ABSTRACT

The geological model, reservoir description, and analysis of recovery factor improvements are
all part of the distinct process that begins with each simulation study. When it comes to
practical reservoir systems, numerical reservoir simulators are the go-to tool for petroleum
engineers seeking reliable solutions. The purpose of this research is to model multi-phase flow
in porous media using a numerical reservoir simulator that is two dimensions in depth. By
defining the fundamental assumptions of the simulator, one can gain a better grasp of the
physical processes and reservoir fluid conditions, which in turn allows for the formulation of
the partial differential equations (PDEs). Now we have a system of coupled, nonlinear partial
differential equations (PDEs). Afterwards, the PDEs are discretized by utilizing the finite-
difference method, which is based on Taylor's series, to transform them into non-linear
algebraic equations. In order to find the solution to the equations involving the distributions of
pressure and fluid saturation in the reservoir, the non-linear algebraic equations that result are
first linearized into linear algebraic equations. This project discusses the aim, objectives, scope
of this study as well as the significance of the study and presents a comprehensive and explicit
look at the context of reservoir simulation and reviewing the works of pioneers in the field of
reservoir simulation, advances and trends in reservoir simulation as well as the gaps identified
in the literature. It then describes the process of designing a reservoir simulator, starting from
the mathematical model to the computer model as well as presenting the methodology used in
this study (Implicit Pressure Explicit Saturation which is considered one of the profound
method of solving multi-phase flow equations. Also included are the outcomes of the created
simulator under the Dirichlet boundary condition. These obtained results were discussed and
appraised while validating these results using the material balance approach and then concludes
by summarizing the results and observations that the pressure saturation variation in a
homogeneous 2D dimensional anisotropic reservoir strongly depends on the reservoir
boundary effects and the initial reservoir condition.
 CHAPTER ONE
INTRODUCTION
1.1 Background to the Study
There are a plethora of applications for reserve estimation, including but not limited to: oil and gas
reserve exploration, development, and production; interest and profit evaluation; investment
decisions in the oil and gas sector; field performance evaluation; and many more. There are
currently five known methods for reserve estimation, ranked from least to most accurate: the
analogy method, the volumetric method, the decline curve method, the material balance method,
and the reservoir simulation method. This work focuses on the latter two methods, which have
become the industry standard for solving reservoir engineering problems.
The goal of reservoir simulation is to create a tool for predicting the performance of hydrocarbon
reservoirs under different operating strategies. This tool is developed by combining mathematical
concepts, computer programming, reservoir engineering, and physics (Aziz, K. and Settari, A.
1979).
The practice of reservoir simulation dates back to the 1930s, when petroleum engineering first
emerged. However, the term "numerical simulation" didn't catch on until the early 1960s, when
predictive methods were transformed into advanced computer programs. These software
applications were revolutionary because they solved massive sets of finite-difference equations
that characterize transient multiphase flow in heterogeneous porous media in two and three
dimensions. Thanks to numerical mathematical methods developed for solving massive systems
of finite-difference equations and the rapid evolution of large-scale, ultra-high speed digital
computers, these computer programs have advanced to a new level.
Because fluid flow in petroleum reservoirs (also called porous media) is typically a very
complicated phenomenon, analytical solutions to mathematical models can only be obtained by
making simplified assumptions about the reservoir's geometry, reservoir characteristics, and
boundary conditions. Because of the non-linearity of the governing equations, the heterogeneity
of the fluid, the irregular shape of a reservoir system, and the complex behavior of multiphase flow,
it is often impossible to develop analytical solutions for practical problems using such
simplifications. Given these restrictions on analytical methods, an alternate strategy is offered:
solving these models using a finite difference method or another appropriate numerical method.
Investment decisions on large exploration and development projects still frequently make use of
numerical reservoir simulation, which is a valuable and important tool. Major oil and gas projects
necessitate decisions like starting secondary and enhanced oil recovery methods, optimizing or
improving field development plans, and determining the commercial viability of underground
reserves. With the development and understanding of reservoir simulation, it is possible to employ
it in the early phases of development, before the hydrocarbon pool is put on production, to avoid
unnecessary expenditures and to effectively carry out proper planning.
When creating a simulator, two main schools of thought emerged: mathematics and engineering.
a) A "mathematical approach" is a set of procedures for deriving and discretizing
PDEs in order to derive the nonlinear algebraic equations. The mathematical
approach was crucial for simulator developers in obtaining the nonlinear algebraic
equations, also called finite-difference equations, and other similar terms.
b) But the engineering method is more in line with the way engineers think and with
the physical definitions of the words used in the flow equations. In the case where
second-order approximations are employed by the mathematical approach,
however, both the engineering and mathematical methods handle boundary
conditions equally accurately.
1.2 Statement of the problem
Reservoir parameters like saturation, pressure, and heterogeneity cannot be determined using
either material balance or decline curve analysis. Assuming average reservoir properties and
treating the reservoir like a tank, the material balance technique evaluates the performance of oil
and gas reservoirs using reservoir engineering calculations. A field's production history can also
be used in decline curve analysis to foretell how well the field will do in the future. Therefore, an
issue arises due to the disregard for the geometry of the reservoir structure.
Therefore, it is necessary to create a reservoir simulator that considers the reservoir's geometry and
provides an estimate of the recoverable reserves within a specified time period. This is particularly
important during the exploration and development phase, when management must decide whether
to invest in the project or not based on economic and risk factors. However, numerical reservoir
simulators utilizing numerical models that account for reservoir geometry and the quantity of
flowing fluids are often necessary because analytical solutions to practical problems in the
reservoir system cannot be developed in many cases owing to the complex behaviors of multiphase
flow, non-linearity of the governing equation, heterogeneity, and the irregular shape of a reservoir
system.

1.3 Aim of the study

To develop a 2-D numerical reservoir simulator for modeling multiphase flow in porous media.

1.4 Objectives of the study

• Assuming multiphase flow, we must formulate the flow equations for a 2-dimensional
reservoir grid block.
• To find the pressure saturation distributions across the grid block using the Implicit Pressure
Explicit Saturation (IMPES) method, which involves discretizing the flow equations that represent
the actual behavior of the flow in the reservoir.
• Using computer codes, create a pressure saturation map for every grid in the reservoir block.

1.5 Significance of the study

The data obtained from this study can be:
1. Based on the project's risk and economic assessment and analysis, reservoir engineers use
this tool to help management make the final investment decision (FID) on whether to start
optimizing the development strategy or to abandon the project.
2. To determine the maximum economic, safe, and ecologically friendly production of
reserves, it is necessary to first estimate the ultimate recovery and then plan the completion
pattern that will work with the well conditions.

1.6 Scope and limitation of the study

This study assumes that the porous medium is homogeneous, with constant permeability and
porosity throughout the reservoir, and focuses solely on developing 2-dimensional cartesian
coordinates for modeling multiphase flow in porous media. This project does not include 1-D or
3-D reservoir simulators.
While many other numerical models have been created over the years, the focus of this project will
be on creating a simulator using the Implicit Pressure Explicit Saturation (IMPES) method.
This study assumes that the porous medium is homogeneous, with constant permeability and
porosity throughout the reservoir, and focuses solely on developing 2-dimensional cartesian
coordinates for modeling multiphase flow in porous media. This project does not include 1-D or
3-D reservoir simulators.
While many other numerical models have been created over the years, the focus of this project will
be on creating a simulator using the Implicit Pressure Explicit Saturation (IMPES) method.

CHAPTER TWO
LITERATURE REVIEW
2.1: Conceptual Review
Simulating the operation of a reservoir system is known as reservoir simulation. This method
involves the use of computer models to forecast the flow of fluids (such as water, gas, and oil)
through rocks with microscopic fissures. The rocks that are found underground and are known to
contain or store oil and gas are known as porous media. The oil industry now uses reservoir
simulation as a standard for solving reservoir engineering problems, and it also helps researchers
and engineers plan the best way to extract gas and oil from rocks. With the use of reservoir
simulation, smart decisions can be made to increase the economic recovery of hydrocarbons from
reservoirs by predicting how the reservoirs will perform in the future.
Reservoir simulation is an involved process that begins with a static model of the geology and ends
with a dynamic model of the reservoir. As can be seen in figure 2.1 below, a reservoir simulator is
built from the ground up with a solid foundation in mathematics and applied science. It all begins
with solving a mathematical model of fluid flow in a specific reservoir-well system using finite
difference equations, then moves on to numerical modeling and computer programming, and
finally produces the simulation software.

Fig 2.1: A schematic diagram of the reservoir simulation process

2.1.1: The History and Evolution of Reservoir Simulation, from the Early Analytical
Methods to the Modern Numerical Techniques and Software.
The history and evolution of reservoir simulation can be divided into three main stages:
From the latter half of the nineteenth century until the middle of the twentieth, we were in the
analytical era. At this point, reservoir engineers modeled fluid flow using analytical solutions to
simplified equations. The concepts of conservation of mass, momentum, and energy, along with
the fluid equations of state, formed the basis of these equations. Among others, Darcy, Buckley-
Leverett, Muskat, Theis, and Fetkovich created some of the most renowned analytical solutions.
The material balance, decline curve, productivity index, and initial hydrocarbons in place were all
approximations that benefited from these solutions. Nevertheless, there were numerous drawbacks
to these methods. For example, they depended on constant boundary conditions, ignored gravity
and capillary effects, failed to account for compositional changes and phase behavior, and assumed
homogeneous and isotropic reservoirs.
Second, starting in the middle of the twentieth century and continuing up to the present day, was
the numerical era. Now that we've gotten to the fun part, reservoir engineers have been using
numerical methods to solve increasingly complicated and realistic fluid flow equations. These
equations followed the same principles as the analytical ones but included additional physical
phenomena like gravity, capillarity, compositional variation, phase behavior, thermal effects,
wellbore dynamics, etc., in addition to account for heterogeneity and anisotropy. Numerical
methods such as finite difference, streamline, finite element, and finite volume were among the
most popular. A system of algebraic or differential equations could be solved using iterative or
direct techniques. These methods needed to discretize the reservoir into a grid of cells or elements,
assign properties to each cell or element, apply boundary and initial conditions, and so on. While
numerical methods improved reservoir modeling in terms of accuracy and flexibility, they came
at the cost of increased data input and computing resources.
Coming in at the tail end of the twentieth century and continuing to this day is the software era,
the third and final stage. Reservoir engineers executed reservoir simulation tasks at this stage using
specialized software packages. Commercial companies, research organizations, or academic
institutions created these software packages. They supplied data management systems,
optimization algorithms, methods for analyzing uncertainty, graphical tools for visualizing data,
and user-friendly interfaces. Eclipse1, CMG2, VIP3, GEM4, tNavigator, INTERSECT, and many
more are among the most widely used software packages for reservoir simulation. Reservoir
engineers now have easier and more efficient access to reservoir simulation thanks to these
software packages, but there were some downsides, like high prices and incompatibilities.
Although numerical solutions to multiphase, multidimensional flow problems were not possible
until the 1950s, digital computing techniques and equipment had already been developed at that
point.

Section 2.1.2: Abstract of Numerical Reservoir Modeling

Computing approximate solutions to the mathematical model depicting fluid flow in porous media
is the goal of numerical reservoir models, which are computer programs that employ numerical
methods. According to Abou et al. (2006), numerical reservoir simulation begins with the
formulation of partial differential equations (PDEs) that describe the physical processes in the real
reservoir system. Then, the PDEs are discretized using one of three prominent methods: the Finite
Difference Approach (FDA), the Integral Formulation Method, or the Finite Element Method. The
reservoir is then discretized into grid blocks using either the block-centred or the point distributed
grid system. Finally, computer codes are written to represent the mathematical model of the flow
process, which is used to build the simulator.

Different kinds of grid systems (2.1.2.1)

In order to find a solution to the nonlinear finite difference equations that represent the fluid flow,
a reservoir is grid-imposed, according to Aziz (1993). The block-centered grid system and the
point distributed grid system are the two most common grid systems used in reservoir simulation.
You can see them in fig 2.6(a) and (b), respectively.

Fig. 2.1.2.1(a): Discretization of a 1-D reservoir using a block centered grid

 Fig. 2.1.2.1(b): Discretization of a 1-D reservoir using a point distributed grid.

In their study, Beckner et al. (2001) compared two grid systems: the block-centered grid and the
point distributed grid. They looked at features like grid cell representation, geometry
representation, heterogeneity representation, numerical accuracy and efficiency, implementation
and data storage, and more. Beckner highlighted specific instances where one grid system was
preferred over the other and outlined the benefits and drawbacks of each grid system so that users
can choose the one that best suits their needs. Nevertheless, there are a number of considerations
that go into selecting between block-centered and point distributed grids, including reservoir
complexity, required accuracy, available computational resources, and simulation goals. It is also
possible to combine the two approaches, as seen in hybrid grids, to achieve a compromise between
accuracy and computational efficiency.

2.1.2.2: Approaches to Reservoir Simulation: The Mathematical Approach as Compared

Against the Engineering Approach.
The two methods offered by Abou-Kassem—the mathematical and the engineering—can be used
to create a simulator (2006). The mathematical method had always been used for discretization
prior to the development of the engineering method. The following procedures are involved in the
mathematical method:
• The constitutive equation, the conservation of mass and momentum, Darcy's law, and other
fundamental principles are used to derive the partial differential equation (PDE) that
describes the fluid flow in the reservoir.
• The reservoir can be discretized using any of the grid systems mentioned earlier.
• Using any of the numerical methods, discretize the resultant PDE into space and time.
Whereas, in the engineering approach, the finite difference equations are derived without going
through the rigor of derivation and discretization of PDEs. This approach involves the following
steps:
• The reservoir is divided into gridblocks or grid points so that the space variable is no longer
an issue.
• Applying the three previously stated principles, we can derive the algebraic flow equations for
gridblock I (or grid point I accounting for the time-dependent variation of the interblock flow terms
and the source/sink term within a given time step.
• The nonlinear algebraic equations are generated by approximating the time integrals in the
resultant flow equation.

The following conclusions can be drawn from analysis of the two approaches according to Abou-
Kassem, et al.
• The engineering approach is closer to the physical meaning of the various terms in the
algebraic flow equations but the mathematical approach can be used to study the analysis
of local truncation errors, consistency, convergence, and stability. Due to this end, it may
be concluded that the mathematical and engineering approaches complement each other.
• The engineering approach rather than the mathematical approach is closer to the engineer's
thinking. While the mathematical approach is to first derive the PDEs, followed by
discretizing the reservoir, and finally discretizing the PDEs, the engineering approach first
discretizes the reservoir, then derives the algebraic flow equations with time integrals, and
finally approximates the time integrals to obtain the same nonlinear algebraic flow
equations.

2.1.3: The Types and Classifications of Reservoir Simulation Models.

Models for the mathematical description and prediction of fluid flow in porous media, like
reservoirs for oil and gas, are known as reservoir simulation models. It is possible to categorize
reservoir simulation models according to a number of factors, including the fluid and rock types
used, the physical and numerical processes involved, and the recovery procedures implemented.
Reservoir simulators were categorized by Ertekin et al. as either gas- or compositional. Reservior
simulators can be further categorized according to recovery methods. For example, thermal-
recovery, conventional-recovery, chemical-flood, and miscible displacement simulators are all
types of simulators that fall into this category. Regardless of the various ways reservoir simulators
are categorized, there are two main types based on the fluid and reservoir descriptions:
compositional simulators and black oil simulators.
The following are descriptions of the many reservoir simulation models available:
Black-oil scenarios Most reservoir fluid models are of this type, since they are the simplest and
most common. They presuppose that the saturated phase characteristics of two hydrocarbon
phases—gas and oil—are solely affected by pressure, and that the proportions of dissolved gas and
vaporized oil in each phase are directly proportional to their saturations. Some three-phase
recovery procedures, primary depletion, waterflooding, and immiscible gas injection are all within
the capabilities of black-oil models. Pressure and fluid characteristics, including oil formation
volume factor Bo, gas formation volume factor Bg, and the solution gas oil ratio, are considered
to be the only variables that can affect mass transfer in black-oil simulators. Rs control the behavior
of PVT and are employed when recovery processes are insensitive to changes in the reservoir
fluids' composition.
When recovery processes are sensitive to changes in the composition of the reservoir fluid, a more
complex and realistic type of model called a compositional model is employed. This includes
scenarios like multiple-contact-miscible processes, pressure-maintenance operations in volatile-
oil and gas-condensate reservoirs, and primary depletion of these reservoirs. Compositional
simulators are typically used to model all of these cases. The foundational premise of
compositional models (simulators) is that the reservoir fluid is composed of multiple components
that can partition into different phases based on their equilibrium ratios. Variation in composition,
phase behavior, displacement into or out of a miscible or nearly-miscible state, improved oil
recovery, and many more phenomena can be captured by compositional models. In contrast to
black-oil simulators, where Bo, Bg, and Rs control the PVT, compositional models use a cubic
equation of state to govern the PVT behavior.
One kind of reservoir fluid model is the thermal model, which takes into consideration the heat
transfer between the reservoir and its environment, the effects of temperature on the fluid and rock
properties, and so on. Injecting steam, in-situ combustion, cold water, geothermal energy, etc., are
all within the capabilities of thermal models.
The reservoir is believed to consist of a single homogeneous and isotropic porous medium in
single-porosity models, which are a subset of reservoir rock models. When there are no or very
little fracture effects in conventional reservoirs, single-porosity models work well.
Reservoir rock models with dual-porosity or dual-permeability assumptions treat the matrix and
fracture as two interconnected porous media. For reservoirs that have been fractured, either
naturally or artificially, and have a significant impact from the fractures, models with dual-porosity
or dual-permeability work well.
Fluid flow, geomechanics, geochemistry, and other physical processes and phenomena can be
represented in reservoir simulation models as coupled models. Changes in stress, compaction,
subsidence, acidization, and other complex reservoir-environment interactions can be handled by
coupled models.
To improve oil recovery, reservoirs can be "flooded" with chemicals, and this process can be
"simulated" using a chemical-flood reservoir simulator. Surfactants, polymers, and alkaline agents
are some of the chemicals that can improve the displacement and sweep efficiency of the reservoir
fluids by lowering the interfacial tension, raising the viscosity, and changing their wettability. The
effects of salinity, adsorption, dispersion, and the complex phase behavior, transport, and reactions
of the chemical species in the reservoir can all be taken into consideration by a chemical-flood
reservoir simulator.
One kind of reservoir simulator is the miscible-displacement model, which can simulate the
injection of a solvent that is either completely or almost miscible with the oil in the reservoir. By
dissipating capillary forces and releasing trapped oil, dissolving in oil can lessen residual oil
saturation and boost recovery factor. The effects of gravity, viscosity, density differences, mass
transfer between phases, and changes in composition can all be taken into consideration by a
miscible-displacement simulator.
2.1.4: The Challenges and Limitations of Reservoir Simulation
As we have seen earlier, reservoir simulation is an important tool for reservoir engineers and
managers to optimize the development and management of oil and gas fields. However, reservoir
simulation also has its own challenges and limitations, such as:

1. Computational cost: Reservoir simulation involves solving a system of partial differential

equations (PDEs) that describe the conservation of mass, momentum, and energy of the
 fluids (oil, gas, and water) in the porous media of the reservoir. These equations are often
nonlinear, coupled, and time-dependent, and they depend on various physical properties of
the fluids and the rock, such as pressure, saturation, density, viscosity, compressibility,
permeability, porosity, etc. Solving these equations on a digital computer requires
discretizing the reservoir domain into a grid or mesh of cells or elements, assigning values
or functions to each cell or element, and applying numerical schemes to approximate the
derivatives and integrals in the PDEs. This process can be very computationally intensive
and time-consuming, especially for large-scale or complex reservoir models.
2. Grid resolution: The accuracy and reliability of the reservoir simulation results depend
largely on the quality and resolution of the grid or mesh used to discretize the reservoir
domain. The grid or mesh should be able to capture the main features and heterogeneities
of the reservoir geology and geometry, as well as the flow dynamics and physics. However,
increasing the grid resolution also increases the computational cost and data input required
for the simulation. Therefore, there is a trade-off between grid resolution and
computational efficiency.
3. Uncertainty quantification: Reservoir simulation is based on a mathematical model that
represents a simplified version of reality. The model parameters, such as fluid and rock
properties, initial and boundary conditions, well locations and types, etc., are subject to
uncertainty due to data scarcity, quality, or variability. Uncertainty quantification is a
process of assessing the impact of uncertainty on the simulation results and their confidence
intervals. Uncertainty quantification can help reservoir engineers and managers to make
better decisions based on risk analysis and scenario planning. However, uncertainty
quantification can also be challenging and costly to perform, as it requires running multiple
simulations with different parameter values or distributions.
4. History matching: History matching is a process of calibrating or adjusting the reservoir
model parameters to match the historical data from the reservoir, such as production rates,
pressures, saturations, etc. It is equally a process of adjusting key properties of the reservoir
model to fit or match the actual historic data. There are several parameters that are varied
either individually or collectively in order to minimize the differences between the
observed data abd the data calculated by the simulator.
2.1.5: Applications of Reservoir Simulation
Reservoir simulation has several uses, including:
The geology, structure, and heterogeneity of a reservoir can be better understood with the use of
reservoir simulation, which integrates data from a variety of sources (seismic, well-log, core,
production, etc.). The primary characteristics and features of a reservoir, including faults, fractures,
shales, channels, porosity, permeability, saturation, and so on, can be better understood through
reservoir characterization. There is less room for error and danger in the reservoir model when the
reservoir is well-characterized.
Refine your reservoir's production performance and recovery potential with the use of reservoir
simulation. This tool allows you to test and compare various scenarios and strategies, including
well locations, types, and completions; production rates and pressures; injection rates and fluids;
artificial lift methods, etc. If the reservoir can be optimized for production, its economic value and
profitability can be maximized.
Simulation of reservoirs can aid in the evaluation and design of various enhanced oil recovery
(EOR) methods and schemes, including thermal, chemical, gas, or microbial injection; flooding
with surfactant or polymer; flooding with low-salinity water; and so on. With EOR, reservoir
displacement and sweep efficiency can be improved, leading to an increase in the oil recovery
factor and an extension of the reservoir life.
To lessen the impact of greenhouse gas emissions and slow the rate of climate change, carbon
capture and storage (CCS) projects inject carbon dioxide (CO2) into either producing or depleted
reservoirs. Reservoir simulation is a useful tool for evaluating and keeping tabs on the safety and
viability of these projects. By amplifying the effects of oil swelling and viscosity reduction, CCS
can also aid in improving oil recovery. To anticipate the reservoir's injectivity, CO₂ storage
capacity, migration, trapping, leakage, and environmental implications, reservoir modeling is a
useful tool.

2.1.6: Benefits of Reservoir Simulation

The benefits of reservoir simulation include:
 1. It can help to improve the understanding of the reservoir geology, structure, and
heterogeneity, by integrating various data sources, such as seismic, well-log, core,
production, etc.
2. It can help to optimize the production performance and recovery potential of the reservoir,
by testing and comparing different scenarios and strategies, such as well locations, types,
and completions; production rates and pressures; injection rates and fluids; artificial lift
methods; etc.
3. It can help to evaluate and design different enhanced oil recovery (EOR) methods and
schemes, such as thermal, chemical, gas, or microbial injection; surfactant or polymer
flooding; low-salinity water flooding; etc.
4. It can help to assess and monitor the feasibility and safety of carbon capture and storage
(CCS) projects, which involve injecting carbon dioxide (CO2) into depleted or producing
reservoirs to reduce greenhouse gas emissions and mitigate climate change.
5. It can help to reduce the uncertainty and risk associated with the reservoir model by using
data assimilation, inversion, history matching, and uncertainty quantification techniques.

2.1.7: The Recent Advances and Trends in Reservoir Simulation.

There are some of the recent advances and trends in reservoir simulation that aim to improve the
quality and efficiency of reservoir modeling and management. Some of them include:
1. Both artificial intelligence (AI) and machine learning (ML) are methods that learn from
data in order to make predictions, decisions, and suggestions based on established patterns,
rules, and correlations. Using AI and ML in reservoir simulation can make many tasks
more accurate, efficient, and robust. These tasks include evaluation of formations,
interpretation of seismic data, geological modeling, history matching, optimization,
forecasting, and production analysis. Using sparse data, AI can create realistic geological
models. It can also reduce the computational cost of reservoir simulation by approximating
complex physics-based models with proxy models.
2. One approach is to use simplified models to mimic the behavior of more complicated reservoir
models. This technique is called proxy modeling. Models grounded in physics, models driven by
data, or a hybrid of the two can all contribute to the development of proxy models. For large-scale
or high-resolution problems, as well as for optimizing and quantifying uncertainty, proxy models
can be utilized to accelerate reservoir simulation. One application of proxy models is the
 optimization of well placement and injection strategies, as well as the execution of rapid sensitivity
analyses.
3. HPC, or high-performance computing, is a method for quickly and efficiently processing large
amounts of data by making use of computer architectures and algorithms that work in parallel. For
difficult problems with big domains, fine grids, complicated physics, or multiple scenarios, HPC
can improve the performance and scalability of reservoir simulation. Using sophisticated solvers
and preconditioners that take advantage of the heterogeneity and parallelism of contemporary
computing platforms, large-scale linear systems resulting from reservoir simulation can be solved
on HPC.
4. Modern approaches to risk analysis and uncertainty quantification incorporate stringent
uncertainty quantification of reservoir properties, production rates, and other variables to
shed light on the possible dangers and unknowns of different production scenarios.

2.1.8: Validating a Reservoir Simulator: Why the need?

The validation of a reservoir simulator is simply the process of testing the numerical codes and it
is the last step in developing a simulator, after which the simulator can be used for practical field
applications. The validation step is mandatory for the following reasons:
1. To make sure that no errors were introduced in the various steps of development or in
computer programming and to fix any errors or issues if any, in order to improve the
simulator's performance.
2. To ensure that the simulator's provide accurate and reliable results.
3. To compare different simulators and see how they perform relative to each other.
4. This validation is distinct from the concept of conducting experiments in support of a
mathematical model. Material balance checks serve as one of the means of validating
reservoir simulators and so validation of the reservoir model is really very important for
the effective use of reservoir simulators.

2.2: Empirical review

When it comes to managing and planning petroleum reservoir exploitation projects, reservoir
simulation is a great decision-support tool (Aanonsen et al., 2009). According to Turgay Ertekin
et al. (2001), reservoir simulation is necessary because petroleum engineers need to be able to
accurately predict how a hydrocarbon reservoir will perform under various operating conditions.
The necessity for this arises from the fact that while investing hundreds of millions of dollars in a
hydrocarbon recovery project, it is essential to evaluate and reduce the risk linked to the chosen
development plan. This risk is exacerbated by factors such as:
1. The regional variations of fluid properties and relative permeability characteristics;
2. The complexity of the hydrocarbon recovery mechanisms;
3. The application of other predictive methods with limitations that may make them
inappropriate;
4. The complexity of the reservoir due to the heterogeneous and anisotropic rock
properties.
According to Turgay Ertekin, et al (2001), the use of reservoir simulation as a predictive tool is
becoming standard in the petroleum industry. A set of algebraic mathematical equations developed
from a set of PDEs with appropriate initial and boundary conditions explains the behaviour of the
reservoir in the reservoir simulation approach. These equations incorporate the most important
physical processes taking place in the reservoir system, including the flow of fluids partitioned
into three phases such as (gas, oil and water) as shown in fig 2.2 below, and the mass transfer
between the various phases.
 Fig 2.2: A reservoir system showing the partitioned three phases.

When calculating fluid flow, a generalized version of Darcy's law accounts for the influences of
capillary, viscous, and gravity forces. Reservoir simulation entails three steps. Among these steps
are choosing a physical reservoir system, creating mathematical models (equations) to describe
that model, and writing code to numerically solve those equations (simulators).
According to Archer (1983), there are two main factors that determine whether a reservoir model
is successful: first, the equations must accurately depict the physical aspects of fluid flow; and
second, the reservoir and well system must be in equilibrium.
the capacity of the grid attributes to depict the reservoir details in one, two, or three dimensions.
Henry Darcy, a French engineer, was the first to empirically formulate how fluids flow generally
in 1856. This is significant because reservoir simulation entails modeling the flow of fluids within
the reservoir into the wellbore as well as how these fluids flow from the reservoir into the grid
blocks of the reservoir. A single fluid's flow rate through a porous medium is given by the formula
that takes the potential gradient, the cross-sectional area normal to the flow direction, and the
viscosity of the fluid into account. Modern descriptions of fluid flow in porous media are based on
this relationship, which is now called Darcy's law. Here is the simplified version of Darcy's law:
v = (-k/μ)∇Φ
where v is the velocity of the flow;
k is the permeability of the rock;
μ is the viscosity of the fluid; and
Φ is the potential gradient which provides the force for fluid flow.
In 1930, researchers conducted experiments to learn more about the movement of immiscible
fluids; their findings showed that when two fluid phases are present, the rock's conductance to both
phases is reduced. A strong function of the phase saturations within the porous medium is the
relative permeability of the rock to each phase of an immiscible two-phase fluid system, as
demonstrated by the works of Wyckoff and Botset. Leverett established a relationship between
capillary pressure and rock porosity and permeability in his work by introducing the notion of the
Leverett J-function. Additionally, he demonstrated and concluded that relative permeability is not
influenced by fluid viscosity but rather by various factors such as the distribution of porosity (pore
size) across the reservoir rock, displacement pressure, pressure gradient, and fluid saturation. In
1955, Rapport derived a single equation for immiscible, incompressible two-phase flow by
combining the three-dimensional partial differential equations. Fagin and Stewart introduced a
general flow model in 1965 called a two-dimensional multiphase reservoir simulator. This model
accurately represented the time-dependent variations in pressure and multiphase saturation.
With the exclusion of capillary effects, West, Garvin, and Sheldon conducted a numerical study
of horizontal, linear, and radial systems driven by gas. They employed a finite difference grid
system and implicit methods to solve the resulting system of linear equations.
Two incompressible, immiscible fluids were modeled by Hovanessian and Fayers for a one-
dimensional inclined flow. Their findings demonstrated that the phase saturations and pressure
distribution were significantly affected by the incorporation of capillary pressure and gravity
forces. Similar to what the Buckley-Leverett method typically achieves at high flow rates, Douglas
et al. similarly investigated a horizontal, one-dimensional system that incorporates capillary
pressure. The outcomes demonstrated that this model yields comparable results.
Models using two-dimensional techniques were found to be extremely unstable in the early studies.
Nevertheless, a stable approach to these two-dimensional problems was introduced by Douglas in
1958 with his work on the "Alternating Direction Implicit Procedure" (ADIP). The ADIP, a
classical finite difference treatment, was used to accurately model the two-dimensional flow of
inimiscible fluids by Douglas, Peaceman, and Ratchford in 1959.
Commonly used to aid the reservoir engineer in describing the reservoir's behavior, the block oil
models provided by Aziz, K. et al. (1979) are shown in figures 2.3(a), (b), and (c) below.
A two-dimensional model for describing areal performance or vertical conformance in a reservoir
is one example of these models. The 2-D areal model, as shown in figure 2.3(a) above, is a useful
tool for describing situations where fluids do not move vertically, such as in a thin sand reservoir.
On the other hand, it can also be applied to thick sands when the permeability is very consistent,
meaning that there is no permeability layering. Figure 2.3(b) shows the 2-dimensional cross-
sectional model, which is commonly used to simulate a slice of a field and is particularly useful
for determining completion intervals and stratification effects.
Figure 2.3(c) shows a three-dimensional model that can take both areal and vertical conformance
into consideration at the same time. Any kind of simulation study is within its capabilities. When
compared to 2-D models, however, the time spent on computer programming and analyzing study
results is significantly higher.
Each of these models has the potential to be a powerful resource for reservoir engineers, but in the
wrong hands, they can yield erroneous and misleading conclusions.

Fig 2.3(a): 2-D areal reservoir model

 Fig 2.3(b): 2-D cross sectional reservoir model

Fig 2.3(c): A 3-D reservoir model

2.2.1: Numerical Methods

When analytical solutions to nonlinear differential equations prove to be intractable, these methods
can be employed to derive more approximate solutions. Those who were early adopters of reservoir
simulation have been dogged in their pursuit of better methods for creating the reservoir model.
As pointed out by Peaceman (1977), the mathematical model of a physical reservoir system is a
collection of partial differential equations (PDEs) coupled with boundary conditions that
adequately depict the physical processes taking place in the reservoir, including fluid flow and
mass transfer between two or three immiscible phases.
In 1953, Bruce et al. introduced the initial numerical approach to determine pressure distributions
in homogeneous reservoirs subjected to unsteady state gas flow. With the advent of new computer
technologies, reservoir simulation has continued to evolve and find new uses. It is common
practice today to employ numerical reservoir simulation when deciding how much to invest in
large-scale exploitation and development projects. Among these choices are the optimization of
field development plans, the determination of the commercial viability of hydrocarbon
accumulations, and the initiation of secondary and enhanced oil recovery methods on large oil and
gas projects based on reservoir simulation results. The three known methods for discretizing any
partial differential equation are the variational method, the integral method, and the Taylor series
method (Aziz and Settari, 1979). One approach yields the variational method, while the other two
lead to the finite-difference difference method. The work at hand, however, will employ the finite-
difference method as a byproduct of the Taylor series method.
2.2.1.2: Finite-Difference Method (FDM)
This approach is the most common method used in reservoir simulation and it was first developed
by A. Thom in the 1920s as used in the work "the method of square" in order to solve nonlinear
hydrodynamic equations. It is a numerical technique used to convert or approximate continuous
ordinary and partial differential equations into discrete form and it was first applied to reservoir
simulation by Prats in 1952. In the finite difference approach, the Taylor-series expansion is used
to define the derivative functions in the governing equation, after which pressure is then calculated
at a fixed point in the center of the reservoir grid block.
In petroleum engineering, the finite difference method is used to solve the diffusivity equation
which is a PDE that describes the flow of fluids in porous media and Ertekin et al, 2001 listed how
the finite difference method can be applied to solve the diffusivity equation. These steps include:
1. Superimpose a finite-difference grid over the reservoir to be modeled. Generally, this
grid can be of two types as used in reservoir simulation. They are the block-centered
grid and the point-distributed grid. In the block-centered grid, for a rectangular
coordinate system, the dimensions of the grid block are first selected before the
placement of points in the central locations of the block, whereas in the point-
distributed grid, the grid points or nodes are selected first (Abou-Kassem et al, 2001).
2. Use the chosen grid system to approximate the spatial derivatives in the continuous
equations. These approximations thus, result in algebraic equations called finite-
difference equations which are obtained by truncating the Taylor-series expansion of
the unknown variable.
The finite-difference formulations to the fluid flow equations require the following:
1. The spatial discretization of the reservoir model into a pattern of grid blocks;
2. Dividing the total simulation time, T into a smaller number of time steps, ∆t.
Abbas Firoozabadi et al (2000) stated some of the advantages and disadvantages of this method as
presented below;

Advantages:
i. It is very simple to use.
ii. It can be easily extended from 1-D, to 2-D, and 3-D as well.
iii. It is compatible with certain aspects of the physics of two- and three-phase flow.

Disadvantages:
i. Grid dependency.
ii. Numerical dispersion.

In 1965, Fagin and Stewart introduced an alternate approach to solving two-phase, two-
dimensional flow problems called the Implicit Pressure Explicit Saturation (IMPES), which was
later adopted by Douglas et al. In 1965, Thomas. E. Gallant introduced a new method called the
Sequential Solution (SS) method to address stability issues with the IMPES solution method.
Nonlinear algebraic equations, according to Ertekin et al. (2001), must be "linearized," or
transformed into linear equations, before they can be solved using standard algebraic methods. It
is possible to solve the simulator equations (finite-difference equations) using any of the methods
used to solve linear equations derived from the finite-difference method after they have been
transformed into linear equations. There are two main categories into which these approaches fall:
direct methods and iterative methods.
One distinguishing feature of direct solution methods is their ability to generate the solution vector
for a specific set of linear equations in a predetermined amount of operations. By applying these
techniques, the initial equations are transformed into simpler, equivalent equations. Some
examples of such techniques are LU (Lower-Upper) decomposition, Gaussian elimination, and
Gauss-Jordan. The direct solver could produce an exact solution after a constant number of
computations if the computer could hold an infinite number of digits, according to Ertekin et al.
(2001). However, due to the finite word length of all real computers, the solution obtained by using
a direct solver will be prone to round-off errors.
Due to the necessity of storing the coefficient matrix and the right vector throughout the solution
process, the direct solver is also constrained by its excessive storage requirements. The high
memory and computational demands of direct solvers make them impractical for large-scale or
sparse systems, despite their reliability and accuracy.
Figure 2.4, taken from Ogbe (2012), Lecture notes on applied reservoir simulation, depicts the
direct solution method.

Fig 2.4: A diagram showing the direct solution method.

As per Ertekin et al. (2001), iterative methods adhere to the iteration process, which begins with a
preliminary solution estimate or guess and is improved successively until it approaches the exact
solution within a reasonable margin. In this category you'll find techniques like the Conjugate-
Gradient Method, Strongly Implicit Procedure (SIP), Jacobi iteration, Gauss-Seidel iteration,
Block SOR (BSOR) methods, and Successive Over-Relaxation (SOR) methods (CGM).
Because they need to do so many iterations, iterative methods are typically slower than direct
methods. On the other hand, their computational storage capabilities are much superior, which is
particularly true for large and sparsely populated coefficient matrices. For some problems, iterative
methods are the best option, and Jaan (2015) summed up the advantages of iterative methods over
direct methods. Included in the edge:
In most cases, iterative procedures are self-correcting; this means that they will fix any rounding
errors or math mistakes that occur in one cycle by themselves in the next cycle.
In many cases, storing the coefficient matrix is completely unnecessary, in contrast to the direct
methods where doing so becomes mandatory.
It is possible to store only the nonzero elements of the coefficient matrix, which makes it possible
to deal with very large matrices.
Also extracted from Ogbe (2012) - Lecture notes on applied reservoir simulation, is fig 2.5 below
shows the Schematic Representation of the Iterative Method.

Fig 2.5: A diagram showing the iterative solution method.

Both mathematical and engineering methods can be used to create a reservoir simulator. According
to Abou-Kassem (2006), the mathematical approach has long been the sole method used because
of the encouragement of intense research in the field of mathematics with the works of IMPES
(Breitenbach et al., 1969), the sequential method, and the adaptive implicit method. This approach
involves deriving and discretizing PDEs in order to obtain nonlinear algebraic equations (Thurnau,
1983).
On the other hand, a novel method was introduced by Abou-Kassem (2006). It utilizes imaginary
wells as boundary conditions and derives the finite difference equations without subjecting them
to the rigorous procedures of partial differential equations and discretization. The "engineering
approach" describes this new way of doing things since it more closely follows the way engineers
think and the physical definitions of the words used in flow equations. If the mathematical method
employs second-order approximations, the two methods can equally accurately handle boundary
conditions; the engineering method is both general and rigorous while being simple. Also, for any
hydrocarbon recovery process, the finite-difference equations obtained by the engineering method
are identical to those obtained by the mathematical method.
The discretized form of the governing equations, according to Farouq Ali (1986), contain no more
than Darcy's law, which describes the volumetric flow rate between any two adjacent blocks.
Combining this observation with an assumption about the time level at which flow terms are
evaluated allowed him to develop the forward-central-difference equation and the backward-
central-difference equation without subjecting undergraduates to the rigorous mathematical
approach often used to teach reservoir simulation. The initial mathematical approach to utilize a
control volume represented by a point at its center was proposed by Ertekin et al. (2001).
Reservoir simulation has made significant contributions to solving the partial differential equations
(PDEs) that describe multi-phase flow. These contributions include:
Handling spatial and temporal nonlinearities (Settari and Aziz, 1975; Coats et al., 1977; Saad, 1989;
Gupta, 1990).
The IMPES, SEQ, fully implicit SS, and adaptive implicit methods were developed for solving
systems of nonlinear PDEs. These methods were first proposed by Fleischenbach et al. (1969),
Spillette et al. (1973), and Coats (1978).
Coming up with sophisticated iterative approaches to solving systems of linear algebraic equations,
like the Block Iterative, Nested Factorization, and Orthomin methods (Behie and Vinsome, 1982;
Appleyard and Cheshire, 1983; Vinsome, 1976).

2.2.2: History Matching

According to Aanonsen et al. (2009), reservoir simulation model incorporates fluid and rock
properties into a mathematical framework to explain fluid flow in porous media. This model is
then used to forecast the performance of the reservoir model under various operating conditions.
However, to improve the predictive capacity of a reservoir model, it is important to incorporate
relevant field data into the model. In petroleum engineering literature, the process of integrating
dynamic data into reservoir models is referred to as history matching.
Reservoir history matching is simply the process of constraining the reservoir model to replicate
the reservoir's actual historical dynamic production data in order to forecast the reservoir's future
production behaviour and reduce ambiguity in the reservoir's description (Chang et al., 2010). Due
to the fact that the amount of accessible independent data is very much less than the number of
variables, history matching is an ill-posed problem. As a result, reservoir models are constructed
using unknown parameters as there exists an infinite number of possible combinations of reservoir
characteristics, each of which results in a reservoir model that fits the data and the mere fact that
the available information about the reservoir is frequently inaccurate and contradictory. It is
equally a process of adjusting key properties of the reservoir model to fit or match the actual
historic data. There are several parameters that are varied either individually or collectively in
order to minimize the differences between the observed data abd the data calculated by the
simulator.
Crichlow (1977) presented the following areas where modifications are usually made whenever
there is a mismatch between the observed and calculated data. They include:
1. Rock data (such as porosity, permeability, saturations, and thickness) modifications.
2. Fluid data (such as PVT data, viscosity, and compressibility) modifications.
3. Relative permeability data.
4. Shift in relative permeability curve (shift in critical saturation data).
5. Individual well completion data (skin effect and bottomhole flowing pressure).
The two basic processes that can be controlled in history matching are as follows:
1. The quantity of fluid in the system at any time and its distribution within the reservoir,
and
2. The movement of fluid within the system under existing potential gradients (Crichlow
1977).
Okotie and Bibobra (2019), presented that the manipulation of these two processes enables the
reservoir engineer to modify any of the afore-mentioned parameters which are the criteria to
history matching. It is compulsory that these modifications of the data reflect good engineering
judgement and be within reasonable limits of conditions existing in that area. History matching
can improve the accuracy and reliability of the reservoir simulation results by reducing the
discrepancy between the model predictions and the actual observations. However, history
matching can also be difficult and time-consuming to perform, as it requires running multiple
simulations with different parameter values or optimization algorithms until a satisfactory match
is achieved. This implies that the total time spent on history matching depends largely on the
expertise of the engineer and how well he or she is familiar with the particular reservoir.

2.3: Gap in reviewed literature.

The following are the gaps in existing literatures reviewed:
There is a wealth of literature outlining the development and progression of reservoir
simulation. On the other hand, more up-to-date and thorough surveys are still required to
address the most current developments and trends in reservoir simulation, including HPC,
proxy modeling, artificial intelligence, machine learning, etc.
Reservoir simulation models have been the subject of substantial research and comparison
across disciplines. However, when it comes to describing and classifying the various
reservoir simulation models, there is still no universal agreement or standardization on
terminology or criteria. Additional models that are more adaptable and cohesive are
required to deal with reservoirs that are complicated and involve numerous physical
processes and phenomena.
Numerous sources have contributed to the development and improvement of algorithms
and methods for solving the governing equations of reservoir simulation. When solving
large-scale or complicated systems of equations, it is still difficult to strike a balance
between accuracy and efficiency. In addition, improved adaptive and robust methods are
still required for reservoir simulation in order to handle nonlinearity and uncertainty.
Fluid behavior in reservoirs can be quite complex, displaying non-ideal phase behavior,
fluid composition variations, and multiphase flow with phase transitions, among other
things. So, there is a need for additional research because it is still difficult to develop
simulation methods that adequately and efficiently model and describe such scenarios.
Various sources have acknowledged and addressed the difficulties and limits of reservoir
simulation. A number of issues, including computational cost, grid resolution, uncertainty
quantification, history matching, and so on, remain challenges to reservoir simulation.
Additionally, better and more reliable results from reservoir simulations still necessitate
more creative and efficient solutions.
 Several sources have proven and assessed the usefulness and advantages of reservoir
simulation. Although reservoir simulation has come a long way, there is still a disconnect
between theory and practice due to the underutilization of certain models and methods.

CHAPTER THREE
METHODOLOGY
3.1. Development of the simulator
In order to build a reservoir simulator, one must first have a thorough comprehension of the fluid
flow processes taking place in the reservoir. Only then can one derive the partial differential
equations that represent these processes, which are too complicated to be solved manually. The
mathematical method and the engineering method are the two main schools of thought when it
comes to creating a reservoir simulator. The reservoir engineering method was first proposed by
Aziz (1993). The following are the main components of this project, which is meant to model
multiphase flow in porous media using a two-dimensional reservoir simulator:
1. Formulate the partial differential equations of the model based on the oil and gas
reservoir characteristics, using the three basic laws governing fluid flow in porous media.
2. Divide the reservoir into grids and then discretize the formulated partial differential
equations in both space and time so as to obtain a linear system of equations.
3. Select an appropriate ordering scheme to obtain an order of coefficients of the linear
system of equations and then compute the pressure and saturation per grid of the reservoir
for each time step and thus generating the pressure and saturation map for each grid
present in the reservoir.
4. Write codes for the system of equations using python programming language.
5. Finally validate the simulator using the material balance method.
Three kinds of model are usually employed in the development of any reservoir simulator
which include the mathematical model, numerical model, and the computer model.

3.2. The Mathematical Model

In other words, this is the main structure or base upon which the reservoir simulator would be
constructed. In this case, the properties of the oil and gas reservoir are described using a system
of partial differential equations (PDEs) along with the right starting and ending points to
approximate the reservoir's behavior. Reservoir simulation is primarily governed by three
primary laws: the conservation of mass law, Darcy's law (the transport equation), and the
constitutive equation, which is also called the equation of state (phase properties such as density,
compressibility and formation volume factor).
3.2.1. Basic Assumptions
The following basic assumptions were made in the formulation of the problem;
1. Negligible gravity effect
2. Uniform grid size
3. Linear fluid flow into the wellbore

❑ Derivation of multiphase flow equation in porous media

𝑞𝑖𝑛 𝑆𝑤 𝑞𝑜𝑢𝑡
𝑆𝑔
𝑆𝑜

Fig 3.1: A realistic representation of a typical black oil system.

In the above figure 3.1, there exists the flow of a black oil fluid in which the oil and water
components are assumed to be immiscible; therefore, there is no mass transfer between these
phases. The gas component is assumed to be soluble in oil but usually not in water; therefore,
mass transfer of the gas component occurs between the oil and gas phases only. The portion of
the gas component dissolved in the oil phase is known as solution gas.
The material balance equation for component c written over a finite control volume of the porous
reservoir over time interval Δ𝑡 is given by the principle of mass conservation which states that
the total mass of fluid leaving and entering the volume element is equal to the mass accumulated
as represented in eqn (3.00) below
The conservation equation, assuming only convective transport is given by:
[[𝜌𝑖 𝑞𝑖 ⌉𝑥 + 𝑞𝑖𝑛 𝐴∆𝑥] − [[𝜌𝑖 𝑞𝑖 ⌉𝑥+∆𝑥 + 𝑞𝑜𝑢𝑡 𝐴∆𝑥] = 𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑠𝑡𝑜𝑟𝑎𝑔𝑒--------------(3.00)
𝑉
Recall that 𝑆𝑜 = 𝑜⁄𝑉 𝑎𝑛𝑑 𝑉𝑜 = 𝑆𝑜 𝑉𝑝 -------------------------------(3.01)
𝑝

𝑉
Also, porosity is defined as ∅ = 𝑝⁄𝑉 → 𝑉𝑝 = ∅𝑉𝑏 𝑎𝑛𝑑 𝑉𝑏 = 𝐴∆𝑥
𝑏

But the mass stored in the pore volume = 𝜌𝑜 𝑉𝑜 . -------------------------(3.02)

By substituting eqn (3.01) for 𝑉𝑜 in eqn (3.02), we then have that
The mass stored in the pore volume = 𝜌𝑜 𝑉𝑜 = 𝜌𝑜 𝑆𝑜 𝑉𝑝 = 𝜌𝑜 𝑆𝑜 ∅𝐴∆𝑥

Therefore, mass rate storage = 𝜕⁄𝜕𝑡 (𝜌𝑜 𝑆𝑜 ∅𝐴∆𝑥)-----------------------(3.03)

Eqn (3.00) is thus written as

[[𝜌𝑖 𝑞𝑖 ⌉𝑥 + 𝑞𝑖𝑛 𝐴∆𝑥] − [[𝜌𝑖 𝑞𝑖 ⌉𝑥+∆𝑥 + 𝑞𝑜𝑢𝑡 𝐴∆𝑥] = 𝜕⁄𝜕𝑡 (𝜌𝑖 𝑆𝑖 ∅𝐴∆𝑥)----------(3.04)
By grouping like terms, we have
By factoring out the common term 𝐴∆𝑥, we have
𝜕(𝜌𝑖 𝑆𝑖 ∅𝐴∆𝑥)
[𝜌𝑖 𝑞𝑖 ⌉𝑥 − 𝜌𝑖 𝑞𝑖 ⌉𝑥+∆𝑥 ] + [𝑞𝑖𝑛 − 𝑞𝑜𝑢𝑡 ]𝐴∆𝑥 = --------------(3.05)
𝜕𝑡
𝜕(𝜌𝑖 𝑆𝑖 ∅𝐴∆𝑥)
[𝜌𝑖 𝑞𝑖 ⌉𝑥 − 𝜌𝑖 𝑞𝑖 ⌉𝑥+∆𝑥 ] = + [𝑞𝑜𝑢𝑡 − 𝑞𝑖𝑛 ]𝐴∆𝑥-----------------(3.06)
𝜕𝑡

Applying Taylor’s series to the left hand side of eqn (3.06), we have
𝜕(𝜌𝑖 𝑞𝑖 )
[𝜌𝑖 𝑞𝑖 ⌉𝑥 − 𝜌𝑖 𝑞𝑖 ⌉𝑥+∆𝑥 ] = −∆𝑥 ----------------------------------------(3.07)
𝜕𝑥

And 𝑞̇ 𝑛𝑒𝑡 = 𝑞𝑜𝑢𝑡 − 𝑞𝑖𝑛 − − − − − − − − − − − (3.08)

Eqns (3.07) and (3.08) can thus be substituted into eqn (3.06) to give
𝜕(𝜌𝑖 𝑞𝑖 ) 𝜕(𝜌𝑖 𝑆𝑖 ∅)
−∆𝑥 =[ + 𝑞̇ 𝑛𝑒𝑡 ] 𝐴∆𝑥------------------------------------(3.09)
𝜕𝑥 𝜕𝑡

By dividing both sides of eqn (3.08) by ∆𝑥 gives

𝜕(𝜌𝑖 𝑞𝑖 ) 𝜕(𝜌𝑖 𝑆𝑖 ∅)
− =[ + 𝑞̇ 𝑛𝑒𝑡 ] 𝐴-------------------------------------------(3.10)
𝜕𝑥 𝜕𝑡

By applying the transport equation which is given as

𝐴𝑘𝑖 𝑑𝑃𝑖
𝑞𝑖 = 𝐴𝑈𝑖 = − ---------------------------------------------------(3.11)
𝜇𝑖 𝑑𝑥

Eqn (3.10) thus becomes

𝜕(𝜌𝑖 𝐴𝑈𝑖 ) 𝜕(𝜌𝑖 𝑆𝑖 ∅)
− =[ + 𝑞̇ 𝑛𝑒𝑡 ] 𝐴---------------------------------------(3.12)
𝜕𝑥 𝜕𝑡

𝜕(𝜌𝑖 𝑈𝑖 ) 𝜕(𝜌𝑖 𝑆𝑖 ∅)
−𝐴 =[ + 𝑞̇ 𝑛𝑒𝑡 ] 𝐴---------------------------------------(3.13)
𝜕𝑥 𝜕𝑡

By dividing both sides of eqn (3.13) by A, eqn (3.13) thus becomes

𝜕 𝜕(𝜌𝑖 𝑆𝑖 ∅)
− 𝜕𝑥 (𝜌𝑖 𝑈𝑖 ) = + 𝑞̇ 𝑛𝑒𝑡 ----------------------------------------(3.14)
𝜕𝑡

The above eqn (3.14) is the continuity equation for oil, gas and water.
However, in vector format, it is expressed as
𝜕
−∇. (𝜌𝑖 𝑈𝑖 ) = 𝜕𝑡 (𝜌𝑖 𝑆𝑖 ∅) + 𝑞̇ 𝑛𝑒𝑡 -----------------------------------------------------(3.15)
 𝜕 𝜕 𝜕
Where ∇= i 𝜕𝑥 + 𝑗 𝜕𝑦 + 𝑘 𝜕𝑧

By applying equation of state (EOS) for both an oil and gas system, we have
𝜌𝑈 = 𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎----------------------------------------------------(3.16)
Therefore, for an oil system the left hand side of eqn (3.14) can be expressed as
𝜌𝑖 𝑈𝑖 = 𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 𝑎𝑡 𝑆𝑇𝐶---------------------------------(3.17)
𝜌𝑖 𝑈𝑖 = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝑉𝑜𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎------------------------------------------------(3.18)
𝑉
Recall that 𝑉𝑜𝑆𝑇𝐶 = 𝛽𝑜 . Therefore, we have
𝑜

𝑉𝑜
𝜌𝑖 𝑈𝑖 = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝛽 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎----------------------------------------------------(3.19)
𝑜

Also, velocity is equal to volume (flow rate) per unit area. Thus, we have
𝑈
𝜌𝑖 𝑈𝑖 = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝛽𝑜 ----------------------------------------------------------------------(3.20)
𝑜

For the right hand side, we have

𝜌𝑖 𝑆𝑖 ∅ = 𝑚𝑎𝑠𝑠 𝑜𝑓 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 𝑎𝑡 𝑆𝑇𝐶-----------------------------------(3.21)
But 𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝑉𝑜𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 𝑎𝑛𝑑 𝑉𝑜 = 𝑆𝑜 𝑉𝑝 . This then implies that eqn (3.21)
can thus be expressed as
𝑉
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝛽𝑜 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒-------------------------------------------------------(3.22)
𝑜

𝑆𝑜 𝑉𝑝
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒-----------------------------------------------------(3.22a)
𝛽𝑜

Recall that porosity is expressed pore volume per unit volume. Therefore eqn (3.22a) can thus be
written as
𝑆 𝑆
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑜𝑆𝑇𝐶 ∗ 𝛽𝑜 ∅ = 𝜌𝑜𝑆𝑇𝐶 ∅ ∗ 𝛽𝑜 ---------------------------------------------------------(3.23)
𝑜 𝑜

The oil flow equation becomes

𝑈 𝜕 𝑆
−∇. (𝜌𝑜𝑆𝑇𝐶 ∗ 𝛽𝑜 ) = 𝜕𝑡 (𝜌𝑜𝑆𝑇𝐶 ∅ ∗ 𝛽𝑜 ) + 𝜌𝑜𝑆𝑇𝐶 𝑞̅𝑜 -------------------------------------------(3.24)
𝑜 𝑜

By factoring out and dividing both sides of the equation by the constant term 𝜌𝑜𝑆𝑇𝐶 , eqn (3.24)
thus becomes
𝑈 𝜕 𝑆
−∇. (𝛽𝑜 ) = 𝜕𝑡 ( ∅ ∗ 𝛽𝑜 ) + 𝑞̅𝑜 -------------------------------------------------------------------(3.25)
𝑜 𝑜

𝑘𝑜 𝑘𝑘𝑟𝑜
Recall that 𝑈𝑜 = − ∇𝑃𝑜 and 𝑘𝑜 = − ∇𝑃𝑜 , eqn (3.25) thus becomes
𝜇𝑜 𝜇𝑜
𝜕 𝑘𝑘𝑟𝑜 𝜕 𝑘𝑘𝑟𝑜 𝜕 𝑆
( 𝐴𝑥 𝛻𝑃𝑜 ) ∆𝑥𝑖 + ( 𝐴𝑦 𝛻𝑃𝑜 ) ∆𝑦𝑖 = 𝜕𝑡 ( ∅ 𝛽𝑜 ) + 𝑞̅𝑜 -----------------------------------------
𝜕𝑥 𝜇𝑜 𝛽𝑜 𝜕𝑦 𝜇𝑜 𝛽𝑜 𝑜

-------------------(3.26)
𝑘𝑘𝑟𝑜
Let 𝜆𝑜 = 𝜇 . Eqn (3.26) thus becomes
𝑜 𝛽𝑜

𝜕 𝑆
∇. (𝜆𝑜 ∇𝑃𝑜 ) = 𝜕𝑡 ( ∅ 𝛽𝑜 ) + 𝑞̅𝑜 ---------------------------------------------------------------------(3.27)
𝑜

For the gas system, the left hand side of eqn (3.14) is expressed as
𝜌𝑖 𝑈𝑖 = 𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 +
𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑜𝑖𝑙 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 +
𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑤𝑎𝑡𝑒𝑟 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 ---------------------------(3.28)
𝑈𝑖 𝜌𝑖 = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑤𝑎𝑡𝑒𝑟 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎--------------------------------------------(3.29)
𝑉𝑔𝑆𝑇𝐶
Recall that 𝑅𝑠𝑜 = 𝑡ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒, 𝑉𝑔𝑆𝑇𝐶 = 𝑅𝑠𝑜 𝑉𝑜𝑆𝑇𝐶
𝑉𝑜𝑆𝑇𝐶

𝜌𝑖 𝑈𝑖 = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑜 𝑉𝑜𝑆𝑇𝐶 𝑜𝑓 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 +
𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑤 𝑉𝑤𝑆𝑇𝐶 𝑜𝑓 𝑤𝑎𝑡𝑒𝑟 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎---------------------------------------------------(3.30)
𝑉 𝑉
𝜌𝑖 𝑈𝑖 = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝛽𝑔 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑜 𝛽𝑜 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑔 𝑜

𝑉
𝑅𝑠𝑤 𝛽𝑤 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎---------------------------------------------------------------------------(3.31)
𝑤

Also, 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 = 𝑣𝑜𝑙𝑢𝑚𝑒 𝑟𝑎𝑡𝑒⁄𝑎𝑟𝑒𝑎 . This implies that

𝑈𝑔 = 𝑉𝑔 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎, 𝑈𝑜 = 𝑉𝑜 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎, 𝑈𝑤 = 𝑉𝑤 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑎𝑟𝑒𝑎
𝑈 𝑈 𝑈
Therefore , 𝜌𝑖 𝑈𝑖 = 𝜌𝑔𝑆𝑇𝐶 [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]-------------------------------------------(3.32)
𝑔 𝑜 𝑤

By considering the right hand side of eqn (3.14), we have

𝜌𝑖 𝑆𝑖 ∅ = 𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 +
𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑜𝑖𝑙 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 +
𝑚𝑎𝑠𝑠 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑤𝑎𝑡𝑒𝑟 𝑎𝑡 𝑆𝑇𝐶 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒------------------------(3.33)
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑑𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑔𝑎𝑠 𝑖𝑛 𝑤𝑎𝑡𝑒𝑟 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒----------------------------------------(3.34)
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑉𝑔𝑆𝑇𝐶 𝑜𝑓 𝑓𝑟𝑒𝑒 𝑔𝑎𝑠 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑅𝑠𝑜 𝑉𝑜𝑆𝑇𝐶 𝑜𝑓 𝑜𝑖𝑙 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑤 𝑉𝑤𝑆𝑇𝐶 𝑜𝑓 𝑤𝑎𝑡𝑒𝑟 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒---(3.35)
𝑉 𝑉
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝛽𝑔 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑜 𝛽𝑜 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝑔 𝑜

𝑉
𝑅𝑠𝑤 𝛽𝑤 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒-----------------------------------------------------------(3.36)
𝑤

• Recall that 𝑉𝑔 = 𝑆𝑔 𝑉𝑝 , 𝑉𝑜 = 𝑆𝑜 𝑉𝑝 , 𝑎𝑛𝑑 𝑉𝑤 = 𝑆𝑤 𝑉𝑝 . Therefore, eqn (3.36)

𝑆𝑔𝑉𝑝 𝑆𝑜 𝑉𝑝
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑜 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒 + 𝜌𝑔𝑆𝑇𝐶 ∗
𝛽𝑔 𝛽𝑜

𝑆𝑤 𝑉𝑝
𝑅𝑠𝑤 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑣𝑜𝑙𝑢𝑚𝑒---------------------------------------------------------(3.37)
𝛽𝑤

𝑝𝑜𝑟𝑒 𝑣𝑜𝑙𝑢𝑚𝑒
Since, 𝑝𝑜𝑟𝑜𝑠𝑖𝑡𝑦 = ⁄𝑣𝑜𝑙𝑢𝑚𝑒 (𝑏𝑢𝑙𝑘) , 𝑡ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒
𝑆 𝑆 𝑆
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ 𝛽𝑔 ∅ + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑜 𝛽𝑜 ∅ + 𝜌𝑔𝑆𝑇𝐶 ∗ 𝑅𝑠𝑤 𝛽𝑤 ∅------------------(3.38)
𝑔 𝑜 𝑤

𝑆 𝑆 𝑆
𝜌𝑖 𝑆𝑖 ∅ = 𝜌𝑔𝑆𝑇𝐶 ∗ ∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]---------------------------------------(3.38a)
𝑔 𝑜 𝑤

By combining both the left hand and right hand equations, the gas flow equation becomes
𝑈 𝑈 𝑈 𝜕 𝑆 𝑆 𝑆
−∇. {𝜌𝑔𝑆𝑇𝐶 [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]} = 𝜕𝑡 (𝜌𝑔𝑆𝑇𝐶 ∗ ∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]) + 𝜌𝑔𝑆𝑇𝐶 𝑞̅𝑔 ---(3.39)
𝑔 𝑜 𝑤 𝑔 𝑜 𝑤

By factoring out and canceling out the constant term 𝜌𝑔𝑆𝑇𝐶 , we have
𝑈 𝑈 𝑈 𝜕 𝑆 𝑆 𝑆
−∇. {𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 } = 𝜕𝑡 (∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]) + 𝑞̅𝑔 ------------------------------(3.40)
𝑔 𝑜 𝑤 𝑔 𝑜 𝑤

𝑘𝑘𝑟𝑖
Recall that 𝑈𝑖 = − (∇𝑃𝑖 + 𝑟𝑖 ∇𝑧). By substituting yield into eqn (3.40), eqn (3.40) becomes
𝜇𝑖

𝑘𝑘𝑟𝑔 𝑘𝑘𝑟𝑜 𝑘𝑘𝑟𝑤 𝜕 𝑆𝑔

∇. {𝛽 (∇𝑃𝑔 + 𝑟𝑔 ∇𝑧) + 𝑅𝑠𝑜 𝛽 (∇𝑃𝑜 + 𝑟𝑜 ∇𝑧) + 𝑅𝑠𝑤 (∇𝑃𝑤 + 𝑟𝑤 ∇𝑧)} = (∅ [𝛽 +
𝑔 𝜇𝑔 𝑜 𝜇𝑜 𝛽𝑤 𝜇𝑤 𝜕𝑡 𝑔

𝑆 𝑆
𝑅𝑠𝑜 𝛽𝑜 + 𝑅𝑠𝑤 𝛽𝑤 ]) + 𝑞̅𝑔 ----------------------------------------------------------------------------(3.41)
𝑜 𝑤

𝑘𝑘
Recall that 𝜆𝑖 = 𝜇 𝛽𝑟𝑖 . This then makes eqn (3.41) to become
𝑖 𝑖

𝜕 𝑆 𝑆
𝛻. {𝜆𝑔 (𝛻𝑃𝑔 + 𝑟𝑔 𝛻𝑧) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 + 𝑟𝑜 𝛻𝑧) + 𝑅𝑠𝑤 𝜆𝑤 (𝛻𝑃𝑤 + 𝑟𝑤 𝛻𝑧)} = 𝜕𝑡 (∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 ]) +
𝑔 𝑜

𝑞̅𝑔 ----------------------------------------------------------------------------------------(3.42)
By neglecting the gravity term 𝑟𝑖 ∇𝑧, the 2D flow equation for the oil and gas flow model is thus
represented as
𝜕 𝜕 𝜕 𝑆 𝑆
{𝜆𝑔 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )} ∆𝑥𝑖 + 𝜕𝑦 {𝜆𝑔 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑦𝑖 = (∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 ]) + 𝑞̅𝑔 -
𝜕𝑥 𝜕𝑡 𝑔 𝑜

------(3.43)
3.3. The Numerical Model
The partial differential equation representing the mathematical model of the reservoir is very
challenging and complex to solve using analytical techniques hence the need then arises to put
the equation in a form that is compatible to the computer in order to easily provide the solutions
to the PDE. Thus, numerical methods were introduced to achieve this. The solutions provided by
these numerical methods often result in approximate solutions rather than the exact solutions.
Basically, there are three basic numerical models used for exist for determining the solutions to
the fluid flow equations and they include the finite difference method, the finite element method
and the finite volume method with the finite difference method being the most widely used
numerical method in reservoir simulation for the mere reason that it can be applied to reservoirs
of varying complexity and the ability for it to also handle irregular grid structures thus indicating
its flexibility.

3.3.1 Finite Difference Method

The discretization process, which originates from knowledge of Taylor's series, is utilized to
transform extremely complicated partial differential equations (PDEs) into discretized flow
equations using this specific type of numerical model. Different types of discretization exist, the
most common of which are spatial and temporal.
3.3.1(a) Spatial discretization
To do this, we must use space as a variable in the fluid flow equation and discretize it in space.
Here, I and j are two-dimensional coordinates for the space variable. The forward difference
approximation, the backward difference approximation, and the central difference approximation
are all useful tools for achieving spatial discretization. Listed below are some of the topics
covered::
1) Forward Difference Approximation: This is derived from Taylor series in one dimension
in the forward direction which is given as
(Δ𝑥)2
𝑃(𝑥 + Δ𝑥) = 𝑃(𝑥) + Δ𝑥𝑓 ′ (𝑥) + 2
𝑓 ′′ (𝑥) + ⋯ ………………………… (3.44)
(Δ𝑥)2
𝑃(𝑥 + Δ𝑥) − 𝑃(𝑥) = Δ𝑥𝑓 ′ (𝑥) + 𝑓 ′′ (𝑥) + ⋯ …………………………… (3.45)
2
𝑃(𝑥+Δ𝑥)−𝑃(𝑥) Δ𝑥
= 𝑓 ′ (𝑥) + 𝑓 ′′ (𝑥) + ⋯
Δ𝑥 2
𝑃(𝑥+Δ𝑥)−𝑃(𝑥) Δ𝑥
𝑓 ′ (𝑥) = − 𝑓 ′′ (𝑥) ………………………………………………. (3.46)
Δ𝑥 2

Therefore,
𝑃(𝑥+Δ𝑥)−𝑃(𝑥)
𝑓 ′ (𝑥) = − 0Δ(𝑥) where 0Δ(𝑥) is the truncation error
Δ𝑥

The derivative at point i is given as

𝜕𝑝 𝑃𝑖+1 −𝑃𝑖
= …………………………………………………………………… (3.47)
𝜕𝑥 Δ𝑥

This can now be implemented in two dimension as

𝜕𝑝 𝑃𝑖+1,𝑗 −𝑃𝑖,𝑗
= --------------------------------------------------------------------------------------(3.48)
𝜕𝑥 Δ𝑥
𝜕𝑝 𝑃𝑖,𝑗+1 −𝑃𝑖,𝑗
= ---------------------------------------------------------------------------------------(3.49)
𝜕𝑦 Δ𝑦

And in time it becomes:

𝜕𝑝 𝑃𝑖,𝑗 𝑛+1 −𝑃𝑖,𝑗 𝑛
= …………………………………………………………………… (3.50)
𝜕𝑡 Δ𝑡

Backward Difference Approximation: This is derived by applying the Taylor series to the
backward direction and is given as
(Δ𝑥)2 ′′
𝑃(𝑥 − Δ𝑥) = 𝑃(𝑥) − Δ𝑥𝑓 ′ (𝑥) + 𝑓 (𝑥) + ⋯
2
(Δ𝑥)2 ′′
′(
𝑃(𝑥 − Δ𝑥 ) − 𝑃(𝑥 ) = −Δ𝑥𝑓 𝑥 ) + 𝑓 (𝑥 ) + ⋯
2
 𝑃(𝑥 − Δ𝑥) − 𝑃(𝑥) Δ𝑥 ′′
= −𝑓 ′ (𝑥) + 𝑓 (𝑥) + ⋯
Δ𝑥 2
𝑃(𝑥)−𝑃(𝑥−Δ𝑥) Δ𝑥
𝑓 ′ (𝑥) = + 𝑓 ′′ (𝑥)
Δ𝑥 2

This implies that

𝑃(𝑥) − 𝑃(𝑥 − Δ𝑥)
𝑓 ′ (𝑥) = + 0Δ(𝑥)
Δ𝑥
Therefore, the derivative at point i is given as
𝜕𝑝 𝑃𝑖 −𝑃𝑖+1
=
𝜕𝑥 Δ𝑥

This can be extended in two dimension as

𝜕𝑝 𝑃𝑖,𝑗 − 𝑃𝑖+1,𝑗
= ------------------------------------------------------------------(3.51)
𝜕𝑥 Δ𝑥
𝜕𝑝 𝑃𝑖,𝑗 − 𝑃𝑖,𝑗+1
= -------------------------------------------------------------------(3.52)
𝜕𝑦 Δ𝑦

And in time it then becomes:

𝜕𝑝 𝑃𝑖,𝑗 𝑛+1 −𝑃𝑖,𝑗 𝑛
= ------------------------------------------------------------(3.53)
𝜕𝑡 Δ𝑡

Central Difference Approximation: The central difference approximation is the most widely used
technique in reservoir simulation and is derived by combining the forward difference and
backward difference as expressed thus:
(Δ𝑥)2 ′′
𝑃(𝑥 + Δ𝑥) = 𝑃(𝑥) + Δ𝑥𝑓 ′ (𝑥) + 𝑓 (𝑥) + ⋯
2
(Δ𝑥)2 ′′
𝑃(𝑥 − Δ𝑥) = 𝑃(𝑥) − Δ𝑥𝑓 ′ (𝑥) + 𝑓 (𝑥) + ⋯
2
2
𝑃(𝑥 + Δ𝑥) − 𝑃(𝑥 − Δ𝑥) = 2Δ𝑥𝑓′(𝑥) + 3 (Δ𝑥)3 𝑓 ′′′ (𝑥)
𝑃(𝑥+Δ𝑥)− 𝑃(𝑥−Δ𝑥) 2
= 𝑓′(𝑥) + 3 (Δ𝑥)2 𝑓 ′′′ (𝑥)
2Δ𝑥

𝑃(𝑥 + Δ𝑥) − 𝑃(𝑥 − Δ𝑥) 2

𝑓 ′ (𝑥) = − (Δ𝑥)2 𝑓 ′′′ (𝑥)
2Δ𝑥 3
𝑃(𝑥+Δ𝑥)− 𝑃(𝑥−Δ𝑥)
𝑓 ′ (𝑥) = + 0Δ(𝑥)2 where 0Δ(𝑥)2 is the truncation error
2Δ𝑥

For space i is given as:

𝜕𝑝 𝑃𝑖+1,𝑗 −𝑃𝑖,𝑗
𝜕𝑥
= 2Δ𝑥
…………………………………………………………….. (3.56)
 In time it becomes:
𝜕𝑝 𝑃 𝑛+1 −𝑃 𝑛
= …………………………………………………………….. (3.57)
𝜕𝑡 Δ𝑡

The central difference scheme is also used to discretize the left-hand side of the PDE in eqn
(3.26) describing fluid flow in the x- direction and y directions as shown below

𝜕 𝑘𝑘𝑟𝑜 𝜕 𝑘𝑘𝑟𝑜 𝑅𝑠𝑜 𝑘𝑦 𝑘𝑟𝑜 𝐴𝑦 𝑃𝑜 𝑖+1,𝑗 − 𝑃𝑜 𝑖,𝑗

( 𝐴𝑥 𝛻𝑃𝑜 ) ∆𝑥𝑖 + ( 𝐴𝑦 𝛻𝑃𝑜 ) ∆𝑦𝑖 = [ 𝛽𝑐 −
𝜕𝑥 𝜇 𝑜 𝛽𝑜 𝜕𝑦 𝜇 𝑜 𝛽𝑜 𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑥𝑖+1⁄
2

𝑅𝑠𝑜 𝑘𝑦 𝑘𝑟𝑜 𝐴𝑦 𝑃𝑜 𝑖+1,𝑗− 𝑃𝑜 𝑖,𝑗 𝑅𝑠𝑜 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖,𝑗+1 − 𝑃𝑜 𝑖,𝑗

𝛽𝑐 ] + [ 𝛽𝑐 −
𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑥𝑖−1⁄ 𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑦𝑖+1⁄
2 2

𝑅𝑠𝑜 𝑘𝑦 𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖,𝑗+1 − 𝑃𝑜 𝑖,𝑗

𝛽𝑐 ] -----------------------------------------------------(3.58)
𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑦𝑖−1⁄
2

The central difference scheme is also used to discretize the left-hand side of the PDE in eqn
(3.43) describing fluid flow in the x- direction and y directions as shown below

𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝜕𝑃𝑔 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝜕𝑃𝑔

(𝛽𝑐 𝛽 𝜇 −𝛽𝑐 𝛽 𝜇 )∆𝑥𝑖
𝑔 𝑔 𝜕𝑥 𝑔 𝑔 𝜕𝑥
𝜕 𝑖+1⁄2 𝑖−1⁄2
{(𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑥𝑖 = +
𝜕𝑥 Δ𝑥𝑖

𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝜕𝑃𝑜 𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝜕𝑃𝑜

(𝛽𝑐 − 𝛽𝑐 )∆𝑥𝑖
𝛽𝑜 𝜇𝑜 𝜕𝑥 𝑖+1⁄ 𝛽𝑜 𝜇𝑜 𝜕𝑥 𝑖−1⁄
2 2
… … … ………………..(3.59)
Δ𝑥𝑖

𝜕𝑃𝑜
Using central differencing for
𝜕𝑥 𝑖+1⁄
2

𝜕𝑃𝑜 𝑃𝑜 𝑖+1,𝑗 −𝑃𝑜 𝑖,𝑗 𝑃𝑜 𝑖+1,𝑗 − 𝑃𝑜 𝑖,𝑗

= = … … … ……………………………………….(3.60)
𝜕𝑥 𝑖+1⁄ 𝑥𝑖+1 −𝑥𝑖 Δ𝑥𝑖+1⁄
2 2

𝜕𝑃𝑜
Also for
𝜕𝑥 𝑖−1⁄
2
𝜕𝑃𝑜 𝑃𝑜 𝑖,𝑗 −𝑃𝑜 𝑖−1,𝑗 𝑃𝑜 𝑖 −𝑃𝑜 𝑖−1
= = … … … ……………….. ..…………………… (3.61)
𝜕𝑥 𝑖−1⁄ 𝑥𝑖 −𝑥𝑖−1 Δ𝑥𝑖−1⁄
2 2

𝜕𝑃𝑔 𝜕𝑃𝑔
Also for and , we have
𝜕𝑥 𝑖−1⁄ 𝜕𝑥 𝑖+1⁄
2 2
𝜕𝑃𝑔 𝑃𝑔 −𝑃𝑔 𝑃𝑔 −𝑃𝑔
𝑖 𝑖−1 𝑖 𝑖−1
= = --------------------------------------------------------------(3.62)
𝜕𝑥 𝑖−1⁄ 𝑥𝑖 −𝑥𝑖−1 Δ𝑥𝑖−1⁄
2 2

𝜕𝑃𝑔 𝑃𝑔 −𝑃𝑔 𝑃𝑔 − 𝑃𝑔
𝑖+1 𝑖 𝑖+1 𝑖
= = ------------------------------------------------------------(3.63)
𝜕𝑥 𝑖+1⁄ 𝑥𝑖+1 −𝑥𝑖 Δ𝑥𝑖+1⁄
2 2

𝜕𝑃𝑜
For the y direction, we have 𝜕𝑦 𝑖+1⁄
2

𝜕𝑃𝑜 𝑃𝑜 𝑖+1,𝑗 −𝑃𝑜 𝑖,𝑗 𝑃𝑜 𝑖+1,𝑗− 𝑃𝑜 𝑖,𝑗

= = … … … ……………………………………….(3.64)
𝜕𝑦 𝑖+1⁄ 𝑦𝑖+1 −𝑦𝑖 Δ𝑦𝑖+1⁄
2 2

𝜕𝑃𝑜
Also for
𝜕𝑦 𝑖−1⁄
2
𝜕𝑃𝑜 𝑃𝑜 𝑖,𝑗 −𝑃𝑜 𝑖−1,𝑗 𝑃𝑜 𝑖 −𝑃𝑜 𝑖−1
= = … … … ……………….. ..…………………… (3.65)
𝜕𝑦 𝑖−1⁄ 𝑦𝑖 −𝑦𝑖−1 Δ𝑦𝑖−1⁄
2 2

𝜕𝑃𝑔 𝜕𝑃𝑔
Also for and , we have
𝜕𝑦 𝑖−1⁄ 𝜕𝑦 𝑖+1⁄
2 2

𝜕𝑃𝑔 𝑃𝑔 −𝑃𝑔 𝑃𝑔 −𝑃𝑔

𝑖 𝑖−1 𝑖 𝑖−1
= = --------------------------------------------------------------(3.66)
𝜕𝑦 𝑖−1⁄ 𝑦𝑖 −𝑦𝑖−1 Δ𝑦𝑖−1⁄
2 2

𝜕𝑃𝑔 𝑃𝑔 −𝑃𝑔 𝑃𝑔 − 𝑃𝑔
𝑖+1 𝑖 𝑖+1 𝑖
= = ------------------------------------------------------------(3.67)
𝜕𝑦 𝑖+1⁄ 𝑦𝑖+1 −𝑦𝑖 Δ𝑦𝑖+1⁄
2 2

By substituting of eqns (3.60), (3.61), (3.62), (3.63), (3.64), (3.65), (3.66) and (3.67) into eqn
(3.59), we have:

𝜕 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑔 𝑖+1 − 𝑃𝑔 𝑖 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑔 −𝑃𝑔

𝑖 𝑖−1
{(𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑥𝑖 = (𝛽𝑐 − 𝛽𝑐 )+
𝜕𝑥 𝛽𝑔 𝜇𝑔 Δ𝑥𝑖+1⁄ 𝛽𝑔 𝜇𝑔 Δ𝑥𝑖−1⁄
2 𝑖+1⁄2 𝑖−1⁄2 2

𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖 𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖

(𝛽𝑐 − 𝛽𝑐 )------------------------(3.68)
𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑥𝑖+1⁄2 𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑥𝑖−1⁄
2

𝜕
By similar computation, 𝜕𝑦 {𝜆𝑔 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑦𝑖 can thus be expressed as
𝜕 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑔 𝑖,𝑗+1 − 𝑃𝑔𝑖,𝑗
{𝜆 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑦𝑖 = (𝛽𝑐 −
𝜕𝑦 𝑔 𝛽𝑔 𝜇𝑔 Δ𝑦𝑖+1⁄
2 𝑖+1⁄2

𝑘𝑥 𝑘𝑟𝑔 𝐴𝑦 𝑃𝑔 −𝑃𝑔 𝑅𝑠𝑜 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖,𝑗+1− 𝑃𝑜 𝑖,𝑗

𝑖,𝑗 𝑖,𝑗−1
𝛽𝑐 ) + (𝛽𝑐 −
𝛽𝑔 𝜇𝑔 Δ𝑦𝑖−1⁄ 𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑦𝑖+1⁄
𝑖−1⁄2 2 2

𝑅𝑠𝑜 𝑘𝑦 𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖,𝑗−1 − 𝑃𝑜 𝑖

𝛽𝑐 )--------------------------(3.69)
𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑦𝑖−1⁄
2

𝜕 𝜕
Therefore, {𝜆𝑔 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )} ∆𝑥𝑖 + 𝜕𝑦 {𝜆𝑔 (𝛻𝑃𝑔 ) + 𝑅𝑠𝑜 𝜆𝑜 (𝛻𝑃𝑜 )}∆𝑦𝑖 =
𝜕𝑥

𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑔 𝑖+1 − 𝑃𝑔 𝑖 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑔 −𝑃𝑔

𝑖 𝑖−1 𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖
(𝛽𝑐 − 𝛽𝑐 ) + (𝛽𝑐 −
𝛽𝑔 𝜇𝑔 Δ𝑥𝑖+1⁄ 𝛽𝑔 𝜇𝑔 Δ𝑥𝑖−1⁄ 𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑥𝑖+1⁄2
2 𝑖+1⁄2 𝑖−1⁄2 2

𝑅𝑠𝑜 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑥 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑔 𝑖+1 − 𝑃𝑔𝑖 𝑘𝑥 𝑘𝑟𝑔 𝐴𝑦 𝑃𝑔 −𝑃𝑔

𝑖 𝑖−1
𝛽𝑐 ) + (𝛽𝑐 − 𝛽𝑐 )+
𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑥𝑖−1⁄ 𝛽𝑔 𝜇𝑔 Δ𝑦𝑖+1⁄ 𝛽𝑔 𝜇𝑔 Δ𝑦𝑖−1⁄
2 2 𝑖+1⁄2 𝑖−1⁄2 2

𝑅𝑠𝑜 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖 𝑅𝑠𝑜 𝑘𝑦𝑘𝑟𝑔 𝐴𝑦 𝑃𝑜 𝑖+1 − 𝑃𝑜 𝑖

(𝛽𝑐 − 𝛽𝑐 ) --------------------------(3.70)
𝛽𝑜 𝜇𝑜 𝑖+1⁄2 Δ𝑦𝑖+1⁄2 𝛽𝑜 𝜇𝑜 𝑖−1⁄2 Δ𝑦𝑖−1⁄
2

𝐾𝑥 𝑘𝑟𝑜𝐴𝑥 𝐾𝑥 𝑘𝑟𝑜𝐴𝑦 𝐾𝑥 𝑘𝑟𝑔𝐴𝑥

Recall 𝑇𝑜𝑥 = 𝛽𝑐 , 𝑇𝑜𝑦 = 𝛽𝑐 , 𝑇𝑔𝑥 = 𝛽𝑐 and
𝑖±1⁄2,𝑗 𝜇𝑜𝐵𝑜 𝑖±1⁄ ,𝑗 𝑖,𝑗±1⁄2 𝜇𝑜𝐵𝑜 𝑖±1⁄ 𝑖±1⁄2,𝑗 𝜇𝑔𝐵𝑔 𝑖±1⁄
2 2 2
𝐾𝑥 𝑘𝑟𝑔𝐴𝑦
𝑇𝑔𝑦 = 𝛽𝑐
𝑖,𝑗±1⁄2 𝜇𝑔𝐵𝑔 𝑖±1⁄
2

By substituting in eqns (3.26) and (3.43), the combined flow for the oil and gas flow model is
thus:

𝑇𝑜𝑥 (𝑝𝑜𝑖+1,𝑗 − 𝑝𝑜𝑖,𝑗 ) + 𝑇𝑜𝑥 (𝑝𝑜𝑖−1,𝑗 − 𝑝𝑜𝑖,𝑗 ) + 𝑇𝑜𝑦 (𝑝𝑜𝑖,𝑗+1 − 𝑝𝑜𝑖,𝑗 ) +

𝑖+1⁄2,𝑗 𝑖−1⁄2,𝑗 𝑖,𝑗+1/2

𝑇𝑜𝑦 (𝑝𝑜𝑖,𝑗−1 − 𝑝𝑜𝑖,𝑗 ) + 𝑇𝑔𝑥 (𝑝𝑔𝑖+1,𝑗 − 𝑝𝑔𝑖,𝑗 ) + 𝑇𝑔𝑥 (𝑝𝑔𝑖−1,𝑗 − 𝑝𝑔𝑖,𝑗 ) +

𝑖,𝑗−1/2 𝑖+1⁄2,𝑗 𝑖−1⁄2,𝑗

𝑇𝑔𝑦 (𝑝𝑔𝑖,𝑗+1 − 𝑝𝑔𝑖,𝑗 ) + 𝑇𝑔𝑦 (𝑝𝑔𝑖,𝑗−1 − 𝑝𝑔𝑖,𝑗 ) + 𝑅𝑠𝑜 𝑇𝑜𝑥 (𝑝𝑜𝑖+1,𝑗 − 𝑝𝑜𝑖,𝑗 ) +
𝑖,𝑗+1/2 𝑖,𝑗−1/2 𝑖+1⁄2,𝑗

𝑅𝑠𝑜 𝑇𝑜𝑥 (𝑝𝑜𝑖−1,𝑗 − 𝑝𝑜𝑖,𝑗 ) + 𝑅𝑠𝑜 𝑇𝑜𝑦 (𝑝𝑜𝑖,𝑗+1 − 𝑝𝑜𝑖,𝑗 ) + 𝑅𝑠𝑜 𝑇𝑜𝑦 (𝑝𝑜𝑖,𝑗−1 − 𝑝𝑜𝑖,𝑗 ) =
𝑖−1⁄2,𝑗 𝑖,𝑗+1/2 𝑖,𝑗−1/2

𝜕 𝑆 𝜕 𝑆 𝑆
[( ∅ 𝛽𝑜 ) + 𝑞̅𝑜 ] + [𝜕𝑡 (∅ [𝛽𝑔 + 𝑅𝑠𝑜 𝛽𝑜 ]) + 𝑞̅𝑔 ]…………………………….. (3.71)
𝜕𝑡 𝑜 𝑔 𝑜
The terms 𝑇𝑜𝑥 and 𝑇𝑔𝑥 represent the transmissibility between block i and i±1 through
𝑖±1⁄2,𝑗 𝑖±1⁄2,𝑗

block boundary 𝑥𝑖±1 and terms 𝑇𝑜𝑦 and 𝑇𝑔𝑦 represent the transmissibilities between
2 𝑖,𝑗±1⁄2 𝑖,𝑗±1⁄2

block j and j±1 through block boundary 𝑦𝑖±1 expressed in STB/D/psi. Transmissibility is a flow
2

term that denotes the ease of fluid transfer between two blocks.

3.3.1(b) Temporal Discretization

This is the process of discretizing the time derivative at the right hand side of the PDE with respect
to the time variable. Two main method is the implicit formulation ( 𝑡 𝑛+1 ) and the explicit
formulation (𝑡 𝑛 ).

3.4 Computer Model

Developing a simulator involves several stages, with the computer model being the most
challenging of the stages with the mathematical model being the building block upon which the
numerical model would be employed to discretize the already developed PDE from the
mathematical model and then used to generate the computer model. The simulator was created
using python codes. The development of the Python code follows the underlisted steps:

1. Input of the fluid properties, rock properties, simulation and well data.
2. Define the already assumed boundary condition to be used in the problem which is the
Dirichlet Boundary condition which is a simple case scenario.
3. Construct a sparse pentadiagonal matrix, single column matrix B and the J vector for the
assumed boundary condition.
4. Use a linear solver to solve for pressure and saturation
5. Update the pressure and saturation for timestep 300, 600 and 1000 days.
6. Using matplotlib, display a pressure saturation distribution map for each gridblock at the
already mentioned timesteps.
 CHAPTER FOUR
RESULTS AND DISCUSSIONS
4.1 Results
This chapter focuses on the validation of the methodology used in this study. The simulator is able
to handle only the Dirichlet boundary condition as that of Neumann and mixed boundary
conditions are beyond the scope of this project. Also presented in this chapter are a surface contour
plot generated in order to represent a pressure distribution map which provides a detailed
visualization of the pressure distribution across the surface. Armed with this information,
management can make sound final investment decisions on whether to invest in the drilling project
or not and to equally ascertain the best location to drill the well.
Table 4.1 The data set used to validate the model.
Reservoir Length, Lx (ft) 6000
Reservoir Width, Ly (ft) 7500
Reservoir Thickness, H (ft) 50
Total Simulation Time, T (in days) 1000
Porosity 0.18
Reservoir permeability in the x direction, kx 1 * E-06
Reservoir permeability in the y direction ky 0.15 * 1 * E-06
Oil relative permeability 0.0294
Gas relative permeability 0.1120
Oil viscosity, µo (cp) 0.9810
Gas viscosity, µg (cp) 0.0260
Solution gas oil ratio, Rs 818.60
Oil formation volume factor, Bo 1.42596
Gas formation volume factor, Bg 0.00064
Initial saturation of gas, Sg 0.605
Initial reservoir pressure 4500 psi
Oil compressibility, Co (/psi) 21 * E-06
Gas Compressibility, Cg (/psi) 40 * E-06
Fig 4.1: Pressure Distribution Map for Dirichlet Boundary Condition at time 0 day
The above pressure map in fig. 4.0 indicates that all four boundaries are assumed to be Dirichlet
boundary conditions, and the block’s initial map is also set at 4500psi.
Fig. 4.2: Pressure Distribution Map for Dirichlet Boundary Condition at time 300 days
Fig. 4.3: Pressure Distribution Map for Dirichlet Boundary Condition at time 600 days
Fig. 4.4: Pressure Distribution Map for Dirichlet Boundary Condition at time 1000 days

4.2 Discussion
For the Dirichlet boundary condition assumed in the problem definition and presented in the
result, conclusions would be drawn and the pressure distribution pattern would be explained
for the assumed boundary condition.
4.1.1 Effect of the Dirichlet Boundary Condition

A pattern is observed from fig 4.1, 4.2, 4.3 and 4.4 and it is observed that there exists a pressure
decline across the reservoir gridblock progressively for time 300 days, 600 days and 1000 days.
The reservoir is subjected to a constant pressure boundary while the reservoir is maintained at an
initial pressure. Once production starts, a rapid pressure decline occurs. This pressure decline
occurs as a result of the reduction of the volume of oil in the reservoir. Also, the saturation of oil
in the reservoir increases because as the pressure declines, gas comes out of the oil to increase the
saturation of the already existing free gas cap present in the oil and gas reservoir. The original
energy of the reservoir comes from the already existing gas cap as the expansion of the free gas
cap forces more oil out of the reservoir thus making more gas to come out of solution thereby
increasing the saturation of the gaseous phase. If production is continued eventually, a point will
reach when the well would start producing gas instead of oil.

CHAPTER FIVE
 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

5.1 Summary of Findings

The numerical modeling of the multi-phase fluid was done by developing a simulator to
determine the pressure and saturation distributions across the reservoir by using the Implicit
Pressure Explicit Saturation (IMPES) scheme which was incorporated during the development of
the simulator. It was found that pressure and saturation varies across the reservoir as production
starts and continues over time. The pressure of oil actually declines throughout the reservoir
whereas saturation of gas increases throughout the reservoir as the gas comes out of oil thus
increasing the saturation of the gas phase when the reservoir is subjected to a constant pressure
boundary condition.

5.2 Conclusion

The following conclusions were drawn from carrying out this study
1. It has been established that the pressure distribution across a reservoir is affected by
boundary conditions and the Dirichlet boundary condition was used as a case study in this
project. An indepth and precise knowledge of the pressure and saturation distribution
across a reservoir is required in order to assist management to make final investment
decision (FID) on whether or not to embark on the oil and gas project and also on the best
area to drill the well in order to maximize oil recovery and production at the minimum
development costs.
2. The best location will be a zone of the area of high pressure and minimum gas saturation
implying a zone of maximum reservoir energy and oil saturation.
5.3 Recommendations

As regards the nature of this project, the following recommendations were made;

1. Future research should incorporate these additional boundary conditions, as well as the 3D
model, since this one only considers the Dirichlet boundary condition and not the Neumann
or mixed ones.
2. Additionally, alternative numerical models like finite volume and finite element methods
could be used to discretize the partial differential equations controlling the fluid flow
processes, in addition to the Finite Difference model that was used to develop the simulator.
Both regular and irregular grids would be adequately represented by these models, in my
firm opinion..

5.4 Contributions to Knowledge

Based on the findings of this study, the optimal drilling zone to achieve maximum oil recovery
from reservoirs with minimal development costs is in zones with low gas saturation and high
pressure.