003 40°c

50°C

60°C
0.002 70°C

Exponential

Langmuir-Freundlich 1
0.001 Langmuir-Freundlich 2
Extrapolation at 70°C
Langmuir 2

Langmuir 1
0
0 2 4 6 8 10 12
Pressure,Mpa

Figure 3.10: Extrapolated and validation at 70°C of shale sample YY33-2 using temperature-dependent Langmuir Volume models

75
 0.006 Temperature Independent Langmuir Models Prediction at 70°C

0.005

0.004
Adsorbed Amount, m3/kg

0.003
30°C

40°c

50°C
0.002
60°C

70°C

0.001 Langmuir-Freudnlich 3

Extrapolation at 70°C Langmuir 3

Bi-Langmuir
0
0 2 4 6 8 10 12
Pressure,Mpa

Figure 3.11: Extrapolation and validation at 70°C of shale sample YY33-2 using temperature-independent Langmuir Volume models

76
3.10 Summary
This chapter introduces the use of temperature-dependent adsorption modelling in
prediction of gas adsorption capacities. By examining different temperature dependent
adsorption models and their predictive performances, the choice of an accurate model to be
used in shale gas reservoir becomes easier and removes error associated when modelling
shale gas adsorption.

This chapter has focused on applying 8 different temperature-dependent adsorption models

to a wide range of shale gas adsorption data sets. Temperature-dependent models have been
further grouped into models that have temperature-dependent Langmuir volume and those
that expresses a constant Langmuir volume. Analysis of temperature-dependent Langmuir
volume showed that the majority of the data were successfully represented when using the
combined Langmuir–Freundlich 2 model, and also the Exponential model. Langmuir 1 and 2
showed the least accuracy for all the data evaluated when used in modelling the gas
adsorption on the experimental shale gas data studied. Temperature-independent Langmuir
volume also proved able successfully to fit the gas adsorption data presented, and in some
cases, when compared with the temperature-dependent Langmuir volume models, it
predicted the gas adsorption most accurately. Validation and extrapolation of all the models
for shale sample YY33-2 showed varying prediction to actual data at 70°C, though models with
temperature-dependent Langmuir volume offered much closer prediction than the
temperature-independent Langmuir volume models. This implies that care and caution must
be exercised when selecting a particular model for representation of gas adsorbed at
temperatures for which no data are available. Temperature-dependent models are important
because they offer an ability to accurately describe gas adsorption at multiple temperatures
and also to be implemented in numerical simulation studies where thermal stimulation is
required for enhanced shale gas recovery process.

Nomenclature
H ads = adsorption enthalpy change for component

b = adsorption equilibrium constant

VL = theoretical maximum Langmuir volume.

77
qso = saturation loading similar to the Langmuir volume

C1 = residual adsorption capacity

p2 = pressure under the condition of T2

n = surface heterogeneity

Eo = characteristic energy of adsorption for standard vapor, J mol

 =similarity coefficient which is dimensionless

A = specific surface area of adsorbent

n = constant in Dubinin-Astakhov equation

 = saturation loading, dimensionless

Vads = adsorbed phase volume

Vads ,0 = adsorbed phase volume

T0 = reference temperature

 = thermal expansion coefficient

f1 = fraction of adsorption site

k1,2
= adsorption equilibrium constant

N ads = amount of adsorbed gas per unit volume adsorbent

N m = Amount of adsorbed gas at monolayer coverage

E1 = Adsorption energy
Vs = theoretical maximum adsorption capacity
DT =Reduction coefficient related to temperature increase
nical = calculated adsorbed concentration
ni exp = experimental adsorption data

N = number of data points for the isotherm

78
 Chapter 4
Material Balance Calculations for Unconventional Gas
Reservoirs Using Temperature-Dependent Gas Adsorption
Models
4.1 Introduction
The material balance equation has been an important tool for making future reservoir
predictions, especially future production forecasts, in conventional reservoirs. It is seen as
one of the most powerful tools used in reservoir engineering (Canel and Rosbaco, 1992).
Material balance models use pressure and production data to determine the volume of
hydrocarbons that is present. They have been used primarily in the estimation of original
hydrocarbons in place, and also in determining expected ultimate recovery and remaining
reserves in a reservoir.

In estimating original gas in place (OGIP) and initial reservoir pressure, several methods,
including those of Kings (1993), Jensen and Smith (1997), Seidel (1999), Ahmed and Roux
(2006), Moghadam et al., (2011) and Firanda (2011), have been proposed. These methods
have utilised the application of the P/Z plot, which requires as the input data cumulative gas
production, average reservoir pressure and the properties of the produced gas (Seidle, 1999).

The success and usefulness of the P/Z plot in conventional reservoirs led to its application in
unconventional reservoirs, such as coal bed methane and shale/tight gas reservoirs. Despite
its usefulness, the P/Z plot may give inaccurate results when applied directly to
unconventional reservoirs such as coal/shale. This is because in its conventional form, it does
not include other sources of gas storage, such as connected reservoirs or adsorption, which
are present in coal/shale reservoirs (Moghadam et al., 2011). This has led to the modification
of P/Z plot in order for it to be suitable for application to unconventional reservoirs, especially
coal / shale gas reservoirs.

Unconventional reservoirs such as coal/shale are characterised by gas adsorption, hence

incorporating adsorption into the derivation of the P/Z method is necessary for accurate
prediction of hydrocarbons in place for such reservoirs. This requires an adsorption model
that can correctly represent the adsorption phenomenon within these reservoirs. Langmuir

79
isotherm represents this phenomenon for the traditional P/Z plot used in unconventional gas
reservoirs. Despite the limitations of Langmuir isotherm, such as adsorption being a function
of only pressure, it remains the only model currently incorporated in most P/Z plots to
evaluate the production performance of unconventional gas reservoirs using material
balance. Several P/Z methodologies have been developed for use in unconventional
reservoirs with the use of classic Langmuir isotherm.

Section 4.2 of this chapter comprises a review of different modifications of the P/Z plot used
to determine the original hydrocarbons in place and the estimation of the initial pressures.
Several of the modified material balance equations for unconventional gas reservoirs can also
be used in the prediction of future gas production.

In section 4.3, an attempt has been made to modify the material balance method used in
unconventional gas reservoirs to include temperature-dependent models. The chapter
concludes with analysis and discussion of the modified methodology used on data from an
unconventional gas reservoir.

4.2 P/Z plot methodology used for unconventional gas reservoirs

4.2.1 King’s Method
King (1993, 1990) was one of the first researchers to modify the P/Z plot for non-volumetric
behaviour of coal seam gas and shale gas reservoirs. With the conventional P/Z plot
developed mainly for use in volumetric reservoirs (where the pore volume of the reservoir is
assumed to be constant, or there is no influx of water into the system), King (1990) offered a
different methodology where the effect of adsorption can be incorporated.

An underlying assumption that was used by King was that there must exist an equilibrium
between the free gas and the adsorbed gas. The desorption process in the reservoir is
described by a strictly pressure-dependent relation. With this assumption, the method could
only be applied when the well was shut in, or for flowing wells that experience rapid
desorption. The key assumptions used by King in developing the material balance for
coal/shale reservoirs are summarised below:

1. Gas is adsorbed primarily in the matrix and free gas exists mostly within the cleats
/fractures

80
 2. The compressibility of water and rock, as well as water production, is considered
3. The coal/shale follows Langmuir isotherm
4. Adsorption is at pseudo-steady state period.

King proposed the material balance equation in the form:

Vb 2i Z scTsc  pi p
Gp   *  *  ……………………………….(4.1)
pscT  Zi Z 

Z
Z*  ………………………………..(4.2)
1  c  pi  p   1  S wavg  
ZRTCE
i p

King also proposed an iterative solution to obtain the bulk volume of the secondary porosity,

which is denoted by Vb 2 .

psc  VL p 
CE    ……………………………………………………………..(4.3)
Z SC RTSC  pL  p 

This refers to the adsorption isotherm in terms of lb-moles/ft3

The average water saturation was defined by King as:

5.615Wp
Savg  ………………………………(4.4)
Vb 2

The Z * consists of two denominator terms: the first represents the free gas in the
cleats/natural fractures, while the second term is the ratio of the adsorbed gas to the free
gas.

According to Seidle (1999), King’s method was found to be suitable only for under- pressured
coal and not for under-saturated coals, and when the method was applied to coal wells, it
was found that erroneous results were obtained. That is, unrealistic results were obtained for
initial pressure and gas in place calculation when King’s method was applied to production
data.

4.2.2 Jensen and Smith Modified Method

A modified method for material balance in unconventional gas reservoirs was proposed by
Jensen and Smith (1997) after King’s method. It offered a more practical basis for evaluation
81
of the estimated recovery and the remaining reserves in the reservoir. Their methodology
used only a few parameters, such as the Langmuir pressure, together with the reservoir
pressures and cumulative production, to make its estimates. Their approach was seen to be
more useful for reservoirs that are highly dominated by adsorption. This meant that water
and free gas that exist in the cleats /natural fractures is considered to be negligible, and the
effect of water saturation was ignored.

Z * was defined as:

Z
Z*  ………………….(4.5)
 BVLdaf 1  a  w pSCTZ
32.037i Z SCTSC  p  pL 

and the modified gas material balance was thus given as:

p pi  G p 
 1   ………………………………(4.6)
p  pL pi  pL  Gi 

4.2.3 Seidle Method – Modified King Method

Seidle (1999) modified King’s method after realising the limitations of his iterative method
when applied to actual production data. By assuming constant water saturation over time,
instead of an averaged water saturation over the drainage area as employed by King, Seidle
(1999) modified the iterative solution proposed by King and offered a direct approach to
calculating initial gas in place, estimated recovery and remaining reserves. Thus Equation 4.4
was no longer needed. Seidle (1999) believed that the water saturation of coal deposits
remains relatively constant over a long period of time, with saturation changes in the order

of 0.35 over a 10-year simulation. The Z * was found to be insensitive to changes in the water

saturation, thus iterations of drainage area had little effect on the value of Z * ; this is because
the ratio of the sorbed gas to the free gas in Equation 4.2 is often larger than the water
saturation. Seidle (1999) verified his approach on production data that were problematic for
the original King method, and found that it provided accurate solutions.

82
4.2.4 Ahmed and Roux Method
Ahmed and Roux (2006) proposed a new, enhanced material balance methodology that
enabled the calculation of the original gas in place, as well as predicting the average reservoir
pressure and future reservoir performance. It removed the reliance on the iterative solution
that was proposed by King (1993, 1990) by providing a simple method of obtaining the initial
gas in place through expressing the material balance as an equation of a straight line.

Gp  G  GF  GA  GR …………………………..………………………………(4.7)

GP = cumulative gas produced, G = gas originally adsorbed, GF = original free gas, GA = gas

currently adsorbed , GR = remaining free gas.

By defining the material balance equation for coal bed methane (CBM) in Equation 4.7 and
defining the terms in the equation, an equation of a straight line can be obtained with
variables x and y .

G  1359.7 AhBGc ……………………………………..(4.8)

GF  7758 Ah 1  S wi  Egi …………………………………….(4.9)

GA  1359.7 AhBV …………………………………(4.10)

BwWp
S wi 1  cw  pi  p   
7758 Ah
GR  …………………………(4.11)
1   pi  p  c f

BwWp Eg   V bp  7758  P  c f  S wi cwi   1  S wi   Eg 

Gp   Ah 1359.7  B  Gc  m  
1   c f P    1  bp  1   c f P   ……..(4.12)
 
7758 Ah 1  S wi  Egi

BwW p Eg
y  Gp  ………………….(4.13)
1   c f P 

83
  V bp  7758 P  c f  Swi cwi   1  Swi  Eg
x  1359.7  B  Gc  m   ……………..(4.14)
 1  bp  1   c f P 

By using historical production data, a plot of x and y will yield a straight line with a slope
and intercept defined respectively as:

m  Ah
a  7758 Ah 1  S wi  Egi

From these equations, the drainage area A can be obtained and subsequently the original
gas in place as:

OGIP  G  GF ………………………………..(4.15)
The usefulness of Equation 4.12 is that it can be used to obtain an estimate of the average
reservoir pressure, and also to make future reservoir production performance predictions.
Ahmed and Roux's (2006) method is applicable to any coal that behaves according to the
Langmuir isotherm equation.

4.2.5 Moghadam et al., 2011 Method

Moghadam et al., (2011) presented an advanced and more rigorous material balance for gas
reservoirs that incorporated many of the modifications by previous authors. Thus, it can
account for water influx, expansion of the formation and of residual liquids in over-pressured
reservoirs, and gas desorption in coal/shale reservoirs. The modified equation has the same
format as the conventional P/Z plot of the material balance equation. According to the
authors, one key advantage of the modified material balance is that it can be used to define
the total compressibility of the system for analysing fluid flow in unconventional gas

reservoirs. They developed a new Z ** that is related to Z * as developed by King (1990) as

follows

Z 
Z **  Z *  *i  ……………………………….(4.16)
 Zi 

p
Or Z **  …………………...(4.17)
 1 p  G  G
  S gi  cwip  cep  cd   i
p
  1  f
 G
 S gi Z Zi  Gf  

This new Z ** replaces Z * in the gas material balance equation as:

84
 p  Gp  pi
 1   ** ……………………………………...(4.18)
Z **  G  Zi

*
Moghadam et al., (2011) argued that the use of p / Z normally showed little resemblance
**
to the conventional p / Z values , but that by using the p / Z , the values obtained are more
similar to those of p / Z .

4.2.6 Firanda Method

Firanda (2011) considered the inclusion of different drive mechanism in the material balance
equation that was introduced by King. These drive mechanisms included water expansion,
rock compaction, connate water expansion and moisture expansion. By so including these,

p / Z * was modified with a new definition of Z* (see Equation 4.22) that included these
terms. The modified material balance equation as introduced by Firanda (2011) can therefore
be used in many aspects of fluid mechanism in a coalbed reservoir. The adsorption term of
the modified formula also included the Langmuir isotherm equation. In terms of the
methodology for predicting the original gas in place and for making future reservoir
performance predictions, Firanda approach is also similar to that of Ahmed and Roux (2006),
in that the material balance was expressed as a straight line with x and y . He proposed two

equations for when the coal is in a saturated condition ( Pres  Pcritical ), Equation 4.20, and

when the coal is at under-saturated condition , ( Pres  Pcritical ) Equation 4.21.

The new material balance equation developed by Firanda is defined as follows

UW = Gmicropores  Gmacropores  Wmacropores  Sconate  M moisture  S pvshrinkage  Winf lux ……….. (4.19)

UW = underground withdrawal, Gmicropores = gas expansion in micro pores , Gmacropores = gas

expansion in macro pores , Wmacropores = water expansion in macro pores , Sconate = connate

water expansion, M moisture = moisture expansion, S pvshrinkage = pore volume shrinkage, Winf lux =

net water influx.

85
    Ah 1  S wi  
  Bg  Bgi  
VP VP
G p Bg   Ah B L i 1  f m  f ash   Ah B L 1  f m  f ash   Bg  
PL  Pi PL  P 
   Bgi 
 VL Pi 
 Ah 1  S gi  S wc    Ah B 1  f m  f ash   Bgi ………..(4.20)

  Bw  Bwi   AhP  S wc cw  c fma     m m fmi 
 P f c  c
P P
 L i
 Bwi  1  fm
 
(We  Wp Bw )

 Ah 1  Swi    Ah 1  S gi  Swc  

Gp Bg  Wp Bw  
 Bgi

  Bg  Bgi   
 Bwi
  Bw  Bwi   AhP  S wccw  c fma   We ….(4.21)

   

z ………….(4.22)
Z* 
ZpscT  BVL 1  S gi  Swc   B  B 
1  c fma P  1  S w  
 Z scTsc  PL  P 
1  f m  f ash  
Bwi
w wi

 VL Pi 
 B 1  f m  f ash   ZiTpsc
PL  P
  P  f m cm  c fmi 
 Z scTsc 1  f m 

4.3 New Method – Temperature-Dependent Adsorption Models

Previous modifications of the material balance equation for unconventional gas reservoirs,
especially for coal and shale, have included the adsorption capability of these resources. The
additional term of gas adsorption has been modelled based on Langmuir isotherm. Langmuir
isotherm describes an equilibrium relationship between the free gas and the adsorbed gas
under isothermal conditions. Thus, gas adsorption has been expressed only as a function of
pressure. Therefore, the use of Langmuir isotherm to account for the adsorption capacity of
shale within the material balance equation will be inadequate, since the effects of varying
temperature are neglected.

Temperature plays an important role in gas adsorption. Hence any adsorption model should
be capable of expressing the adsorption as a function of both pressure and temperature. The
majority of gas in shale gas reservoirs is from adsorbed gas, and an ample knowledge of gas
adsorption behaviour over a wide range of pressure and temperature is needed. According to
Gasparik et al., (2015), sorption isotherms need to be measured over an extended range of
pressures greater than 20 MPA and temperatures greater than 100 °C in order for the
isotherm to be considered at in-situ reservoir conditions typical of shale gas reservoirs.
Langmuir model fails to provide such knowledge, as adsorption can be modelled only at a
single temperature.

86
Since gas adsorption is also a function of temperature, geothermal gradients will contribute
substantially to the adsorption capacity of these gases. An example is the black Warrior basin,
where there is temperature variation from about 26.85 C to 51.85 C within a 0.3-1.8 km depth
range (Charoensuppanimit et al., 2015; Pashin and McIntyre, 2003).

Furthermore, a key assumption of Langmuir isotherm is the homogeneity of the adsorbent.

However, this may not be suitable or true for shale gas systems, since different materials such
as clay minerals and kerogen may contribute to gas adsorption on shale (Lu et al., 1995).

Over the years, several researchers have modified the classic Langmuir model to include a
temperature term that makes it appropriate to describe the adsorption capacity of shale/coal.
Lu et al., (1995) proposed the use of bi-Langmuir model to describe the adsorption of gas at
several temperatures. One term of the model describes gas adsorption on clay minerals, while
the other accounts for gas adsorption on kerogen. The model, therefore, is suitable for non-
homogeneous adsorbents, especially in the Devonian shale where two mineral compositions,
of clay and kerogen, are said to be mainly responsible for gas storage. Ye et al., (2016) also
proposed a variation of the classic Langmuir isotherm equation by introducing an exponential
relation that expresses the Langmuir volume ( VL ) as dependent on temperature. Thus, the

constant Langmuir volume ( VL ) as expressed in the Langmuir equation, is replaced by a VL

that is a function of temperature.

Based on application of the models to different sets of shale gas data, it has been established
that modified Langmuir model with temperature dependency can accurately describe shale
gas adsorption under various pressures and temperatures (Fianu et al., 2018; Lu et al., 1995;
Ye et al., 2016).

Bi-Langmuir model is expressed as:

   E1    E2  
 f1k1T exp  RT  p 1  f1 1 k2T exp  RT  p 
1/2 1/2

V  VL        ……………....(4.23)
1  k T 1/2 exp   E1  p  E 
1 1 k2T 1/2 exp  2  p 
 1 1   
 RT   RT 

 E   E 
b1  k1T 1/2 exp  1  , b2  k2T 1/2 exp  2 
 RT   RT 

87
All variables with the exception of pressure (P) and temperature (T) are obtained by matching
the adsorption data using regression analysis.

Hence,

 f b p 1  f1  b2 p 
V  VL  1 1   ……………………..(4.24)
1  b1 p 1  b2 p 

Exponential model for adsorption is given as:

Vs exp( DT T )  p
V ………………………(4.25)
1  p

T
  1/ c , c 
B
A exp  
T 

The coefficients A and B are obtained by matching the adsorption data through regression
analysis.

To incorporate Equations 4.23 (Bi-Langmuir model) and 4.25 (Exponential model) into the
material balance equation, Ahmed and Roux's (2006) approach has in this research been used,
expressing the material balance as an equation of a straight line with x and y . This method
has the advantage of avoiding the use of an iterative process, as employed by the other
methods, in solving the resulting equations. It offers the advantage of minimizing costs
associated with obtaining pressure survey data, since the average reservoir pressures can
easily be obtained by using the initial reservoir pressure, Langmuir isotherm and cumulative
gas production. Table 4.1 summarises the different methods used in the application of MBE
in unconventional gas reservoirs including the new methodology developed in this thesis.

88
Table 4.1: Summary of methods used in applications of MBE in unconventional gas
reservoirs

Method Description Advantages/Limitations

King’s Method ( 1993) Modified P/Z plot for Only suitable for under-
unconventional reservoirs pressured coal and not
like coal/shale gas under-saturated coal.
reservoirs. Effect of Iterative solution made
adsorption was introduced calculation tedious.
into the methodology. Langmuir Isotherm used in
prediction of gas
adsorption.
Jensen and Smith Modified Modified Kings Use of Langmuir isotherm
Method (1997) Methodology with a more meant only single
practical based evaluation temperature could be used
of the estimated recovery for adsorption calculations
and remaining reserves.
Neglected the effect of
water saturation in their
solution. Suitable for
reservoirs with high
adsorption rate
Seidle Method –Modified Another modification of Avoided the iterative
King Method (1999) Kings methodology but with solution adopted by King in
the assumption of constant its calculation. Langmuir
water saturation over time isotherm used to account
instead of average water for gas adsorption.
saturation
Ahmed and Roux Method Expressed the material Method is applicable to any
(2006) balance as an equation of a coal/shale that behaves
straight line to enable according to Langmuir

89
 calculations of original gas isotherm. Limited to only a
in place, and also prediction single temperature for
of average reservoir evaluation of adsorption
pressure
Moghadam Method (2009) Similar but more rigorous Used to define total
and advanced form of the compressibility of the
material balance proposed system for analysing fluid
by King. flow in unconventional gas
reservoirs. Adsorption was
described by Langmuir
isotherm
Firanda Method (2011) Introduced different drive Offered similar
mechanisms in the material methodology used by
balance equation proposed Ahmed and Roux by
by King. These included expressing material balance
water expansion, rock as a straight line. Adsorption
compaction, connate water isotherm used is Langmuir
expansion and moisture isotherm.
expansion.
New Approach (This Similar approach adopted Avoids iterative solution of
Research) by Ahmed and Roux by Kings approach. Adopts
expressing material balance temperature-dependent gas
as a straight line. Average adsorption model such as
reservoir pressure as well as Bi-Langmuir and
future performance of the Exponential model in its
reservoir can be obtained. methodology to account for
gas adsorption at several
temperatures.

90
4.4 Methodology
From Equation 4.7, the produced gas can be expressed as

Gp  G  GF  GA  GR ……………………..(4.26)

GA which is the gas adsorbed can now be expressed as a function of both pressure and

temperature. Inserting both the Bi–Langmuir and exponential formulae into GA yields the

following equations respectively:

 f b p 1  f1  b2 p 
GA  V  VL  1 1   …………….(4.27)
1  b1 p 1  b2 p 

Vs exp( DT T )  p
GA  V  ………………………..(4.28)
1  p

From Equation 4.10, V can be replaced with either one of the temperature-dependent
models. For Bi-Langmuir model, the modified material balance equation can therefore be
expressed as:

BI-Langmuir model

   f b p 1  f1  b2 p   
1359.7  B  Gc  VL  1 1   
BwW p Eg    1  b1 p 1  b2 p   
Gp   Ah    7758 Ah 1  S wi  Egi ……(4.29)
1   c f P 
  f wi wi   
 7758    P c  S c  1  S 
wi  gE 
 
 1   c f P  

Neglecting rock and fluid compressibility:

   f b p 1  f1  b2 p   
1359.7  B  Gc  VL  1 1   
G p  BwWp Eg  Ah    1  b1 p 1  b2 p     7758 Ah 1  S  E …….(4.30)
wi gi
 
 7758  P  c f  S wi cwi   1  S wi   Eg 

Equation 4.29 can be expressed as an equation of a straight line ,i.e y  mx  c , where

91
 BwW p Eg
y  Gp  ……………………..(4.31)
1   c f P 

  f b p 1  f1  b2 p   7758  P  c f  Swi cwi   1  Swi  Eg

x  1359.7  B  Gc  VL  1 1    …….(4.32)

  1  b1 p 1  b2 p   1   c f P 

Neglecting rock and fluid compressibility:

y  Gp  BwWp Eg …………………….(4.33)

  f b p 1  f1  b2 p  
x  1359.7  B  Gc  VL  1 1 
    7758  P  c f  S wi cwi   1  S wi   Eg ………(4.34)
  1  b1 p 1  b2 p  

When x is plotted against y , this will yield a straight line with slope m and intercept a .

a  7758 Ah 1  S wi  Egi

m  Ah

Exponential Model
For the Exponential model, the modified material balance can be expressed as:

  Vs exp( DT T )  p  
1359.7  B  Gc   
BwW p Eg   1  p  
Gp   Ah   7758 Ah 1  S wi  Egi
………(4.35)

1   c f P    7758  P  c f  S wi cwi   1  S wi   Eg 
 

 1   c f P  


Neglecting rock and fluid compressibility:

  Vs exp( DT T )  p  
1359.7  B  Gc   
G p  BwWp Eg  Ah   1  p    7758 Ah 1  S wi  Egi …….(4.36)
 7758  P c  S c  1  S  E 
   f wi wi   wi  g 

Equation 4.35 Can be expressed as an equation of a straight line i.e. y  mx  c , where

BwW p Eg
y  Gp  …………..(4.37)
1   c f P 

92
  V exp( DT T ) p  7758  P  c f  Swi cwi   1  Swi  Eg
x  1359.7  B  Gc  s  …….(4.38)
 1  p  1   c f P 

Neglecting rock and fluid compressibility:

y  Gp  BwWp Eg ……..(4.39)

 V exp( DT T )  p 
x  1359.7  B  Gc  s   7758  P  c f  S wi cwi   1  S wi   Eg …….(4.40)
 1  p 

4.4.1 Average Reservoir Pressure

Ahmed and Roux's (2006) method of expressing the material balance as an equation of a
straight line with variables x and y can be used in estimating the average reservoir pressure
based on historical production data alone. The proposed method by Ahmed and Roux (2006)
assumed a single layer reservoir system with a knowledge of the initial reservoir pressure,
Langmuir relationship and an initial gas in place.

The material balance equation can be expressed in terms of G as:

bp
G p  Wp Eg  G  1359.7  B Ah Vm ……………(4.41)
1  bp

The initial gas in place at the initial reservoir pressure of pi is given as:

G  1359.7  B Ah  Gc …………….(4.42)

bp
Where Gc  Vm
1  bp

Combining Equation 4.41 and Equation 4.42 and rearranging results in:

 p   1  bpi   1 
        G p  BwWp E p    1  0 ………………..(4.43)
 pi   1  bp   G 

Equation 4.23 can be solved iteratively for the average reservoir pressure using Newton

Raphson method. By assuming an average reservoir pressure, denoted as pold , a new

pressure pnew can be easily calculated from the expression

93
 f  pold 
pnew  pold  ……………………………………….(4.44)
f \  pold 

Where

 p   1  bpi   1 
f  pold           G p  BwWp E p    1 ………………..(4.45)
 pi   1  bp   G 

 1  1  bp  bpold 1  bpi   198.6 BwWp

f \  pold     i
 2
 ……………….(4.46)
 i 
p 1  bpold  pi 1  bpold   ZTG

198.6 p
Eg  scf/bbl…………………….(4.47)
ZT

6
Convergence of the iteration is said to be achieved once pold  pnew  10 and a solution is

obtained, otherwise the iteration is continued on by setting pold to pnew .

The above methodology for obtaining the average reservoir pressure is used only when the
adsorption process is represented by Langmuir isotherm. By incorporating a temperature-
dependent adsorption model, Equation 4.45 and Equation 4.46 can be modified as follows:

Bi-Langmuir model:

 
 
f  pold   
1   f1b1 p  1  f1  b2 p   1  G  198.6 pBwWp  ……..(4.48)
 f b p  1  f1  b1 pi    p  1
   1  b1 p 1  b2 p  G  ZT 
 1 1 i    
 1  b1 pi  1  b2 pi 

 f1b1

f1b12 p

1  f1  b2  1  f 2  b22 p 
 2 
 1  b1 p 1  b1 p 2 1  b2 p 1  b2 p   198.6 BwWp
f  pold  
\
 ……………(4.49)
 f1b1 pi 1  f1  b2 pi  GZT
  
 1  b2 pi
 
1 b p
1 i

Exponential Model

 p   1   pi   1 
f  pold           G p  BwWp E p    1 ……………………….(4.50)
 pi   1   p   G 

94
  1  1   p  bpold 1   pi   198.6 BwWp
f \  pold     i
 2
 ………………(4.51)
 i 
p 1   pold  pi 1   pold   ZTG

4.4.2 Prediction of Reservoir Production Performance

Future reservoir production performance can be predicted d with the modified methodology,
which includes the temperature-dependent gas adsorption models. A finite difference
scheme can be adopted with the modification to predict future reservoir performance. The
modified material balance equation can be expressed in the following form, neglecting water
and rock compressibility after finite difference approximation.

Bi-Langmuir model:

 bp b p 
G p  BwWp Eg  G  a1  1  2   a2  Egi  Eg  …………………(4.52)
1  b1 p 1  b2 p 

and expressing it in a finite difference scheme after differentiation is:

 b  p n  p n 1  b  p n  p n 1  
G pn 1  G pn   BwnWpn Egn  Bwn 1Wpn 1Egn 1   a1     a2  Egn  Egn 1  …..(4.53)
1 2

 1  b p n 1 2 1  b2 p  
n 1 2 
 1

Exponential Model

 p 
G p  BwWp Eg  G  a1    a2  Egi  Eg  ……………….(4.54)
1   p 

   p n  p n 1  
G n 1
 G  B W E  B W
n n n n n 1 n 1
E n 1
  a1    a2  Egn  Egn 1  …….(4.55)
 1   p  
n 1 2 
p p w p g w p g

a1  1359.7 AhBVL and a2  7758 Ah 1  S w 

The steps needed to carry out a prediction of future reservoir performance are outlined by
the algorithm (see Figure 4.1), and the following equations are also useful:

The gas saturation equation is given as

95
 Bwn 1W pn 1
1  Swi    pi  p n 1
 c  cw S wi  
7758 Ah
f
n 1
S  ……………….(4.56)
1   pi  p c 
g n 1
f 

krg n 1
The relative permeability ratio of at the gas saturation S g can be used to estimate the
krw

gas water ratio as

n 1
n 1
krg   w Bw 
(GWR)    …………………………………….(4.57)
krw   g Bg 

96
Figure 4.1: Algorithm for predicting future production using temperature-dependent models

97
  GWR    GWR 
n 1 n

G pn11  G pn 
2
Wpn 11  Wpn  …………………(4.58)

Gas and water flow rates are calculated respectively as:

0.703hk  krg  p pwf 

n 1 n 1
n 1
Q  ……………………………(4.59)
 r  
g

T   g Z  ln  e   0.75  s 
  rw  
avg

n 1 n 1
n 1
k    g Bg 
Q   rw    Qgn 1 ………………………………(4.60)
k 
  w Bw 
w
 rg 

The average gas flow rate as reservoir pressure declines from p n to p n1 is

Qgn  Qgn 1
 Qg  avg

2
……………………………………………..(4.61)

The incremental time needed for the incremental gas production during the pressure drop is

G p G pn1  G pn
t   …………………………………………………..(4.62)
Q  g avg Q g avg

4.5 Results and Discussion

To verify the application of the modified method, simulation data provided by King (1990),
Ahmed and Roux (2006) and Seidle (1999) have been adopted to confirm the initial gas in
place and other historical production data. King (1990) used finite –difference generated data
to validate his method as opposed to actual field data. This synthetic data has been used by
other researchers in material balance calculations for unconventional gas reservoirs and have
also been used throughout this thesis. The reservoir can be described by a 2-D areal model
that contains a single coal well. The reservoir is described as being homogeneous with the
well-draining a 320-acre coal deposit. Initial water saturation is reported as 0.95 with an initial
reservoir pressure of 1500 psia. The estimated initial gas in place is 12.763 Bcf with actual
production data given in Appendix 4. The adsorption capacity of this reservoir is expressed by
the Langmuir isotherm, with Langmuir parameters given in Appendix 4, Table 4.8.
98
To obtain the parameters for the adsorption models with temperature dependence, gas
adsorption has to be expressed at different temperatures, and a regression analysis
performed on the data to obtain the relevant model parameters. Since no data have been
provided relating to the experimental adsorption capacity of the coal reservoir at several
temperatures, obtaining the model parameters for the temperature-dependent models was
a challenge. To obtain the model parameters for the temperature-dependent adsorption
model, a similar method to that of Wang (2016) and Yue et al., (2015) was adopted, where
unknown independent parameters in Equations 4.24 and 4.25 are determined from Langmuir
isotherm curve at provided temperature conditions by nonlinear regression (see Figures 4.2,
4.3 and 4.4). In the case where multiple gas adsorption data at different temperatures are
available, temperature-dependent models could easily be applied to obtain model
parameters to be used in simulation studies.

360 564.67°R

340

320
V, Scf/ton

300

280

260

240
560 660 760 860 960 1060 1160 1260 1360 1460 1560
Pressure,Psi

Data Langmuir Isotherm

Figure 4.2: Regression with Langmuir isotherm parameters (data from Ahmed and Roux,
2006)

99
 360

340
564.67°R
320
V, Scf/ton

300

280

260

240
560 660 760 860 960 1060 1160 1260 1360 1460 1560
Pressure,Psi

Data Exponential

Figure 4.3: Shale gas adsorption capacity at selected temperature using Exponential Model
(data from Ahmed and Roux, 2006)

Bi-Langmuir Model
350

340

330 564.67°R
320

310
V, Scf/ton

300

290

280

270

260

250
560 760 960 1160 1360 1560
Pressure,Psi

Data Bi-Langmuir

Figure 4.4: Shale gas adsorption capacity at selected temperature using Bi-Langmuir Model
(data from Ahmed and Roux, 2006)
100
Tables 4.2 and 4.3 show the model parameters obtained for both Bi-Langmuir and
Exponential models.

Table 4.2: Bi-Langmuir parameters

VL scf/ton 428.49912

k1 1/psi 0.0657363

k2 1/psi 0.0630964

E1 cal/mol -0.036275

E2 cal/mol -0.012802

f1 0.97

Table 4.3: Exponential Parameters

Vs scf/ton 491.88

DT 1/°F 0.000244

A °F/psi 0.0655498

B °F 0.0199987

101
Using the above data, the OGIP has been determined for both temperature-dependent
models. By neglecting the formation and water compressibility, Table 4.4 shows the case of
Bi-Langmuir model and Exponential model.

Table 4.4: Expressing MBE as a straight line without compressibility

Bi-Langmuir model Exponential

p psia Langmuir Gp Wp
V
MMscf MMscf
y x y Mscf x
scf/ton MMscf
1500 345.097 0 0 0 0 0 0
1315 335.90 265.086 0.15749 348.19 19195.6 338.91 19811.92
1021 316.23 968.41 0.290238 1084.6 65137.81 1071.635 66336.73
814.4 296.53 1704.033 0.368292 1819.5 111001.7 1806.609 112541.5
664.9 277.33 2423.4 0.425473 2530.8 155607.3 2518.803 157240.5
571.1 262.14 2992.901 0.464361 3092.7 190849.8 3081.53 192388.9

A plot of x and y will yield a straight line on a Cartesian scale (see Figures 4.5 and 4.6). The
slope of the straight line can then be used to determine the drainage area and subsequently
the initial gas in place. A slope of 15,946 acre-ft. and 15,933 acre-ft. was obtained for Bi-
Langmuir and Exponential models respectively, at the current reservoir temperature. Table
4.5 shows the corresponding drainage area and OGIP.

102
 3500
Slope (acre-ft.)
3000 New Method 15946
Ahmed and Roux Method 15900
2500

2000
Y-MMscf

1500

1000

500

0
0 50000 100000 150000 200000 250000
X=2311.49(345.1-V)-3.879Eg
Bi-Langmuir Ahmed and Roux Method

Figure 4.5: MBE as a straight line for Bi-Langmuir model without compressibility

103
 3500 Slope (acre-ft.)
New Method 15933
3000 Ahmed and Roux Method 15900

2500

2000
Y-MMscf

1500

1000

500

0
0 50000 100000 150000 200000 250000
X=2311.49(345.1-V)-3.879Eg
Exponential

Figure 4.6: MBE as a straight line for Exponential model without compressibility

3500
Slope ( acre-ft)
3000 New Method 16000
Ahmed and Roux Method 15957
2500

2000
Y-MMscf

1500

1000

500

0
0 50000 100000 150000 200000 250000
X
Bi-Langmuir Ahmed and Roux Method

Figure 4.7: MBE as a straight line for Bi-Langmuir model with compressibility
104
 3500

Slope ( acre-ft)
3000
New Method 15989
Ahmed and Roux Method 15957
2500
Y- MMscf

2000

1500

1000

500

0
0 50000 100000 150000 200000 250000
X

Exponential Ahmed and Roux Method

Figure 4.8: MBE as a straight line for Bi-Langmuir model with compressibility

4.5.1 OGIP and Average Pressure

Firstly, by neglecting rock and fluid compressibility, the values of the OGIP obtained for both
Bi-Langmuir and Exponential models are 12.757 bscf and 12.742 bscf respectively (See Table
4.5). Compared with Ahmed and Roux’s method, which gave OGIP as 12.72 bscf, the new
methodology of incorporating temperature-dependent models into material balance
calculation resulted in a much closer estimation of the OGIP obtained from the field data as
12.763 bscf.

However, the inclusion of compressibility calculation resulted in a slight over-prediction of

the original gas in place, as shown in Table 4.6. Although using Bi- Langmuir model in the
material balance calculation resulted in the correct estimation of the drainage area of 320
acres, the OGIP was slightly higher than actual data reported. Ahmed and Roux’s method
under-predicted the OGIP when rock and fluid compressibility were included in the
calculation (See Table 4.6).

105
Furthermore, an estimate for the initial gas in place can be made at different temperatures
of the reservoir, something which cannot be done when Langmuir isotherm is used in the
calculation of the adsorption potential of the reservoir. Figure 4.9 shows estimated gas in
place when both temperature-dependent models are used with decreasing gas in place as the
reservoir temperature increases. Thus, it can be concluded that at a much higher reservoir
temperature, the original gas in place in a CBM/shale gas reservoir will be much lower than
when the reservoir temperature is low. This is because the contribution by gas adsorption will
be smaller at higher temperatures, since gas adsorption is exothermic.

Table 4.5: OGIP determination using BI-Langmuir model and Exponential model without
compressibility

Method Slope (acre- ft) Drainage area (acre) OGIP (Bscf)

Data 320 12.763
Ahmed and Roux Method 15900 318 12.72
Bi-Langmuir 15946 318.92 12.757
Exponential Model 15933 318.66 12.742

Table 4.6: OGIP determination using BI-Langmuir model and Exponential model with
compressibility

Method Slope (acre- ft) Drainage area (acre ) OGIP (Bscf)

Data 320 12.763
Ahmed and Roux Method 15957 319 12.76
Bi-Langmuir 15946 320 12.800
Exponential Model 15989 319.78 12.791

106
Figure 4.9: OGIP at different temperatures using Bi-Langmuir and Exponential model

107
Using Equations 4.48 and 4.50, the average reservoir pressure for the reservoir can be
determined by using only the historical cumulative production values. An excellent pressure
match has been obtained for both cases of temperature-dependent models (see Figures 4.10
and 4.11). The pressure match results are in agreement with both the reported data and
Ahmed and Roux’s method. This shows that the new methodology is capable of predicting
average reservoir pressure using historical production data.

1600

1400

1200

1000
Pressure, Psia

800

600

400

200

0
0 500 1000 1500 2000 2500 3000 3500 4000
Time, days

Data Bi-langmuir Ahmed and Roux method

Figure 4.10: Pressure match for the new method using Bi-Langmuir model

108
 1600

1400

1200

1000
Pressure,Psia

800

600

400

200

0
0 500 1000 1500 2000 2500 3000 3500 4000
Time, days

Data Exponential Ahmed and Roux method

Figure 4.11: Pressure Match for the new method using Exponential Model

109
4.5.2 Reservoir performance prediction
Both temperature-dependent models have been used to estimate production performance
of the well. The validity of the modified methodology with temperature-dependent models
has been tested against the cumulative gas production. The performed simulation in owed
for different pressure steps to be used, and it was found that the choice of smaller pressure
steps resulted in a much more accurate match to the production data. See Figures 4.12 and
4.13 for a match of the simulated results with those of the cumulative production data. Using
Bi-Langmuir model and Exponential model in the methodology showed an excellent match
with the reported cumulative gas production when a time step size of 20 psi was chosen.
However, using real pressure time steps showed higher total prediction for both models. This
result is also in congruence with results reported by Ahmed and Roux (2006).

4000

3500

3000
Cummulative Production ,MMscf

2500

2000

1500

1000

500

0
400 600 800 1000 1200 1400 1600
Pressure,Psia

Pressure drop of 20 psi Real Data Using real pressure time steps Ahmed and Roux method

Figure 4.12: Predicted total production versus reservoir pressure for Bi-Langmuir model

110
 4000

3500

3000
Cummulative Production ,MMscf

2500

2000

1500

1000

500

0
400 600 800 1000 1200 1400 1600
Pressure, Psia

Pressure drop of 20 psi Actual data Using real data time steps Ahmed and Roux method

Figure 4.13: Predicted total production versus reservoir pressure for Exponential model

111
4.6 Summary
Temperature-dependent models have been introduced into the material balance equation
for unconventional gas reservoirs, such as CBM and shale gas reservoirs. These reservoirs
have a substantial portion of their gas in place coming from gas adsorption. The limitations of
the use of Langmuir isotherm within the material balance equation make it necessary to
introduce gas adsorption models that can predict adsorption as a function of both pressure
and temperature. Two different temperature-dependent models, namely, Bi-Langmuir model
and Exponential model, have been introduced and subsequently incorporated into the
material balance equation. By modifying the material balance to include a temperature-
dependent gas adsorption model, a new model was developed and applied to available data.
The following conclusions are made based on the results of validation of the model:

1. Accurate adsorption potential of the reservoir can be modelled once several gas
adsorption data are available at several temperatures. Langmuir isotherm, although
useful, fails to model gas adsorption at several temperatures. Results compared with
earlier methodologies, like Ahmed and Roux, demonstrated excellent predictions, and
in some cases a closer match to reported gas in place.
2. Extrapolation of adsorption capacity can be made at actual reservoir temperature with
temperature-dependent gas adsorption models, thereby giving an accurate
representation of adsorption capacity that can ultimately improve the estimation of
gas in place values.
3. More accurate estimation of reservoir performance can be made with the use of
temperature-dependent gas adsorption models in material balance calculations once
the adsorption capacity of the reservoir is expressed at different temperatures.

112
Nomenclature
p = pressure

T = temperature

G p = produced gas

G = gas originally adsorbed

GF = original free gas

GA = adsorbed gas currently

GR = remaining free gas

 = porosity

Z = gas compressibility factor

Qg = gas flow rate

c f = compressibility of the formation

Swi = initial water saturation

b = Langmuir constant

Bw = water formation volume factor

Gc = gas content at critical desorption pressure

Qw = water flow rate

Vs = theoretical maximum adsorption capacity

DT = reduction coefficient

V = gas content at pressure p

113
A = drainage area

h = average thickness

VL = Langmuir Volume

R = universal gas constant

E g = gas expansion factor

E gi = initial gas expansion factor

Sw = water saturation

cwi = water compressibility

Wp = produced water

S g = gas saturation

GWR = gas water ratio

B = bulk density

Gi = initial gas in place

krg = relative permeability of gas

krw = relative permeability of water

u g = viscosity of gas

uw = viscosity of water

(Qg )avg = average gas rate

Bg = gas formation volume factor

114
 Chapter 5
Numerical Simulation of Gas Adsorption in Shale Gas
Reservoirs

5.1 Introduction
Accurate simulation and modelling of shale gas reservoirs is deemed crucial for efficient
exploitation of these resources. Obtaining realistic results for resource estimation and
performance predictions has a significant impact on the economics of the operating
companies and all interested parties. Integrating all the unique characteristics of shale gas
reservoirs within a single reservoir simulator for accurate predictions of future performance
is considered a difficult task. For many years now, various researchers have tried to tackle
some of these challenges, which include, but are not limited to, how the natural fractures are
simplified and represented in a simulator, the transport of gas within the matrix and fractures,
adsorption and desorption phenomena within the shale gas system, and how the fractures
are propagated within the shale formation upon hydraulic fracturing.

Naturally fractured reservoirs are characterized by the presence of two distinct porous media:
the matrix, and fractures. Naturally fractured reservoirs have been referred to as a dual
porosity system because of the two porous media that are present (Barenblatt et al., 1960).
The matrix feeds fluid locally to the fractures, and the fractures form a continuous system
connected to the well.

In the dual porosity model, which was further modified by Warren and Root (1963), the matrix
does not contribute directly to the wellbore. The system is seen as an orthogonal set of
intersecting fractures and sugar cubic matrix blocks (Figure 5.1) and by using differential
equations, analytical solutions could be obtained for well test analysis.

115
Figure. 5.1: Sugar cube model by warren and root

Warren and Root (1963) assumed that flow from the matrix to the fracture occurs under a
transfer function with Darcy characteristics, and also that flow occurs under pseudo-steady
state conditions in the matrix blocks with a single value assigned to the pressure in the blocks.
The pressure differential between the matrix and the fractures, therefore, determines the
mass transfer rate of the fluid. Thus, the inter-porosity flow has been described by two main
mechanisms, which are the pseudo-steady state and the transient flow. Warren and Root
(1963) predicted that on a semi-log plot of the test data, two parallel straight lines will
develop. The slope of the parallel lines represents the flow capacities of the formation,
whereas the vertical separation of the lines represents the storage capacity of the fractures.
This is shown in Figure 5.2

116
Figure 5.2: Pseudo-steady state matrix flow pressure build-up curve (Warren and Root,
1963)

Odeh (1965) developed a simplified model with mathematical equations that described the
unsteady state behavior of fractured reservoirs. He concluded from his studies that there was
no difference between fractured reservoirs and homogeneous reservoirs on the basis of
drawdown and build-up curves using field-measured data. This, according to Kazemi (1969),
contradicted the results of Warren and Root (1963), but with a smaller block dimensions and
a higher permeability, their results remains valid. Odeh (1965) used a similar model as Warren
and Root (1963), although his results did not show the two parallel straight lines as depicted
in Fig. 5.2.

The dual porosity model by Warren and Root (1963) was extended to include transient flow
in the matrix block by Kazemi (1969). The main distinction for Kazemi (1969) was the use of
transient flow within the matrix, instead of assuming a pseudo-steady state as was done by
Warren and Root (1963) . Kazemi (1969) also used the slap model (sheets of parallel fracture
sets, Figure 5.3) to describe the reservoir. By considering a direct flow to the well, he found
that similar results to Warren and Root (1963) were obtainable without affecting the results

117
significantly way, except for a smooth transitional zone which occurs due to the non-
permanent flow regime of the fluid flow from the matrix to the fractures. Thus deviations will
occur in the transitional period when the rock matrix is described by a pseudo-steady state
regime.

Figure 5.3: Dual porosity idealisation (Kazemi, 1969)

de Swaan (1976) expanded on the dual porosity model by using transient flow as the basis of
flow from the matrix to the fractures. He defined his transient model by the intrinsic
properties of the matrix and the fractures, as opposed to the bulk properties of the matrix
and fractures used by both Warren and Root (1963) and Kazemi (1969). Both a slab matrix
shape model and a spherical matrix block solution were also presented in the same study.

Serra et al., (1983) also used the transient model to describe the fluid flow from the matrix to
the fractures, and based their model on the use of intrinsic properties rather than bulk
properties. Serra et al., (1983) model was very similar to de Swaan (1976), except for the use
of a different storativity ratio and inter-porosity coefficient.

Najurieta (1980) showed in a simplified description the pressure behavior of a naturally

fractured reservoir based on appropriate solutions of the deSwaan differential equation. In

118
his study, he deduced equations that described the transitional period, taking into account
the unsteady state behavior of the matrix. He also showed that the behavior of a uniformly
fractured reservoir can be fully described by four parameters, each of which is a function of
two or more of the five basic reservoir parameters (fracture and matrix porosity, fracture and
matrix permeability, and fracture spacing).

The method of solving for the system of equations that characterises shale gas reservoirs have
been predominately finite difference approximation with several researchers choosing to
implement the solution based on popular methods like newton’s method or by specially
formulated solutions. Wu et al., (2013) proposed a generalised mathematical model and
numerical approach for unconventional gas reservoirs. The model and numerical scheme
were based on generalised flow models using unstructured grids. A fully implicit scheme was
adopted to solve the discrete nonlinear equations using newton iteration method. They
utilised a control volume or integrated finite difference concept to discretise the domain
space. Their approach to solving the system of equations that described the shale reservoir
was to compute the spatial discretisation, time discretisation and finally using iterative
approaches to solve the nonlinear algebraic equations.

Sun et al., (2012) introduced a new technique for the numerical solution of various partial
differential equations (PDE) that governs the flow and transport phenomena in porous media.
The authors introduced the use of shifting matrices to transfer information between cell-
centers and face center and vice versa. Since many programming languages are inefficient
when it comes to looping sequences, the technique eliminates the use of loops within the
body of the code and therefore reduces the CPU time. Finite difference method was used for
the implementation of their study however it was recognised that others like finite volume or
finite element may equally be used.

An equation type approach for the numerical solution of partial differential equations that
governs transport phenomena in porous media was developed by Sun et al., (2012b).
According to the authors, there is no need to reduce the number of governing equations by
any mathematical manipulations. In their approach, the physics of the governing equation is
separated from the solver and direct solution is obtained without any iterations. They called

119
this method an equation-type method or experimenting pressure method. One other
advantage for using this method is that the boundary conditions are easily implemented.

In this chapter, a review of the concept of dual porosity and dual permeability in shale gas
reservoirs is conducted and a mathematical model have been developed to describe this
concept with newton iterations implemented to solve the nonlinear PDE. A temperature-
dependent gas adsorption model is then incorporated into the simulator to allow for the
examination of thermal stimulation as an enhanced gas recovery mechanism in shale gas
reservoirs.

5.2 Mathematical Formulation of Isothermal Model in Shale Gas

Reservoir
Mathematical formulations are required in order to model the flow of gas through the
nanopore networks of shale reservoirs. Several considerations need to be made concerning
flow through shale, such as flow due to pressure forces, diffusion, and slippage effect. These
need to be included in the mathematical formulation to ensure accurate representation of
gas flow through shale. Table 5.2.1 show the flow regimes as a function of Knudsen number.
The concept of dual continuum is used in the simulation of shale gas, due to the presence of
both matrix and fractures. Free gas exists primarily in the fractures, whereas in the matrix,
both free gas and adsorbed gas coexist (See Fig. 5.4). Gas transfer occurs through pressure
difference between the matrix and the fractures, and eventually to the wellbore for
production (Guo et al., 2015). The following assumptions are made for the purposes of
developing the mathematical model for this study. Some of the assumptions will change for
the purpose of introducing thermal application in shale gas reservoirs.

1. A single gas component is assumed to be present in our reservoir (methane) and only
one phase flow is considered
2. Gravity effect and heterogeneity on the gas flow is ignored
3. The gas present is considered to be ideal, with a gas deviation factor equalled to one
4. Isothermal conditions of flow are assumed throughout the reservoir
5. Formation rocks are considered incompressible and porosity is considered constant
throughout the formation
6. Gas viscosity is also considered constant and does not change with pressure depletion

120
 7. Langmuir isotherm is used to describe the gas adsorption and desorption phenomena
in the reservoir, with instantaneous equilibrium achieved at any pressure.

Figure 5.4 Gas storage in shale (Guo et al., 2015)

Table 5.2.1: flow regimes as a function of Knudsen number (Lee and Kim 2016)

Knudsen Number (Kn) Flow regime

0-10-3 Darcy flow (No slip)
10-3-10-1 Slip Flow
10-1-101 Transition flow

101- ∞ Free –Molecular flow

The continuity equation for every grid block can be expressed in a general format as

dM
 .   u   Q ……………………………………………………… (5.1)
dt

M = mass accumulation term

 = density

u= velocity

121
Q = the source term

t = time.

The fracture –matrix interaction in shale gas reservoirs can be modelled either by using a dual
continuum approach or by discrete fracture. A continuum approach, such as dual porosity
and dual permeability, has been adopted over the discrete fracture model because it is less
rigorous and not as computationally intensive compared to the discrete fracture model. Two
mass balance equations can be obtained for both matrix and fracture systems (Warren and
Root, 1963, see Figure 5.5).

122
Figure 5.5: Transport of gas through shale towards production (Modified from Guo et al.,
2015)

 dM 
    .  u  m   Q m ………………………………………………(5.2)
 dt m

123
 dM 
    .   u   f   Q  f ………………………………………… (5.3)
 dt  f

Where subscript m and f refers to the matrix and fracture respectively.

5.2.1 Mass Accumulation Term

The mass accumulation term can be expressed in a general form as

M    S g  g ……………………………………………………………………… (5.4)

For a gas reservoir, S g = 1, which refers to the fraction of pore volume occupied by the gas

phase and subscript g refers to gas phase.

Considering the matrix system of shale gas reservoirs, the mass accumulation can be written
for both free gas and adsorbed gas as follows:

M free    S g  g ……………………………………………………………… (5.5)

M ads   1    qads ……………………………………………………… (5.6)

Where qads is gas adsorption volume per unit bulk volume; this can be represented by the
use of Langmuir isotherm.

 s M g VL pm
qa  ……………………………………………………………(5.7)
Vstd PL  pm

Vstd = mole volume under standard conditions

PL = Langmuir pressure

VL =Langmuir volume

 s = density of shale

124
pm = pressure for the matrix

Combining the above equations, the mass accumulation for the matrix can be represented as

M    g  1    qads  …………………………………………………………….(5.8)

With the corresponding partial differential form given as

dM    g   1    qads 
  ……………………………………………………(5.9)
dt t t

Using the equation of state (EOS), the density of the gas can be expressed as

m pM 
g    p ………………………………………… (5.10)
V ZRT Z

For an ideal gas,

 g   p ………………………………………………………….. (5.11)

qads qads pm    s M VL pm  pm MPLVL s pm ……….(5.12)

 *   * 
t pm t pm  Vstd PL  pm  t Vstd  PL  pm  t
2

Therefore, the accumulation term for the matrix can be expressed as

 dM     g   1    qads   1  m  MPLVL s  pm

     m  2  …………………(5.13)
 dt m t t  Vstd  PL  pm   t

Due to the presence of only free gas within the fracture system, the mass accumulation can
be expressed as

 dM     g  p
     f  f ………………………………………………. (5.14)
 dt  f t t

5.2.2 Flow Vector Term

Gas flow in shale gas reservoirs is distinguished by flowing media such as the matrix and
fracture system (Guo et al., 2015). The flow occurs from the matrix to the fractures and then
subsequently to the wellbore for production (see fig 5.5).

125
Assuming the gravity term is ignored in a gas system due to low density of the fluid, Darcy’s
law can be written without the gravity term

1
u K .  p  ……………………………………………. (5.15)


K = permeability tensor.

 1 
.   u   .    K p   ……………………………………(5.16)
  

However, due to the nanopores network in shale, conventional Darcy law cannot be used to
represent the flow process in the matrix. Gas transport in nanopores includes mechanisms
such as Knudsen diffusion, viscous flow and slip flow (see Fig. 5.6)

Figure 5.6: Gas flow mechanism in a nano pore. Red dots are Knudsen diffusion while blue
dots are viscous flow. (Guo et al, 2015)

When the Knudsen number is less than one, flow is normally described as viscous flow that is
that the mean free path of the gas is smaller compared to the pore diameter. The viscous
flow caused by pressure gradient can be described by Darcy law (Kast and Hohenthanner,
2000).

Knudsen diffusion is more likely to occur when the pore diameter is small enough so that the
mean free path of the gas is close to the pore diameter, that is, that the Knudsen number will
be greater than 1. In such circumstances, the collision between the gas molecules and the

126
wall surfaces dominates. The gas mass flow can be expressed by the Knudsen diffusion (Kast
and Hohenthanner, 2000).

However, slip flow occurs in low permeability formations or under very low-pressure
conditions. It is therefore important to take into account the Klinkenberg effect (Klinkenberg,
1941; Sakhaee-Pour and Bryant, 2012)

The apparent permeability is expressed as

 b 
km  kmi 1  m  ……………………………………………………….(5.17)
 pm 

16  8RT   8 RT    2 
0.5

bm        1 ..........................(5.18)
3r   M   M  r  

Therefore, the flow vector term for the matrix can now be expressed as

 k   k 
.   u m  .   g  m pm    .   pm  m pm   ………………………..(5.19)
       

Viscous flow and Knudsen diffusion occurs in the fracture and hence has been included by
Guo et al., (2015). Considering these two flow terms, the mass flux can be represented by the
summation of these mechanisms.

f kf  D
J f  J fv  J fk   p f  f kf p f …………………………….(5.20)
 pf

k fi  f  bf 
Jf  1   p f ………………………………………………….(5.21)
g  p f 

Expressing Equation 4.1.21 in the form of the conventional Darcy flow equation, the fracture
apparent permeability can be expressed as

 b 
k f  k fi 1  f  ……………………………………………….(5.22)
 p f 


Dkf 
bf  …………………………………………………………………. (5.23)
k fi
127
 4k fi  RT
Dkf  ………………………………………….(5.24)
k fi 2M
2.8284
f

p f = fracture pressure

k fi = initial fracture permeability

k f = apparent fracture permeability

b f = Klinkenberg coefficient for the fracture system

Dkf = Knudsen diffusion coefficient for the fracture system

 f = fracture porosity

Finally, combining all the above equations results in the flow vector term being defined for
the fracture as

 k   k 
.   u  f  .   f  f p f    .   p f  f p f   ……………….(5.25)
       

5.2.3 Source and Sink Term

In shale gas reservoirs, the dominating factor that ensures the production of shale gas is the
fracture system. The matrix system also plays an important role in production, as fluids are
transferred from the matrix and consequently into the fracture for production. It is therefore
critical that the interaction between the matrix and fracture system is well modelled in the
case of a dual porosity continuum system. Kazemi et al., (1976) and Warren and Root (1963)
proposed a method to help calculate the cross-flux that results between the matrix and the
fracture by assuming that flow occurs in a pseudo-steady state with a confined boundary. The
transfer of gas between the matrix and the fracture is therefore represented as

km  g  pm  p f 
T ……………………………………………..(5.26)


128
  1 1 1  …………………………………………………………….(5.27)
  4  
 L 2 L 2 L 2 
 x y z 

The fracture spacing in the x, y and z directions are represented by Lx , Ly , Lz respectively.

 is the crossflow coefficient

Since fluid will be flowing out from the matrix into the fractures, the sink term for the matrix
can be described as

Qm  T ……………………………………………………….(5.28)

By producing from a vertical well in the fracture system, the model developed by Aronofsky
and Jenkins (1954) is

kf f 
qp 
  re 
 p f  pwf  …………….(5.29)
ln    s  Dq
 rw 

For a well located at the corner,    / 2 , and for a production well located in the centre,
  2 .

The drainage radius can be defined where k x  k y as

re  0.14 2  x    y   ………………………………………….. (5.30)

2 2
 

x, y refers to the length of the grid.

The source/sink term for the fracture system can therefore be expressed as

Q f  T  qq ……………………………………………………………………….. (5.31)

Finally, to characterise gas flow through shale gas reservoirs, the partial differential equation
model can be expressed as:

For a matrix system:

129
 1  m  MPLVL s  pm  .  p  km p    T
m  2   m m  …………………. (5.32)
 Vstd  PL  pm   t   

Or

 1  m  MPLVL s  pm  .   kmi  pm  bm  p    T

m  2    m  ………………..(5.33)
 Vstd  PL  pm   t     

For a fracture system:

p  k 
 f  f  .   p f  f p f    T  qq ……………………………………… (5.34)
t    

or

p   k fi  p f  b f  
 f  f  .   p f  p f    T  qq ……………………………………. (5.35)
t    
  

The boundary conditions considered for this study are:

Initial conditions: pm  x, y, t  |t 0  p f  x, y, t  |t 0  pi …………………………….(5.36)

 p 
Boundary conditions for matrix: Fm .n |1  0  |1  0  ……………………………..(5.37)
 n 

 p 
Boundary conditions for fracture: Ff .n | 2  0  | 2  0  , p f  x, y, t  | 2  pw
 n 
…………………..(5.38)

130
5.3 Discretization and Implementation
In reservoir simulation, the finite difference approach is the most widely adopted numerical
technique used. Other numerical techniques include the finite element and finite volume
method. The Finite difference approach offers great flexibility when handling a nonlinear
partial differential equation. The set of governing equations can be converted into both a
spatial and time discrete form. 1-D finite difference method have been applied on the system
of equations in this chapter for the purpose of showing the formulation and derivation of the
final discretised equations, however the actual simulation was achieved with a 2-D
formulation as shown in Appendix 5. The Central difference scheme is used for spatial
discretization of both the first and second order derivative terms, as follows:

p pi 1  pi 1
 …………………….(5.39)
x 2x

 2 p pi 1  2 pi  pi 1
 …………………(5.40)
x 2 x 2

For the time derivative, a backward difference scheme is adopted, which gives rise to an
implicit backward difference method.

p pin1  pin
 ………………………… (5.41)
t t

Applying the above difference schemes to the systems of equations 5.33 - 5.38 results in the
following approximation:

Matrix system:

 1  m  MPLVL s  pm  .   kmi  pm  bm  p     km  g  pm  p f 

m  2    m 

 Vstd  PL  pm   t     g

131
    n 1   n 1 2
 n 1  U i 1  U i 1  2U i( n 1) 
n 1 ( n 1) ( n 1)
B U  U n
U  U
f  C   i i
  A 
i 1 i 1
  AU  
  
n 1 2  t (2x)  (x) 2
i

 PL  U i      
U ( n 1)
 U i 1  2U i( n 1) 
( n 1)

  DU i U i  Vi   0
n 1 n 1 n 1
 E  i 1
 (  x ) 2

…………(5.42)

pm  U , p f  V , m  C ,  m  L L s  B ,  mi  A , pm  U , p f  V ,
1   MP V  k
Where
Vstd 
kmi
 D , Abm  E
g

Fracture system:

p   k fi  p f  b f    km  g  pm  p f 
 f  f  .    p f    qq …………….(5.43)
t     
   g

2
 V n 1  Vi n  1  Vi n11  Vi n11  ( n 1)  Vi 1  Vi 1  2Vi
n 1 n 1 n 1
  Vi n11  Vi n11  2Vi n 1 
f1  K  i   A    AV    F  
t 2x  (x) 2 (x) 2
i
      
 DVi n 1 U in 1  Vi n 1   0

………………….. (5.44)

k fi kmi
Where pm  U , p f  V ,  f  K ,   A1 , pm  U , p f  V ,  D , qq  0
 g

A1b f  F ,

5.3.1 Newton Method

The newton iterative method is used to solve systems of equation involving nonlinear partial
differential equations. The Newton Method is known to give second order convergence or
quadratic convergence. The idea of a quadratic convergence means fewer iterations may be
required, compared to other methods that might have much lower convergence rate before
a solution is reached. This method is thus less complicated when compared with other
methods, and it can also be easily applied to a variety of other problems. However, one key

132
problem associated with the Newton method has to do with the fact that convergence might
not always be achieved, and it is highly sensitive to the initial guess used.

For a system of nonlinear equations with N unknowns, they can be represented by a

multidimensional function as

f1 1 , 2 ,...N   0
f 2 1 , 2 ,...N   0
.
…………………… (5.45)
.
.
f3 1 , 2 ,...N   0

By denoting the unknowns as   and giving an initial solution guess to be   0  , the Taylor
 
series for the system of equation can be written

f1
 
f1 1 , n ,...N   f1 1 0 , 2 0 ,...N  0 
1
0
1

f1  0 f1  0
 2  ...  N  ...  0
2 N
.
.
.
f 2

f 2 1 , n ,...N   f 2 1 0 , 2 0 ,...N  0  1
 0
1

f 2  0 f 2  0
 2  ...  N  ...  0
2 N
.
.
.
f3

f3 1 , n , ...N   f3 1 0 , 2 0 ,...N  0   1
 0
1

f f …………………….. (5.46)
 0  0
 3 2  ...  3 N  ...  0
2 N

133
The matrix form of Equation 5.46 can be written as

 f1  0 f1  0 f1 0 

  2
...
N   1   f1 1 , n ,...N  
 1    
 f 2  0 f 2  0 f 2  0     
  2
...
N    
 1      
.   2     f1 1 , n ,...N   ………….. (5.47)
  .   
. 
 .   
.    
  .   
 f N  0 f N  0 f N  0     f  ,  ,...  
 
N 
 1
...
 
  1 1 n

 2 N 
N

i  i  i  0 Represents the change in the solution from one iteration to the next.

J  is the Jacobean matrix for the Newton method containing the partial derivatives from

Equation 5.45 above.

 f1  0 f1  0 f1  0 

  ... 
2 N
 1 
 f 2  0 f 2  0 f 2  0 
  ... 
2 N
 1 
 J   .  ………………………….. (5.48)
 
. 
. 
 
 f N  0 f N  0
...
f N  0 
 1 2 N 
 

Hence the updated formula for Newton’s method is as follows

 J         f  
n n n
……………………………….(5.49)

Where        
n 1  n
   
n

Finally, the algorithm for the implementation of Newton’s method for a system of nonlinear
equations can be represented in the following steps.

134
Step 1: Guess an initial root:     
0

Step 2: Determine the Jacobian matrix for the system of equations:  J 

 n
Step 3: Solve  J        f   ideally iterative solver must be used with convergence
n n

checks implemented within this step

Step 4: Update results by introducing a relaxation factor  :

 
n 1  n

      
 n 1  n
   
Step 5: Repeat Steps 3 and 4 until convergence.

Computing the Jacobian for the Matrix results in

B
C
f 2 B U  U  n 1 n
 P U  n 1 2
 U n 1  U in11  2U in 1  2 A n 1 2E
   A  i 1  Ui 
i i L i

U i t  P  U n 1  t (x)  (x) (x) 2 ----

3 2 2
L i 
 D U in 1  Vi n 1   DU in 1

-Diagonal ……………………(5.50)

f 1  U n 1  U n 1  U n 1 E
  A  i 1 2 i 1   A i 2  ------- Upper diagonal ………….. (5.51)
U i 1 2  (x)  (x) (x) 2

f 1  U n 1  U in11  U in 1 E
 A  i 1   A  ----------Lower diagonal………….(5.52)
U i 1 2  (x) 2
 ( x ) 2
( x ) 2

Computing the Jacobian for the Fractures results in

f1 K  V n 1  Vi n11  2Vi n 1  2 A1 ( n 1) 2F

  A1  i 1  V   D (U in 1  Vi n 1 )
Vi t  (  x ) 2
 ( x ) 2 i
( x ) 2
-------- Diagonal
 DVi n 1

……….. (5.53)

f1 1 A1 A1
2  i 1
V  Vi 1  
n 1 n 1 F
 V n 1  -------------Upper diagonal …………. (5.54)
Vi 1 2 (x) (x) 2 i
(x) 2

135
 f1 1 A1 A1

Vi 1 2 (x) 2 Vi
n 1
1  Vi
n 1
1  
(x)
V n 1 
2 i
F
(x) 2
-------------Lower diagonal ………… (5.55)

The above-noted Jacobean for the shale matrix and fracture PDE can now be used to set up
Jacobean matrices in Equation 5.48.

5.3.2 Solution Method of Coupled Nonlinear PDE

There are two main approaches to solving a set of coupled partial differential equations
(PDEs): the segregated solution approach and the coupled solution approach (Mazumder,
2016). The segregated solution approach has been adopted for the solution of the two
nonlinear PDEs involving the matrix and fracture of the shale rock. This method involves
solving each individual PDE sequentially and in isolation. To address the coupling between
the two PDEs, the sequence of the solution is normally enclosed within an outer loop. Each
inner iteration is continued until the PDEs have attained partial convergence. The main
advantage of using this method over the coupled solution approach is its modularity
(Mazumder, 2016). That is, the code written for the solution of the two PDEs can easily be
extended to include other PDEs. For instance, the isothermal code written for the two
nonlinear PDEs can be extended to include heat transfer. Fig. 5.7 shows the schematic
representation of the segregated solution approach.

136
Figure 5.7: schematic representation of segregated solution approach for coupled nonlinear
PDEs

5.3.3 Validation of Numerical Method with Analytical method

Engineering predictions and decisions are sometimes based on computational models that
are deemed robust and have also been validated. Computational models’ results are
sometimes checked for reliability and consistency by validating the results with either a
physical model or actual field data. This allows for decisions related to engineering processes
to be made with confidence.

Guo et al., (2015) noted that there are no real field data that are similar to the theoretical
case developed above for the governing mathematical equations; it is expected that the
analytical results can be compared with the numerical simulation to verify the accuracy of the
results. Guo et al., (2015) adopted the approach of validating the numerical method by
simplifying Equation 5.42 into a one dimensional, linear, horizontal and steady state flow
condition (See Equation 5.56) . The results obtained can be compared with the analytical
solution derived by Wu et al., (1998) for a 1D steady –state gas transport. Thus, in this
research, similar approach to Guo et al., (2015) have been adopted to validate the numerical
method.

137
  k   P  b  P 
   0 ……………………….(5.56)
x   x 

The boundary conditions given for the solution of the steady-state gas transport are:

A constant mass injection rate qm per unit cross-sectional area is imposed at the inlet ( x  0
), and

Gas pressure is kept constant at the outlet ( x  L ). The analytical solution is given by Wu et
al., (1998) as

 2qmu g  L  x  
p( x)  b  b 2   p( L) 2  2b  p( L)    …………………. (5.57)
 k  

The results from numerical simulation of Equation 5.56 (See Figure 5.8 ) show a clear and
exact match to the analytical solution developed by Wu et al. (1998), implying that the model
is cable of representing the flow conditions in a shale gas reservoir.

138
 60

50

40
Gas Pressure, bar

30

20

10

0
1 2 3 4 5 6 7 8 9 10
Distance, m

Analytical solution Numerical solution

Figure 5.8: Comparison of numerical and analytical solutions for steady state gas flow in a
finite linear system.

139
5.4 Results and Discussion
5.4.1 Simulation Study I
Bustin et al., (2008) and Guo et al., (2015) presented numerical data that could be used for
numerical examples involving the flow of gas in shale gas reservoirs. Table 5.1 provides the
rock properties and fluid properties used in the simulation. Gas production is considered from
the fracture systems due to the higher permeability compared with the matrix systems. A
single well is drilled in the middle of the domain, as illustrated in Fig 5.9, to be a producer
well. A 2-D reservoir model with dimensions of 160ft x 160ft has been discretized into 51 x 51
uniform grids. It is assumed that only a single gas (methane) phase is present and that the
gas adsorption and desorption obey the Langmuir isotherm with constant temperature. The
duration of the numerical simulation has been set to 2.7 years (1000 days). All the simulation
in chapter 5 and 6 have been conducted using high level programming language MATLAB.
MATLAB is an efficient programming tool for solving numerical problems especially non-linear
partial differential equations compared to other programming language. No commercial
simulator has been used throughout this study.

Figure 5.9: Schematics of numerical experiment with single well production

140
Table 5.1: Parameters used in the simulation model

Parameter Value
Reservoir depth 5463 ft
Pressure gradient 0.54 psi/ft
Temperature 579.6 °R
Matrix initial permeability 0.04 mD
Fracture initial permeability 10 mD
Matrix porosity 0.05
Fracture porosity 0.001
Initial reservoir pressure 1508 psia
Bottomhole pressure 500 psia
Molecular Weight (CH4) 0.016 lb/lb mole
Standard gas volume 0.7910 scf/mol
Langmuir pressure 300 psia
Langmuir volume 0.0448 gmol/lb
Shale rock density 159.19 lb/scf
Initial gas viscosity 0.0102 cp
Wellbore radius 0.328 ft
Fracture spacing 0.656 ft

Figs. 5.10 and 5.11 show the pressure distribution in the matrix and fractures after a period
of 1000 days of production. It can be observed that pressures in both cases have reached the
value of bottomhole pressure. The pressure distribution shows a slight decrease towards the
centre of the domain where the well is located (see Figs. 5.10-5.11). Production rate and
eventual cumulative production over the period of simulation are shown in Figs. 5.12 and
5.13.

141
Figure 5.10: Fracture pressure at the end of simulation (2.7 years)

142
Figure 5.11: Matrix pressure at the end of simulation (2.7 years)

The gas production rate results show an initially high production followed by a gradual decline
in production with time, over the life of the field (see Fig. 5.12). The effect of several reservoir
parameters has been evaluated to show how they impact on gas production in shale gas
reservoirs. These include the effect of matrix and fracture porosity and permeability.

143
 30000

25000
Production rate,Scf/day

20000

15000

10000

5000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Figure 5.12: Gas production rate for a shale gas model

1000000
cummulative Production rate, Scf/day

900000
800000
700000
600000
500000
400000
300000
200000
100000
0
0 100 200 300 400 500 600 700 800 900 1000
Time,days

Figure 5.13: Cummulative production of gas for a shale gas model

5.4.2 Effect of Gas Adsorption

To illustrate the effect of gas adsorption on production rate and cumulative production,
simulation studies were conducted, first by considering only the free gas present in the
system. Thus, the total gas in place was considered to be made up of only the free gas. The
144
second simulation was conducted by incorporating an adsorption term defined by Langmuir
isotherm. This meant that the total gas in place was now a combination of both free gas and
adsorbed gas. Simulation results in Figs. 5.14 and 5.15 show the effect of considering the
production rate and cumulative production in the system with and without adsorption. It can
be observed that with the inclusion of gas adsorption defined by Langmuir isotherm, the
production rate increases significantly, whereas without the adsorption term, production
could be underestimated. Thus it can be concluded that gas adsorption in shale gas reservoirs
will have a great effect on overall production from the reservoir. It is important to note that
there was a significant contribution from desorption at the early stage of production from
Figure 5.14 due to the choice of a much lower bottomhole pressure. The lower bottomhole
pressure meant more gas could be desorbed and consequently produced compared in a real
case where bottomhole pressure might be higher and may take considerable time to drop to
the economic shut in pressure. In such instance, the influence of gas desorption will be felt at
later stages of production. A higher pressure value might have resulted in a solution closer to
the actual case , however due to the numerical set up of this study and the fact that newton’s
method requires a selection of an initial guess closer to the actual solution, it meant that
choosing any other pressure value may have resulted in convergence issues causing the
simulation not to reach full convergence.

145
 60000

50000
Production rate, Scf/day

40000

30000

20000

10000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

With Adsorption Without Adsorption

Figure 5.14: Gas production rate with and without adsorption

2000000
cummulative Production rate, Scf/day

1800000
1600000
1400000
1200000
1000000
800000
600000
400000
200000
0
0 100 200 300 400 500 600 700 800 900 1000
Time,days

With Adsorption Without Adsorption

Figure 5.15: Cumulative production rate with and without adsorption

146
5.4.3 Effect of Matrix and Fracture Porosity
To study the effect of porosity on production from shale gas reservoirs, different porosity
values have been chosen for both the matrix and fracture porosity. Porosity defines the
amount of gas that can be stored in the reservoir. For shale gas reservoirs, the matrix porosity
has a greater influence on production than the fracture porosity. Figs. 5.16 and 5.17 shows
that an increase in porosity leads to increase in production rate, and Figs. 5.16-5.19 show that
matrix porosity has a significant effect on shale gas production, compared with fracture
porosity. This difference confirms the idea of the dual porosity model being incorporated into
the shale gas model where the matrix system is the main storage space for the gas.

80000
70000
Production rate,Scf/day

60000
50000
40000
30000
20000
10000
0
0 100 200 300 400 500 600 700 800 900 1000
Time,days

Matrix Porosity 5% Matrix Porosity 20%

Figure 5.16: Effect of matrix porosity on gas production rate

147
 3000000
Cummulative production, Scf/day
2500000

2000000

1500000

1000000

500000

0
0 100 200 300 400 500 600 700 800 900 1000
Time,days

Matrix Porosity 5% Matrix Porosity 20%

Figure 5.17: Effect of matrix porosity on cumulative production

60000

50000
Production rate,Scf/day

40000

30000

20000

10000

0
0 100 200 300 400 500 600 700 800 900 1000
Time , days

Fracture Porosity 0.1% Fracture Porosity 10%

Figure 5.18: Effect of fracture porosity on gas production rate

148
 2500000
Cummulative Production, Scf/day
2000000

1500000

1000000

500000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Fracture Porosity 0.1% Fracture Porosity 10%

Figure 5.19: Effect of fracture porosity on cumulative gas production

5.4.4 Effect of Matrix and Fracture Permeability

In shale gas reservoirs, permeability has a great influence on production results. The effect of
matrix permeability is different from the effect of fracture permeability. Matrix permeability
is usually very small and hence will have the least effect on production results, whereas in
shale gas reservoirs, fracture permeability contributes greatly to production from the
reservoirs. Figs. 5.20-5.23 show that as the fracture and matrix permeability increases,
production results also increase. The results from Figs. 5.20-5.23 also show that the main
contributing channel towards production is from the fracture system.

149
 60000

50000
Production rate,Scf/days

40000

30000

20000

10000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Matrix Permeability 0.04mD Matrix Permeability 0.1mD

Figure 5.20: Effect of matrix permeability on production rate

2500000
Cummulative Production, Scf/day

2000000

1500000

1000000

500000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Matrix permeability 0.04mD Matrix permeability 0.1mD

Figure 5.21: Effect of matrix permeability on cumulative production

150
 300000

250000
Production rate,Scf/day

200000

150000

100000

50000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Fracture Permeability 10mD Fracture Permeability 50mD

Figure 5.22: Effect of fracture permeability on gas production

7000000
Cummulative Production, Scf/day

6000000

5000000

4000000

3000000

2000000

1000000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Fracture Permeability 10mD Fracture Permeability 50mD

Figure 5.23: Effect of fracture permeability on cumulative gas production

151
5.5 Heat Transfer through a Porous Media
By the first law of thermodynamics, the total energy of a system is conserved; that is, the
energy stored in a volume is equal to the energy that enters the volume, minus the amount
of energy that leaves the volume. The internal energy of a system is therefore conserved and
is equal to the sensible component of the thermal energy. The change in thermal energy can
be described as the summation of the transport of energy by flow, change of energy by
conduction and heat sources and the surroundings. The mathematical form of the above
description can be seen as

d
dt 
 edx    qadv .nds    qcond .nds   QT dx ………………………………………….. (5.58)
  

e is the internal thermal energy density

n is the outward unit normal vector

QT is the energy production term in J/ (s.m3)

QT  Qsource  Qconv ……………………………………………………………………. (5.59)

Qconv  ah Tm  T 

a is the specific surface area

h is the heat transfer coefficient

Qconv is the rate of energy transfer per volume of the exchange from the solid to the fluid

Applying the divergence theorem on the surface integral, the volume integral can be
expressed as

 de 
  dt  .q

adv  .qcond  QT dx  0 …………………………………………. (5.60)


de
  .qadv  .qcond  QT  0 …………………………………………………. (5.61)
dt

152
By inserting the heat flux expressions:

de
  .  eV   .  k T   QT  0 ……………………………………………… (5.62)
dt

qcond  kT T ……………………………………………(5.63)

qadv  eV ………………………………………………………….. (5.64)

V Intrinsic volume flux

kT is thermal conductivity

The change in thermal energy density can be expressed as

e   c p T ---------------------------------------------------------------(5.65)

This is similar to the definition of the specific heat capacity, where the thermal energy

required to raise the temperature of a volume with mass M mass by T degrees (Lampe, 2013)
is given as

U  M mass c p T ……………………………………………. (5.66)

c p is the specific heat capacity.

By replacing e in Equation 5.62 with Equation 5.65, and using the velocity term v , the
following expression can be obtained:

T
 c p   c p v.T   c pT .v  .  kT T   QT  0 …………………………. (5.67)
t

v, actual velocity is related to the intrinsic velocity by the expression

v  V ……………………………………………………………………….(5.68)

Rearranging and applying the continuity equation results in the expression

T
 c p   c p v.T  .  kT T   QT ………………………………… (5.69)
t

153
Equation 5.69 can therefore be assumed to be the energy equation for the gas phase. Similar
derivation can be obtained for the matrix system except that no advective term is applied and

the scale factor is now 1    .

The two energy equations can therefore be written explicitly as

For matrix:

Tm
1    mc p,m  1    .  kT ,mTm   1    q '' m  ah Tm  Tg  …………..(5.70)
t

For the gas:

Tm
   g c p , g    .  kT , g Tg    g c p , g T .qg   q '' g  ah Tm  Tg  ……………......(5.71)
t

Where

qg is the velocity vector of the gas phase.

q'' is the heat source term.

By assuming a local thermal equilibrium, the energy equation can easily be simplified at the

walls of the pores by setting the relation Tm  Tg  T , and by summing Equations 5.70 and

5.71, the following expression is obtained:

T
 c p final t
 .   K final T    g c p , g qg .T  Q " ………………………………………….(5.72)

Where

 c p final
 1    m c p ,m   g c p , g ………………………………. (5.73)

K final  1    kT ,m   kT , g ……………………………………….(5.74)

Q"  1    qm"   q"g ……………………………………………………………………. (5.75)

Equation 5.73 is similar to the volumetric heat capacity at constant pressure, Equation 5.74 is
equivalent to thermal conductivity, and Equation 5.75, is the overall heat source.

154
According to Wang (2016), the heat capacity of most formations is at least 100 times larger
than that of the gas phase, and since formation porosity of a shale formation is normally less

than 3%, Equations 5.73-5.75 are dominated by the terms 1     m c p ,m , 1    kT ,m and

1    qm" respectively. Also, since the flow rate of the gas is limited by the low permeability

of shale formation, and considering that the gas has low density, the term  g c p , g qg .T can

be neglected in Equation 5.72.

Therefore, we can simplify equation 5.72 as the following:

T
 m c p ,m 
t
 .  kT ,mT   Q" ………………………………………….(5.76)

Equations 5.42, 5.43 and 5.76 can therefore be combined and with appropriate initial and
boundary conditions, a complete description of the system coupled with heat can be
achieved.

 1  m  MPLVL s  pm  .   kmi  pm  bm  p    T

m  2    m 
 ………………..(5.77)
 Vstd  PL  pm   t    

p   k fi  p f  b f  
 f  f  .   p f  p f    T  qq ……………………………………. (5.78)
t    
  

T
 m c p ,m 
t
 .  kT ,mT   Q" ………………………………………….(5.79)

Note: Assumptions have been made to simplify the gas properties and flow equations. In
general, gas flow in deep pressurized reservoirs does not follow the ideal gas law, and the
variations of pressure around gas production wells are too large to use constant properties.

155
5.5.1 Fully Implicit Method for the Heat Equation
Since the heat equation introduced in this chapter is a linear PDE, use can be made of the fully
implicit discretization method to solve for the temperature distributions within the reservoir.
The discretized form of the heat equation can be written as:

Ti ,nj1  Ti ,nj  Ti n1,1j  2Ti ,nj1  Ti n1,1j Ti ,nj11  2Ti ,nj1  Ti ,nj11  Qin, j
k    ………….(5.80)
t  x 2
x 2
 c
  m p

k t k t
Sx  , and Sy 
x 2 y 2

Rearranging terms will result in

Qin, j
 S xTi n1,1j  S xTi n1,1j  S yTi ,nj11  S yTi ,nj11  1  2S x  2S y  Ti ,nj1  Ti ,nj  ………… (5.81)
mc p

And finally, temperature distribution can be estimated as

Qin, j t
Ti ,nj   S xTi n1,1j  S xTi n1,1j  S yTi ,nj11  S yTi ,nj11
mc p
Ti ,nj1  …………….(5.82)
1  2S x  2S y 

5.5.2 Temperature –Dependent Variables

The introduction of temperature-dependent variables is needed in order to investigate the
process of thermal stimulation applications in shale gas reservoirs. Thus, several temperature-
dependent variables would have to be defined to enable the coupling of the previous
governing equations for both the matrix and fracture systems with the heat equation. The
key variables to be explored include the gas adsorption capacity and the viscosity of the gas,
both of which are temperature dependent.

The most widely used adsorption model is the Langmuir isotherm. This model is able to
describe the relationship between pressure and the adsorbed gas content as a function of
pressure. The key disadvantage of this model is that it does not account for the effect of
temperature on the adsorption of gas onto the shale matrix. To account for the effect of

156
temperature, several modifications of the Langmuir isotherm have been adopted and used in
the coupling process.

The modified Langmuir model can be expressed as follows:

s M g  bp 
1) V   VL  ………………….(5.83)
Vstd  1  bp 

Where b  b0 exp  H RT 

Freundlich isotherm has been combined with Langmuir isotherm to account for the
heterogeneity of the adsorbent (Fianu et al., 2018; Helminen et al., 2000). The variable n

describes the degree of heterogeneity of the adsorbent. A value of n greater than one
represents a highly heterogeneous adsorbent. The combined Langmuir-Freundlich model is:

s M g  bp n 
2) V   L
V  …………….(5.84)
 1  bp
n
Vstd 

Finally, to account for temperature dependency of the Langmuir volume VL , Ye et al., (2016)

proposed an exponential model. The exponential model is expressed as:

 
 
s M g  exp   DT T  p  …………… (5.85)
3) V  Vs
Vstd     B   
 p   T  A exp     
   T    
 

Where Vs , DT , A and B are parameters to be determined by fitting the experimental

adsorption data.

5.5.3 Gas Viscosity

Well established correlations have been used to derive gas viscosity. One of the most popular
and comprehensive studies on the viscosities of natural gases was carried out by Lee et al.,
(1966). Their correlation can be used to calculate gas viscosities at pressures from 0.69 to
55.16MPa and temperatures from 310 to 455 K. This correlation is given below as:


u g  K exp X  Y  ……………………….(5.86)

157
Where

K
 7.77  0.0063M  T 1.5
122.4  12.9M  T

1914.5
X  2.57   0.0095M
T

Y  1.11  0.04 X

5.5.4 Z Factor Correlation

There are several equations of states that can be solved to obtain the Z-factor or correlations.
In this study, Hall and Yarborough’s correlation has been used to obtain the Z-factor.

1
t
Tpr

A  0.06125te1.2(1t )
2

B  14.76t  9.76t 2  4.58t 2

C  90.7t  242.2t 2  42.4t 2

D  2.18  2.82t

y  y 2  y3  y 4
 APpr   By 2  Cy D  0 ……………………….(5.87)
(1  y) 3

T
Tpr 
Tpc

P
Ppr 
Ppc

APpr
z
y

158
5.6 Simulation Study II
With the introduction of temperature dependent variables, it is now possible to investigate
the application of thermal stimulation in simulation study I. However, to do this, certain
assumptions that were made for simulation study I would have to be altered. In particular,
the model changes from an isothermal condition to a non-isothermal condition. Several
temperature-dependent forms of the Langmuir isotherm are considered, and temperature
effects on density and viscosity are also considered. Temperature effects increase the
complexities in reservoir modelling and simulation because changes in the formation
temperature have multiple effects on adsorption capacities and real gas properties, and can
also induce thermal stress (Wang, 2016). The thermal properties used for the simulation of
study II are shown in Table 5.2.

Table 5.2 Thermal properties used in simulation study II

Thermal Properties Values

Thermal Stimulation temperature 700°R
Formation-heat capacity 1000 J/K/Kg
Formation-heat conductivity 4 W/m/K

Under simulation study II, two different scenarios have been considered. Firstly, the
adsorption/desorption relationship has been modelled using modified Langmuir model with
temperature-dependent parameters to compare simulations with and without thermal
stimulation. Secondary, different temperature-dependent gas adsorption models have been
compared to understand their application in predicting future gas productions. An
exponential model with temperature-dependent Langmuir volume and the Langmuir-
Freundlich model have been considered. The simulation run has been conducted for 1000
days.

The reason for the choice of these two models stems from earlier studies in chapter 3, where
several temperature-dependent adsorption models were applied to a wide range of shale gas
datasets. In those studies, it was observed that Exponential model and Langmuir-Freundlich

159
model proved most accurate for a variety of the shale data, thus, including both models in
the shale gas model to evaluate the thermal stimulation.

In scenario 1, Figs. 5.24 and 5.25 show the temperature distribution within the formation
after 10 days and 1000 days of production. It can be observed that the heat propagation has
increased from the initial temperature throughout the reservoir over the course of the
production period. The gas production rate increased in the case of thermal stimulation,
compared with when no thermal stimulation was applied. As has been mentioned earlier, the
modified Langmuir model with a temperature-dependent gas adsorption model have been
used in the simulation studies to represent the adsorption potential as a function of both
pressure and temperature. The rate of propagation is, however, dependent on the thermal
properties of the rock and the stimulation temperature that is used. A higher stimulation
temperature will no doubt improve the temperature propagation throughout the formation.
A much higher thermal conductivity will also imply an increase in the rate of heat propagation
in the formation.

Figure 5.24: Temperature profile after 10 days of heating

160
Figure 5.25: Temperature profile after 1000 days of heating

70000

60000
Production rate, scf/day

50000

40000

30000

20000

10000

0
0 200 400 600 800 1000 1200
Time, days

Without thermal stimulation With thermal stimulation

Figure 5.26: effect of thermal stimulation on production rate

161
 6000000

5000000
Production, scf/day

4000000

3000000

2000000

1000000

0
0 200 400 600 800 1000 1200
Time, days

Without thermal stimulation With thermal stimulation

Figure 5.27 Effect of thermal stimulation on cumulative production

In scenario 2, the production profile from Figs. 5.28-5.29 shows the different production
results using different temperature-dependent gas adsorption models. The results show that
the choice of temperature -dependent model used in the simulation studies results in a
different production recovery profile. This is because different temperature dependent
models would extrapolate different adsorption capacities at different temperatures, as can
be seen in Fig. 5.30. Temperatures have been extrapolated to 700°R, and from Fig. 5.30 it can
be seen that the Exponential model predicted a much higher gas adsorption rate than the
temperature-dependent forms of the Langmuir model and Freundlich model. The predictions
of different gas adsorption at different temperatures affect the overall gas production from
the system. As can be observed from Figs. 5.29 and 5.30, overall gas production rate and
cumulative production over 1000 days was higher for the Exponential model compared with
the temperature-dependent forms of Langmuir model and Langmuir-Freundlich model. There
was a percentage decrease of about 0.008 % in the cumulative production of Langmuir model
when compared with the Exponential model. Exponential model showed an increase of
0.032% in cumulative production compared with Bi-Langmuir model. Langmuir Model
showed a 0.024% increase in overall cumulative production, compared with Bi-Langmuir
162
model. From Figure 5.30, it can be observed that exponential model predicted higher
adsorption rate even at temperature of 700°R, compared with modified Langmuir and
Langmuir–Freundlich models. This higher rate in gas adsorption at that temperature meant
that more gas could be produced at the end of simulation. The fact that different models
predicted different gas production at the end of simulation study suggests that care must be
taken when choosing the model to use in reservoir simulation studies. Validation of models
with multiple temperature as well as validation of extrapolated temperature for gas
adsorption is needed before choosing a particular model to be used in predicting future gas
production in shale gas reservoirs.

0.04

0.035

0.03
Adsorbed gas, scf/lb

0.025

0.02 Data @ 579.8°R

0.015 Exonential @700°R

0.01 Temp.Depend Langmuir @ 700°R

0.005 Temp.Depend Freundlich @ 700°R

0
0 200 400 600 800 1000 1200 1400 1600
Pressure ,Psia

Figure5.28: Gas adsorption at extrapolated temperature from 579.8°R to 700°R

163
Table 5.3: Best fitted parameters for Temperature dependent models at 700°R (Based on
AAD%)

Exponential Model Langmuir-Freundlich Model Langmuir

Model(Temperature
Dependent)
Parameters Value Parameters Value Parameter Value
Vs gmol/lb 0.05 VL gmol/lb 0.044811876 V gmol/lb 0.044799994
L

Dt 0.00004 bo 1/psi 7.42932E-05 bo 1/psi 0.000536127

A 7.75E-02 H KJ/mol -23667.6696 H KJ/mol -11706.5652
B 20.39 n 0.99

58500

58000

57500
Production rate , scf/day

57000

56500

56000

55500

55000

54500

54000
0 100 200 300 400 500 600 700 800 900 1000
Time, days

Langmuir model TD Exponential model Langmuir-Freundlich model

Figure 5.29: Prediction of production rate by different temperature-dependent adsorption

models

164
 5000000
Cummulative production, scf/day

4000000

3000000

2000000

1000000

0
0 100 200 300 400 500 600 700 800 900 1000
Time, days
Langmuir model Exponential model Langmuir-Freundlich model

Figure 5.30: Cumulative production prediction by different temperature dependent

adsorption models

5.7 Summary
In this chapter, a mathematical model for simulation of shale gas reservoirs is developed. The
model includes non-Darcy flow that occurs within the matrix of the shale. Two simulation
studies were conducted, first to evaluate the effect of including adsorption in numerical
simulation of shale gas reservoirs, and second to introduce the concept of temperature-
dependent adsorption models in mathematical models for shale gas reservoirs. The results
from simulation study I showed the following:

 Gas production increased as a result of simulation with adsorption rather than

without adsorption. This showed that gas adsorption can affect the rate of gas
production
 Matrix porosity and fracture permeability have considerable influence in gas
production. Thus, an increase in these two factors will cause an overall increase in gas
production
 Fracture porosity and matrix permeability had minimal influence on gas production
in shale gas reservoirs. Any considerable increase will result in increase in the overall
gas production.

165
In simulation study II, temperature-dependent gas adsorption models have been used in
mathematical formulation to account for thermal stimulation in shale gas reservoirs.
Different temperature-dependent gas adsorption models, including modified Langmuir
temperature-dependent model, Exponential model with temperature-dependent
Langmuir volume and combined Langmuir – Freundlich model, were included in the
simulation studies to investigate future gas production. The results showed that:

 Thermal stimulation has the advantage of increasing cumulative gas production

at the end of simulation window, compared with when no thermal stimulation
was employed
 Different temperature dependent gas adsorption models predicted different
cumulative gas production over the course of simulation
 There was a 0.032% increase in cumulative production using Exponential model
compared with Bi-Langmuir model, and a decrease of 0.008% in cumulative
production of Langmuir model when compared with the Exponential model.
Langmuir model showed a 0.024% increase in cumulative production, compared
with Bi-Langmuir model
 Validation of all the models with published data at multiple temperatures is
necessary to ensure the most accurate model is used in simulation studies.

166
Nomenclature
M = mass accumulation term

 = density

u = velocity

Q = source term

t = time.

m= matrix

f = fracture

S g = fraction of pore volume occupied by the gas phase

qads = gas adsorption volume per unit bulk volume

Vstd = mole volume under standard conditions

PL = Langmuir pressure

VL = Langmuir volume

s = density of shale

pm = pressure for the matrix

K = the permeability tensor

p f = fracture pressure

k fi = initial fracture permeability

k f = apparent fracture permeability

b f = Klinkenberg coefficient for the fracture system

167
Dkf = Knudsen diffusion coefficient for the fracture system

 f = fracture porosity

 = crossflow coefficient

b = Klinkenberg factor

p( L) = gas pressure at the outlet

L = length from the inlet to the outlet

x = task location along the gas transport path

k = intrinsic permeability

 = compressibility factor

qm = air injection rate

u g = dynamic viscosity

e = internal thermal energy density

n = outward unit normal vector

QT = energy production term in J/(s.m3)

a = specific surface area

h = heat transfer coefficient

Qconv = rate of energy transfer per volume of the exchange from the solid to the fluid

V = intrinsic volume flux

kT = thermal conductivity

c p = specific heat capacity

168
qg = velocity vector of the gas phase.

q'' = heat source term.

169
 Chapter 6
Thermal Stimulation of Shale Gas Reservoirs Using
Microwave Heating

6.1 Introduction
Thermal stimulation strategies have been used by several researchers as a way of improving
oil recovery by increasing formation temperature. They have grown in prominence especially
in unconventional reservoirs, such as heavy oil and shale oil. Recently, thermal stimulation
has also been applied to coal bed methane reservoirs. The different kinds of thermal
stimulation that are currently being used in unconventional reservoirs include cyclic steam
injection, in-situ combustion and steam assisted gravity drainage (SAGD). Thermal stimulation
as a means of enhanced oil recovery can account for about 60% of oil production in our world
today (Chekhonin et al., 2012).

Though several studies have been conducted on how to improve heavy oil or shale oil
recovery by elevating the formation temperature, very few have considered the
implementation of similar thermal techniques as a way of enhancing gas recoveries (Lin et al.,
2015).

In recent years , there have been developments in the use of coupled electromagnetic-
thermal modelling to quantify and visualize microwave heating; however, these
developments have been focused primarily on food , wood and minerals, with few
applications in shale and coal (Li et al., 2017; Pitchai et al., 2016)

Wang (2016) demonstrated that the efficiency of thermal stimulation largely depends on the
gas adsorption and rock properties of the formation. Thus by altering the gas
adsorption/desorption behaviour through thermal stimulation, more gas could be recovered
from hydraulically fractured shale gas reservoirs. Wang et al., (2015) also used similar
techniques in coal bed methane reservoirs, by investigating the application of thermal
stimulation of hydraulically fractured coalbed methane reservoir. Altering the gas adsorption
and desorption behaviour through thermal stimulation in coal bed methane (CBM) has also
resulted in increased gas recoveries. Studies conducted by Salmachi and Haghighi (2012)

170
showed that as the formation temperature increased through a process of hot water injection
for a period of 2 years , the gas adsorption/ desorption process was altered . This led to an
increase in methane production by 58% in 12 years of production. They argued that thermal
treatment in CBM results in the breakdown of bonds between the gas molecules and surfaces,
which ultimately alters the adsorption/desorption properties of the coal.

In- situ combustion has also been studied (Chapiro and Bruining, 2015) as a means of
improving the permeability of shale gas reservoirs. Permeability is a key factor in improving
the production from tight / shale gas reservoirs. An increase in permeability would cause the
natural gas to flow more easily to the surface and hence recoveries would improve. The
authors argued that the combination of hydraulic fracturing and horizontal drilling allowed
more gas to be easily extracted from shale; however, the impact of this combination on the
environment suggests that alternative techniques needs to be explored. Similar studies have
also suggested that thermal stimulation of shale gas reservoirs can result in permeability
improvements, just as with oil shales (Busch and Amann-Hildenbrand, 2013; Sanmiguel et al.,
2002). According to Jamaluddin et al., (2000), thermal treatments in tight gas reservoirs
enhance the permeability of the formation by vaporizing the capillary blocked water,
dehydrating the clay bound water and creating thermally induced micro-fractures. Their
studies demonstrated increase in permeability of 50% or more for core samples that were
exposed to heat treatments.

In order to study shale gas interactions during microwave heating, a fully coupled
electromagnetic-thermal model has been developed for this research. The development of
such a coupled electromagnetic-thermal model enables the understanding of microwave
heating with the mechanical model of shale gas considering non-Darcy effects. The effect of
microwave frequency on temperature evolution and also on production have been studied.
Free gas evolution during microwave heating, and also overall production with and without
microwave heating, has been investigated.

171
6.2 Microwave Heating
There are three main kinds of electrical heating that are most used in the oil and gas industry
as a stimulation strategy for improved production. These are low frequency electric heating,
inductive heating where an alternating current is used, and high frequency heating in the form
of microwaves (Sahni et al., 2000). Microwaves are electromagnetic waves that have
frequency range between 300 MHz and 300 GHz with wavelength of 1 mm to 1m (Denny,
2007). The applications of microwave technology are varied, ranging from food processing
and pharmaceuticals to wood processing and ceramics. In the petroleum industry, it is often
seen as an enhanced recovery method targeting mostly unconventional resources such as
heavy oil and tar sands. Microwave application is preferred where the medium is made up of
polar molecules rather than non-polar molecules, which cannot absorb the microwave
energy. Water content in the formation plays a key role in microwave heating since the high
frequency waves act on the water molecules, causing them to heat up (through dipole
rotation or molecular rotation causing friction as a result of the rotation); the heat is
transmitted towards the surrounding formation. The microwave antenna can be placed
either downhole (see Figure 6.1) or an electromagnetic wave generator is placed at the
surface and transmitted downhole into the formation.

Figure 6.1: Schematics of microwave heating in a shale reservoir

172
Microwave heating is sometimes referred to as dielectric heating. Dielectric heating refers
mainly to heating by electric field due to the presence of dipoles in polar molecules. The
heating process is mainly influenced by the dielectric properties of the medium, which in turn
depend on the electrical properties of the medium. Other factors that control the
temperature distribution within the medium are the heat capacity and thermal conductivity
of the material.

Conventional heating methods such as conduction, convention and radiation are the main
heating sources for both the internal and external medium, but usually at much slower
heating speeds compared with heating by microwave. Instead of heat transfer like most
conventional heating, the energy is converted from electromagnetic energy to thermal
energy (Sun et al., 2016).

The use of microwave heating in oil and gas reservoirs is mostly limited to heavy oil
production, with limited application to or studies about gas production in shale gas reservoirs
(Mutyala et al., 2010). According to Mutyala et al., (2010), commercial application of
microwave heating is lacking due to high initial setup costs and the uncertainty surrounding
its potential. However, they went further to argue that if it could be demonstrated that the
potential of microwave heating significantly increases production and ultimately profits, then
microwave heating could be a useful application in the future for the petroleum industry.

Applications of microwave heating began in the 1970’s (Eskandari et al., 2014) with different
researchers mainly involved in researching the effect of field specification on the performance
of microwave heating. Numerical modelling studies of microwave heating involve highly
coupled nonlinear problems that combine several fields of physics and material science (Sun
et al., 2016). An attempt is often made to replace the conventional heat source with a
microwave source. Rybakov et al., (2013) argue that where such a replacement can be
justified economically, the microwave technology can replace the conventional heat source
either fully or partially.

For increasing the formation temperature, microwaves can be generated either at the surface
or downhole (See Figure 6.1). The surface microwaves are directed to the formation via
waveguides similar to optical fibres, whereas for downhole microwave generation, electricity

173
is sent downhole with the help of electric cables and the microwaves propagated through the
formation (Kamal et al., 2011).

Thermal stimulation is particularly useful with heavy oil and bitumen, where the temperature
of the formation is raised to the point where the viscosity of the fluid is drastically reduced.
In most of these unconventional resources where thermal stimulation has been applied, the
most common techniques are steam injection, steam assisted gravity drainage (SAGD), THAI
and THAI – CAPRI, with very few applications of electrical heating. The use of electric heating
is particularly useful where steam cannot be used because of the formation depth or
excessive heat losses, and in cases where there exist thief zones (Sierra et al., 2001).

Microwave technology has recently been applied to shale oil (Hascakir and Akin,2009), but as
yet, no application in shale gas reservoirs has been reported in a full field demonstration. The
purpose of this chapter is to develop a microwave heating model as a source of thermal
stimulation by coupling with the fluid flow equations developed in the previous chapter.

The application of microwave heating in the petroleum industry is often targeted at specific
problems, such as reducing oil viscosity in heavy oil reservoirs. Early applications of electrical
heating employed in heavy oil have been summarised by Sierra et al., (2001) and are
summarised in the table below.

Table 6.1: Electrical heating employed in heavy oil (Sierra et al., 2001)

Date Test Site Project description Results

1985 Ardmore, Oklahoma Very viscous oil6-15 APi. Starting with 18-20°C the
Several temperature temperature within 3 feet rose
observation wells were to greater than 105°C within
drilled within 4-15’ of test 40 days of heating
well. Electrode energized by
50 KVA Power Unit
1988 Frog Lake 11D-15-56- Low frequency Resistive 2-3 times production increase
3 W4M heating using avg. 30KW
1989 Schoonebeck 280 Low frequency Resistive Production increase 13 to 31
heating using avg. 60KW m3/day
91-88 Lashburn A1-11-48- Mix of resistive heating and Peak simulation ratio of 3.75
25-W3M tubing heating application while tubing heating yielded
power: 30-50 KW marginal improvement

174
 91-88 Rio-Panon Field Multiwell reservoir heating Average doubling of
project. 100-140 KW used production is reported
1997 St. Paul, Alberta Induction heating avg. Production increase by 1.5
power 10KW times
1998 Jenner Field, Alberta Induction heating avg. Marginal decrease in water cut
power 30KW
1998- pelican Strap well induction heating Not successful
99
1998 Elk Point Induction heating avg. Marginal success
power 20KW
1998- Bahrain Oilfield Induction heating with Tripling of the production is
2001 power 5-8 KW reported

Microwave heating has been shown to be effective in increasing the overall gas production
when compared with primary depletion mechanisms in highly viscous and low permeability
reservoirs. By conducting simulation studies, Sahni et al., (2000) modelled the application of
electromagnetic heating in two different reservoirs of thin sand zones separated by shale
layers and a low permeability reservoir with moderate viscous oil. They showed that in some
cases, electromagnetic heating could be a substitute for steam injection, or could be used in
combination with steam to improve oil production. For example, it could be used as a pre-
heating mechanism to create preferential pathways before steam is injected, especially in tar
sand. This, according to Sahni et al., (2000), can reduce heat loses during steam injection and
improve its performance. The results of Sahni et al., (2000) showed an 80% increase in
cumulative production over primary depletion when two 60 Kw microwave sources were
placed around the antenna within a year, with temperature rising by 300°F.

A numerical simulation study was conducted by Pizarro and Trevisan (1990) to simulate
electric heating in heavy oils, with the aim of improving recovery. The results were then
validated with analytical solutions and a comparison made with the field test that was
conducted. Results from the simulation studies point to electrical heating as a well stimulation
process, since the heating effect observed was concentrated on the near well region. A
reasonable match with the numerical simulation was also observed for the field test data from
Rio Panon in Brazil. The field test showed an increase in production from 1.2 Bbls/day to 10
Bbls/day after applying 300 Kw of power for 70 days in a viscous oil.

175
Kasevich et al., (1994) concluded from several field tests they conducted that high frequency
waves in the form of radio frequency heating performs as is expected from theoretical
applications. In one such test, temperature rose to about 220 °F around a borehole of
formation temperature of 90°F after stimulating the formation with radio frequency for 40
hours. Thus, they demonstrated through both low and high-power tests that the use of radio
frequency as a thermal stimulation is a viable alternative to conventional heating methods.

Ovalles et al., (2018) carried out simulation studies to investigate the impact of dielectric
heating in a heavy oil reservoir using a shaped dipole antenna. The results from simulation
showed an increase in oil production when dielectric heating is applied in a heavy oil reservoir.
They also demonstrated that peak production is increased when the radio frequency heating
is started before the start of production.

Besides the use of microwave heating in heavy oil production, it has also been used in
removing moisture in the near wellbore area. Water blockage during drilling, completion and
stimulation can seriously affect gas well deliverability if not treated. For instance, moisture
invasion can result in a decrease in permeability, thereby inhibiting the wells from producing
to their maximum capacity. Studies conducted by Wang et al., (2015) demonstrated the
effectiveness of using microwave heating to increase gas relative permeability as well as
decreasing water saturation through the heated zone in the reservoir. This resulted in an
increase in gas well productivity.

Denny (2007), Jamaluddin et al., (1995) and Li et al., (2006) conducted experiments using
microwave heating as a way of reducing water blockage caused by fluid invasion during
drilling in a sandstone gas reservoir. The aim of their studies was to investigate the feasibility
of microwave heating for removing excess water. The results from the experiment conducted
by using MW heating (2450 MHz, 600 W, 40 minutes) showed the composition of the samples
remained the same, while water saturation decreased rapidly. Kamal et al., (2011) also
demonstrated the use of microwave heating in addressing the problem of liquid loading in
gas reservoirs. The target of the experiment was to use dielectric heating to reduce the fluid
density, thereby allowing the gas to flow naturally to the surface. The result of the experiment
showed that fluid density decreased and gas and vapour flow occurred naturally to the
surface.

176
Several advantages can be obtained from the use of microwave heating over conventional
electric heating as well as other thermal stimulation methods. Some of the challenges of
thermal stimulation using fluid injection, like steam, can be easily overcome by the use of
microwave heating. These include low initial fluid injection, steam override, and difficulty in
establishing communication paths between wells. Hiebert et al., (1986) argue that these
problems hardly occur in microwave heating. They argue, for instance, that since no fluid is
involved in microwave heating, the problem of low initial fluid injection can be easily avoided;
also, the penetration depth of microwaves normally occurs around the near wellbore and
deeper into the formation, allowing for a larger portion of the reservoir to be stimulated.
Finally, production can easily be obtained during or immediately after the microwave is
injected, provided sufficient formation pressure is available.

Other advantages of microwave heating over conventional heating methods include quicker
start-up and stopping, non-contact, speed and volumetric features, a higher level of safety
and automation, and energy conversion instead of heat transfer; also, it is environmentally
friendly (Farag et al., 2012; Ge et al., 2013; Li et al., 2017)

Even though at present there are no commercial applications of microwave heating in shale
gas reservoirs, different energy transfer methods from the source to the formation mean that
microwave energy is more beneficial than other conventional thermal processes (Mutyala et
al., 2010). According to Mutyala et al., (2010) , conventional issues, such as long heating
periods coupled with energy loss, can be easily minimised when microwave heating is used.

Microwave application can have several drawbacks. Over exposure of electromagnetic waves
can cause damage to personnel and hence proper insulation is required for all the tubing,
casing and wellhead. The use of microwave heating can also be more costly than the same
amount of steam energy (Hiebert et al., 1986).

One major drawback of using microwave heating is the problem of high temperature
localisation around the microwave antenna. This can easily cause damage, especially to the
encasing material, resulting in poor energy distribution, with the heat being conducted
instead of being electromagnetically transported as a wave (Bientinesi et al., 2013;
Muhammad, 2017). Bientinesi et al., (2013), however, showed that the presence of shale
quartz can limit the increase in temperature around the antenna, by investigating
177
temperature distribution, oil displacement and the effectiveness of “low lossy” quartz sand
in reducing temperature around the casing.

6.3 Modelling microwave heating

The heating process using microwaves can be modelled either analytically or numerically.
Analytical solutions to microwave heat transfer has been performed by Hossan and Dutta
(2011) and Pincombe and Smyth (1991). However, due to the complexities associated with
Maxwell’s equation, almost all of the microwave heating process is solved using numerical
techniques. Several researchers have modelled the process of microwave heating using
numerical techniques such as the finite difference time domain method (FDTD) that was
proposed by Yee (1966).

Microwave heating can be solved by coupling the heat and mass transfer equation with a
source term. Rather than solving Maxwell’s equation for the electromagnetic field, Lambert’s
law, expressing power as an exponential decay, could also be used to describe the source
term (Campañone and Zaritzky, 2005; Zhou et al., 1995) with the expression

P  Po e 2 d  ………………………………………..(6.1)

Po = surface power

d = maximum distance measured from the surface

 = attenuation factor, which is a function of the dielectric constant of the material.

Where Maxwell’s equation is used, the source term is determined by the electromagnetic
(EM) power that is dissipated in the material due to its dielectric losses. This can be expressed
(Torres and Jecko, 1997; Zhao et al., 2011) by

  

1 D  E
P E.  D. ……………………………..(6.2)
2 t t 
 

And for the case of steady state time harmonic EM fields, this may yield

178
 1
P   o '' E 2 ……………………………………………..(6.3)
2

P is the absorbed power per unit volume (W/m3)

Where  is the angular frequency, o is the vacuum permittivity,  '' refers to the

imaginary part of the materials permittivity, E is the amplitude of E . The microwave power
absorbed can also be represented as a function of the electric field (Chaiyo and Rattanadecho,
2013; Klayborworn et al., 2013):

P  2 f  o r '  tan   E 2 ……………………………………………(6.4)

Where r' represents relative dielectric constant,  o is the permittivity of free space and

tan  is the tangential loss coefficient.

Farag et al., (2012) describe the equation given by Lambert’s law as representing the local
value of the power at a certain distance from the surface of the heated material, while
Equation 6.3 represents generated power.

6.3.1 Maxwell equations

There are basically four of governing equations that can be used to characterised
electromagnetism, known collectively as the Maxwell equation. These four equations are
derived from the work of other researchers, but were combined together to describe
electromagnetism by Maxwell in 1873. They include Gauss’s law, Faraday’s law and Ampere’s
law. Maxwell’s equations are expressed in terms of vector components of electric and
magnetic field together with their source quantities J and  . The differential form of
Maxwell’s equation can be expressed as follows:


 B
 E   ……………………….(6.5)
t

179
Equation 6.5 is known as Faraday’s law and it states that a circulating electric field induces a
time varying magnetic field, whilst a time varying magnetic field induces a circulating electric
field.


 D

 H  J  , …………………………….(6.6)
t

According to Amperes law (Equation 6.6), a circulating magnetic field will induce current and
a time varying electric field.

Gauss’s law is defined by Equation 6.7, which shows that an electric field will diverge on
positive charges and converge on negative charges, whilst Equation 6.8 shows that the net
magnetic flux out of a region is zero. Thus, the magnetic field will always form loops.


. D   ……………………………………………….(6.7)


. B  0 , ……………………………………………(6.8)

   
E  E  x, y, z, t  refers to the Electric filed, H  H  x, y, z, t  is the magnetic field,
   
D  D  x, y, z, t  is the electric flux density, B  B  x, y, z, t  is the magnetic flux density,
 
J  J  x, y, z, t  is the electric current density,     x, y, z , t  electric charge density.

6.3.2 Dielectric properties and constitutive relations

One key factor that controls the interaction between the material and electric field is the
dialectic properties of the material. The dielectric constant is the measure of how the electric
fields interact with the material. It can also be defined as the permittivity relative to the
permittivity of free space. The permittivity (  ) measures how well the material stores electric
energy, whereas the permeability (  ) refers to how well the material stores magnetic
energy.

   o r …………………………….(6.9)

  o r ………………………………….(6.10)
180
 o = 8.854187817 1012 F/m , r  1.256637061106 H/m

 r = relative permittivity

r = relative permeability.

A constitutive relationship can be established between the electric field and the electric flux
density and the dielectric constant as

 
D  t    E …………………………………..(6.11)

 
B  t    H ……………………………………..(6.12)

For linear, isotropic and dispersive material

 
D  t     t  * E  t  …………………………………….(6.13)

The relative permittivity is made up of two parts, real and imaginary parts. The real part refers
to the dielectric constant  ' , whereas the imaginary part refers to dielectric loss factor  " .

 r   'r  j "r ……………………………………….(6.14)

Where the dielectric properties of a porous material is a function of temperature , this can be
further defined as (Klayborworn et al., 2013)

 'r T    "rf T   1     'rp …………………………………………(6.15)

 "r T    "rf T   1     "rp ………………………………………….(6.16)

The loss tangent coefficient can be written as

 "r  T 
tan   ' ……………………………………….(6.17)
 r T 

By assuming no charges or current, the Maxwell equations for linear and isotropic material
can be written as

181
 
 B
 E   …………………………………………(6.18)
t


D

 H  …………………………………………. (6.19)
t


. D  0 ……………………………………………(6.20)


. B  0 ……………………………………………… (6.21)

   
Where D  t    E , B  t    H

To solve Equations 6.18-6.21 , Yee (1966) proposed a method called finite difference time
domain (FDTD) as a three dimensional solution to Maxwell’s equations. By substituting the
constitutive relations into Maxwell’s equations, we can obtain the following final set of
equations:


 H
  E   …………………………………(6.22)
t


 E
 H   ……………………………………..(6.23)
t

 
.   H   0 …………………………………………..(6.24)
 

 
.   H   0 ……………………………………………..(6.26)
 

182
6.3.3 Transverse Electromagnetism (TEM), Transverse Electric (TE) and
Transverse Magnetic (TM)
There are three categories of electromagnetic waves according to the longitudinal
components of the electric field ( Ez ) and the magnetic field ( H z ). These three categories

are known as the transverse electromagnetic, transverse electric and transverse magnetic
(Figure 6.2). Where the Ez  0 and H z  0 , both E and H are said to be transverse to the

direction of propagation. When this happens, the electromagnetic wave is normally referred
to as a TEM wave. In the transverse magnetic mode, Ez  0 and H z  0 . The magnetic field

is transverse to the direction of propagation. In transverse electric mode, Ez  0 while

H z  0 .The electric field is transverse to the direction of propagation. TE mode have been
chosen as the mode of propagation for this work.

Figure 6. 2: TEM, TE and TM

6.4 Finite Difference Time Domain (FDTD) Method

The finite difference time domain method has been widely used in solving most
electromagnetic problems due to its simplicity. It was first proposed by Yee (1966) but has
subsequently been improved by other researchers. It has been successfully applied in a
variety of problematic situations, such as antennas, electromagnetic absorption in the human
body exposed to radiation, and microchip circuits (Akram and Jasmy, 2012; Rathi et al., 2012;
Toroğlu and Sevgi, 2014). Since the exact solution of Maxwell’s equations is impossible,
183
numerical methods such as FDTD can be applied. For the time-dependent Maxwell’s
equations, discretization is achieved using central difference approximations to space and
time, with resulting finite difference equations solved using a leapfrog approach. The curl
equations are written in FDTD form on a Yee cell, as shown in Figure 6.3. The electric and
magnetic fields are staggered in time so that the electric field will exist at integer time steps

of 0, t , 2t,.. and the magnetic field at half time steps of t , t  t , 2 t  t ,...

2 2 2

Figure 6.3: The Yee cell

In the main loop of the FDTD equation, an update equation is derived and the resulting
equations are solved using the finite difference equations for the future time values of each
field. The curl equations are defined in Equations 6.27 and 6.28.

~ ~ ~ ~
i , j , k 1
E | i , j 1,k  Ez |t i , j ,k E y |t  E y |t i , j ,k
 z t 
E i , j ,k
C x t
|
y z
~ ~ ~ ~
E | i , j ,k 1  Ex |t i , j ,k Ez |t i 1, j ,k  Ez |t i , j ,k
 x t 
E i , j ,k
C | ……………………………………….(6.27)
y t
z x
~ ~ ~ ~
E y |t i 1, j ,k  Ez |t i , j ,k Ex |t i , j 1,k  E y |t i , j ,k
 
E i , j ,k
C z t
|
x y

And also
184
 H z |it ,j ,kt 2  H z |it ,j t1,2k H y |it ,j ,kt 2  H y |it ,j ,kt 21
 
H i , j ,k
C |
x t  t
2 y z
H x |ti ,j ,kt 2  H x |ti ,j ,kt 21 H z |it ,j ,kt 2  H z |it 1,tj ,2k
 
H i , j ,k
C |
y t  t
…………………………………..(6.28)
2 z x
H y |t t 2  H y |it 1,tj ,2k
i , j ,k
H x |t t 2  H x |ti ,j t1,2k
i , j ,k

 
H i , j ,k
C |
z t  t
2 x y

6.4.1 Courant stability criterion

For numerical stability, the appropriate time step needs to be chosen according to stability
conditions. In general, the Courant condition is given as

x
t  …………………………………………………….(6.29)
co d

x is the grid size, with d  1, 2 or 3 for one, two- and three-dimensional problems
respectively. The time step t is however normally chosen as

x
t  (Sullivan, 2000)
2co

6.4.2 Update Equations

Update equations can now be written for solving the future time values of the fields
associated with Ez and H z modes. (See appendix 6 for derivation of update equations)

 c t  E
H x |ti ,j t 2  H x |ti ,j t 2    o i , j  C x |it , j
  xx | 
 c t  E
H y |ti ,j t 2  H y |ti ,j t 2   o i , j  C y |it ,j t 2
  yy | 
  ………………………………(6.30)
~ ~
D z |it ,j t  D z |it , j   co t  C z |ti , j
H

~  1  ~ i, j
Ez |ti ,j t    Dz |t t
  zz |
i, j


Likewise, Hz mode is derived from the following update equations:

185
  c t  E i, j
H z |ti ,j t 2  H z |ti ,j t 2    o i , j  C z |t
  zz | 
~ ~
D x |it ,j t  D z |it , j   co t  C x |ti ,jt
H

2
~ ~
D y |it ,j t  D z |it , j   co t  C y |it ,jt
H
…………………………………….(6.31)
2

~  1  ~ i, j
Ex |ti ,j t   i, j 
Dx |t t
  xx | 
~  1  ~ i, j
E y |ti ,j t  
  yy |i , j  y t t
D |
 

6.4.3 Microwave penetration depth

In microwave heating, an efficient energy process is highly dependent on the penetration
depth of the sample. According to Sun et al., (2016), the penetration depth remains a useful
parameter to quantify the heating efficiency and uniformity within a sample heated by
microwave. This can normally be observed in very low efficiency of heating in samples that
have large size but shallow penetration depth. The penetration depth, d , is normally defined
as the distance from the surface to the place at which the wave power or magnitude of the
field strength drops to e 1 (=0.37) of its value at the surface ( See Figure 6.4), whereas power
penetration depth , Dp , is half the value of the field penetration depth (Peng et al., 2010; Sun

et al., 2016). There is a negative correlation between the frequency and the penetration
depth. That is, when the frequency is increased, the penetration depth also decreases. This
implies that materials with a high capability to convert microwave energy to heat will have a
very low penetration depth. In microwave heating, a sample dimension needs to be
approximately equal to the penetration depth, since any penetration depth smaller than the
sample dimension will imply very limited microwave energy, and hence only a small portion
of the sample could end up being heated.

The penetration depth can be expressed (Peng et al., 2012, 2010; Sun et al., 2016)as

1/2
2c

 1  tan 2  tan 2   tan 2   tan 2    tan 2   tan 2    1
1/2
d  2  Dp 
'      
  r r 
'

…(6.32)

186
Since shale is highly non-magnetic, we can ignore the magnetic effects, and hence the

2c
penetration depth of the microwave can be expressed as d 
 
1/2
  ' 1  (tan   ) 2  1 
 
……………………(6.33)

Where c is the velocity of light, tan  is the loss tangent.

Figure 6.4: Energy deposition in a lossy formation. (Liu et al., 2018)

Figure 6.5 shows the penetration depth of three different recommended industrial
frequencies, of 0.915 GHz , 2.45 GHz and 5.8 GHz. It can be observed that a higher frequency
of 5.8 GHz resulted in the lowest power penetration depth of the sample. According to Sun
et al., (2016), once the frequency is set, any material with a high capability of converting
microwave energy to heat tends to have a low penetration depth, while low-loss materials
have a high penetration depth.

187
Figure 6.5: Effect of microwave frequency of 0.915, 2.45 and 5.8 GHz on power penetration
depth

6.5 Numerical Modelling of a 2D Reservoir

In order to investigate the application of microwave technology in shale gas reservoirs, a 2D
reservoir model was simplified with a waveguide in Transverse Electric (TE) mode. A
microwave waveguide is located at the centre of the model. The microwave frequency is set
at 2.45 GHz. The initial temperature of the reservoir is 579.6 ° R. The reservoir is divided into
a 21 x 21 matrix block with dimensions of 50 ft. x 50 ft. A summary of all computational
parameters is taken from literature, experimental data or from our estimations, and is given
in Table 6.2.

188
Table 6.2: Parameters used in simulation study

Parameters Value Reference

Microwave frequency 2.45 GHz (Abdulrahman and Meribout, 2014)
Dielectric Constant 3 (Josh et al., 2012)
Lost Factor 0.2 (Al-Harahsheh et al., 2009)
Heat capacity of shale (J/K/kg) 1000 (Wang et al., 2017)
Thermal Conductivity (W/M/K) 4 (Wang et al., 2017)
Initial reservoir Pressure (psia) 1508.39 (Guo et al., 2015)
Initial reservoir temperature(°R) 579.6 Assumption
Bottom hole pressure (psia) 500 Assumption

To solve for the gas production associated with microwave heating in shale gas reservoirs, a
coupled electromagnetic–thermal model has been developed for a shale gas reservoir. The
process of coupling all of the equations is described using Figure 6.6.

1. Firstly, the electric field from Maxwell’s equation is solved using finite difference time
domain (FDTD) and a transverse electric (TE) formulation
2. The power dissipated is calculated from the solution of Maxwell’s equation in step 1
3. Temperature computations within the reservoir are calculated using power obtained
from Maxwell’s equation as a heat source
4. The pressure field of matrix and fracture is determined from shale gas equations
5. Production from the reservoir is calculated.

189
Figure 6.6: Combined electromagnetic, thermal and production algorithm for a shale gas
reservoir

190
Figure 6.7: Finite difference time domain (FDTD) algorithm for solving electric field with a
perfectly matched layer (PML) as boundary condition

191
6.6 Results and Discussion
6.6.1 Effect of microwave heating on gas production
The effect of microwave heating on production rate and cumulative production can be
observed in Figures 6.8 and 6.9. From Figure 6.8, production rate increases considerably
where microwave heating is applied, compared with no microwave heating. This is because
the application of microwave heating ensures an increase in temperature of the formation,
which consequently increases desorption of the adsorbed gas. The contribution from the
adsorbed gas therefore enhances the production rate from the reservoir. After 360 days,
microwave heating contributes to an increase of 25% in overall cumulative production.

3000

2500
Production rate, Mscf

2000

1500

1000

500

0
0 50 100 150 200 250 300 350 400
Time, days
With Microwave heating Without Microwave heating

Figure 6.8: Cumulative production with and without microwave heating

192
 90
85
80

Production rate, Mscf

75
70
65
60
55
50
45
40
0 50 100 150 200 250 300 350 400
Time,days

With Microwave heating Without Microwave heating

Figure 6.9: Production rate with and without microwave heating

6.6.2 Temperature distribution

Figure 6.10a-6.10c shows temperature distributions in the reservoir following microwave
heating at different heating time periods: 90 days, 180 days and 360 days. Temperature is
highest at the centre where the microwave radiation is emitted and gradually decreases at
surrounding nodes. As the heating time increases and reaches 360 days, it can be observed
that the central node where the microwave pulse is excited has the highest temperature.
Also, the average temperature throughout the reservoir rises as the microwave heating
continues around the central node where the waveguide is located.

193
Figure 6.10a: Temperature distribution after 90 days of microwave heating

Figure 6.10b: Temperature distribution after 180 days of microwave heating

194
Figure 6.10c: Temperature distribution after 360 days of microwave heating

6.6.3 Effect of microwave frequency

Frequency ranges that are normally suggested for industrial microwave processing include
915 MHz, 2450 MHz and 5800 MHz (Li et al., 2017; Peng et al., 2011). Different frequency
ranges used in microwave heating can affect the temperature differential within the material
or, as in this case, the reservoir formation. The choice of different frequencies also has a
significant effect on production in shale gas reservoirs. In order to demonstrate the effect of
microwave heating, three frequency ranges of 915 MHz, 2450 MHz and 5800 MHz have been
chosen, and this resulted in different temperature profiles, as shown in Figures 6.11-6.13.
Critical observation of Figures 6.11-6.13 shows that higher microwave frequency results in
relatively higher temperature profile. However, the increase in temperature for higher
microwave frequency is only observed at the surface of the material, whereas for lower
frequencies, temperature rises more than the higher frequencies at the inner section of the
sample. This is because the penetration depth for high microwave frequency is much lower
than the penetration depth of lower microwave frequency (Peng et al., 2011). The
penetration depth is, however, strongly dependent on the dielectric properties of the
material, which is also heavily dependent on microwave frequency and temperature. Thus,
there would be poor heating efficiency at the higher frequency range of 5800 MHz and 2450
MHz compared with 915 MHz, since energy would be easily lost to the surroundings. The
195
greater energy efficiency at the lower frequency range of 915 MHz would lead to an increase
in production and ultimately higher cumulative production, compared with a high frequency
of 2450 MHz, as seen in Figures 6.14 and 6.15. Production rate is highest when frequency is
lowest at 915 MHz, and lowest at the high frequency of 5800 MHz as seen from Figure 6.14.
In Figure 6.14, overall cumulative production is highest at a frequency of 915 MHz, compared
with 2450 MHz and 5800 MHz

Figure 6.11: Temperature distribution at 5800 MHz at time 360 days

196
Figure 6.12: Temperature distribution at 2450 MHz at time 360 days

197
Figure 6.13: Temperature distribution at 915 MHz at time 360 days

100
90
80
Production rate , Mscf

70
60
50
40
30
20
10
0
0 50 100 150 200 250 300 350
Time, days

2450 MHz 5800 MHz 915 MHz

Figure 6.14: Production rate at frequency of 915, 2450 and 5800 MHz

198
 3500

3000
Production rate , Mscf

2500

2000

1500

1000

500

0
0 50 100 150 200 250 300 350 400
Time, days

2450 MHz 5800 MHz 915 MHz

Figure 6.15: Cumulative production at frequency of 915, 2450 and 5800 MHz

6.6.4 Effect of microwave heating on free gas content

According to Mariotte’s law, free gas content can be predicted (Liu et al., 2012, Li et al., 2017)
by

 PTs
Vg  ……………………………………. (6.34)
s s
PT

199
Where Vg is the free gas content , P is the gas pressure, T is the reservoir temperature ,  s

is the density of shale , Ts and Ps represents the temperature and pressure at standard

conditions. From Figures 6.16 and 6.17, we can observe the evolution of the gas content with
and without microwave heating. Reliance on normal pressure depletion mechanisms, without
any heating, shows very small changes in the gas content. We can observe that gas content
around the location of the wellbore is lower than that further away from the wellbore.
However, after a period of 360 days of microwave heating, there is a considerable drop in the
gas content around the wellbore and waveguide. Also, the gas drainage is much higher further
away from the wellbore with microwave heating, compared with the case of no microwave
heating. This is because the areas affected by the microwave heating result in desorption of
the gas. Therefore, as the temperature rises, there is a consequent decrease in the gas
content.

Figure 6.16: Free gas content evolution with microwave heating at 360 days

200
Figure 6.17: Free gas content without microwave heating at 360 days

6.7 Conclusion:
Thermal stimulation in shale gas reservoirs could be invaluable in increasing gas production.
The idea behind thermal stimulation is to accelerate desorption of the adsorbed gas by
altering the desorption behaviour through a rise in formation temperature. To do this, a fully
coupled electromagnetic–thermal reservoir model has been developed to investigate the
impact of thermal stimulation on shale gas recovery, and several factors have been
investigated to assess their impact on the effectiveness of thermal stimulation. Microwave
heating has been particular chosen as the heat source. In order to couple the heat source
from the microwave to the shale system, Maxwell equations that describe electromagnetism
have been solved using a finite difference time domain method (FDTD). This methodology
involved solving for the electric field using a leapfrog approach. A transverse electric (TE)
mode has been chosen as the propagating mode. Several conclusions can be deduced from
the results of the simulation undertaken in this chapter:

201
 Firstly, the coupled electromagnetic –thermal model for shale gas reservoirs predicted
a more production, compared with a scenario where no microwave heating was used.
Cumulative production increased by 25% at the end of 360 days when microwave
heating was used. The simulation results show that thermal stimulation by microwave
can indeed promote shale gas recovery.
 Secondly, the temperature profile at the location of the waveguide showed the
highest heat distribution compared with surroundings further away from the heat
source. The choice of microwave frequency also had an effect on the overall
production and heat distribution in the formation. The highest temperature was
observed for frequencies of 2450 MHz and 5800 MHz, compared with 915 MHz
However, penetration depth was low for frequency ranges of 2450 and 5800 MHz,
compared with 915 MHz. Low penetration depth meant only a small portion of the
formation was heated and consequently, production rate was low for high frequency
ranges, compared with a frequency of 915 MHz
 Finally, simulation results showed decreasing free gas content in the matrix block
when microwave heating was applied, compared with no microwave heating. This can
be attributed to the rise in temperature at the surrounding waveguide, which causes
desorption of the adsorbed gas around the thermally stimulated areas.

202
Nomenclature
Po = surface power

d = maximum distance measured from the surface

 = attenuation factor
P = absorbed power per unit volume (W/m3)

 = angular frequency,

 o = vacuum permittivity,
 '' = imaginary part of the materials permittivity,

E = amplitude of E

 r ' = relative dielectric constant

 o = permittivity of free space
tan  = tangential loss coefficient

 r = relative permittivity

r = relative permeability

c = velocity of light


E  refers to the Electric filed


H  is the magnetic field


D  is the electric flux density


B  is the magnetic flux density


J  is the electric current density

  electric charge density

Vg = free gas content

203
T is the reservoir temperature

 s is the density of shale

Ts = temperature at standard conditions

Ps = pressure at standard conditions

204
 Chapter 7
Summary, Conclusion and Future Work

7.1 Conclusion:
This thesis presents novel coupled electromagnetic –thermal model with microwave heating
taking into consideration the non-Darcy flow within the shale matrix and fractures while also
developing new methodology for the inclusion of temperature dependent gas adsorption
models in material balance calculations. The main results are summarised as:

1. A statistically robust method of sum of normalised error approach have been adopted
when selecting error function to use in gas adsorption modelling. This ensures optimal
selection of adsorption model
2. Several variation of temperature dependent adsorption models have been developed
using isosteric heat of adsorption and applied to shale gas data
3. Designing new methodology for the inclusion of temperature dependent adsorption
models in material balance calculations for unconventional gas reservoirs
4. Evaluating thermal stimulation strategy as enhance gas recovery using microwave
heating as the main heat source to improve overall cumulative production.

Based on the achievements outlined above, a detailed summary is provided in the following
section.

7.1.1 Review and Modelling in Shale Gas Reservoirs (Chapter 2)

Several single- and multi- component shale gas adsorption models commonly used in
estimating gas adsorption were reviewed, and applied to both shale and activated carbon to
evaluate their accuracy and effectiveness in predicting gas adsorption:

1. Langmuir isotherm is the most common single adsorption model applied in most
commercial simulators, despite its limitations.
2. Vacancy solution model proved to be more successful in predicting single gas
adsorption when applied to the New Albany shale dataset.
3. Due to very limited studies on multi-component adsorption in shales, activated carbon
is commonly used as an adsorbent for binary gas adsorption in shale gas.

205
 4. Vacancy solution model was most successful in predicting binary gas adsorption,
compared to ideal adsorbed solution and extended Langmuir model.
5. A statistically robust method of analysing error functions such as sum of normalised
error (SNE) have been used adsorption modelling. This has enabled the selection of
adsorption models with the most accurate representation of gas adsorption
phenomenon in shale gas reservoirs.

7.1.2 Temperature Dependent Gas Adsorption Modelling in Shale gas

Reservoirs (Chapter 3)
This thesis introduced several modified version of Langmuir models to account for
temperature variation in the shale gas reservoir. Both temperature-dependent Langmuir
volume and non-temperature-dependent Langmuir volume have been included in modelling
to evaluate their accuracy in modelling shale gas reservoirs:

1. Data from several literature sources have been collected and for the first time,
temperature-dependent gas adsorption models have been used to evaluate their
accuracy in predicting shale gas adsorption.
2. Exponential model showed the most accuracy for all the data sets analysed , followed
by Langmuir-Freundlich model 2.
3. Very little difference was observed between models that have temperature-
dependent Langmuir volume and those with non-temperature-dependent Langmuir
volume.
4. Introduction of temperature-dependent models enables the extrapolation of
temperatures for which data about gas adsorption is unavailable
5. Validation and extrapolation showed different predictions for all three-key
temperature-dependent adsorption models in the form of Bi-Langmuir, Exponential
and Langmuir-Freundlich models.
6. Due to differences in gas adsorption prediction upon extrapolating to higher
temperatures, caution is advised when choosing a particular temperature-dependent
adsorption model to be used in numerical simulation studies involving thermal
stimulation.

206
7.1.3 Material Balance Calculations for Unconventional Gas Reservoirs Using
Temperature Dependent Gas Adsorption Models (Chapter 4)
New analytical methodology has been developed, similar to that of Ahmed and Roux, to be
used in material balance analysis for unconventional gas reservoirs. The methodology has
been validated using data from available literature sources:

1. Temperature-dependent gas adsorption models can be used in place of Langmuir

isotherm to evaluate gas in place calculations
2. Accurate reservoir performance analysis can be done with temperature-dependent
models. Gas in place calculations, pressure predictions and future productions can be
accurately obtained using this method.
3. Comparison of the results of Ahmed and Roux with the new methodology in predicting
reservoir performance showed slight variations, but overall, both models showed a
good match with actual reservoir data.
4. The new methodology developed in this study has the added advantage of predicting
gas adsorption at several temperatures where adsorption data is unavailable.

7.1.4 Numerical Simulation of Gas adsorption in Shale Gas Reservoirs

(Chapter 5)
A mathematical model for shale gas reservoirs has been developed including dual porosity
and dual permeability concept. The mathematical model accounts for non-Darcy flow in
the shale matrix. Simulation studies with and without gas adsorption have been
evaluated. Finally, the heat equation is coupled with the shale gas model to evaluate
thermal stimulation.

1. Including gas adsorption in reservoir simulation studies shows an increase in gas

production, compared with simulation studies in which no adsorption model is used.
2. Matrix porosity and fracture permeability have the greatest influence on production
in shale gas reservoirs, compared with matrix permeability and fracture porosity.
3. Gas production increased with a thermally enhanced gas recovery process, compared
with when no thermal stimulation is applied to the reservoir.
4. The choice of adsorption model used in thermal stimulation studies can result in
different production profiles, since extrapolation at higher temperatures results in
different adsorption capacity predictions.
207
 5. Validation of adsorption models with actual data at multiple temperatures is
necessary to ensure accurate simulation.

7.1.5 Thermal stimulation of shale Gas Reservoirs using Microwave Heating

(Chapter 6)
Thermal stimulation of shale gas reservoirs using microwave heating involves solving
Maxwell’s equation. The finite difference time domain method has been used to obtain the
electric field needed to calculate the power source. This is then coupled to the heat equation
as a source term to obtain the temperature distributions in the reservoirs.

1. Gas production involving microwave heating resulted in an increase of about 25% in

total production at the end of one year.
2. A higher temperature profile could be observed at the centre of the reservoir where
the microwave waveguide is located. Further away from the waveguide, lower
temperature distributions were obtained.
3. A lower frequency range resulted in much higher penetration depth and consequently
a higher production profile, compared with a higher frequency range, which resulted
in a much higher temperature distribution but lower penetration depth
4. A higher temperature profile near the waveguide showed much greater drainage of
free gas content than when no microwave heating was employed.

7.2 Future Work

Considerable amount of work was made in this study to evaluate the contribution of gas
adsorption in shale gas reservoirs; however, further attempts could be made to improve the
overall understanding of modelling and simulation of shale gas reservoirs some of which are
recommended below as future work:

1. Use of artificial neural networks in predicting gas adsorption could help to improve on
the adsorption predictions in shale gas reservoirs. Neural networks have been applied
in many areas of the petroleum industry and they have proven to be very accurate in
use. The use of artificial neural networks to predict gas adsorption of single
component, multi-component and temperature dependency of the adsorbent could
prove invaluable in obtaining accurate estimates in modelling shale gas reservoirs. To
achieve this, a database of gas adsorption in shale gas reservoirs is essential, because

208
 training the network to be able to predict accurately the adsorption capacities relies
heavily on having available data. The main challenge will be with obtaining reliable
shale gas multi-component adsorption data for such training and prediction by the
network.
2. The coupled electromagnetic–thermal model developed for this thesis only assumed
a single phase; however, most shale gas reservoirs have more than a single-phase fluid
present. Hence developing a two-phase model with both gas and water could help to
improve the understanding of reservoir simulation when predicting gas recovery. Also,
since the gas present in shale gas is mostly multi-component gas, the temperature-
dependent adsorption model used in simulation could be replaced with a multi-
component model.
3. The effect of heterogeneity should also be investigated. Throughout this thesis, the
reservoir is assumed to be homogeneous. This is not necessarily true in a shale gas
reservoir, and as such, using a heterogeneous model to understand the thermal
stimulation application could be useful.
4. A Temperature-dependent form of permittivity should be used in microwave heating
of shale gas reservoirs. It is well established that the permittivity of a substance is a
function of temperature. Hence to improve any simulation involving microwave
heating, the temperature-dependent form of permittivity should be used instead of a
constant value. However, the functional form of this dependency is not readily
available unless experiments are carried out to determine this.
5. Finally, consideration of hydraulic fracturing within the shale gas reservoir will move
this conceptual study a step closer to actual representation of shale gas production.
Hydraulic fracturing in shale gas reservoir is essential for gas production and hence
further studies should incorporate this concept to simulate potential production from
shale gas reservoirs.

209
 References
Abdulrahman, M. M., & Meribout, M. (2014). Antenna array design for enhanced oil
recovery under oil reservoir constraints with experimental validation. Energy, 66, 868-
880.
Ahmadpour, A., Wang, K., & Do, D. D. (1998). Comparison of models on the prediction of
binary equilibrium data of activated carbons. AIChE Journal, 44(3), 740-752.
Ahmed, T. H., Centilmen, A., & Roux, B. P. (2006, January). A generalized material balance
equation for coalbed methane reservoirs. In SPE Annual Technical Conference and
Exhibition. Society of Petroleum Engineers.
Akram, G., & Jasmy, Y. (2012). Simulation of the finite difference time domain in two
dimension. World Academy of Science, Engineering and Technology, 63, 623-627.
Al-Harahsheh, M., Kingman, S., Saeid, A., Robinson, J., Dimitrakis, G., & Alnawafleh, H.
(2009). Dielectric properties of Jordanian oil shales. Fuel Processing Technology, 90(10),
1259-1264.
Allen, S. J., Gan, Q., Matthews, R., & Johnson, P. A. (2003). Comparison of optimised
isotherm models for basic dye adsorption by kudzu. Bioresource Technology, 88(2),
143-152.
Anderson, D. M., Nobakht, M., Moghadam, S., & Mattar, L. (2010, January). Analysis of
production data from fractured shale gas wells. In SPE unconventional gas conference.
Society of Petroleum Engineers.
Aronofsky, J. S., & Jenkins, R. (1954). A simplified analysis of unsteady radial gas
flow. Journal of Petroleum Technology, 6(07), 23-28.
Arri, L. E., Yee, D., Morgan, W. D., & Jeansonne, M. W. (1992, January). Modeling coalbed
methane production with binary gas sorption. In SPE rocky mountain regional meeting.
Society of Petroleum Engineers.
Barenblatt, G. I., Zheltov, I. P., & Kochina, I. N. (1960). Basic concepts in the theory of
seepage of homogeneous liquids in fissured rocks [strata]. Journal of applied
mathematics and mechanics, 24(5), 1286-1303.
Bazan, R. E., Bastos-Neto, M., Staudt, R., Papp, H., Azevedo, D. C. S., & Cavalcante Jr, C. L.
(2008). Adsorption equilibria of natural gas components on activated carbon: pure and
mixed gas isotherms. Adsorption Science & Technology, 26(5), 323-332.
Bell, G. J., & Rakop, K. C. (1986, January). Hysteresis of methane/coal sorption isotherms.
In SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers.
Bhatia, S. K., & Ding, L. P. (2001). Vacany solution theory of adsorption revisited. American
Institute of Chemical Engineers. AIChE Journal, 47(9), 2136.
Bientinesi, M., Petarca, L., Cerutti, A., Bandinelli, M., De Simoni, M., Manotti, M., &
Maddinelli, G. (2013). A radiofrequency/microwave heating method for thermal heavy
oil recovery based on a novel tight-shell conceptual design. Journal of Petroleum
Science and Engineering, 107, 18-30.
210
Busch, A., & Amann-Hildenbrand, A. (2013). Predicting capillarity of mudrocks. Marine and
petroleum geology, 45, 208-223.
Bustin, A. M., Bustin, R. M., & Cui, X. (2008, January). Importance of fabric on the production
of gas shales. In SPE unconventional reservoirs conference. Society of Petroleum
Engineers.
Campanone, L. A., & Zaritzky, N. E. (2005). Mathematical analysis of microwave heating
process. Journal of Food Engineering, 69(3), 359-368.
Canel, C. A., & Rosbaco, J. (1992, January). Compositional Material Balance: Its Application
to the Development of an Oil and Gas Field with Retrograde Condensation. In SPE Latin
America Petroleum Engineering Conference. Society of Petroleum Engineers.
Chaiyo, K., & Rattanadecho, P. (2013). Numerical analysis of heat–mass transport and
pressure build-up in 1D unsaturated porous medium subjected to a combined
microwave and vacuum system. Drying Technology, 31(6), 684-697.
Chalmers, G. R., & Bustin, R. M. (2008). Lower Cretaceous gas shales in northeastern British
Columbia, Part II: evaluation of regional potential gas resources. Bulletin of Canadian
Petroleum Geology, 56(1), 22-61.
Chapiro, G., & Bruining, J. (2015). Combustion enhance recovery of shale gas. Journal of
Petroleum Science and Engineering, 127, 179-189.
Chareonsuppanimit, P., Mohammad, S. A., Robinson Jr, R. L., & Gasem, K. A. (2012). High-
pressure adsorption of gases on shales: Measurements and modeling. International
Journal of Coal Geology, 95, 34-46.
Charoensuppanimit, P., Mohammad, S. A., & Gasem, K. A. (2016). Measurements and
modeling of gas adsorption on shales. Energy & Fuels, 30(3), 2309-2319.
Charoensuppanimit, P., Mohammad, S. A., Robinson Jr, R. L., & Gasem, K. A. (2015).
Modeling the temperature dependence of supercritical gas adsorption on activated
carbons, coals and shales. International Journal of Coal Geology, 138, 113-126.
Chekhonin, E., Parshin, A., Pissarenko, D., Popov, Y., Romushkevich, R., Safonov, S., ... &
Stenin, V. (2012). When rocks get hot: thermal properties of reservoir rocks. Oilfield
Review, 24(3), 20-37.
Chen, J., Loo, L. S., & Wang, K. (2011). An ideal absorbed solution theory (IAST) study of
adsorption equilibria of binary mixtures of methane and ethane on a templated
carbon. Journal of Chemical & Engineering Data, 56(4), 1209-1212.
Chen, L., Jiang, Z., Liu, K., Ji, W., Wang, P., Gao, F., & Hu, T. (2017). Application of Langmuir
and Dubinin–Radushkevich models to estimate methane sorption capacity on two shale
samples from the Upper Triassic Chang 7 Member in the southeastern Ordos Basin,
China. Energy Exploration & Exploitation, 35(1), 122-144.
Chen, Q., Tian, Y., Li, P., Yan, C., Pang, Y., Zheng, L., ... & Meng, X. (2017). Study on shale
adsorption equation based on monolayer adsorption, multilayer adsorption, and
capillary condensation. Journal of Chemistry, 2017.

211
Chen, S., Jin, L., & Chen, X. (2011). The effect and prediction of temperature on adsorption
capability of coal/CH4. Procedia Engineering, 26, 126-131.
Cipolla, C. L., Lolon, E. P., Erdle, J. C., & Rubin, B. (2010). Reservoir modeling in shale-gas
reservoirs. SPE reservoir evaluation & engineering, 13(04), 638-653.
Clarkson, C. R. (2003). Application of a new multicomponent gas adsorption model to coal
gas adsorption systems. SPE Journal, 8(03), 236-251.
Clarkson, C. R., & Bustin, R. M. (2000). Binary gas adsorption/desorption isotherms: effect of
moisture and coal composition upon carbon dioxide selectivity over
methane. International Journal of Coal Geology, 42(4), 241-271.
Clarkson, C. R., Bustin, R. M., & Levy, J. H. (1997). Application of the mono/multilayer and
adsorption potential theories to coal methane adsorption isotherms at elevated
temperature and pressure. Carbon, 35(12), 1689-1705.
Clarkson, C. R., Bustin, R. M., & Levy, J. H. (1997). Application of the mono/multilayer and
adsorption potential theories to coal methane adsorption isotherms at elevated
temperature and pressure. Carbon, 35(12), 1689-1705.
Clarkson, C. R., Bustin, R. M., & Levy, J. H. (1997). Application of the mono/multilayer and
adsorption potential theories to coal methane adsorption isotherms at elevated
temperature and pressure. Carbon, 35(12), 1689-1705.
Clarkson, C. R., & Haghshenas, B. (2013, April). Modeling of supercritical fluid adsorption on
organic-rich shales and coal. In SPE unconventional resources conference-USA. Society
of Petroleum Engineers.
Cochran, T. W., Kabel, R. L., & Danner, R. P. (1985). Vacancy solution theory of adsorption
using Flory‐Huggins activity coefficient equations. AIChE journal, 31(2), 268-277.
Curtis, J. B. (2002). Fractured shale-gas systems. AAPG bulletin, 86(11), 1921-1938.
Danner, R. P., & Wenzel, L. A. (1969). Adsorption of carbon monoxide‐nitrogen, carbon
monoxide‐oxygen, and oxygen‐nitrogen mixtures on synthetic zeolites. AIChE
Journal, 15(4), 515-520.
de Swaan O, A. (1976). Analytic solutions for determining naturally fractured reservoir
properties by well testing. Society of Petroleum Engineers Journal, 16(03), 117-122.
Denney, D. (2007). Cleaning up water blocking in gas reservoirs by microwave
heating. Journal of Petroleum Technology, 59(04), 76-81.
Do, D. D. (1998). Adsorption analysis: equilibria and kinetics(Vol. 2, pp. 1-18). London:
Imperial college press.
Do, D. D., & Do, H. D. (1997). A new adsorption isotherm for heterogeneous adsorbent
based on the isosteric heat as a function of loading. Chemical engineering
science, 52(2), 297-310.
Eskandari, S., Jaghargh, E. M., & Rasaei, M. R. (2014). 2D Numerical Simulation and Analysis
of Microwave Heating in Heavy Oil Reservoirs. Energy Sources, Part A: Recovery,

212
 Utilization, and Environmental Effects, 36(18), 1961-1971.
Farag, S., Sobhy, A., Akyel, C., Doucet, J., & Chaouki, J. (2012). Temperature profile
prediction within selected materials heated by microwaves at 2.45 GHz. Applied
Thermal Engineering, 36, 360-369.
Fianu, J., Gholinezhad, J., & Hassan, M. (2018). Comparison of temperature-dependent gas
adsorption models and their application to shale gas reservoirs. Energy & fuels, 32(4),
4763-4771.Abdulrahman, M.M., Meribout, M., 2014. Antenna array design for
enhanced oil recovery under oil reservoir constraints with experimental validation.
Energy 66, 868–880. doi:10.1016/j.energy.2014.01.002
Ahmadpour, A., Wang, K., Do, D.D., 1998. Comparison of models on the prediction of binary
equilibrium data of activated carbons. AIChE J. 44, 740–752.
doi:10.1002/aic.690440322
Ahmed, T., Roux, B., 2006. A Generalized Material Balance Equation for Coalbed Methane
Reservoirs. SPE J. SPE 102638.
Akram, G., Jasmy, Y., 2012. Simulation of the Finite Difference Time Domain in Two
Dimension 6, 717–721.
Al-Harahsheh, M., Kingman, S., Saeid, A., Robinson, J., Dimitrakis, G., Alnawafleh, H., 2009.
Dielectric properties of Jordanian oil shales. Fuel Process. Technol. 90, 1259–1264.
doi:10.1016/j.fuproc.2009.06.012
Allen, S.J., Gan, Q., Matthews, R., Johnson, P.A., 2003. Comparison of optimised isotherm
models for basic dye adsorption by kudzu. Bioresour. Technol. 88, 143–152.
Anderson, D.M., Nobakht, M., Moghadam, S., Mattar, L., 2010. SPE 131787 Analysis of
Production Data from Fractured Shale Gas Wells 23–25.
Aronofsky, J.S., Jenkins, R., 1954. SIMPLIFIED ANALYSIS of UNSTEADY RADIAL GAS FLOW.
Arri, L., Yee, D., Morgan, W., Jeansonne, M., 1992. Modeling coaled methane production
with binary gas sorption. SPE Rocky Mt. Reg. Meet. SPE 24363. doi:10.2523/24363-MS
Arri, L.E., Yee, D., Morgan, W.D., Jeansonne, M.W., 1992. Modeling coaled methane
production with binary gas sorption. SPE Rocky Mt. Reg. Meet. SPE 24363.
doi:10.2523/24363-MS
Barenblatt, G.I., Zheltov, I.P., Kochina, I.N., 1960. Basic concepts in the theory of seepage of
homogeneous liquids in fissured rocks [strata]. J. Appl. Math. Mech. 24, 1286–1303.
Bazan, R.E., Staudt, R., Papp, H., Azevedo, D.C.S., Cavalcante, C.L., 2008. Adsorption
Equilibria of Natural Gas Components on Activated Carbon : Pure and Mixed Gas
Isotherms 2, 323–332.
Bell, J.G., Rakop, C. karen, 1986. Hysteresis of Methane/Coal Sorption Isotherms. SPE 15454.
Bhatia, S.K., Ding, L.P., 2001. Vacancy Solution Theory of Adsorption Revisited 47, 9–11.
Bientinesi, M., Petarca, L., Cerutti, A., Bandinelli, M., De Simoni, M., Manotti, M., Maddinelli,
G., 2013. A radiofrequency/microwave heating method for thermal heavy oil recovery
213
 based on a novel tight-shell conceptual design. J. Pet. Sci. Eng. 107, 18–30.
doi:10.1016/j.petrol.2013.02.014
Busch, A., Amann-Hildenbrand, A., 2013. Predicting capillarity of mudrocks. Mar. Pet. Geol.
45, 208–223. doi:10.1016/j.marpetgeo.2013.05.005
Bustin, A.M.M., Bustin, R.M., Cui, X., 2008. Importance of Fabric on the Production of Gas
Shales. SPE Unconv. Reserv. Conf. doi:10.2118/114167-MS
Campañone, L.A., Zaritzky, N.E., 2005. Mathematical analysis of microwave heating process.
J. Food Eng. 69, 359–368. doi:10.1016/j.jfoodeng.2004.08.027
Canel, C.., Rosbaco, J., 1992. Compositional Material Balance: Its Application to the
Development of an Oil and Gas Field with Retrograde Condensation 117–128.
Chaiyo, K., Rattanadecho, P., 2013. Numerical analysis of heat-mass transport and pressure
buildup of unsaturated porous medium in a rectangular waveguide subjected to a
combined microwave and vacuum system. Int. J. Heat Mass Transf. 65, 826–844.
doi:10.1016/j.ijheatmasstransfer.2013.06.066
Chalmers, G.R.L., Bustin, R.M., 2008. Low Cretaceous gas shales in northeastern British
Columbia, Part II: Evaluation of regional potential gas resources. Bull. Can. Pet. Geol.
56, 22–61. doi:10.2113/gscpgbull.56.1.22
Chapiro, G., Bruining, J., 2015. Combustion enhance recovery of shale gas. J. Pet. Sci. Eng.
127, 179–189. doi:10.1016/j.petrol.2015.01.036
Chareonsuppanimit, P., Mohammad, S.A., Robinson, R.L., Gasem, K.A.M., 2012a. High-
pressure adsorption of gases on shales: Measurements and modeling. Int. J. Coal Geol.
95, 34–46. doi:10.1016/j.coal.2012.02.005
Chareonsuppanimit, P., Mohammad, S.A., Robinson Jr, R.L., Gasem, K.A.M., 2012b. High-
pressure adsorption of gases on shales: Measurements and modeling. Int. J. Coal Geol.
95, 34–46.
Charoensuppanimit, P., Mohammad, S.A., Gasem, K.A.M., 2016. Measurements and
Modeling of Gas Adsorption on Shales. Energy {&} Fuels 30, 2309–2319.
doi:10.1021/acs.energyfuels.5b02751
Charoensuppanimit, P., Mohammad, S.A., Robinson, R.L., Gasem, K.A.M., 2015. Modeling
the temperature dependence of supercritical gas adsorption on activated carbons,
coals and shales. Int. J. Coal Geol. 138, 113–126. doi:10.1016/j.coal.2014.12.008
Chekhonin, E., Parshin, A., Pissarenko, D., Popov, Y., Romushkevich, R., Safonov, S.,
Spasennykh, M., Chertenkov, M. V., Stenin, V.P., 2012. When Rocks Get Hot: Thermal
Properties of Reservoir Rocks. Oilf. Rev. Autumn 3, 20–37.
Chen, J., Loo, L.S., Wang, K., 2011. An Ideal Absorbed Solution Theory ( IAST ) Study of
Adsorption Equilibria of Binary Mixtures of Methane and Ethane on a Templated
Carbon 1209–1212.
Chen, L., Jiang, Z., Liu, K., Ji, W., Wang, P., Gao, F., Hu, T., 2017. Application of Langmuir and
Dubinin–Radushkevich models to estimate methane sorption capacity on two shale
214
 samples from the Upper Triassic Chang 7 Member in the southeastern Ordos Basin,
China. Energy Explor. Exploit. 35, 122–144. doi:10.1177/0144598716684309
Chen, Q., Tian, Y., Li, P., Yan, C., Pang, Y., Zheng, L., Deng, H., Zhou, W., Meng, X., 2017.
Study on Shale Adsorption Equation Based on Monolayer Adsorption , Multilayer
Adsorption , and Capillary Condensation 2017.
Chen, S., Jin, L., Chen, X., 2011. The effect and prediction of temperature on adsorption
capability of coal/CH 4. Procedia Eng. 26, 126–131. doi:10.1016/j.proeng.2011.11.2149
Cipolla, C., Lolon, E., Erdle, J., Rubin, B., 2010. Reservoir Modeling in Shale-Gas Reservoirs.
SPE Reserv. Eval. Eng. 13, 23–25. doi:10.2118/125530-PA
Clarkson, C.R., 2003. Application of a New Multicomponent as Adsorption Model to Coal Gas
Adsorption Systems. SPE J. 8, 236–251. doi:10.2118/78146-PA
Clarkson, C.R., Bustin, R.M., 2000. Binary gas adsorption r desorption isotherms : effect of
moisture and coal composition upon carbon dioxide selectivity over methane. Int. J.
Coal Geol. Bustinr Int. J. Coal Geol. 42, 241–271.
Clarkson, C.R., Bustin, R.M., Levy, J.H., 1997. Application of the mono/multilayer and
adsorption potential theories to coal methane adsorption isotherms at elevated
temperature and pressure. Carbon N. Y. 35, 1689–1705. doi:10.1016/S0008-
6223(97)00124-3
Clarkson, C.R., Haghshenas, B., 2013. Modeling of Supercritical Fluid Adsorption on Organic-
Rich Shales and Coal. SPE Unconv. Resour. Conf. 1–24. doi:10.2118/164532-MS
Clarkson, C.R., Haghshenas, B., others, 2013. Modeling of supercritical fluid adsorption on
organic-rich shales and coal, in: SPE Unconventional Resources Conference-USA.
Cochran, T.W., Danner, R.P., 1985. Vacancy Solution Theory of Adsorption Using Flory-
Huggins Activity Coefficient Equations 31, 268–277.
Curtis, J.B., 2002. Fractured shale-gas systems. Am. Assoc. Pet. Geol. Bull. 86, 1921–1938.
doi:10.1306/61EEDDBE-173E-11D7-8645000102C1865D
Danner, R.P., Wenzel, L.A., 1969. Adsorption of carbon monoxide-nitrogen, carbon
monoxide-oxygen, and oxygen-nitrogen mixtures on synthetic zeolites. AIChE J. 15,
515–520. doi:10.1002/aic.690150410
de Swaan O., A., 1976. Analytic Solutions for Determining Naturally Fractured Reservoir
Properties by Well Testing. doi:10.2118/5346-PA
Denny, D., 2007. Cleaning Up Water Blocking in Gas Reservoirs by Microwave Heating 0–2.
Do, D.D., 1998. Adsorption Analysis : Equilibria and Kinetics.
Do, D.D., Do, H.D., 1997. A new adsorption isotherm for heterogeneous adsorbent based on
the isosteric heat as a function of loading. Chem. Eng. Sci. 52, 297–310.
doi:10.1016/S0009-2509(96)00418-6
Eskandari, S., Jaghargh, E.M., Rasaei, M.R., 2014. 2D numerical simulation and analysis of
microwave heating in heavy oil reservoirs. Energy Sources, Part A Recover. Util.
215
 Environ. Eff. 36, 1961–1971. doi:10.1080/15567036.2011.557702
Farag, S., Sobhy, A., Akyel, C., Doucet, J., Chaouki, J., 2012. Temperature profile prediction
within selected materials heated by microwaves at 2.45GHz. Appl. Therm. Eng. 36,
360–369. doi:10.1016/j.applthermaleng.2011.10.049
Fianu, Gholinezhad, J., Hassan, M., 2018. Comparison of Temperature Dependent Gas
Adsorption Models and their Application to Shale Gas Reservoirs. Energy & Fuels
acs.energyfuels.8b00017. doi:10.1021/acs.energyfuels.8b00017
Firanda, E., 2011. The Development of Material Balance Equations for Coalbed Methane
Reservoirs.pdf 1–14.
Fitzgerald, J.E., Robinson, R.L., Gasem, K.A.M., 2006. Modeling high-pressure adsorption of
gas mixtures on activated carbon and coal using a simplified local-density model.
Langmuir 22, 9610–9618. doi:10.1021/la060898r
Foo, K.. ., Hameed, B.H., 2009. Insights into the modeling of adsorption isotherm systems.
Chem. Eng. J. 156, 2–10. doi:10.1016/j.cej.2009.09.013
Frantz, J., Sawyer, W., MacDonald, R., Williamson, J., Johnston, D., Waters, G., 2005.
Evaluating Barnett Shale Production Performance Using an Integrated Approach. Proc.
SPE Annu. Tech. Conf. Exhib. 1–18. doi:10.2523/96917-MS
Gasparik, M., Gensterblum, Y., Ghanizadeh, A., Weniger, P., Krooss, B.M., 2015. High-
Pressure/High-Temperature Methane-Sorption Measurements on Carbonaceous Shales
by the Manometric Method: Experimental and Data-Evaluation Considerations for
Improved Accuracy. SPE J. 20, 790–809. doi:10.2118/174543-PA
Ge, L., Zhang, Y., Wang, Z., Zhou, J., Cen, K., 2013. Effects of microwave irradiation
treatment on physicochemical characteristics of Chinese low-rank coals. Energy
Convers. Manag. 71, 84–91. doi:10.1016/j.enconman.2013.03.021
Gil, A., Grange, P., 1996. Application of the Dubinin-Radushkevich and Dubinin-Astakhov
equations in the characterization of microporous solids.
Guo, C., Wei, M., Liu, H., 2015. Modeling of gas production from shale reservoirs considering
multiple transport mechanisms. PLoS One 10, 1–24. doi:10.1371/journal.pone.0143649
Guo, S., 2013. Experimental study on isothermal adsorption of methane gas on three shale
samples from Upper Paleozoic strata of the Ordos Basin. J. Pet. Sci. Eng. 110, 132–138.
doi:10.1016/j.petrol.2013.08.048
Guo, W., Hu, Z., Zhang, X., Yu, R., Wang, L., 2017. Shale gas adsorption and desorption
characteristics and its effects on shale permeability. Energy Explor. Exploit. 35, 463–
481. doi:10.1177/0144598716684306
Guo, W., Xiong, W., Gao, S., Hu, Z., Liu, H., YU, R., 2013. Impact of temperature on the
isothermal adsorption/desorption of shale gas. Pet. Explor. Dev. 40, 514–519.
doi:10.1016/S1876-3804(13)60066-X
Hall, F.E., Zhou, C., Gasem, K.A.M., Robinson, R.L., Yee, D., 1994. Adsorption of Pure
Methane, Nitrogen, and Carbon Dioxide and Their Binary Mixtures on Wet Fruitland
216
 Coal. SPE 29194.
Hartman, R.C., Labs, W., Ambrose, R.J., Holding, R., 2011. Shale Gas-in-Place Calculations
Part II ─ Multi-component Gas Adsorption Effects. Soc. Pet. Eng.
Heller, R., Zoback, M., 2014. Journal of Unconventional Oil and Gas Resources Adsorption of
methane and carbon dioxide on gas shale and pure mineral samples. J. Unconv. OIL
GAS Resour. 8, 14–24. doi:10.1016/j.juogr.2014.06.001
Helminen, J., Helenius, J., Paatero, E., Turunen, I., 2000a. Comparison of sorbents and
isotherm models for NH3-gas separation by adsorption. AIChE J. 46, 1541–1555.
doi:10.1002/aic.690460807
Helminen, J., Helenius, J., Paatero, E., Turunen, I., 2000b. Comparison of sorbents and
isotherm models for NH3-gas separation by adsorption. AIChE J. 46, 1541–1555.
doi:10.1002/aic.690460807
Hiebert, A.D., Vermeulen, F.E., Chute, F.S., Capjack, C.E., 1986. Numerical Simulation Results
for the Electrical Heating of Athabasca Oil-Sand Formations. SPE Reserv. Eng. 1, 76–84.
doi:10.2118/13013-PA
Ho, Y.S., 2004. Selection of optimum sorption isotherm. Carbon N. Y. 42, 2115–2116.
doi:10.1016/j.carbon.2004.03.019
Ho, Y.S., Porter, J.F., Mckay, G., 2002. DIVALENT METAL IONS ONTO PEAT : COPPER , NICKEL
AND LEAD SINGLE COMPONENT SYSTEMS 1–33.
Hossan, M.R., Dutta, P., 2011. Analytical solution for temperature distribution in microwave
heating of rectangular objects, in: ASME 2011 International Mechanical Engineering
Congress and Exposition. pp. 623–628.
Ho, Y. S., & McKay, G. (1998). A comparison of chemisorption kinetic models applied to
pollutant removal on various sorbents. Process safety and environmental
protection, 76(4), 332-340.
Hu, H., 2014. Methane adsorption comparison of different thermal maturity kerogens in
shale gas system. Chinese J. Geochemistry 33, 425–430. doi:10.1007/s11631-014-0708-
9
Hwang, K.S., Dae Ki Choi, Sung Yong Gong, Sung Yong Cho, 1997. Adsorption and thermal
regeneration of methylene chloride vapor on an activated carbon bed. Chem. Eng. Sci.
52, 1111–1123. doi:10.1016/S0009-2509(96)00470-8
Ibrahim, M., U, S.C., Wattenbarger, R.A., Texas, A., 2006. SPE 100836 Analysis of Rate
Dependence in Transient Linear Flow in Tight Gas Wells.
Ibrahim, M., Wattenbarger, R.A., 2003. Determination of OGIP for Tight Gas Wells — New
Methods 1–14.
Jamaluddin, A.K.M., Bennion, D.B., Thomas, F.B., Ma, T.Y., 2000. Application of heat
treatment to enhance permeability in tight gas reservoirs. J. Can. Pet. Technol. 39, 19–
24.

217
Jamaluddin, A.K.M., Vandamme, M., Mann, B.K., 1995. Formation Heat Treatment (FHT): A
State-of-the-art Technology For Near-wellbore Formation Damage Treatment.
doi:10.2118/95-67
Javadpour, F., Fisher, D., & Unsworth, M. (2007). Nanoscale gas flow in shale gas
sediments. Journal of Canadian Petroleum Technology, 46(10).
Jensen, D., Smith, L.K., 1997. A practical approach to coalbed methane reserve prediction
using a modified material balance technique, in: Proceedings, International Coalbed
Methane Symposium, Tusaloosa, Alabama, USA.
Ji, W., Song, Y., Jiang, Z., Wang, X., Bai, Y., Xing, J., 2014. Geological controls and estimation
algorithms of lacustrine shale gas adsorption capacity: A case study of the Triassic
strata in the southeastern Ordos Basin, China. Int. J. Coal Geol. 134–135, 61–73.
doi:10.1016/j.coal.2014.09.005
Josh, M., Esteban, L., Delle Piane, C., Sarout, J., Dewhurst, D.N., Clennell, M.B., 2012.
Laboratory characterisation of shale properties. J. Pet. Sci. Eng. 88–89, 107–124.
doi:10.1016/j.petrol.2012.01.023
Kamal, M., Nikhil, G., Patwardhan, S.D., Spe, A.F., 2011. Gas Well Deliquification Using
Microwave Heating 27–29.
Kapoor, A., Ritter, J.A., Yang, R.T., 1990. An Extended Langmuir Model for Gas Mixtures on
Heterogeneous Surfaces 660–664.
Kapoor, A., Ritter, J.A., Yang, R.T., 1989. On the Dubinin-Radushkevich Equation for
Adsorption in Microporous Solids in the Henry’s Law Region 1118–1121.
Kapoor, A., Yang, R.T., 1989. Correlation of equilibrium adsorption data of condensible
vapours on porous adsorbents. Gas Sep. Purif. 3, 187–192.
doi:http://dx.doi.org/10.1016/0950-4214(89)80004-0
Kasevich, R.S., Price, S.L., Faust, D.L., Fontaine, M.F., 1994. Pilot testing of a radio frequency
heating system for enhanced oil recovery from diatomaceous earth. SPE 69th Annu.
Tech. Conf. Exhib. 105–119.
Kast, W., Hohenthanner, C.., 2000. Mass transfer within the gas-phase of porous media. Int.
J. Heat Mass Transf. 43, 807–823. doi:10.1016/S0017-9310(99)00158-1
Kaul, B.K., 1984. Correlation and Prediction of Adsorption Isotherm Data for Pure and Mixed
Gases 711–716.
Kazemi, H., 1969. Pressure Transient Analysis of Naturally Fractured Reservoirs with Uniform
Fracture Distribution. doi:10.2118/2156-A
Kazemi, H., Jr, L.S.M., Porterfield, K.L., Zeman, P.R., 1976. Numerical Simulation of Water-Oil
Flow in Naturally Fractured.
King, G.R., 1993. Material-Balance Techniques for Coal-Seam and Devonian Shale Gas
Reservoirs With Limited Water Influx. SPE Reserv. Eng. 8, 67–72. doi:10.2118/20730-PA
King, G.R., 1990. Material Balance Techniques for Coal Seam and Devonian Shale Gas

218
 Reservoirs Spe 20730, 181–192.
Klayborworn, S., Pakdee, W., Rattanadecho, P., Vongpradubchai, S., 2013. Effects of material
properties on heating processes in two-layered porous media subjected to microwave
energy. Int. J. Heat Mass Transf. 61, 397–408.
doi:10.1016/j.ijheatmasstransfer.2013.02.020
Klinkenberg, L.J., 1941. THE PERMEABILITY OF POROUS MEDIA TO LIQUIDS AND GASES
"f.
Koresh, J., 1982. Study of molecular sieve carbons: The Langmuir modelin ultramicroporous
adsorbents. J. Colloid Interface Sci. 88, 398–406. doi:10.1016/0021-9797(82)90268-5
Kuila, U., Prasad, M., 2013. Specific surface area and pore-size distribution in clays and
shales. Geophys. Prospect. 61, 341–362. doi:10.1111/1365-2478.12028
Kumar, K.V., 2007. Optimum sorption isotherm by linear and non-linear methods for
malachite green onto lemon peel. Dye. Pigment. 74, 595–597.
doi:10.1016/j.dyepig.2006.03.026
Kumar, K.V., 2006. Comparative analysis of linear and non-linear method of estimating the
sorption isotherm parameters for malachite green onto activated carbon. J. Hazard.
Mater. 136, 197–202. doi:10.1016/j.jhazmat.2005.09.018
Kumar, K.V., Porkodi, K., 2007. Mass transfer, kinetics and equilibrium studies for the
biosorption of methylene blue using Paspalum notatum. J. Hazard. Mater. 146, 214–
226.
Kumar, K.V., Porkodi, K., 2006. Relation between some two- and three-parameter isotherm
models for the sorption of methylene blue onto lemon peel. J. Hazard. Mater. 138,
633–635. doi:10.1016/j.jhazmat.2006.06.078
Kumar, K.V., Porkodi, K., Rocha, F., 2008. Isotherms and thermodynamics by linear and non-
linear regression analysis for the sorption of methylene blue onto activated carbon:
Comparison of various error functions. J. Hazard. Mater. 151, 794–804.
doi:10.1016/j.jhazmat.2007.06.056
Lampe, V., 2013. Modelling Fluid Flow and Heat Transport in Fractured Porous Media.
Master thesis 113.
Lane, H.S., Lancaster, D.E., Watson, A.T. ed, 1991. Characterizing the Role of Desorption in
Gas Production from Devonian Shales. Energy Sources 13, 337–359.
doi:10.1080/00908319108908993
Lane, H.S., Watson, A.T., Lancaster, D.E., 1989. Identifying and Estimating Desorption From
Devonian Shale Gas Production Data. SPE Annu. Tech. Conf. Exhib. doi:19794-MS
Lee, A., Gonzalez, M., Eakin, B., 1966. The viscosity of natural gases. J. Pet. Technol. 18, 997–
1000. doi:http://dx.doi.org/10.2118/1340-PA
Li, G., Meng, Y.-F., Tang, H., -, 2006. Clean Up Water Blocking in Gas Reservoirs by
Microwave Heating : Laboratory Studies. SPE J. 1–8.

219
Li, H., Lin, B., Yang, W., Hong, Y., Wang, Z., 2017. A fully coupled electromagnetic-thermal-
mechanical model for coalbed methane extraction with microwave heating. J. Nat. Gas
Sci. Eng. 46, 830–844. doi:10.1016/j.jngse.2017.08.031
Lin, Y., Wang, H., He, S., Nikolaou, M., 2015. Increasing Shale Gas Recovery through Thermal
Stimulation : Analysis and an Experimental Study.
Lu, X.-C., Li, F.-C., Watson, T.A., 1995. Adsorption Studies of Natural Gas Storage in Devonian
Shales.
Luffel, D., Guidry, K., Curtis, J., 1996. Development of Laboratory and Petrophysical
Techniques for Evaluating Shale Reservoirs, GRI-95/049. ed. Gas Reseaarch Institute.
Ma, D., Zhang, J., Bai, J., Zhang, H., 2014. Thermodynamic Characteristics of Adsorption-
Desorption of Methane in 3 # Coal Seam of Sihe 782–794.
Malek, A., Farooq, S., 1996. Comparison of isotherm models for hydrocarbon adsorption on
activated carbon. AIChE J. 42, 3191–3201. doi:10.1002/aic.690421120
Manik, J., Ertekin, T., Kohler, T.E., 2002. Development and validation of a compositional
coalbed simulator. J. Can. Pet. Technol. 41, 39–45. doi:10.2118/02-04-03
Markham, E.C., Benton, A.F., 1931. The Adsorption of Gas Mixtures by Silica 469, 7–8.
Mavor, M.J., Owen, L.B., Pratt, T.J., 1990. Measurement and Evaluation of Coal Sorption
Isotherm Data. SPE 20728.
Mazumder, S., 2016. Numerical methods for partial differential equations : finite difference
and finite volume methods.
Mengal, S.A., Wattenbarger, R.A., 2013. Accounting For Adsorbed Gas in Shale Gas
Reservoirs. SPE Conf. 25–28. doi:10.2118/141085-MS
Mengal, S.A., Wattenbarger, R.A., 2011. Accounting For Adsorbed Gas in Shale Gas
Reservoirs. SPE Conf. 25–28. doi:10.2118/141085-MS
Moghadam, S., Jeje, O., Mattar, L., 2011. Advanced gas material balance in simplified
format. J. Can. Pet. Technol. 50, 90–97. doi:10.2118/139428-PA
Muhammad, R.A., 2017. Numerical Simulation of heavy oil and bitumen recovery and
upgrading techniques. University of Nottingham.
Mutyala, S., Fairbridge, C., Paré, J.R.J., Bélanger, J.M.R., Ng, S., Hawkins, R., 2010. Microwave
applications to oil sands and petroleum: A review. Fuel Process. Technol. 91, 127–135.
doi:10.1016/j.fuproc.2009.09.009
Najurieta, H.L., 1980. A Theory for Pressure Transient Analysis in Naturally Fractured
Reservoirs. doi:10.2118/6017-PA
O’Brien, J., Myers, A., 1985. Rapid calculations of multicomponent adsorption equilibria
from pure isotherm data. Ind. Eng. Chem. … 1188–1191. doi:10.1021/i200031a049
Odeh, A.S., 1965. Unsteady-State Behavior of Naturally Fractured Reservoirs.
doi:10.2118/966-PA
220
Ovalles, C., Vaca, P., Okoniewski, M., Dieckmann, G., Pasalic, D., Dunlavey, J., 2018.
Numerical simulation of dielectric heating in a heavy oil reservoir using a shaped dipole
antenna. Soc. Pet. Eng. - SPE Canada Heavy Oil Tech. Conf. CHOC 2018 2018–Janua.
Pan, Z., Connell, L.D., 2007. A theoretical model for gas adsorption-induced coal swelling.
Int. J. Coal Geol. 69, 243–252. doi:10.1016/j.coal.2006.04.006
Pashin, J.C., McIntyre, M.R., 2003. Temperature-pressure conditions in coalbed methane
reservoirs of the Black Warrior basin: Implications for carbon sequestration and
enhanced coalbed methane recovery. Int. J. Coal Geol. 54, 167–183.
doi:10.1016/S0166-5162(03)00034-X
Peng, Z., Hwang, J.-Y., Park, C.-L., Kim, B.-G., Onyedika, G., 2012. Numerical analysis of heat
transfer characteristics in microwave heating of magnetic dielectrics. Metall. Mater.
Trans. A Phys. Metall. Mater. Sci. 43. doi:10.1007/s11661-011-1014-3
Peng, Z., Hwang, J., Mouris, J., Hutcheon, R., Huang, X., 2010. Microwave penetration depth
in materials with non-zero magnetic susceptibility. ISIJ Int. 50, 1590–1596.
doi:10.2355/isijinternational.50.1590
Pincombe, A.H., Smyth, N.F., 1991. Microwave heating of materials with low conductivity.
Proc. R. Soc. Lond. A 433, 479–498.
Pitchai, K., Chen, J., Birla, S., Jones, D., Subbiah, J., 2016. Modeling microwave heating of
frozen mashed potato in a domestic oven incorporating electromagnetic frequency
spectrum. J. Food Eng. 173, 124–131. doi:10.1016/j.jfoodeng.2015.11.002
Pizarro, J.O.S., Trevisan, O.V., 1990. Electric heating of oil reservoirs. Numerical simulation
and field test results. J. Pet. Technol. 1320–1326.
Porter, J.F., McKay, G., Choy, K.H., 1999. The prediction of sorption from a binary mixture of
acidic dyes using single- and mixed-isotherm variants of the ideal adsorbed solute
theory. Chem. Eng. Sci. 54, 5863–5885. doi:10.1016/S0009-2509(99)00178-5
Rajput, H.V., 2016. Development of Compositional Simulator for Liquid-Rich Shale
Reseervoirs. The Pennsylvania State Uninversity.
Rathi, V., Shrivastava, P.K., Pokhariya, H.S., 2012. 1-D Implementation of Maxwell ’ s
Equations in MATLAB to Study the Effect of Absorption Using PML. Int. J. Electron. Syst.
64–69.
Raut, U., Famá, M., Teolis, B.D., Baragiola, R.A., 2007. Characterization of porosity in vapor-
deposited amorphous solid water from methane adsorption. J. Chem. Phys. 127, 1–6.
doi:10.1063/1.2796166
Reda, D.C., 1987. Slip-flow experiments in welded tuff, in: Coupled Processes Associated
with Nuclear Waste Repositories.
Ren, W., Tian, S., Li, G., Sheng, M., Yang, R., 2017. Modeling of mixed-gas adsorption on
shale using hPC-SAFT-MPTA. Fuel 210, 535–544. doi:10.1016/j.fuel.2017.09.012
Rexer, T.F.T., Benham, M.J., Aplin, A.C., Thomas, K.M., 2013. Methane Adsorption on Shale
under Simulated Geological Temperature and Pressure Conditions\nThomas F. T.
221
 Rexer,† Michael J. Benham,‡ Andrew C. Aplin,§ and K. Mark Thomas. Energy & Fuels
27, 3099−3109.
Richter, E., Schutz, W., Myers, A.L., 1989. Effects of Adsorption Equation on Prediction of
Multicomponent Adsorption Equilibria By The Ideal Adsorbed Solution Theory. Chem.
Eng. Sci. 44, 1609–1616.
Ross, D.J.K., Bustin, R.M., 2007. Shale gas potential of the Lower Jurassic Gordondale
member, northeastern British Columbia, Canada. Bull. Can. Pet. Geol. 55, 51–75.
doi:10.2113/gscpgbull.55.1.51
Ross, D.J.K., Marc Bustin, R., 2009. The importance of shale composition and pore structure
upon gas storage potential of shale gas reservoirs. Mar. Pet. Geol. 26, 916–927.
doi:10.1016/j.marpetgeo.2008.06.004
Ruthven, D.M., 1984. Principles of Adsorption and Adsorption Processes.
Rybakov, K.I., Olevsky, E.A., Krikun, E. V., 2013. Microwave sintering: Fundamentals and
modeling. J. Am. Ceram. Soc. 96, 1003–1020. doi:10.1111/jace.12278
Sahni, A., Kumar, M., Knapp, R.B., 2000. Electromagnetic Heating Methods for Heavy Oil
Reservoirs. SPE/AAPG West. Reg. Meet. 2000 1–10. doi:10.2118/62550-MS
Sakhaee-Pour, A., Bryant, S.L., 2012. Gas Permeability of Shale.
Sakurovs, R., Day, S., Weir, S., Duffy, G., 2007. Application of a modified Dubinin-
Radushkevich equation to adsorption of gases by coals under supercritical conditions.
Energy & fuels 21, 992–997. doi:10.1021/ef0600614
Salmachi, A., Haghighi, M., 2012. Feasibility Study of Thermally Enhanced Gas Recovery of
Coal Seam Gas Reservoirs Using Geothermal Resources. Energy & Fuels 26, 5048–5059.
doi:10.1021/ef300598e
Sandoval, D.R., Yan, W., Michelsen, M.L., Stenby, E.H., 2017. Modeling of Shale Gas
Adsorption and its Influence on Phase Equilibrium. Ind. Eng. Chem. Res.
acs.iecr.7b04144. doi:10.1021/acs.iecr.7b04144
Sanmiguel, J.E., Mallory, D.G., Mehta, S.A., Moore, R.G., 2002. Formation heat treatment
process by combustion of gases around the wellbore. J. Can. Pet. Technol. 41, 70–76.
doi:10.2118/02-08-05
Seidle, J.P., 1999. A Modified p/Z Method for Coal Wells. SPE SPE 55605.
Seidle, J.P., Arri, L.E., 1990. Use Of Conventional Reservoir Models For Coalbed Methane
Simulation. CIM/SPE Int. Tech. Meet. doi:10.2118/21599-MS
Serra, K., Reynolds, A.C., Raghavan, R., 1983. New Pressure Transient Analysis Methods for
Naturally Fractured Reservoirs. doi:10.2118/10780-PA
Sierra, R., Tripathy, B., Bridges, J.E., Farouq Ali, S.M., 2001. Promising Progress in Field
Application of Reservoir Electrical Heating Methods. SPE Int. Therm. Oper. Heavy Oil
Symp. 1–17. doi:10.2118/69709-MS
Sing, K.S., Everett, D.H., Haul, R.A., Mouscou, L., Piorotti, R., Rouquerol, J., Siemieniewska,
222
 T., 1985. Reporting Phyisiosorption Data for Gas/Solid Systems with special reference
to the determination of surface area and porosity. Int. UNION PURE Comm. COLLOID
Surf. Chem. Incl. Catal. * Report. PHYSISORPTION DATA GAS / SOLID Syst. with Spec.
Ref. to Determ. Surf. Area Porosity 57, 603–619.
Singh, V.K., 2013. Overview of Material Balance Equation (MBE) in Shale Gas & Non-
Conventional Reservoir. SPE Middle East Oil Gas Show Conf. doi:10.2118/164427-MS
Sreńscek-Nazzal, J., Narkiewicz, U., Morawski, A.W., Wróbel, R.J., Michalkiewicz, B., 2015.
Comparison of Optimized Isotherm Models and Error Functions for Carbon Dioxide
Adsorption on Activated Carbon. doi:10.1021/acs.jced.5b00294
Stevenson, M.D., Pinczewski, W. V, Somers, M.L., Bagio, S.E., 1991. Adsorption/Desorption
of Multicomponent Gas Mixtures at In-Seam Conditions. SPE 23026.
Sullivan, D.M., 2000. Electromagnetic Simulation Using the FDTD Method. IEEE Press Ser. RF
Microw. Technol. doi:10.1109/9780470544518
Sun, J., Wang, W., Yue, Q., 2016. Review on microwave-matter interaction fundamentals
and efficient microwave-associated heating strategies. Materials (Basel). 9.
doi:10.3390/ma9040231
Sun, S., Salama, A., & El Amin, M. F. (2012a). Matrix-oriented implementation for the
numerical solution of the partial differential equations governing flows and transport in
porous media. Computers & Fluids, 68, 38-46.

Sun, S., Salama, A., & El-Amin, M. F. (2012b). An equation-type approach for the numerical
solution of the partial differential equations governing transport phenomena in porous
media. Procedia Computer Science, 9, 661-669.

Suwanayuen, S., Danner, R.P., 1980. Vacancy Solution Theory of Adsorption from Gas
Mixtures. AIChE J. 26, 76–83.
Szepesy, L., Illes, V., 1963. Adsorption of gases and gas mixtures. III. Investigation of the
adsorption equilibria of binary gas mixtures. Acta Chim. Hung. Tomus 35, 245.
Talu, O., Myers, A.L., 1988. Rigorous Thermodynamic Treatment of Gas Adsorption. AIChE J.
34.
Talu, O., Myers, A.L., Gabitto, J.F., McCoy, B.J., Fitch, B., Buscall, R., White, L., Dixon, D.C.,
White, L.R., Landman, K.A., 1988. Letters to the editor. AIChE J. 34, 1931–1936.
doi:10.1002/aic.690341124
Tang, X., Ripepi, N., Luxbacher, K., Pitcher, E., 2017. Adsorption Models for Methane in
Shales : Review , Comparison , and Application 10787–10801.
doi:10.1021/acs.energyfuels.7b01948
Tien, C., 1994. Adsorption calculations and modeling. Butterworth-Heinemann.
Toroğlu, G., Sevgi, L., 2014. Matlab codes for First and Second Order EM Differential
Equations. IEEE Antennas Propag. Mag. 56.

223
Torres, F., Jecko, B., 1997. Complete FDTD analysis of microwave heating processes in
frequency-dependent and temperature-dependent media. IEEE Trans. Microw. Theory
Tech. 45, 108–117. doi:10.1109/22.552039
Wang, H., 2016. A Numerical Study of Thermal-Hydraulic-Mechanical (THM) Simulation with
the Application of Thermal Recovery in Fractured Shale Gas Reservoirs. SPE Reserv.
Eval. Eng. 0, 0–34. doi:10.3837/tiis.0000.00.000
Wang, H., Merry, H., Amorer, G., Kong, B., 2015. Enhance Hydraulic Fractured Coalbed
Methane Recovery by Thermal Stimulation. SPE/CSUR Unconv. Resour. Conf. c.
doi:10.2118/175927-MS
Wang, H., Rezaee, R., Saeedi, A., 2015. The Interaction of Reservoir Properties and
Microwave Heating – An Experimental and Numerical Modelling Study of Enhanced Gas
Recovery (EGR). Procedia Earth Planet. Sci. 15, 542–548.
doi:10.1016/j.proeps.2015.08.092
Wang, Y. dou, Wang, X. ying, Xing, Y. fan, Xue, J. kang, Wang, D. sheng, 2017. Three-
dimensional numerical simulation of enhancing shale gas desorption by electrical
heating with horizontal wells. J. Nat. Gas Sci. Eng. 38, 94–106.
doi:10.1016/j.jngse.2016.12.011
Wang, Y., Tsotsis, T.T., Jessen, K., 2015. Competitive Sorption of Methane/Ethane Mixtures
on Shale: Measurements and Modeling. Ind. Eng. Chem. Res. 54, 12187–12195.
doi:10.1021/acs.iecr.5b02850
Wang, Z., 2016. Analyzing the Adaption of Different Adsorption Models for Describing the
Shale Gas Adsorption Law m, 1921–1932. doi:10.1002/ceat.201500617
Warren, J.E.E., Root, P.J.J., 1963. The Behavior of Naturally Fractured Reservoirs. Soc. Pet.
Eng. J. 3, 245–255. doi:10.2118/426-PA
Wong, Y.C., Szeto, Y.S., Cheung, W.H., McKay, G., 2004. Adsorption of acid dyes on chitosan
- Equilibrium isotherm analyses. Process Biochem. 39, 693–702. doi:10.1016/S0032-
9592(03)00152-3
Wu, Y.-S., Pruess, K., Persoff, P., 1998. Gas Flow in Porous Media with Klinkenberg Effects.
Transp. Porous Media 32, 117–137. doi:10.1023/A:1006535211684
Wu, Y. S., Li, J., Ding, D., Wang, C., & Di, Y. (2014). A generalized framework model for the
simulation of gas production in unconventional gas reservoirs. Spe Journal, 19(05), 845-
857.
Yang, F., Xie, C., Xu, S., Ning, Z., Krooss, B.M., 2017. Supercritical Methane Sorption on
Organic-Rich Shales over a Wide Temperature Range.
doi:10.1021/acs.energyfuels.7b02628
Yang, R.T., 1997. Gas Separation by Adsorption Processes, Chemical Engineering. Imperial
College Press.
Ye, Z., Chen, D., Pan, Z., Zhang, G., Xia, Y., Ding, X., 2016. An improved Langmuir model for
evaluating methane adsorption capacity in shale under various pressures and

224
 temperatures. J. Nat. Gas Sci. Eng. 31, 658–680. doi:10.1016/j.jngse.2016.03.070
Yee, K., 1966. Numerical solution of initial boundary value problems involving Maxwell’s
equations in isotropic media. Antennas Propagation, IEEE Trans.
doi:10.1109/TAP.1966.1138693
Yu, W., Sepehrnoori, K., 2014. Simulation of gas desorption and geomechanics effects for
unconventional gas reservoirs. Fuel 116, 455–464. doi:10.1016/j.fuel.2013.08.032
Yu, W., Sepehrnoori, K., Patzek, T.W., 2015. Modeling Gas Adsorption in Marcellus Shale
With Langmuir and BET Isotherms. SPE J. doi:10.2118/170801-PA
Yu, W., Sepehrnoori, K., Patzek, T.W., 2014. SPE-170801-MS Evaluation of Gas Adsorption in
Marcellus Shale 27–29.
Yue, L., Wang, H., He, S., Nikolaou, M., 2015. Increasing Shale Gas Recovery through
Thermal Stimulation : Analysis and an Experimental Study.
Zhang, L., Li, D., Wang, L., Lu, D., 2015. Simulation of Gas Transport in Tight / Shale Gas
Reservoirs by a Multicomponent Model Based on PEBI Grid 2015.
Zhang, T., Ellis, G.S., Ruppel, S.C., Milliken, K., Yang, R., 2012. Effect of organic-matter type
and thermal maturity on methane adsorption in shale-gas systems. Org. Geochem. 47,
120–131. doi:10.1016/j.orggeochem.2012.03.012
Zhao, X., Huang, K., Yan, L., 2011. Review of Numerical Simulation of Microwave Heating
Process. INTECH Open Access Publ. 27–28. doi:10.5772/13387
Zhou, L., Puri, V.M., Anantheswaran, R.C., Yeh, G., 1995. Finite element modeling of heat
and mass transfer in food materials during microwave heating — Model development
and validation. J. Food Eng. 25, 509–529. doi:10.1016/0260-8774(94)00032-5
Zhu, W., Hrabanek, P., Gora, L., Kapteijn, F., Moulijn, J.A., 2006. Role of Adsorption in the
Permeation of CH 4 and CO 2 through a Silicalite-1 Membrane. Ind.Eng.Chem.Res.
doi:10.1021/ie0507427

225
Appendix 2

Table 2.8: Excess adsorption data of Methane and Nitrogen in Albany Shale
(Chareonsuppanimit et al., 2012)

Methane Nitrogen Carbon dioxide

Pressure Excess adsorption Pressure Excess Pressure Excess
(MPa) (mmol/g) (MPa) adsorption (MPa) adsorption
(mmol/g) (mmol/g)
1.45 0.0138 1.47 0.0012 1.7 0.0479
2.85 0.0253 2.86 0.0052 3.06 0.0715
4.23 0.0316 4.23 0.0083 4.76 0.0916
5.63 0.0352 5.62 0.0109 5.66 0.0985
6.99 0.0374 6.99 0.0116 6.96 0.1085
8.36 0.0386 8.37 0.0133 8.26 0.1136
9.76 0.0395 9.77 0.0145 9.75 0.1179
11.12 0.0397 11.14 0.0147 11.4 0.115
12.52 0.0412 12.56 0.0147 12.6 0.0942

Table 2.9: single component Adsorption data for Methane and Ethane on Activated Carbon
(Valenzuela and Myers 1989)

Ethane Methane
P(Kpa) n(mmol/g) P(Kpa) n(mmol/g)
0.56 0.2432 16.0253 0.1919
2.3465 0.547 27.3577 0.2972
2.5331 0.485 31.464 0.3061
4.5729 0.8049 34.6371 0.3654
4.7329 0.8018 46.5294 0.4556
7.706 1.0642 48.0626 0.464
9.5325 1.169 54.5687 0.5216
11.7723 1.3238 62.608 0.5725
15.1987 1.4979 75.6336 0.6662
17.2385 1.6063 83.4462 0.7103
24.3446 1.8785 83.6329 0.6876
28.4642 1.9802 94.0587 0.8089
32.0373 2.1167 99.8715 0.8143
37.0635 2.2399 112.8304 0.8785
43.9563 2.4108 124.8694 0.9424
50.0624 2.5402
54.2621 2.6169
60.4615 2.72
226
 68.3275 2.8458
72.5938 2.8967
85.0728 3.0658
85.2327 3.0497
101.9913 3.2393
113.257 3.3576

Table 2.10: Binary gas Mixture of Ethane and Methane on activated carbon at temperature
of 293.15K (Valenzuela and Myers 1989)

P(Kpa) y1 x1 n(mmol/g)
(Ethane) (Ethane)
101.191 0.066 0.605 1.3456
99.7249 0.083 0.658 1.4116
102.485 0.245 0.852 2.0652
99.7249 0.251 0.859 2.054
99.7249 0.489 0.941 2.6255
102.538 0.519 0.953 2.7692
99.7515 0.731 0.975 2.9887

Example calculation of Sum of normalised error

1. Select an isotherm model. Example langmuir isotherm

2. Select an error function to use . example is SSE
3. Using excel solver, the parameters for SSE are

VL 0.052580101
b 0.319189112

4. using these parameters, we can obtain the error values for the rest of the other error
functions
SSE 1.5951E-05
ARE 4.444640058
SAE 0.009828207
MPSD 15.11924434
Hybrid 0.011474011

5. Now find the parameters for the other error functions

SSE ARE SAE MPSD Hybrid

VL 0.052580101 0.051587 0.045547 0.052739 0.05
b 0.319189112 0.337697 0.610559 0.305701 0.4

227
 6. Complete a table using the parameters for a specific error function to compute the rest of
the other error functions

SSE ARE SAE MPSD Hybrid MAX

SSE 1.5951E-05 1.65E-05 7.39E-05 1.73E-05 2.48E-05 7.39E-05
ARE 4.444640058 4.367862 8.846628 4.590383 5.226215 8.846628
SAE 0.009828207 0.009114 0.014963 0.010677 0.009225 0.014963
MPSD 15.11924434 14.85807 30.0934 15.61501 17.77791 30.0934
Hybrid 0.011474011 0.012915 0.068296 0.010742 0.023323 0.068296

7. Find the maximum error value for each individual row

MAX
SSE 7.39E-05
ARE 8.846628
SAE 0.014963
MPSD 30.0934
Hybrid 0.068296

8. Compute the ratio of each column using the maximum values

SSE ARE SAE MPSD HYBRID

SSE 0.215958065 0.222738 1 0.23412 0.335855
ARE 0.502410632 0.493732 1 0.518885 0.590758
SAE 0.656839031 0.609083 1 0.713589 0.616516
MPSD 0.502410632 0.493732 1 0.518885 0.590758
Hybrid 0.168004102 0.189107 1 0.15728 0.341501

9. Add all the values from each column. This becomes the SNE for the individual error function

SSE ARE SAE MPSD HYBRID

SNE 2.045622464 2.008392 5 2.142759 2.475388

10. The minimum value becomes the optimal error function (Bold) to use in calculations

SSE ARE SAE MPSD HYBRID

SNE 2.045622464 2.008392 5 2.142759 2.475388

228
Table 2.11: Non-linear Langmuir isotherm parameters

SSE ARE SAE MPSD HYBRID

Methane
VL 0.05258 0.05158 0.04554 0.0527 0.0545
b 0.31918 0.33769 0.61055 0.3057 0.2811
CO2
VL 0.139438 0.167666 0.152714 0.167668 0.14103
b 0.38098 0.24297 0.31488 0.24297 0.35303
Nitrogen
VL 0.03127 0.0573 0.0295 0.0573 0.0628
b 0.0815 0.0346 0.0924 0.0346 0.0288

Table 2.12: Non-linear BET isotherm parameters

SSE ARE SAE MPSD HYBRID

Methane Po  46.18 Mpa
Vm 0.0351 0.0355 0.0346 0.0362 0.0363
C 30.3761 30.447 33.0086 28.768 24.386
CO2 Po  12.84 Mpa
Vm 0.002484 0.001766 0.001763 0.001766 0.002421
C 12.220 6.6377 13.408 6.6379 341.769
Nitrogen Po  104.05 Mpa
Vm 0.0223 0.0327 0.02221 0.0327 0.03459
C 12.32225 6.4517 13.184 6.4518 5.4748

Table 2.13: Non-linear Dubinin –Astakhov isotherm parameters

SSE ARE SAE MPSD HYBRID

Methane
Wo 0.071649 0.069431 0.06104 0.06943 0.0787
D 2.46547 2.517542 2.16262 2.51754 2.4096
m 0.935167 0.913509 0.60568 0.91350 1.0465
CO2
Wo 0.11932 0.13277 0.129899 0.1327 0.1198
D 2.16333 2.55759 2.78645 2.5415 1.96157
m 0.694976 1.10380 0.98100 1.09688 0.69851
Nitrogen
Wo 0.070201 0.205516 0.06991 0.205517 0.12384
D 1.624661 1.800114 2.138955 1.800114 1.44084
229
 m 1.118434 2.017611 1.42275 2.017614 1.331438

Table 2.14: Non-linear Vacancy Solution Model isotherm parameters

SSE ARE SAE MPSD HYBRID

Methane
n1 0.04687 0.046 0.04652 0.046469 0.04628
b1 0.0008 0.00609 0.0066 0.003909 0.0039
1v 0.0428 0.3676 0.3468 0.0901 0.08905
v1 1.359807 0.9440 1.0706 1.6863 1.7015
CO2
 0.59801 0.51319 0.50157 0.51319 0.43465
n1

b1 0.016474 0.028472 0.035681 0.02847 0.0357

1v 0.03407 0.01539 0.0086 0.01539 0.01168
v1 5.0910 5.4004 6.2924 5.4004 5.9118
Nitrogen
 0.020314 0.019184 0.019127 0.019264 0.01926
n1

b1 0.000157 0.000248 0.00025 0.000255 0.00025

1v 0.0324 0.0225 0.0230 0.0233 0.0233
v1 1.4750 1.7980 1.7845 1.8005 1.810

Table 2.15: Langmuir isotherm parameters and for methane and ethane adsorption

SSE ARE SAE MPSD HYBRID

Methane
VL 2.5881 2.5544 2.4963 2.4963 2.5215
b 0.0045 0.0046 0.0048 0.0048 0.0047
Ethane
SSE ARE SAE MPSD HYBRID
VL 3.83889 3.4501 3.8474 3.4500 3.5981
b 0.04272 0.0578 0.0415 0.0578 0.0525

230
Table 2.16: Vacancy Solution Model isotherm parameters for methane and ethane
adsorption

SSE ARE SAE MPSD HYBRID

Methane
n1 0.046875 0.046 0.0465 0.0464 0.0462
b1 0.000896 0.00609 0.0066 0.0039 0.00393
1v 0.042834 0.3676 0.3468 0.0901 0.0890
v1 1.3598 0.9440 1.0706 1.6863 1.7015
Ethane
SSE ARE SAE MPSD HYBRID
 5.3946 4.6175 4.5928 4.5721 4.5721
n1

b1 0.1071 0.2053 0.2171 0.2043 0.2043

1v 0.0274 0.0122 0.0124 0.0126 0.0126
v1 4.6475 5.325 5.3512 5.2764 5.2764

231
Appendix 3

Table 3.5: Best fitted parameters for the Langmuir 1 model on the studied samples

Sample Reference a1 m3/kg a2 K.m3/kg b0 1/Pa H KJ/mol R2

Green River Formation Zhang et al 2012 2.53E-02 4.28E-04 3.44E-04 15.61E+03 0.99

Woodford Shale Zhang et al 2012 3.44E-04 4.28E-04 8.10E-04 14.4E+03 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 4.43E-02 4.27E-04 8.99E-03 8.29E+03 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 5.69E-02 4.28E-04 2.92418E-05 23.7E+03 0.64

Ordos Basin YY33-2 Ji et al 2014 2.46E-03 1.09E+00 1.12799E-05 27.5E+03 0.94

Ordos Basin YY34-1 Ji et al 2014 5.27E-03 2.48E-03 6.26832E-10 53.1E+03 0.92

Ordos Basin A Guo 2014 2.08E-03 3.48E-04 3.01428E-07 39.7E+03 0.79

Ordos Basin B Guo 2014 4.93E-03 4.43E-04 2.78823E-06 30.8E+03 0.89

Ordos Basin C Guo 2014 4.93E-03 4.43E-04 3.26045E-06 30.8E+03 0.91

232
 CSW2 Lu et al 1995 1.69E-03 8.25E-04 2.66E-04 16.3E+03 0.98

Antrim-7 Lu et al 1995 4.56E-03 8.30E-04 3.22E-04 17.4E+03 0.99

Table 3.6: Best fitted parameters for Langmuir 2 model on the studied samples

Sample Reference qso m3/kg q1 K b0 1/Pa H KJ/mol R2

Green River Formation Zhang et al 2012 0.025 1.33 3.52E-04 15.5E+03 0.99

Woodford Shale Zhang et al 2012 0.026 1.30 7.14E-04 14.8E+03 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.044 0.96 9.01E-03 8.28E+03 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 0.056 1.20 2.97092E-05 23.7E+03 0.64

Ordos Basin YY33-2 Ji et al 2014 0.006 5.31 1.36743E-06 32.8E+03 0.94

Ordos Basin YY34-1 Ji et al 2014 0.006 5.31 8.38465E-07 32.8E+03 0.83

Ordos Basin A Guo 2014 0.002 0.22 3.61198E-07 38.9E+03 0.79

Ordos Basin B Guo 2014 0.004 0.25 4.26618E-06 30.0E+03 0.89

233
 Ordos Basin C Guo 2014 0.004 0.25 5.13922E-06 30.0E+03 0.94

CSW2 Lu et al 1995 0.001 1.50 3.29E-04 15.8E+03 0.98

Antrim-7 Lu et al 1995 0.004 1.52 3.67E-04 17.0E+03 0.99

Table 3.7: Best fitted parameters for the Langmuir-Freundlich 1 model on the studied samples

Sample Reference a1 m3/kg a2 K. m3/kg b0 1/Pa H KJ/mol n R2

Green River Formation Zhang et al 2012 0.019 7.32 2.93E-03 8.82E+03 0.77 0.98

Woodford Shale Zhang et al 2012 0.026 0.32 1.22E-03 13.2E+03 0.98 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.041 0.30 7.79E-03 8.70E+03 1.04 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 0.022 129.57 9.05E-04 9.36E+03 0.50 0.68

Ordos Basin YY33-2 Ji et al 2014 0.002 1.36 1.08495E-05 27.5E+03 0.73 0.97

Ordos Basin YY34-1 Ji et al 2014 0.002 3.91 2.86225E-05 22.2E+03 0.42 0.94

234
 Ordos Basin A Guo 2014 0.004 4.20 8.54037E-05 18.0E+03 0.28 0.94

Ordos Basin B Guo 2014 0.006 19.62 6.1835E-05 15.5E+03 0.47 0.95

Ordos Basin C Guo 2014 0.006 20.13 6.48377E-05 15.6E+03 0.42 0.91

CSW2 Lu et al 1995 0.006 24.48 8.41994E-05 9.20E+03 0.63 0.98

Antrim-7 Lu et al 1995 0.012 94.71 3.85E-04 5.59E+03 0.54 0.98

Table 3.8: Best fitted parameters for the Langmuir-Freundlich 2 model on the studied samples

Sample Reference qso m3/kg q1 K b0 1/Pa H KJ/mol n R2

Green River Formation Zhang et al 2012 0.0241 0.04 2.53E-04 16.5E+03 1.04 0.99

Woodford Shale Zhang et al 2012 0.0275 0.05 1.41E-03 12.9E+03 0.97 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.0422 0.04 5.66E-03 9.55E+03 1.06 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 3.8324 70.08 3.88872E-05 11.2E+03 0.48 0.69

Ordos Basin YY33-2 Ji et al 2014 0.0028 243.53 5.98157E-07 35.6E+03 0.85 0.97

235
 Ordos Basin YY34-1 Ji et al 2014 0.0028 243.54 4.07482E-07 35.6E+03 0.85 0.93

Ordos Basin A Guo 2014 0.0092 2.99 3.56432E-05 22.1E+03 0.30 0.94

Ordos Basin B Guo 2014 0.0345 3.29 3.61044E-05 18.7E+03 0.49 0.95

Ordos Basin C Guo 2014 0.0996 3.21 1.78E-04 11.8E+03 0.42 0.96

CSW2 Lu et al 1995 0.0049 2.57 1.38E-03 9.31E+03 0.71 0.97

Antrim-7 Lu et al 1995 0.0075 2.57 1.69E-03 11.7E+03 0.74 0.99

Table 3.9: Best fitted parameters for the Exponential model on the studied samples

Sample Reference Vs std m3/kg Dt 1/K A K1/2/Mpa B K R2

Green River Formation Zhang et al 2012 0.026 0.000279 0.54 16.76 0.99

Woodford Shale Zhang et al 2012 0.028 0.00020 0.64 24.44 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.044 0.00007 1.42 0.65 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 0.043 0.000002 1.42 36.56 0.69

236
 Ordos Basin YY33-2 Ji et al 2014 0.01 0.01100 2.51 7.61 0.98

Ordos Basin YY34-1 Ji et al 2014 0.009 0.0168 2.88 11.43 0.98

Ordos Basin A Guo 2014 0.005 0.0174 3.57 32.04 0.93

Ordos Basin B Guo 2014 0.007 0.00673 1.03 29.01 0.90

Ordos Basin C Guo 2014 0.005 0.00575 1.58 33.43 0.96

CSW2 Lu et al 1995 0.001 0.00181 0.907 3.67 0.98

Antrim-7 Lu et al 1995 0.005 0.00371 1.58 0.859 0.99

Table 3.10: Best fitted parameters for the Bi- Langmuir model on the studied samples

Sample Reference Nm std k1 1/Mpa k2 1/Mpa  E1 J/mol E2 J/mol f1 R2

m3/kg

Green River Formation Zhang et al 2012 0.0284 3.00E-04 2.06E+00 6.20E-07 3.04E-03 0.10 0.93

Woodford Shale Zhang et al 2012 0.0315 3.00E-04 2.92E+00 6.18E-07 3.04E-03 0.13 0.96

237
 Barnett Shale Tarrant A3 Zhang et al 2012 0.0511 3.00E-04 3.42E+00 6.31E-07 3.04E-03 0.12 0.98

Barnett Shale Blakeley 1 Zhang et al 2012 0.0413 1.39E-03 1.08E+01 2.29E+04 3.05E-03 0.77 0.99

Ordos Basin YY33-2 Ji et al 2014 0.0061 3.00E-04 9.09E+00 8.12E-07 3.04E-03 0.15 0.75

Ordos Basin YY34-1 Ji et al 2014 0.0058 2.75E-03 1.15E-05 2.44E+03 3.46E+04 0.07 0.89

Ordos Basin A Guo 2014 0.0023 2.65E-03 2.28E-07 2.37E+03 4.89E+04 0.12 0.81

Ordos Basin B Guo 2014 0.0047 2.65E-03 8.07E-07 2.37E+03 4.27E+04 0.10 0.84

Ordos Basin C Guo 2014 0.0046 3.00E-04 6.48E+00 8.21E-07 3.04E-03 0.16 0.78

CSW2 Lu et al 1995 0.0140 1.50E-03 4.17E-04 1.08E+04 2.62E+04 0.97 0.98

Antrim-7 Lu et al 1995 0.0249 4.58E-03 1.10E-03 8.65E+03 2.35E+04 0.90 0.99

Table 3.11: Best fitted parameters for the Langmuir 3 model on the studied samples

Sample Reference qso m3/kg b0 1/Pa H KJ/mol R2

Green River Formation Zhang et al 2012 0.025 1.20E-04 18.3E+03 0.98

238
Woodford Shale Zhang et al 2012 0.026 8.10E-04 14.4E+03 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.044 8.86E-03 8.33E+03 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 0.056 2.96E-05 2.37E+03 0.64

Ordos Basin YY33-2 Ji et al 2014 0.005 1.22E-07 40.0E+03 0.97

Ordos Basin YY34-1 Ji et al 2014 0.005 1.12E-09 52.0E+03 0.93

Ordos Basin A Guo 2014 0.002 5.81E-06 31.7E+03 0.75

Ordos Basin B Guo 2014 0.005 3.90E-05 23.2E+03 0.87

Ordos Basin C Guo 2014 0.004 3.94E-05 25.0E+03 0.95

CSW2 Lu et al 1995 0.001 3.20E-04 15.8E+03 0.98

Antrim-7 Lu et al 1995 0.004 3.22E-04 17.4E+03 0.99

239
Table 3.12: Best fitted parameters for the Langmuir-Freundlich 3 model on the studied samples

Sample Reference qso m3/kg b0 1/Pa H KJ/mol n R2

Green River Formation Zhang et al 2012 0.024 2.56E-04 16.4E+03 1.04 0.99

Woodford Shale Zhang et al 2012 0.025 5.69E-04 15.4E+03 1.04 0.99

Barnett Shale Tarrant A3 Zhang et al 2012 0.042 5.84E-03 9.48E+03 1.06 0.99

Barnett Shale Blakeley 1 Zhang et al 2012 3.051 4.82E-05 11.9E+03 0.48 0.68

Ordos Basin YY33-2 Ji et al 2014 0.006 1.25E-06 33.6E+03 0.83 0.98

Ordos Basin YY34-1 Ji et al 2014 0.006 8.41E-07 33.7E+03 0.83 0.93

Ordos Basin A Guo 2014 0.011 3.78E-05 21.5E+03 0.29 0.94

Ordos Basin B Guo 2014 0.035 3.51E-05 18.7E+03 0.49 0.95

Ordos Basin C Guo 2014 0.103 1.69E-04 11.8E+03 0.42 0.96

CSW2 Lu et al 1995 0.007 6.50E-04 10.0E+03 0.68 0.98

Antrim-7 Lu et al 1995 0.005 9.48E-04 14.0E+03 0.84 0.99

240
Appendix 4
Table 4.7: production data (Ahmed and Roux, 2006)

Time (days ) G p (MMscf) Wp (MMscf) p (psia)

0 0 0 1500
730 265.086 0.15749 1315
1460 968.41 0.290238 1021
2190 1704.033 0.368292 814.4
2920 2423.4 0.425473 664.9
3650 2992.901 0.464361 571.1

Table 4.8: Coal properties (Ahmed and Roux, 2006)

Langmuir’s pressure constant b  0.00276 psi-1

Langmuir volume constant VL  428.5 scf/ton
Average bulk density B  1.70 g/cm3
Average thickness h  50 ft
Initial water saturation Swi  0.95
Drainage area A  320 acres
Initial pressure pi  1500 psia
Critical pressure pd  1500 psia
Temperature T  105 F

Initial gas content Gc  345.1 scf/ton

Formation volume factor Bw  1.00 bbl/STB
Porosity   0.01

Water compressibility cw  3 106 psi-1

Formation compressibility c f  6  106 psi-1

241
Appendix 5

Reda (1987) carried out an experimental study of the welded tuff at Yucca Mountain, with the following properties used in the verification of
the analytical model given below:

b (Klinkenberg factor) = 7.6 105 Pa

p( L) is the gas pressure at the outlet = 1.0 105 Pa.s

L is the length from the inlet to the outlet = 10 m

x is the task location along the gas transport path

k is the intrinsic permeability = 5.0 1019 m2

 is the compressibility factor = 1.8 105 Pa1m1

qm is the air injection rate= 1.0 106 kg/s

u g is the dynamic viscosity= 1.84 105 Pa.s

242
Discretizing the systems of equations using finite difference method

Matrix:

 1  m  MPLVL s  pm  .   kmi  pm  bm  p     km  g  pm  p f 

m  2    m 

 Vstd  PL  pm   t     g

 1  m  MPLVL s  pm     kmi pm pm   kmibm pm    km pm  pm  p f 

m  2  
 Vstd  PL  pm   t x   x  x  g

m  C , 1  m  MPLVL  s kmi

kmi
B ,  A , pm  U , p f  V ,  D , Abm  E
Vstd  ug

 B  U   U U 
C  2
  AU E   DU U  V 
  PL  U   t x  x x 

 B  U  U 
2
 2U  2U
C  2
 A   AU 2  E 2   D U  V 
  PL  U   t  x  x x

Discretizing using backward Euler for time and central difference for space
243
    U n 1  U n  U in1,1j  U in1,1j 
2
U in, j 11  U in, j 11 
2
U (in1,1)j  U i(n1,1)j  2U i(,nj1)  U (in,j1)1  U i(,nj1)1  2U i(,nj1) 
B
f  C     A    A

  AU i , j 

  AU i , j 
i, j i, j

n 1 n 1
n 1 2  
       (y ) 2
2
P  U  t  
 (2 x ) 
 
 (2 y ) 
 
 ( x ) 
  
 L i , j 
U (in1,1)j  U i(n1,1)j  2U i(,nj1)  U (in,j1)1  U i(,nj1)1  2U i(,nj1) 
  DU i , j U i , j  Vi , j 
n 1 n 1 n 1
E  E
 (x) 2
  (y ) 2


Now to compute the Jacobian (Partial derivative for the Jacobian Matrix)

B
C
2 B U  U  P  U in, j 1 
n 1 2
f
n
 U in1,1 j  U in1,1j  2U in, j 1   U in, j 11  U in, j 11  2U in, j 1  2A 2A
   A  A  U in, j 1  U in, j 1
i, j i, j L
 
  
 ( x)
U i , j t  PL  U in, j 1  t   
3 2 2 2 2
 ( x )   ( y )  ( y ) -----Diagonal

 D U in, j 1  Vi ,nj1   DU in, j 1

2E 2E
 
(x) 2
( y ) 2

1 U 
n 1
f  U n 1 U in, j 1 E ------- Upper diagonal (East)
  A  i 1, j 2 i 1, j   A 
U i 1, j 2  (x)  (x) 2
(x) 2

1 U  U n 1 
n 1
f U n 1 E ---------- (North)
  A  i , j 1 2 i , j 1   A i , j 2 
U i , j 1 2  (y ) 
 (y ) (y ) 2

1 U 
n 1
f  U n 1 U in, j 1 E ----------Lower diagonal (West)
 A  i 1, j 2 i 1, j   A 
U i 1, j 2  (x)  (x) 2
(x) 2

1  U i , j 1  U i , j 1 
n 1 n 1
f U in, j 1 E ------------ (South)
 A   A 

U i , j 1 2  (y ) 2
(y ) 2
(y ) 2

244
FRACTURE:

p   k fi  p f  b f    km  g  pm  p f 
 f  f  .    p f    qq
t     
   g

k fi kmi
 f  K ,   A1 , pm  U , p f  V ,  D , qq  0
 ug

V   1 V 
K   A (V  b f )   DU U  V 
t x  x 

V  V   2V  2V
2

K  A   AV 2  Ab f  DU U  V   0
t  x  x x 2

A1b f  F ,

V  V   2V  2V
2

K  A   AV 2  F 2  DU U  V   0
t  x  x x

Discretization:

245
 2 2
 Vi ,nj1  Vi ,nj   Vi n1,1j  Vi n1,1j   Vi ,nj11  Vi ,nj11  ( n 1)
 Vi n1,1j  Vi n1,1j  2Vi ,nj1  ( n 1)
 Vi ,nj11  Vi ,nj11  2Vi ,nj1 
K
   A    A    AVi , j    AVi , j  
  t   2  x   2  y   (  x ) 2
  (  y ) 2

 Vi 1, j  Vi 1, j  2Vi , j 
n 1 n 1 n 1
 Vi , j 1  Vi , j 1  2Vi , j 
n 1 n 1 n 1

  DVi , j U i , j  Vi , j   0
n 1 n 1 n 1
F    F 
 (x) 2  (y ) 2
   

2 2
 Vi ,nj1  Vi ,nj   Vi n1,1j  Vi n1,1j   Vi ,nj11  Vi ,nj11  ( n 1)
 Vi n1,1j  Vi n1,1j  2Vi ,nj1  ( n 1)
 Vi ,nj11  Vi ,nj11  2Vi ,nj1 
fi , j  K    A    A    AVi , j    AVi , j  
 t 2x 2y (x) 2 (y ) 2
         
 Vi n1,1j  Vi n1,1j  2Vi ,nj1   Vi ,nj11  Vi ,nj11  2Vi ,nj1 
  DVi , j U i , j  Vi , j 
n 1 n 1 n 1
F    F 
 (x) 2
(y ) 2
   

Now to compute the Jacobian (Partial derivative for the Jacobian Matrix)

f K  V n 1  Vi n1,1j  2Vi ,nj1   Vi ,nj11  Vi ,nj11  2Vi ,nj1  2 A1 ( n 1) 2 A1 ( n 1) 2F 2F

  A1  i 1, j   A1
   V  V    D(U in, j 1  Vi ,nj1 )  DVi ,nj1 --------
Vi , j t  (  x ) 2   (  y ) 2  ( x ) 2 i, j
(  y ) 2 i, j
( x ) 2
( y ) 2
   
Diagonal

f 1 A1 A1
2  i 1, j i 1, j 
n 1 n 1 F
 V  V  V n1  -------------Upper diagonal (South)
Vi 1, j 2 (x) (x) 2 i, j
(x) 2

f 1 A1 A1
2  i , j 1 i , j 1 
n 1 n 1 F
 V  V  V n1  ---------------- (North)
Vi , j 1 2 (x) (y) 2 i, j
(y) 2

246
 f 1 A1 A1
2  i 1, j
V  Vi 1, j  
n 1 n 1 F
 V n1  -------------Lower diagonal
Vi 1, j 2 (x) (x) 2 i, j
(x)2

f 1 A1 A1

Vi , j 1 2 (y) 2 V n 1
i , j 1  V n 1
i , j 1  
(y)
V n1 
2 i, j
F
(y)2
--------------- (South)

247
Appendix 6

Solving the Maxwell Equations using FDTD

By assuming no charges or current, the Maxwell equations for linear, isotropic, on-
dispersive material can be written as

 B
 E  
t

 D
 H 
t

. D  0

. B  0

Where
 
D t    E
 
B t    H

To solve these equation, Yee et al proposed a method called finite difference time domain
(FDTD) as a three dimensional solution to Maxwell’s equations.

By substituting the constitutive relations into the Maxwell’s equations we can obtain the
following set of Maxwell’s equation

 H
  E  
t

 E
 H  
t

 
.   H   0
 

 
.   H   0
 

248
Since the exact solutions of maxwells equations is impossible , numerical methods such as
FDTD can be applied. For the time dependent maxwells equations, discretization is achieved
using central difference approximations to space and time with resulting finite difference
equations solved using a leapfrog approach. The curl equations are written in FDTD form on
a yee cell as shown below.The electric and magnetic field are staggered in time so that the
electric field will exist at integer time steps of 0, t , 2t,.. and the magnetic field at half
time steps of t , t  t , 2 t  t ,...
2 2 2

~ ~ ~ ~
i , j , k 1
E | i , j 1,k  Ez |t i , j ,k E y |t  E y |t i , j ,k
 z t 
E i , j ,k
C x t
|
y z
~ ~ ~ ~
E | i , j ,k 1  Ex |t i , j ,k Ez |t i 1, j ,k  Ez |t i , j ,k
 x t 
E i , j ,k
C y t
|
z x
~ ~ ~ ~
E y |t i 1, j ,k  Ez |t i , j ,k Ex |t i , j 1,k  E y |t i , j ,k
 
E i , j ,k
C z t
|
x y

And now

 xx |i , j ,k H x |t t 2  H x |t t 2
i , j ,k i , j ,k


E i , j ,k
C x t
|
co t
 yy |i , j ,k H y |ti ,j ,kt 2  H y |ti ,j ,kt 2

E i , j ,k
C y t
|
co t
 zz |i , j ,k H |
i , j ,k
z t t 2  H z |it ,j ,kt 2

E i , j ,k
C z t
|
co t

H z |it ,j ,kt 2  H z |it ,j t1,2k H y |it ,j ,kt 2  H y |it ,j ,kt 21
 
H i , j ,k
C |
x t  t
2 y z
i , j , k 1
H | i , j ,k
x t t 2 H | x t t 2 H | i , j ,k
z t t 2  H z |it 1,tj ,2k
 
H i , j ,k
C |
y t  t
2 z x
H y |t t 2  H y |it 1,tj ,2k
i , j ,k
H | i , j ,k
x t t 2  H x |ti ,j t1,2k
 
H i , j ,k
C |
z t  t
2 x y

By normalizing the magnetic field , we can obtain similar order of magnitude for the both
the E and H fields.

~
O 
H H ,
O

249
 

 r   H
~

 E  
co t
 
~
   E
 H   r
co t

If we normalise the electric field


~
O  1 
E E E
O no

~
1  
D D  co D
 o o

By substituting the normalised fields, we can rewrite the Maxwell’s equation as


~
 1 D
 H 
co t
 
~  H
 E   o

co t
 
~ ~
D  r E

Expansion of the curl equation and only concentrating on 2-D solution of the equation, our
resulting equations can be summarised below

250
 ~ ~
 Ez  E y 1  H x H y H z 
    xx   xy   xz 
y z co  t t t 
 
~ H ~
 Ex  Ez
~
1 H x H y H z 
 E   o
     yx   yy   yz 
co t z x co  t t t 
~ ~
 Ey  Ex 1 H x H y H z 
     zx   zy   zz 
x y co  t t t 

~
H z H y 1  Dx
 
y z co t

~ ~
1 D
 H x H z 1  Dy
 H   
co t z x co t
~
H y H 1  Dz
 x 
x y co t

~ ~ ~ ~

 
Dx   xx Ex   xy E y   xz Ez
~ ~ ~ ~ ~ ~
D  r E Dy   yx Ex   yy E y   yz Ez
~ ~ ~ ~
Dz   zx Ex   zy E y   zz Ez

By assuming only diagonal tenors

~ ~
 Ez  E y 1 H x 
     xx
y z co  t 
~ ~
 Ex  Ez 1 H y 
     yy 
z x co  t 
~ ~
 Ey  Ex 1 H z 
     zz 
x y co  t 

251
 ~ ~
Dx   xx Ex
~ ~
Dy   yy E y
~ ~
Dz   zz Ez

The curl terms can now be calculated separately

~ ~
 xx H x  Ez  E y
  
E
C x
co t y z
~ ~
 yy H y  Ex  Ez
  
E
C y
co t z x
~ ~
 zz H z  Ey
 Ex
  
E
C z
co t x y

H z H y
 
H
C x
y z
H x H z
 
H
C y
z x
H y H x
 
H
C z
x y

Approximation using the finite difference

~ ~ ~ ~
i , j , k 1
Ez |t i , j 1,k  Ez |t i , j ,k E y |t  E y |t i , j ,k
 
E i , j ,k
Cx t
|
y z
~ ~ ~ ~
Ex |t i , j ,k 1  Ex |t i , j ,k Ez |t i 1, j ,k  Ez |t i , j ,k
 
E i , j ,k
C y t
|
z x
~ ~ ~ ~
E y |t i 1, j ,k  Ez |t i , j ,k Ex |t i , j 1,k  E y |t i , j ,k
 
E i , j ,k
C z t
|
x y

And now

252
  xx |i , j ,k H x |t t 2  H x |t t 2
i , j ,k i , j ,k


E i , j ,k
C x t
|
co t
 yy |i , j ,k H y |ti ,j ,kt 2  H y |ti ,j ,kt 2

E i , j ,k
C y t
|
co t
 zz |i , j ,k H z |t t 2  H z |t t 2
i , j ,k i , j ,k


E i , j ,k
C z t
|
co t

H z |it ,j ,kt 2  H z |it ,j t1,2k H y |it ,j ,kt 2  H y |it ,j ,kt 21
 
H i , j ,k
C |
x t  t
2 y z
H x |ti ,j ,kt 2  H x |ti ,j ,kt 21 H z |it ,j ,kt 2  H z |it 1,tj ,2k
 
H i , j ,k
C |
y t  t
2 z x
H | i , j ,k
y t t 2  H y |it 1,tj ,2k H | i , j ,k
x t t 2  H x |ti ,j t1,2k
 
H i , j ,k
C |
z t  t
2 x y

And finally
~ ~
1 D x |it ,j ,kt  D x |it , j ,k

H i , j ,k
C |
x t  t
2 co t
~ ~
1 D y |it ,j ,kt  D y |it , j ,k

H i , j ,k
C |
y t  t
2 co t
~ ~
1 D z |it ,j ,kt  D z |ti , j ,k

H i , j ,k
C |
z t  t
2 co t

 E |
~ ~
D | i , j ,k
x t   xx | i , j ,k i , j ,k
x t

  E |
~ ~
i , j ,k i , j ,k i , j ,k
D | y t yy | y t

  E |
~ ~
i , j ,k i , j ,k i , j ,k
D | z t zz | z t

Reduction to 2 dimension

For 2D,   0 meaning no wave propagation in the z direction.

z

253
 ~ ~
E | i , j 1  Ez |t i , j  xx |i , j H x |t t 2  H x |t t 2
i, j i, j
|  z t
E i, j
| 
E i, j
C x t
y C x t
co t
~ ~
Ez |t i 1, j  Ez |t i , j  yy |i , j H y |ti ,j t 2  H y |ti ,j t 2
| 
E i, j
| 
E i, j
C y t
x
C y t
co t
~ ~ ~ ~
E y |t i 1, j  Ez |t i , j Ex |t i , j 1  E y |t i , j  zz |i , j H |
i, j
z t t 2  H z |ti ,j t 2
| 
E i, j
|  
E i, j
C z t
x y
C z t
co t

~ ~
1 D x |it ,j t  D x |it , j
H z |ti ,j t 2  H z |ti ,j t1 2 
H i, j

C
H i, j
|
x t  t
 C |
x t  t
2 co t
2 y
~ ~
H z |ti ,j t 2  H z |ti 1,tj 2 1 D y |it ,j t  D y |it , j

H i, j

H i, j
C |
y t  t
2 x
C |
y t  t
2 co t
H y |it ,j t 2  H y |it 1,tj 2 H x |ti ,j t 2  H x |ti ,j t1 2 ~ ~

C
H i, j
|   1 D z |it ,j t  D z |it , j

H i, j
z t  t
2 x y C |
z t  t
2 co t

 E |
~ ~
D |   xx |
i, j
x t
i, j i, j
x t

  E |
~ ~
Dy |ti , j yy |
i, j i, j
y t

  E |
~ ~
i, j i, j i, j
D | z t zz | z t

We can notice that the Maxwell’s equations have been decoupled into two distinct sets of
equations of Ez and H z .

254
 ~ ~
E | i , j 1  Ez |t i , j
|  z t
E i, j
C x t
y
~ ~
Ez |t i 1, j  Ez |t i , j
C | 
E i, j
y t
x
 xx | H x |t t 2  H x |t t 2
i, j i, j i, j

 
E i, j
Cx t
|
co t
 yy |i , j H y |ti ,j t 2  H y |it ,j t 2
| 
E i, j
C y t
co t
H y |it ,j t 2  H y |it 1,tj 2 H x |ti ,j t 2  H x |it ,j t1 2
 
H i, j
C |
z t  t
2 x y
~ ~
1 D z |it ,j t  D z |it , j

H i, j
C |
z t  t
2 co t

 
~ ~
Dz |ti , j   zz |i , j Ez |it , j

For the H z

~ ~
E | i , j 1  Ez |t i , j
|  z t
E i, j
C x t
y
~ ~
Ez |t i 1, j  Ez |t i , j
C | 
E i, j
y t
x
 zz | H z |t t 2  H z |t t 2
i, j i, j i, j

C z |t   co
E i, j

t
H z |ti ,j t 2  H z |ti ,j t1 2

H i, j
C |
x t  t
2 y
H | i, j
z t t 2  H z |it 1,tj 2

H i, j
C |
y t  t
2 x
~ ~
1 D z |it ,j t  D z |ti , j

H i, j
C |
z t  t
2 co t

 
~ ~
Dx |ti , j   xx |i , j Ex |ti , j

  E
~ ~
Dy |ti , j yy |i , j i, j
|
y t

255
Update equations can now be written for solving the future time values of the fields
associated with the Ez and H z modes.

 c t  E
H x |ti ,j t 2  H x |ti ,j t 2    o i , j  C x |it , j
  xx | 
 c t  E
H y |ti ,j t 2  H y |ti ,j t 2   o i , j  C y |it ,j t 2
  yy | 
 
~ ~
D z |it ,j t  D z |it , j   co t  C z |ti , j
H

~  1  ~ i, j
Ez |ti ,j t    Dz |t t
  zz |
i, j

Likewise Hz mode

 c t 
H z |ti ,j t 2  H z |ti ,j t 2    o i , j  C zE |ti , j
  zz | 
~ ~
D x |it ,j t  D z |it , j   co t  C x |ti ,jt
H

2
~ ~
D y |it ,j t  D z |it , j   co t  C y |it ,jt
H

~  1  ~
Ex |ti ,j t   i, j 
Dx |ti ,j t

 xx  |
~  1  ~ i, j
E y |ti ,j t  
  yy |i , j  y t t
D |
 

256
257